Modelling Natural Resource Responses to Climate Change (MONARCH)
A Local Approach
Development of a conceptual and methodological framework for universal application
Edited by P.M. Berry, P.A. Harrison T.P. Dawson and C.A. Walmsley ii
This report should be referenced as:
Berry, P.M., Harrison, P.A., Dawson, T.P. and Walmsley, C.A. (Eds.) (2005). Modelling Natural Resource Responses to Climate Change (MONARCH): A Local Approach. UKCIP Technical Report, Oxford. iii
CONTENTS
Executive summary...... vii
Chapters:
1. Introduction, project aims and methods ...... 1 P.M Berry, P.A. Harrison, J.E. Hossell and G.J. Masters 1.1 Introduction...... 1 1.2 Aims...... 2 1.3 Methods...... 2 1.4 Project partners ...... 8 1.5 References...... 9
2. Bioclimatic classification and case study selection...... 13 J.E. Hossell, A.E. Riding and P.A. Harrison Summary...... 13 2.1 Introduction...... 13 2.2 Bioclimatic classification method...... 14 2.3 Bioclimatic classification results...... 14 2.4 Conservation characterisation...... 19 2.5 Identification of climatically sensitive areas...... 22 2.6 Selected case study areas ...... 27 2.7 Selection of habitats and species...... 27 2.8 Bioclimatic classification of the case study areas...... 28 2.9 Uncertainty in future climate change projections ...... 30 2.10 Discussion and conclusions ...... 36 2.11 References...... 36 Chapter 2 Annex...... 37
3. Species distribution and dispersal modelling ...... 43 P.A. Harrison, R.G. Pearson, T.P. Dawson, S. Freeman, J.E. Hossell, H. Lyons, P. Scholefield and P.M. Berry Summary...... 43 3.1 Introduction...... 43 3.2 SPECIES model...... 44 3.3 Downscaled SPECIES model ...... 46 3.4 Land cover change scenarios ...... 49 3.5 Dispersal model ...... 52 3.6 Discussion and conclusions ...... 56 3.6 References...... 59 Chapter 3 Annex...... 63 iv
4. Implications for the Composition of Species Communities...... 68 G. Masters and N. Ward Summary...... 68 4.1 Introduction...... 68 4.2 The species Arriver and Leaver conceptual models ...... 68 4.3 The application of the Arriver and Leaver conceptual models ...... 72 4.4 Discussion...... 73 4.5 References...... 73
5. Impacts on coastal environments...... 74 G.E. Austin and M.M. Rehfisch Summary...... 74 5.1 Introduction...... 74 5.2 Methods...... 75 5.3 Suffolk coastline case study...... 85 5.4 Discussion and conclusions ...... 96 5.5 References...... 98
6. Impacts for the Hampshire case study area ...... 102 J.E. Hossell, G.E. Austin, P.M. Berry, N. Butt, H.Q.P. Crick, S. Freeman, P.A. Harrison, G.J. Masters, A. Morrison, M.M. Rehfisch, P. Scholefield, N. Ward And I. Wilde Summary...... 102 6.1 Introduction...... 103 6.2 Bioclimatic classification...... 103 6.3 Land cover changes...... 107 6.4 Beech hangers ...... 110 6.5 Wet heathland...... 128 6.6 Estuarine waterbirds...... 140 6.7 Discussion and conclusions ...... 144 6.8 References...... 145
7. Impacts for the Central Highlands case study area ...... 148 P.M. Berry, N. Butt, H.P.Q. Crick, S. Freeman, P.A. Harrison, J.E. Hossell, G. Masters, P. Scholefield and N. Ward Summary...... 148 7.1 Introduction...... 149 7.2 Bioclimatic classification...... 149 7.3 Land cover changes...... 159 7.4 Caledonian pine woodland...... 159 7.5 Montane/upland heath...... 173 7.6 Discussion and conclusions ...... 185 7.7 References...... 187
v
8. Impacts for the Snowdonia case study area...... 189 G.J. Masters, P.M. Berry, J.E. Hossell, N. Ward, S.N. Freeman, A.N. Banks, N. Butt, H.Q.P. Crick, P.A. Harrison and A. Morrison Summary...... 189 8.1 Introduction...... 190 8.2 Bioclimatic classification...... 191 8.3 Land cover changes...... 195 8.4 Conservation monitoring in Snowdonia ...... 196 8.5 Upland oak woodland...... 198 8.6 Montane/upland heath...... 212 8.7 Discussion and conclusions ...... 230 8.8 References...... 231
9. Impacts for the Cuilcagh/Pettigo Peatlands case study area ...... 237 P.M. Berry, G.M. Masters, J.E. Hossell, N. Butt, S. Freeman, P.A. Harrison and N. Ward Summary...... 237 9.1 Introduction...... 237 9.2 Bioclimatic classification...... 238 9.3 Land cover changes...... 239 9.4 Species and dispersal modelling ...... 240 9.5 Implications for species community composition...... 246 9.6 Discussion and conclusions ...... 249 9.7 References...... 249
10. Conclusions and future research...... 251 P.M. Berry, G.J. Masters, P.A. Harrison, J.E. Hossell, H.Q.P. Crick, S.N. Freeman, N.E. Ellis, H.G. Orr, C.A. Walmsley and O. Watts 10.1 Introduction...... 251 10.2 Methodological developments...... 251 10.3 Key impacts on species in the case study areas...... 252 10.4 Uncertainties and limitations of the models and data inadequacies ...... 255 10.5 The way forward ...... 260 10.6 Conclusions...... 262 10.7 References ...... 263
MONARCH 2 Report – Executive Summary vii ______
Executive Summary
Introduction to the MONARCH project
MONARCH (Modelling Natural Resource Responses To Climate Change) is a phased investigation into the impacts of climate change on the nature conservation resources of Britain and Ireland. It is an important step towards understanding the complex interactions between climate change, species and habitats. The first phase of MONARCH developed a modelling technique that predicts areas of suitable climate for species at a 10-kilometre square resolution, and provides a guide to the future distribution of a species within Britain and Ireland. MONARCH 2 sought to develop this approach at the local and regional scale, downscaling the model to a 1-kilometre square resolution within areas of up to 2,500 square kilometres. The original species climate-space modelling approach was expanded to consider the role of land cover in influencing species’ distributions, the ability of species to disperse in response to climatic and land cover changes, and also the potential effect of these responses on the structure and function of ecosystems.
This downscaled and expanded methodology was tested within case study areas chosen for their sensitivity to climate change, importance for nature conservation, geographical spread across Britain and Ireland and data availability. The four case study areas selected were Hampshire, Central Highlands, Snowdonia and a cross-border area around Cuilcagh/Pettigo in Ireland.
The MONARCH 2 methodology
The MONARCH 2 methodology followed five stages: 1. Relevant data sets on observed species’ distributions, climate, soil, land cover and statutory conservation sites were obtained at multiple scales and incorporated into a geographical information system (GIS). 2. Four case study areas were selected for testing models of species’ distribution and dispersal at the local scale. This involved the identification of climatically sensitive areas using the bioclimatic classification developed in MONARCH 1 and the identification of key habitats and species of conservation interest within them. 3. Four or five species within each of two key habitats were selected per case study in conjunction with local stakeholders (except the Irish study which concentrated solely on peatland). The species selection process ensured the inclusion of a range of interacting dominant, rare and recruitment species. 4. Models for simulating the response of species’ distributions and dispersal to climate and land cover change at the landscape scale were developed for application in terrestrial environments. For the coastal environment, the models that predict the effect of habitat change resulting from sea-level rise (previously developed under MONARCH 1) and those that predict the effect of winter weather on the redistribution of waterbirds were integrated. 5. The species distribution models were run under a range of future climate and land cover change scenarios. Species dispersal and community impact models were developed to assess species’ movements into and out of the selected habitats and to determine potential implications for species composition of the communities.
Four climate change scenarios for Britain and Ireland have been developed for three time slices on behalf of the UK Climate Impacts Programme (UKCIP), the latest, published in 2002, known as the UKCIP02 scenarios. In MONARCH 1 the UKCIP98 scenarios were used as an input to the modelling. In MONARCH 2 the more recent UKCIP02 scenarios have been adopted, focusing on the
viii MONARCH 2 Report – Executive Summary ______scenarios for the 2020s and 2050s. The Low and High emissions scenarios were used in MONARCH 2 to capture the range of potential outcomes of future greenhouse gas emissions on British and Irish climate. The UKCIP02 scenarios are based on a single climate model, the Hadley Centre Regional Climate Model (HadRM3).
Bioclimatic classification
The bioclimatic classification, originally developed in MONARCH 1, was revised using the UKCIP02 2050s High scenario to highlight geographical areas where the bioclimate would change most by 2050. These areas sensitive to climate change were combined with data on areas of nature conservation importance to select four case study areas for use in testing the MONARCH 2 methodology. The bioclimatic classification results showed that:
• Within Britain and Ireland, 42% of the land area was predicted to have a future bioclimate unlike any current bioclimate classification category for Britain or Ireland. In Britain, Snowdonia, Sussex, Kent, the South Essex coast, Lanarkshire, mid-west coastal Scotland and the western Grampians showed the greatest difference in bioclimate between the baseline classification and the 2050s High climate change scenario. In Ireland, Fermanagh, Galway and Connemara showed patterns of bioclimate that differed greatly from that presently experienced across the region. This culminated in the selection of four case study areas in regions where bioclimate was predicted to be outside the current bioclimate classification and where there were habitats of high conservation interest. • In each case study area, the effects of the 2050s Low and High emissions scenarios were examined using the 5-kilometre square UKCIP02 data. Classification at this finer resolution highlighted differences between the two UKCIP baseline climate datasets that were particularly apparent in the upland case study areas, in that a number of the squares classified using the UKCIP02 baseline, did not fit into the UKCIP98 baseline classification. These differences may reflect the lower number of stations and different interpolation routine used in the UKCIP98 dataset.
Species’ distribution and dispersal modelling
Four models were tested in each of the four case study areas to assess the impacts of climate and land cover change on species’ distributions. 1. The SPECIES model, developed in MONARCH 1, was used to simulate changes in the potential climate space of species at a 5-kilometre square resolution and provide a guide to future species distribution at the Britain and Ireland scale. 2. The SPECIES model was developed further in MONARCH 2 for use within the case study areas. It was downscaled to operate at a 1-kilometre square resolution and forecast changes in a species’ climate space (and suitable land cover) at the local to regional scale. 3. A land cover model was developed to create scenarios of land cover change consistent with the UKCIP02 climate change scenarios to refine the species distribution forecasts. 4. A species dispersal model was developed to investigate the ability of species to track changes in their potential distributions under the climate and land cover change scenarios.
MONARCH 2 Report – Executive Summary ix ______
Results from testing the models in the four case study areas showed that:
• The climatic sensitivity of the case study areas was not always reflected in the predictions from the SPECIES model. This may be because the dominant and relatively robust species selected to characterise the habitats are widespread with no range margins in or near the case study areas. • While widespread and dominant species are predicted to continue to find suitable climate space in the future within the case study areas, losses in climate space for some dominant species are forecast for East Anglia and Central England. The northern range margins of species were projected to move northwards, while southern margins retreated in response to climate change. • The addition of land cover in the downscaled SPECIES model improved the simulation of current species’ distributions at the national scale generally, by constraining the simulated current suitable climate space. This also caused the response of the species’ distribution to climate change to be suppressed in some predictions when compared with outputs from the original SPECIES model. • The land cover change model was most successful in simulating the current distributions of particular land cover classes where climate and soil conditions provided strong restrictions on their locations (e.g. peat or calcareous based land covers). Results for land covers, such as broad-leaved mixed woodland and inland marshes in Ireland, where their presence reflects the strong influence of management on their current locations, were generally poor. • The dispersal model successfully simulated a range of dispersal abilities for species, with trees showing little expansion due to their long time to reproductive maturity, while species with long distance dispersal capabilities and high mean and/or maximum dispersal distance showed the greatest rates of expansion. However, there is a shortfall on accurate data required for the dispersal model and the effect of this upon model predictions has not been assessed.
• Generally, species showed limited dispersal rates that are inadequate to match the predicted changes in their suitable space. A few species, however, may be able to track changing climate within the case study areas.
Modelling the implications for the composition of species communities
The SPECIES and dispersal modelling identified which species would potentially move into or out of a habitat, thus providing the inputs into the Arriver and Leaver conceptual models. Results from testing these two models in the four case study areas showed that:
• The models provided a mechanism for assessing the impact of species’ movement on ecosystem structure and function. For example, hawfinch (Coccothraustes coccothraustes) may decline in Hampshire due to loss of suitable environmental space. However, due to niche overlap with other species, a decline or loss of hawfinch from the beech hanger habitat may have only a small impact on the species composition and structure of this habitat. In contrast, yellow- necked mouse (Apodemus flavicollis) is predicted to become more widespread within the beech hangers and more widely in Hampshire by colonising new woodlands. • In situations where species already present in a habitat gain climate space in a case study area, there could be a change in their position in the community. This is illustrated by bracken (Pteridium aquilinum) and western gorse (Ulex gallii) in the montane/upland heath habitat of Snowdonia. The downscaled SPECIES model predicts that these species may spread and threaten to occupy habitat at higher altitudes where montane species of nature conservation interest are currently important.
x MONARCH 2 Report – Executive Summary ______
Waterbirds and coastal habitats
Two models were used to predict the effects of climate change and sea level rise on the distribution and numbers of over-wintering waterbirds in Britain and Ireland. The modelling approach was assessed using the estuaries of the Suffolk coastline and then applied in the Hampshire case study area. Results showed that:
• Average minimum temperatures on the muddy estuaries of the east coast were found to explain a proportion of the variation in the distribution of six of the seven species considered. This supports the hypothesis that changes in weather patterns over the past three decades have resulted in a broad scale redistribution of waders over-wintering in Britain from western coastal estuaries to those on the east coast. • Sea-level rise was simulated to have little impact on waterbird numbers in the estuaries studied because the coast is protected by hard defences, which are likely to be maintained. Thus, there will be no substantial re-alignment of the estuary extents and consequentially the values of the morphological variables used by those models were largely unchanged under future scenarios. • The models examining the impacts of sea level rises and changes in winter temperatures (under the various UKCIP02 scenarios) showed that the estuaries of the Suffolk and Hampshire coastlines should retain the capacity to hold numbers of waders similar to those at present.
Model limitations and assumptions
MONARCH 2 models have been developed to help guide those involved in nature conservation policy and practice to appreciate the potential effects of climate change on species and their communities. Nonetheless, model predictions should be interpreted with due caution and should be viewed as first approximations indicating the potential magnitude and broad pattern of future impacts, rather than as accurate simulations of future species’ distributions. There are important limitations to the predictive capacity of the simulation models used in MONARCH 2 including:
• The MONARCH 2 modelling framework has been designed to take best advantage of the available data at different spatial resolutions. However, the project identified some serious shortfalls in available data of observed species’ distributions for Europe and the case study areas. Base errors arising from data limitations are therefore unavoidable. Data reliability was also an issue with the Land Cover Map 2000, while data compatibility issues occurred in the British and Irish climate and soils data, as well as between the UKCIP 1998 and 2002 scenarios, and the Land Cover Map 2000 and Phase 1 Habitat Survey. • The MONARCH 2 methodology also assumed that land cover is a suitable surrogate for habitat availability. The model of land cover change only took account of the impact of climatic factors on the distribution of land cover classes. It did not consider any economic or social drivers that may affect the location or management of such land cover types. It also assumed that land cover types excluded from the analysis are not significantly affected by climate change. The land cover change model and the bioclimatic classification were developed using data for only Britain and Ireland. There is a need to extend the range of climatic conditions over which both models were trained using European data in order to capture the full range of conditions that may exist in the future. • Biotic interactions between species, such as competition, predation and symbiosis with other species, were not considered in the models of species’ distribution. Applying these models at the Britain and Ireland scale, where climatic influences on species’ distributions are shown to be dominant, can minimize the impact of biotic interactions.
MONARCH 2 Report – Executive Summary xi ______
• The models only examine the impacts of averaged climatic changes. The increased frequency and magnitude of extreme events such as droughts or storms may have significant impacts on species’ distributions that MONARCH is unable to fully take into account.
• Predicting adaptive changes by species in response to climate change has not been accounted for within the MONARCH 2 modelling framework. Thus, model predictions can be viewed with greater confidence for species not expected to undergo significant evolutionary change over the next 50 years. • The Arriver/Leaver models are currently limited in that they only assess changes in species composition and are often data limited. This species’ interaction approach is not a direct measure of ecosystem function, like productivity, but it can inform on possible shifts or changes of community and ecosystem function.
Case study areas
The conclusions drawn from the modelling outputs for the four case study areas are summarised below. These should be interpreted within context of the limitations and assumptions of modelling at this local scale that MONARCH 2 has identified.
Hampshire
1. Hampshire is predicted to experience bioclimatic conditions more akin to those of continental Europe by the 2050s.
2. Despite this climatic change, the modelling suggests that the selected wet heath and beech hanger species will show little response to climate change. Suitable climate space is predicted to remain in Hampshire for the modelled Beech Hanger and wet heathland plants, although there is likely to be some loss for hawfinch and curlew (Numenius arquata).
3. In terms of land cover, grassland shows a large reduction in extent across southern England, including Hampshire under future climate change scenarios.
4. The dispersal model shows a limited extension in range for beech (Fagus sylvatica) and ash (Fraxinus excelsior) due to a combination of the relatively short timescale considered and their long time to reproductive maturity. Dog’s mercury (Mercurialis perennis), yellow-necked mouse and the wet heathland plants, however, seem able to expand their distribution.
5. On the estuaries, the models show little change in oystercatcher (Haematopus ostralegus) numbers, but if the number of redshank (Tringa totanus) wintering in the UK increases, then populations in Hampshire are likely to grow.
Central Highlands
1. The bioclimatic classification suggests that by the 2050s there will be a significant shift in pattern of climate across the region, with a larger area associated with upland instead of montane climatic conditions.
2. The land cover change scenarios reflected the climatic changes, predicting a significant decrease in suitable climate space for montane habitat under the 2050s High scenario, and also a total loss of suitable climate space for dwarf shrub land cover in the study area. In contrast, there may be a large increase in the area suitable for neutral grassland.
xii MONARCH 2 Report – Executive Summary ______
3. The SPECIES model indicated no change in extent for many of the Caledonian pine woodland and upland/montane heath species. Therefore, dominant species are expected to remain within the Black Wood of Rannoch. When land cover change was incorporated for stiff sedge (Carex bigelowii), then the loss of montane habitat led to a decrease in suitable area.
4. The trees showed limited dispersal, although silver birch (Betula pendula) had the potential to disperse greater distances due to its shorter time to reproductive maturity. The hairy wood ant (Formica lugubris), which is currently rare in the case study area, also apparently has considerable potential for dispersal under future climate.
5. Climate and disturbance induced changes affecting the canopy and ground-flora will have feedbacks that may alter the composition and structure of the woodland community.
Snowdonia
1. The bioclimatic classification identified Snowdonia as one of the areas where the climate is predicted to show greatest difference between the baseline bioclimate and the future 2050s High climate.
2. The land cover model forecasts that suitable climate space for bog and deep peat, fen, marsh and swamp will increase by the 2050s, while that for open and dense dwarf shrub and neutral and acid grassland will show small losses in their extent.
3. Sessile oak (Quercus petraea) and the selected montane/upland heath species are not expected to lose climate space within Snowdonia, but both bluebell (Hyacinthoides non-scripta) and pied flycatcher (Ficedula hypoleuca) become close to the range margin and could do so.
4. The dispersal model indicates that the upland oak woodland plants show limited opportunity for dispersal, while the montane/upland heath species could disperse more widely.
5. If bracken (Pteridium aquilinum) and western gorse (Ulex gallii) colonise higher altitudes, as the modelling work suggests, they will have significant impacts on the composition of upland heath, while the upland heath dominant heather (Calluna vulgaris) may colonise and dramatically alter the composition of the montane community.
Cuilcagh/Pettigo peatlands
1. The Cuilcagh/Pettigo peatlands are predicted to experience little bioclimatic change by the 2050s, although some of the area would become warmer.
2. This is likely to limit change in land cover with the current dominant species likely to remain so.
3. The climate change scenarios predict that Ireland will become somewhat less suitable for all the modelled species, apart from white beak-sedge (Rhynchospora alba) and bracken.
4. The addition of Corine land cover in the downscaled model did not provide a good match between the simulated climate space and current distributions of species and so the dispersal model was not run.
MONARCH 2 Report – Executive Summary xiii ______
Next phase of research (MONARCH 3)
MONARCH 2 has developed and tested the MONARCH 1 SPECIES model at a finer resolution, added ancillary components of land cover, dispersal and community impact, and applied these to four case study areas. However, limitations in the downscaled procedures, particularly data limitations, preclude their widespread application. The third and final phase of MONARCH development will refine, speed up and improve the interpretative potential of MONARCH modelling as a universally applicable tool for assessing changes in species’ distribution due to climate change for Britain and Ireland. MONARCH 3 will therefore be implemented at the 5-kilometre square resolution and take experience gained from MONARCH 2 forward as follows: • The SPECIES model will be automated, so that a greater range of species can be investigated, both to inform further our understanding of potential climate change impacts and to contribute to the Biodiversity Action Plan review process. • A less complex approach to the integration of land cover constraints to species’ distributions, utilizing simple land-cover ‘masks’ will be integrated with the automated SPECIES model. • The model will be validated through hindcasting using good long-term datasets for twelve species. • The range of uncertainty relating to the various emission scenarios, the choice of global climate model and natural climatic variation will be examined to assess their impacts on the MONARCH outputs using six species.
Conclusions
MONARCH 2 has sought to advance the science and understanding of the potential impacts of climate change on biodiversity by downscaling the resolution of the modelled climate and land cover suitability surfaces to the 1-kilometre square scale. It has also extended the scope of modelling to consider the role of land cover, the potential for dispersal and impacts on community structure. The original objective was to predict likely changes to species distributions at a (1-kilometre) landscape- scale, because it is more appropriate to land managers seeking to consider climate change. However, MONARCH 2 has discovered that reducing the scale of application of climate space and associated modelling to a local or regional level requires data, in particular species distribution data, that is frequently unavailable or of an inadequate quality to enable the models to be run effectively. In addition, many other factors, such as competitive interactions between species, are important at a local scale but they are not readily integrated into the models, while the inclusion of land cover appears to have masked the impact of climate. In essence, MONARCH 2 tested the capacity for downscaling the climatic envelope modelling approach and concluded that, owing to limitations at the finer scale, it was more appropriate to provide generic indications of likely changes within exemplar species for a full range of taxa at the Britain and Ireland scale, rather than further develop the downscaled approach. MONARCH 3 will therefore refine the modelling at the Britain and Ireland scale as in the original MONARCH model, but with added refinements and a faster procedure that will enable a better assessment of the potential changes in species distributions across Britain and Ireland under a changing climate.
MONARCH 2 Report – Chapter 1 1 ______1 Introduction, project aims and methods
P.M BERRY, P.A. HARRISON, J. E. HOSSELL and G.J. MASTERS
1.1 Introduction
Climate change has continued to be high on the global research agenda and the most recent Intergovernmental Panel on Climate Change scenarios suggest that by 2100 global temperatures could increase by between 1.4 and 5.8oC relative to 1990 (IPCC, 2001a). These increases could have a significant effect on biodiversity at the global scale (e.g. IPCC, 2001b; McCarty, 2001; Thomas et al., 2004) and much observational evidence suggests that species are already being affected by climate change, as well as by other human activities (e.g., Walther et al, 2002; Parmesan and Yohe, 2003). These changes also will affect the ability of nations to fulfil their obligations under the UN Convention on Biological Diversity (CBD), and in 2002 the Conference of Parties to the CBD affirmed the need for enhanced co-operation between the United Nations Framework Convention on Climate Change (UNFCCC), the CBD and the United Nations Convention on Combating Desertification. An Ad Hoc Technical Expert Group was set up to examine interlinkages between biodiversity and climate change mitigation and adaptation. This group used MONARCH 1 (Harrison et al., 2001) as one of the case studies illustrating the need for harmonization of climate change mitigation and adaptation activities, and biodiversity considerations (Secretariat of the CBD, 2003). The need to incorporate climate change into conservation strategy is being increasingly recognized at the global level, as well as nationally and more regionally, but adaptation needs to be based on sound scientific knowledge, such as the MONARCH project aims to provide.
The projected level of global climate changes, and its impacts, are not uniform, but vary regionally. Even within Britain and Ireland there is still significant regional variation such that under some scenarios there may be a difference of 2.5oC in mean temperature changes and 40% in precipitation (Hulme, et al, 2002). MONARCH 1 had provided indicative results of the impacts of climate change on selected species and habitats in Britain and Ireland, but in order for these to inform conservation management and policy at a more local scale, there was a need for the outputs to be downscaled. The development of the Hadley Centre regional climate model (HadRM3), which was used to produce the UKCIP02 climate change scenarios at a 50 km resolution, and which has been downscaled to a 5km resolution, enabled further advancement of the impact modelling research in MONARCH 2. There was also an interest to explore the inclusion of other factors and processes, such as habitat availability, dispersal and ecosystem functioning, in order to provide a more integrated approach.
Since the completion of MONARCH 1, a number of sectoral and sub-UK regional studies have been completed under the umbrella of the UK Climate Impacts Programme (UKCIP). The potential impacts of climate change have been explored for agriculture, looking not just at impacts but also at the need to adapt and exploit opportunities (IWE, 2002), gardens (Bisgrove and Hadley, 2002), grassland ecosystems, and water demand (Downing et al., 2003). The first marine study covering all of Britain and Ireland, MarClim, is important because it is using long-term (50 year) datasets to examine the response of inter-tidal species to changes in climate and it will use the UKCIP02 scenarios to investigate what this means for the future distributions of marine species. The Engineering and Physical Sciences Research Council (EPSRC) and UKCIP also have been promoting research into climate impacts on the built environment, including transport and utilities, given the long time-scale of the infrastructure involved (UKCIP and EPSRC, 2003). A generic climate change impacts risk assessment methodology also has been developed to aid decision making, in the context of climate change (Willows and Connell, 2003).
In Britain and Ireland, recent climate change impacts research of biodiversity interest has included amphibians (Beebee, 2002) and plant phenology (Fitter and Fitter, 2002), butterflies (Hill et al., 2002) moths (Conrad et al., 2002), other insect herbivores (Masters et al., 1998; Masters and Brown, 2001), and range changes of birds (Wilson et al., 2002). The implication of climate change impacts for nature
2 MONARCH 2 Report – Chapter 1 ______reserves in Great Britain and their future viability also has been explored (Dockerty et al., 2003). Other work of relevance to the climate change debate is the impacts of the 1995 drought on insects and plants (Buckland et al., 1997; Morecroft et al., 2002).
1.2 Aims
The MONARCH project is a phased investigation into the impacts of climate change on the natural conservation resources of the British Isles. It uses an integrated methodology linking established impact models to coherent climatological classifications (Harrison et al., 2001a) and provides a valuable framework for studying the response of the key biodiversity elements to human-induced climate change. The first phase of the MONARCH project focussed on understanding broad-scale responses at the international (Britain and Ireland) scale. However, a downscaled MONARCH approach was sought to investigate the impacts of climate change in areas of conservation importance at the landscape level and so help inform the development of climate change adaptation policy and practice at this level.
Thus, the second phase of the MONARCH project aimed to advance methodologies at fine spatial resolutions in order to capture processes and interactions that are specific to ecosystems at the local landscape, scale, namely:
• land use change and interactions with species’ distributions, • species’ dispersal in relation to fragmentation and loss of habitats, • integration of the above to provide integrated management approaches to conservation.
The first research module of Phase 2 (MONARCH 2.1) focused on the development of a conceptual and methodological framework for universal application anywhere in Britain and Ireland.
The scientific objectives of MONARCH 2.1 were to:
1. Define criteria for the selection of case study areas based on data availability and site sensitivity criteria. 2. Predict future land use changes. 3. Advance models for simulating the potential climate and suitability space of species at the regional scale. 4. Integrate models of species dispersal with predictions of altered species’ suitability space to provide scenarios for species distributions. 5. Develop a conceptual framework of the implications of species dispersal and redistribution for the composition of species communities.
The second module (MONARCH 2.2) aimed to test the methodology within the case study areas identified in MONARCH 2.1. The areas chosen were Hampshire, Central Highlands, Snowdonia National Park and Cuilcagh/Pettigo, a cross-border area in Ireland.
1.3 Methods
An overview of the conceptual and methodological framework is presented in Figure 1. The foundation for each case study was the creation of a GIS database at a fine spatial resolution containing all necessary environmental data sets, observed species’ distributions and climate change scenarios for selected time steps into the future. The GIS database provided the inputs to a set of integrated models of species’ climate space, species’ dispersal and land use/land cover change options. Models for simulating the response of species’ distributions and dispersal to climate and land cover change at the landscape scale were developed for application in terrestrial environments. For the coastal environment, models that predict the effect of habitat change resulting from sea-level rise
MONARCH 2 Report – Chapter 1 3 ______
(previously developed under MONARCH 1) and models that predict the effect of winter weather on the redistribution of waterbirds were integrated. The impacts models were run with a range of future climate and land cover change scenarios. Model results indicating species’ movements into and out of the selected habitats were used to inform conceptual models of community composition. Four case study areas were selected for testing models of species’ climate space and dispersal at the landscape scale. This involved the identification of climatically sensitive areas using the bioclimatic classification developed in MONARCH 1 and their characterisation in terms of conservation attributes. Two habitats and four or five species for each habitat were selected per case study area in conjunction with local stakeholders. Selection was constrained by a protocol that ensured a range of interacting dominant, rare, flagship and recruitment species were chosen.
Figure 1: Conceptual and methodological framework for MONARCH 2
GIS database
Current environmental Species’ Climate scenarios at selected datasets distributions future time steps
LAND USE SPECIES DISPERSAL /COVER MODELS MODELS CHANGES
Potential area of Potential area of Potential area of suitable climate achievable migration suitable habitat
Potential future species’ distribution
Species Composition
Assessment of adaptive responses
4 MONARCH 2 Report – Chapter 1 ______
1.3.1 Data
1.3.1.1 Climatic data
Baseline climate
For Great Britain, monthly climate data for all years from 1961 to 2000 were available for 26 variables at a 5km x 5km spatial resolution from UKCIP (Hulme et al., 2002). Period-mean data from 1961-90 was used within MONARCH 2 to ensure continuity with previous studies. Six variables were directly extracted from the UKCIP baseline climatology: mean, minimum and maximum temperature, precipitation, sunshine hours and wind speed. For wind speed, data were only available for the years 1969 to 2000, so a period-mean for 1969-1990 was computed.
For Ireland, monthly climate data for all years from 1961 to 2000 were available for eight variables at a 5km x 5km spatial resolution from the British-Irish Council (Jenkins et al., 2003). Individual years from 1961 to 1990 were directly extracted from the BIC database and used to compute a period-mean climatology for the same six variables.
Monthly values of solar radiation were derived from the sunshine hours data using the method of Rietveld (1978). This method is based on Prescott’s (1940) modification of the original Ångström (1924) formula relating solar radiation to sunshine hours, astronomical daylength, the amount of short-wave radiation received at the top of the atmosphere and two coefficients describing the percentage of short-wave radiation reaching the earth’s surface or being absorbed by clouds on a completely cloud covered day.
Monthly values of potential evapotranspiration were calculated using the method of Penman (1948). The Penman equation is utilised in many biological models and has been used widely throughout the world with generally satisfactory results (FAO, 1986). It consists of two terms: the energy (radiation) term and the aerodynamic (wind and humidity) term. The energy term describes the relationship between evaporation rates and the net flux of radiant energy at the surface. The aerodynamic term describes the external aerodynamic resistance between the surface (open water or vegetative canopy) and the air, assuming that evaporation is proportional to the vapour pressure deficit of the air and an empirical function of wind speed. The method was originally calibrated for southeast England, but it has proved to be successful over a wide range of climatic regimes (Jones et al., 1990). The method is physically-based (although it contains an empirical wind function) and, therefore, provides realistic responses to climate change.
For the coastal waterbird modelling, five weather data elements (monthly averages of mean minimum temperature, average precipitation, average wind speed, days of frost, days of snow/sleet) were required to calculate monthly regional averages. UKCIP02 baseline data were used to derive weather variables for Britain and Irish data were supplied by the Irish Meteorological Office.
Climate change scenarios
Climate change scenarios (UKCIP02) were available at a 5km x 5km spatial resolution from UKCIP (Hulme et al., 2002). The scenarios are based on the Hadley Centre high resolution regional climate model (HadRM3). Four scenarios have been constructed from HadRM3 which reflect differences in greenhouse gas emissions (Low emissions, Medium-low emissions, Medium-high emission and High emissions). These are based on the IPCC SRES emission scenarios (B1, B2, A2 and A1FI, respectively). The UKCIP02 scenarios differ from UKCIP98 in that they use the same climate sensitivity (HadCM3 = 3oC) for all scenarios, rather than scaling the high and low scenarios to upper and lower limits. It should also be noted that the UKCIP02 scenarios are based on a single climate model and, hence, do not cover any scientific or modelling uncertainties. This is discussed further in Chapter 2.
MONARCH 2 Report – Chapter 1 5 ______
Scenarios have been created for three 30-year periods centred on the 2020s (2011 to 2040), the 2050s (2041-2070) and the 2080s (2071-2100). In MONARCH2, the Low and High emissions scenarios were utilised for the 2020s and 2050s time-slices. These show an annual warming rate of about 0.1 to 0.3oC per decade for the Low emissions scenario, and about 0.3 to 0.5oC per decade for the High emissions scenario, depending on the region (Hulme et al., 2002). There is greater summer warming in the southeast than the northwest and greater warming in the summer and autumn than in winter and spring. Little change (or a slight drying) is projected for annual total precipitation changes. However, winters are wetter by 5 to 15% for the Low emissions scenarios, and by more than 30% for the High emissions scenario, whilst summers are drier by up to 50% by the 2080s for some regions and scenarios.
The daily weather datasets generated by the LARS weather generator have also been produced for four of the UKCIP02 scenarios (Low and High emissions for the 2020s and 2050s time-slices). Projected changes in mean temperature and precipitation and wet and dry spell length from HadRM3 were used to adjust the interpolated parameter files of LARS-WG according to the UKCIP02 scenarios and generate 50 years of daily weather for each time slice and for each 5 km grid cell (Semenov and Barrow, 1997).
1.3.1.2 Topography data
The 1 km x 1 km GTOPO elevation data was used across all of Britain and Ireland. Although the 50m Ordnance Survey elevation dataset for Great Britain was held by UKCIP and was available for local analysis of case studies in lowland areas, it was not used, as this level of detail was not required.
1.3.1.3 Soils data
Data for the available water-holding capacity (AWC) of the soil were obtained for England and Wales from the Soil Survey and Land Research Centre (SSLRC) on a 5 km x 5 km grid via UKCIP. The Macaulay Land Use Research Institute (MLURI) AWC data for Scotland were only available at the 10 km resolution. This spatial resolution was artificially increased to 5 km x 5km to match the resolution of the climate data and other soils data sets by simply replicating the AWC value for a 10km2 grid cell in the four 5km2 grid cells it encompassed. The required soils variables did not exist for all of Ireland, so, in MONARCH 1, a soil map containing information on soil names from the General Soil Map of Ireland (Gardiner and Radford, 1980) was digitised and a conversion routine was developed to estimate values of AWC using a 10 km x 10 km grid (Harrison et al., 2001b). This process was repeated in MONARCH 2 to estimate values of AWC on a 5km x 5km grid based on the original digitised soil polygons.
For the land cover modelling, the National Soil Map at a 1km x 1km resolution was used to estimate AWC for bare soil at 30cm depth. This data was then summarised to 10km2 cells based on the average of the available water capacities for agricultural land only.
1.3.1.4 Land cover data
The Land Cover Map 2000 (Centre for Ecology and Hydrology) at the 1 km x 1 km resolution was used to provide the inputs into the land cover change model (Chapter 3) and downscaled SPECIES model for Britain (Chapter 4). It is derived from a computer classification of satellite scenes, obtained mainly from Landsat satellites. 16 target classes were identified, to be mapped with 90% accuracy, and these were sub-divided into 27 sub-classes in order that broad habitats could be identified. Although LCM2000 covers Northern Ireland, CORINE data were used for the whole of the Irish case study area in order to have cross-border comparability. CORINE is based on the visual interpretation of satellite imagery and the allocation of land cover types to one of 44 standard land cover classes, thus it is not directly comparable to LCM2000, but LCM2000 has been attributed with CORINE designations.
6 MONARCH 2 Report – Chapter 1 ______
1.3.1.5 Species distribution data
Observed European distributions were obtained from Hulten (1959), Meusel et al. (1965; 1978; 1992) and Jalas and Suominen (1972-91) for plants, Tolman (1997) for butterflies, and Mitchell-Jones et al. (1999) for mammals. The European distribution data for birds were supplied electronically through the European Bird Census Council, but can be found in The EBCC Atlas of European Breeding Birds (Hagemeijer and Blair, 1997). All the British and Irish distributions, except for birds, were provided by the Environmental Records Centre, CEH Monks Wood. Data describing the British distributions of birds were taken from the New Atlas of Breeding Birds in Britain and Ireland (Gibbons et al., 1993). Species distributions at a 1km x 1km spatial resolution were required for characterising the initial population centres for the species dispersal model within the four case study areas. Observed data were only available for Hampshire (from the Hampshire Biodiversity Information Centre) and Cuilcagh/Pettigo (from Environment and Heritage Service, Northern Ireland). Data for Snowdonia were derived from a combination of the Phase 1 Habitat Survey of Wales (which includes the Upland Survey), various other survey records (e.g. some Carex bigelowii sites) and expert opinion. Observed data were available at the 5km x 5km spatial resolution for the Central Highlands from recorders of the Botanical Society of the British Isles. These were downscaled to a 1km x 1km resolution using expert opinion combined with the Land Cover Map 2000. The exact details are explained in the relevant chapters. A similar approach was used in the Republic of Ireland.
1.3.1.6 Conservation data
Conservation data were required to characterise the conservation interest of the bioclimate classes and case study areas, and to assist in the interpretation of model outputs. Data on designated sites was collated in MONARCH 1 (Harrison et al, 2001a) and this was supplemented by data from JNCC. The data consisted of:
• All UK sites designated as Sites of Special Scientific Interest (SSSI) or Areas of Special Scientific Interest in Northern Ireland (ASSI). • Sites in both the UK and Ireland declared as National Nature Reserves. • Sites in both the UK and Ireland designated under EU directives as Special Areas of Conservation (SAC) or Special Protection Areas (SPA). • Sites in the UK identified under the Ramsar Convention. • National Parks and Natural Heritage Areas in Ireland.
1.3.2 Models
1.3.2.1 Bioclimatic classification
The bioclimatic work in MONARCH 2 aimed to provide a means to select those areas of the Britain and Ireland with high conservation interest that were most sensitive to climate change. The work followed the same method of bioclimatic classification that had been developed in MONARCH 1, but took the technique further to allow the classification to be re-assessed under the UKCIP98 2050s High climate change scenario. The areas found to be most sensitive to this level of climate change were examined with respect to their conservation interest. Several aspects of the potential impact of climate change on species and habitats were considered in determining which areas would be suitable for further analysis as case study regions within MONARCH 2.2.
The UKCIP02 baseline and 2050s Low and High scenarios were also classified into the MONARCH 2 bioclimatic classification for the four case study areas to examine the effect of the improvement in spatial resolution of these data sets on the pattern of bioclimatic classes at this local level. A number of differences were found between the baseline climate data at the 10km and 5km levels. The possible causes of these differences were explored.
MONARCH 2 Report – Chapter 1 7 ______
The pattern of local bioclimate classes and their sensitivity to the UKCIP02 scenarios was related also to the distribution of the designated sites across the study areas and to the frequency and type of monitoring undertaken on the habitats and species being modelled in the case study areas. This analysis aimed to identify possible datasets/sites for monitoring and detecting climate change impacts.
1.3.2.2 Species climate space and dispersal models
At the continental scale, climate is expected to be the dominant factor affecting the distribution of species and, therefore, the components of habitats (Pearson et al., 2002). Past observations and results of current modelling studies have shown that species, despite having apparently similar initial distributions, respond individually to climate change (Huntley et al., 1995; Cannon, 1998; Bale et al., 2002), leading to totally new distributions and habitat composition. The SPECIES (Spatial Estimator of the Climate Impacts on the Envelope of Species) model was used to predict the changes in climate space and hence the potential geographical distribution of a wide range of species. SPECIES uses a complex computer simulation program (neural network) to characterise the current distribution and current climate space of species in Europe and to estimate their potential re-distribution under alternative climate change scenarios in Britain and Ireland (Pearson et al., 2002).
A number of integrated algorithms in SPECIES undertake a pre-processing of input climatic and soils data to derive bioclimatic variables of relevance to species’ requirements. The model integrates these data to predict the potential distribution of species through the characterisation of bioclimatic envelopes. The model is trained using existing empirical data on the European distributions of species to enable a Europe-wide climate space to be characterised that captures the climatic range of future scenarios for Britain and Ireland. Once a network is trained and validated at the European scale, it is then applied at a finer spatial resolution in Britain and Ireland.
Although climate may be the prime factor influencing species distribution at the broad-scale, other factors, such as habitat availability and dispersal ability, may be more important at regional scales. To study the combined effects of climate change and habitat fragmentation (as driven by changes in land cover) on species distributions, the SPECIES model was downscaled to 10km2. Outputs from the continental scale climate-driven neural network are used as inputs to a second neural network, along with 10km2 land cover data (Chapter 3.3), to generate regional scale suitability surfaces for species (Pearson et al., 2004), which are applied at the 1km2 scale in the dispersal model. This provides an insight into the roles of climate and land cover as determinants of species’ distributions and enables predictions of distributions under scenarios of changing climate and land cover type to be explored. Land cover change scenarios were developed using a generalised additive model based on climate, soil and environmental variables and the land cover classes. Using a logistic link function, the presence or absence of the classes that were considered to be impacted by climate change were predicted, based on baseline conditions for the LCM2000 classes for England, Wales and Scotland, and Corine land cover classes for Ireland. The UKCIP02 climate change scenarios then were applied to the models.
The ability of species to track changes in the regional suitability surfaces simulated by the downscaled SPECIES model will be dependent on their individual dispersal mechanisms. Thus, the modelled suitability surfaces were coupled with a spatially-explicit cellular automata simulation of species dispersal in changing environments. The model simulates the stochastic dispersal of species in three basic steps: (i) survival, (ii) within-cell population dynamics, and (iii) dispersal (Pearson and Dawson, 2004). Long distance dispersal is incorporated within the model enabling investigation of the potential for species to migrate rapidly under future climate change. Output from this hierarchy of models is a map of the probability of occurrence for each species, based on dispersal characteristics, climatic suitability and land use for future scenarios. These maps form an input to conceptual models of ecosystem functioning, enabling statements to be made about which species might be lost from, or gained by, a habitat.
8 MONARCH 2 Report – Chapter 1 ______
1.3.2.3 A descriptive model of the implications for species communities
Climate change will affect the composition of habitats, including the species and functional type richness and diversity, and also the species and functional type relative abundance. Habitat level impacts were qualitatively estimated by examining the type and strength of species interactions, and also species functional type within a community. Of prime interest was the consequence of species arriving or leaving a community, due to climate change, upon habitat composition and structure. A conceptual framework was developed where a species interaction matrix, detailing trophic relations, within each habitat provided the basis of two conceptual models: the Arriver, which examined the consequences of a species arriving (e.g. range expansion) in a community, and the Leaver, which examined the consequences of a species leaving (e.g. local extinction) a community.
1.3.2.4 Waterbird and coastal habitat models
Wintering waterbirds have started to re-distribute themselves as a result of climate change, tending to winter more to the north and east of Britain and Ireland as winters have become milder. The numbers on estuaries are also a function of the morphology of an estuary, which determines the sedimentary profiles and hence feeding conditions for the birds. As part of MONARCH 2, the BTO developed a modelling protocol that provides predictions of changes in numbers. This was developed with respect to the estuaries on the internationally important Suffolk coast. Predictions were based on two-stage models.
Stage 1: Predictions of bird numbers the region would hold based on climate. Stage 2: Predictions of the bird numbers the region could hold based on habitat availability following changes in estuary morphology due to sea-level rise.
The model was also applied to four estuaries in Hampshire, as a contribution to the assessment of impacts of climate change on biodiversity of the area.
1.4 Project partners
1.4.1 Research Team
Environmental Change Institute Dr. Pam Berry, Dr. Terry Dawson, Dr. Paula Harrison, University of Oxford Richard Pearson, Nathalie Butt Zoology Department (Co-ordination, SPECIES and dispersal models) South Parks Road Oxford OX1 3PS.
ADAS Ecology Group Dr. Jo Hossell, Dr. Alison Riding, Dr. Hester Lyons, Wolverhampton Research Team Dr. Catherine Bradshaw, Paul Scholefield, Woodthorne Bethan Clemence Wergs Road (Bioclimatic classification, land use/land cover Wolverhampton change) WV6 8TQ
British Trust for Ornithology Dr. Humphrey Crick, Dr. Mark Rehfisch, The Nunnery Dr. Graham Austin, Dr. Steve Freeman Thetford (Water bird model, SPECIES and dispersal Norfolk models for birds) IP24 2PU
MONARCH 2 Report – Chapter 1 9 ______
CABI Bioscience Dr. Greg Masters, Nicola Ward, Imogen Wilde Bakeham Lane (Habitat level interpretation) Egham Surrey TW20 9TY
1.4.2 Funders
MONARCH 2 was funded by English Nature (lead), Countryside Council for Wales, Environment Agency, Environment and Heritage Service, Forestry Commission, Joint Nature Conservation Committee, National Parks and Wildlife Service (Republic of Ireland), National Trust, Royal Society for the Protection Birds, Scottish Executive, Scottish Natural Heritage, UK Climate Impacts Programme, Welsh Assembly Government, Woodland Trust. The Research Team would like to thank these organisations for their intellectual input and practical support.
1.5 References
Ångström, A. (1924). Solar and terrestrial radiation. Quarterly Journal of the Royal Meteorological Society, 50, 121.
Bale, J. S., Masters, G. J., Hodkinson, I. D., Awmack, C., Bezemer, T. M., Brown, V. K., Butterfield, J., Buse, A., Coulson, J. C., Farrar, J., Good, J. E. G., Harrington, R., Hartley, S., Jones, T. H., Lindroth, R. L., Press, M. C., Symmioudis, I., Watt, A. S and Whittaker, J.B. (2002). Herbivory in global change research: direct effects of rising temperatures on insect herbivores. Global Change Biology, 8, 1-16.
Beebee, T.J.C. (2002). Amphibian phenology and climate change. Conservation Biology, 16(6), 1454.
Bisgrove, R. and Hadley, P. (2002), Gardening in the Global Greenhouse. UKCIP Technical Report, 139 pp.
Buckland SM, Grime JP, Hodgson JG and Thompson K (1997). A comparison of plant responses to the extreme drought of 1995 in northern England. Journal of Ecology 85, 875-882.
Cannon, R. J. C. (1998). The implications of predicted climate change for insect pests in the UK, with emphasis on non-indigenous species. Global Change Biology, 4, 785-796.
Secretariat of the CBD (2003). Interlinkages between Biological Diversity and Climate Change. Advice on the Integration of Biodiversity Considerations into the Implementation of the United Nations Framework Convention on Climate Change and its Kyoto Protocol. CBD Technical Series No. 10.Montreal, Secretariat of CBD.
Conrad, K. F., I. P. Woiwod, and Perry, J.N. (2002). Long-term decline in abundance and distribution of the garden tiger moth (Arctia caja) in Great Britain. Biological Conservation, 106(3), 329-337.
Dockerty, T., A. Lovett, and Watkinson, A. (2003). Climate change and nature reserves: examining the potential impacts, with examples from Great Britain. Global Environmental Change 13(2), 125- 135.
Downing, T.E, Butterfield, R.E., Edmonds, B., Knox, J.W., Moss, S., Piper, B.S. and Weatherhead, E.K. (and the CCDeW project team) (2003). Climate Change and the Demand for Water, Research Report, Stockholm Environment Institute Oxford Office, Oxford.
10 MONARCH 2 Report – Chapter 1 ______
FAO (1986). Early Agrometeorology Crop Yield Assessment. Plant Production and Protection Paper 73, Food and Agricultural Organisation of the United Nations, Rome, Italy, 150 pp.
Fitter, A. H. and Fitter, R. S. R. (2002). Rapid changes in flowering time in British plants. Science, 296, 1689-1691.
Gardiner, M.J. and Radford, T. (1980). Ireland: General Soil Map, 2nd Edition. Teagasc (formerly An Foras Taluntais), Dublin.
Gibbons, D.W. Reid, J.B. and Chapman R.A. (eds.) (1993). The new atlas of breeding birds in Britain and Ireland: 1988-1991. T & A D Poyser, London.
Hagemeijer, W.J.M. and Blair, M. (1997) (Eds.) The EBCC Atlas of European Breeding Birds: Their Distribution and Abundance. T.& A.D. Poyser, London.
Harrison, P.A., Berry, P.M. and Dawson, T.P. (Eds.) (2001a). Climate Change and Nature Conservation in the Britain and Ireland: Modelling natural resource responses to climate change (the MONARCH project). UKCIP Technical Report, Oxford.
Harrison, P.A., Dawson, T.P, Viles, H. A., Austin, G. E. and Berry, P.M. (2001b). The MONARCH project: Study aims and methods. In: Harrison, P.A., Berry, P.M. and Dawson, T.P. (Eds.) (2001a) Climate Change and Nature Conservation in the Britain and Ireland: Modelling natural resource responses to climate change (the MONARCH project). UKCIP Technical Report, Oxford.
Hill, J.K., Thomas, C.D., Fox, R., Telfer, M.G., Willis, S.G., Asher, J. and Huntley, B. (2002). Proceedings of the Royal Society Biological Sciences Series B, 269(1505), 2163-2171.
Hulme, M., Jenkins, G.J., Lu, X., Turnpenny, J. R., Mitchell, T.D., Jones, R.G., Lowe, J., Murphy, J.M., Hassell, D., Boorman P., McDonald, R. & Hill, S. (2002). Climate Change Scenarios for the United Kingdom: The UKCIP02 Scientific Report. Tyndall Centre for Climate Change Research, School of Environmental Sciences, University of East Anglia, Norwich, UK, 120 pp.
Hulten, E. (1959). The Amphi-Atlantic Plants and their Phytogeographical Connection. Almqvist & Wiksell, Stockholm, 340 pp.
Huntley, B., Berry, P.M., Cramer, W. and McDonald, A.P. (1995). Modelling present and potential future ranges of some European higher plants using climate response surfaces. Journal of Biogeography, 22, 967-1001.
Institute of Water and Environment (2002). DEFRA Climate Change Impacts and Adaptations Research Programme (CC03) Project Summaries Report 1987-2002. UKCIP Report, Oxford.
Intergovernmental Panel on Climate Change (2001a). Climate Change 2001: The Scientific Basis. Intergovernmental Panel on Climate Change, Cambridge University Press, Cambridge.
Intergovernmental Panel on Climate Change (2001b). Climate Change 2001: Impacts Adaptation and Vulnerability. Summary for Policymakers. Intergovernmental Panel on Climate Change, Cambridge University Press, Cambridge.
Intergovernmental Panel on Climate Change (2002). Climate Change and Biodiversity. Intergovernmental Panel on Climate Change, Technical Paper V.
Jalas, J. and Suominen, J. (1972-91). Atlas Florae Europaeae. Vols. 1-9, Societas Biologica Fennica Vanamo, Helsinki.
MONARCH 2 Report – Chapter 1 11 ______
Jenkins, G., Cooper, C., Hassell, D. and Jones, R. (2003). Scenarios of climate change for islands within the BIC region. British-Irish Council, www.british-irishcouncil.org/climatechange.
Jones, P.D., Marsh, R. and Farmer, G. (1990). Calculation of Potential Evapotranspiration over the European Community Countries. Progress Report to the Commission of the European Communities, Joint Research Centre - Ispra, Contract number 3578-88-12 ED ISP GB, Climatic Research Unit, University of East Anglia, Norwich, UK, 29 pp.
Masters, G.J. & Brown, V.K. (2001) Effects of Experimental Manipulation of Climate on Calcareous Grassland Plants and Invertebrates. In: Impacts of Climate Change on Wildlife, eds. R.E. Green, M. Harley, M. Spalding & C. Zöckler, RSPB publication on behalf of EN, WWF-UK, UNEP WCMC & RSPB, pp. 57-59.
Masters, G.J., Brown, V.K., Clarke, I.P., Whittaker, J.B. & Hollier, J.A. (1998) Direct and indirect effects of climate change on insect herbivores: Auchenorrhyncha (Homoptera). Ecological Entomology, 23, 45-52.
McCarty, J.P. (2001). Ecological consequences of recent climate change. Conservation Biology, 15, 320-326.
Meusel, H., Jäger, E. and Weinert, E. (1965). Vergleichende Chorologie der Zentraleuropäischen Flora Vol 1. Gustav Fischer, Jena.
Meusel, H., Jäger, E., Weinert, E. and Rauschert, S.T. (1978). Vergleichende Chorologie der Zentraleuropäischen Flora Vol 2. Gustav Fischer, Jena.
Meusel, H., Jäger, E. and Weinert, E. (1992). Vergleichende Chorologie der Zentraleuropäischen Flora Vol 3. Gustav Fischer, Jena.
Mitchell-Jones, J.A., Amori, G., Bogdanowicz, W., Krystufek, B., Reijinders, P.J.H., Spitzenberger, F., Stubbe, M., Thissen, J.B.M., Vohralík, V. and Zima, J. (1999). The atlas of European mammals. T. & A.D. Poyser, London.
Morecroft, M. D., C. E. Bealey, Howells, O., Rennie, S. and Woiwod, I.P. (2002). Effects of drought on contrasting insect and plant species in the UK in the mid-1990s. Global Ecology and Biogeography, 11, 7-22.
Parmesan, C. and Yohe, G. (2003). A globally coherent fingerprint of climate change impacts across natural systems. Nature, 421, 37-42.
Pearson, R.G. and Dawson, T.P. (2004). Modelling long-distance plant dispersal across fragmented landscapes. In preparation.
Pearson, R.G., Dawson, T.P. and Lui, C. (2004). Modelling species distributions in Britain: a hierarchical integration of climate and land-cover data. Ecography: in press. Pearson, R.G., Dawson, T.P., Berry, P.M. and Harrison, P.A. (2002). SPECIES: a spatial evaluation of climate impact on the envelope of species. Ecological Modelling, 154, 289–300.
Penman, H.L. (1948). Natural evaporation from open water, bare soil, and grass. Proceedings of the Royal Society, London Series A, 193, 120-146.
Prescott, J.A. (1940). Evaporation from a water surface in relation to solar radiation. Transactions of the Royal Society of Southern Australia, 64, 114-118.
12 MONARCH 2 Report – Chapter 1 ______
Rietveld, M.R. (1978). A new method for estimating the regression coefficients in the formula relating solar radiation to sunshine. Agricultural Meteorology, 19, 243-252.
Semenov M.A. and Barrow. E.M. (1997) Use of a stochastic weather generator in the development of climate change scenarios Climatic Change, 35,397-414
Semenov M.A., Brooks R.J. (1999) Spatial interpolation of the LARS-WG stochastic weather generator in Great Britain. Climate Research 11,137-148
Thomas, C.D., Cameron, A., Green, R.E., Bakkenes, M., Beaumont, L.J., Collingham, Y.C., Erasmus, B.F.N., Ferreira de Siqueira, M., Grainger, A., Hannah, L., Hughes, L., Huntley, B., van Jaarsveld, A.S., Midgley, G.F., Miles, L., Ortega-Huerta, M.A., Peterson, A.T., Phillips, O.L. and Williams, S.E. (2004). Extinction risk from climate change. Nature, 427, 145-148.
Tolman, T. (1997). Butterflies of Britain and Europe. Collins Field Guide. Harper Collins, London 320 pp.
UKCIP and EPSRC (2003). Building Knowledge for a Changing Climate. UKCIP Summary report, 21 pp.
Walther, G-R., Post, E., Convey, P., Menzel, A., Parmesan, C., Beebee, T.J.C., Fromentin, J-M., Hoegh-Guldberg, O. and Bairlein, F. (2002). Ecological responses to recent climate change. Nature, 416, 389-395.
Willows, R. and Connell, R. (2003). Climate Adaptation: Risk, uncertainty and decision-making. UK Climate Impacts Programme, Technical Report, 154pp.
Wilson, A.M., Henderson, A.C.B. and Fuller, R.J. (2002). Status of the Nightingale Luscinia megarhynchos in Britain at the end of the 20th Century with particular reference to climate change. Bird Study, 49(3), 193-204.
MONARCH 2 Report – Chapter 2 13 ______2 Bioclimatic classification and case study selection
J.E. HOSSELL, A.E. RIDING and P.A. HARRISON
Summary
This chapter describes the creation of a baseline bioclimatic classification using the UKCIP98 10km data and its manipulation using the 2050s High scenario data to highlight areas of Britain and Ireland that show greatest bioclimatic sensitivity to climate change. In particular, South-east England, Snowdonia, the Central Highlands and north-west Ireland were identified as regions where the 2050s High bioclimate would be most different from the baseline. This information has been combined with data on the distribution of sites and habitats of conservation interest to enable the selection of four case study areas for use in testing the downscaled methodology developed in MONARCH 2. Around 10 species have been selected within each case study area to test the new downscaled SPECIES and dispersal models developed in MONARCH 2).
The publication of the UKCIP02 climate scenarios (Hulme et al., 2002), during the lifetime of the project, provided data at a higher spatial resolution (5km) than the previous UKCIP98 scenarios (10km resolution). However, transferring to this new dataset has also required that the original bioclimatic classification derived using the UKCIP98 data (Hulme and Jenkins, 1998) be recalibrated in the four selected locations to use the more detailed 2002 scenario data. This recalibration has highlighted some differences between the two baseline datasets. It is unclear (without repeating the national level bioclimatic classification with the UKCIP02 data) what effect this would have had on the original selection of the four case study areas. Certainly the sunshine hours data (calculated from cloud cover) are considerably different between the two datasets. The possible reasons for the discrepancies between the two dataset baselines have been considered (see study area chapters) but the difference in the derivation of the two datasets means that it is not possible to directly attribute the disparity to a single cause.
The uncertainty surrounding climate change scenario projections is also considered in this chapter through analysis of available scenario data from other Global Climate Models. The assessment indicates that UKCIP02 scenarios present a slightly warmer and drier summer pattern of change than the other four models considered and hence may represent a more severe set of impact results than studies based on other climate models.
2.1 Introduction
In MONARCH 1, a bioclimatic classification was devised to identify areas of similar climatic conditions, which were then characterised according to their nature conservation interest. The aim of the classification was to discriminate between areas where similar species and habitats existed under different climatic conditions, on the basis that their response to climate change may not be uniform across the full range of a species’ or habitat’s distribution. Studies of species response at the margins of their distribution, for example, have shown a greater sensitivity to climate change than populations within the core of their range (e.g. Parmesan et al., 1999).
In MONARCH 2 this work has been taken a step further in using the classification to explore the bioclimatic sensitivity of Britain and Ireland to climate change. The identification of four areas that are projected to experience relatively large climatic change has enabled the species impact models (see Chapters 3 and 4) to be tested in areas where significant but, depending on the site, contrasting types of species responses may be expected. The method used to select the species to be studied in the testing of the species impact models is also described and the species selected for each area are listed.
14 MONARCH 2 Report – Chapter 2 ______
The selection of the areas was made on the basis of the UKCIP98 scenario data. With the publication of the UKCIP02 datasets, the bioclimatic classification method was used to reclassify the climate within the case study areas using these newer scenario data. The differences between the UKCIP98 and 02 baseline datasets have been examined and the UKCIP02 scenarios have been compared with scenarios constructed from other global climate models to explore uncertainty in projections of future climate.
2.2 Bioclimatic classification method
This section describes how that bioclimatic classification method was developed for the identification of detailed study areas and how the classification was then recalibrated to use the UKCIP02 data for these selected areas. Throughout this chapter the suffix 98 will be used to refer to the bioclimatic classification derived from the UKCIP98 data, whilst 02 will be used for the classification using the UKCIP02 datasets. The term “emission scenarios” is used to indicate the UKCIP02 climate change data in order to distinguish the use of these data from the UKCIP98 scenarios.
The variables used in the Baseline98 classification of MONARCH 2 vary slightly from those used to define the classification developed in MONARCH 1. The difference is due to the change in climate variables available with the UKCIP02 scenarios, since it was considered desirable to select only variables that were available or could be calculated from both the UKCIP98 and UKCIP02 datasets. The variables used are listed in Table A2.1 of the annex to this Chapter.
The derivation of the bioclimatic classification involved three steps: 1. Principal Components Analysis (PCA) to reduce down the number of variables in the analysis and derive independent, uncorrelated variables. 2. Hierarchical clustering routine using the principal component data to establish the classes and the natural “break points” i.e. the optimal number of classes. The clustering process used was the HCLUST algorithm within S-Plus 4 (Mathsoft, 1997). The number of classes within the final classification was selected to provide a similar level of resolution to the MONARCH 1 classification, which had 21 classes, but since a slightly different variable set had been used some variation in the optimal number of classes was expected. 3. Discriminant function analysis to establish relationships between the classification and the original climate variables so that the climate change scenario data may be used to determine class membership. The process used was the General Discriminant Function Analysis option within Statistica v6.0 (Statsoft Inc, 2001). The module allowed the original climate dataset for the Baseline98 classification to be used to define the climatic boundaries for each of the classes. This enabled the class membership of a second dataset to be predicted based on these Baseline98 conditions. For the selection of the detailed study areas, the 2050s High scenario from the UKCIP98 dataset were classified into the Baseline98 classes in this way. The baseline classification of the case study areas was then recalibrated using the UKCIP02 5km baseline and the effects of the 2050s Low and High emission scenarios were examined for these areas.
Steps 1 and 2 of the method are the same as those used to develop the bioclimatic classification in MONARCH 1 (see Hossell et al., 2001; Hossell et al., 2003 for more detail on these methods).
2.3 Bioclimatic classification results
2.3.1 Principal Components Analysis (PCA)
It is clear from Figure 2.1 that the east to west contrast in moisture levels provides a defining structure to the climate of Britain and Ireland. The north to south pattern in the temperature variables accounts for the second greatest element of the climate variation across the countries. Table 2.1 shows the climatic variables related to each of the first four principal components, which may be considered as new, independent variables defined by the analysis.
MONARCH 2 Report – Chapter 2 15 ______
Figure 2.1: The pattern of the first four factors of the principal components analysis.
Table 2.1: Relationship of PCA factors to original variables for the Baseline98 climate data. Factor 1 Factor 2 Factor 3 Factor 4 Precipitation Jan-Dec Mean Temperature Jan-May, Mean Windspeed Jan-Dec PET Jan-Nov Sept-Dec HER Jan-Dec Min. Temperature Jan-Dec Spring Precipitation Growing Degree Days Abs. Min. Temperature Growing Season Length HER = Hydrologically Effective Rainfall; PET = Potential Evapo-Transpiration.
In order to determine how climate change may affect the relative climatic pattern across Britain and Ireland, the same PCA was also undertaken for the UKCIP98 2050s High scenario. The results of this analysis are shown in Table 2.2. The 2050s High scenario produces an extra principal component and suggests a more complex pattern of climate variation across the regions. It is also interesting to note that climate change results in a switch between the first and second principal components. Under this climate change scenario temperature becomes a more important distinguishing variable than rainfall across Britain and Ireland.
16 MONARCH 2 Report – Chapter 2 ______
Table 2.2: Variables most strongly related to the first four factors of the UKCIP98 2050s High PCA. Factor 1 Factor 2 Factor 3 Factor 4 Mean Temperature Jan-Jun, Precipitation Jan-Dec Mean Windspeed Jan-Dec PET Feb-Nov Sept-Dec Min. Temperature Jan-Dec HER Jan-Dec Sunshine hours Jan Spring Precipitation Growing Degree Days Abs. Min. Temperature Growing Season Length HER = Hydrologically Effective Rainfall; PET = Potential Evapo-Transpiration.
2.3.2 Hierarchical clustering
The clustering analysis places squares into all possible cluster combinations between one class (all squares in the same class) and 3785 classes (all squares in a separate class). Within this range the choice of the number of classes is dependent upon the detail required in the classification and the positioning of natural breaks. These breaks show where splitting a class into two greatly reduces the within class variation of the classification and so may be viewed as a measure of uniformity within classes. In looking for a class solution at a similar level of detail to the MONARCH 1 classification, the analysis of natural points shows that 21, 23 and 26 classes would all be possible options. Out of these three options, the greatest increase in the uniformity of class members occurs with the change from 25 to 26 clusters. Hence, the 26 class “solution” was selected for the Baseline98 classification. Figure 2.2 shows the distribution of the 26 classes in the Baseline98 classification.
Figure 2.2: The pattern of classes for the Baseline98 bioclimatic classification.
MONARCH 2 Report – Chapter 2 17 ______
The difference between the number and membership of classes for the MONARCH 1 and 2 classifications is to be expected, due to the difference in the number and type of climate variables used in the two classifications. The increased number of classes in the Baseline98 “solution” is also valuable since it provides better discrimination between the bioclimate of south central England and central Ireland, which gain several of the extra 5 classes.
2.3.3 Discriminant function analysis
The class membership provided for the Baseline98 classification was added to the dataset of climate variables and the discriminant function analysis run to reclassify the data based on these variables. The analysis provided an 82.77% agreement with the classification produced using the clustering analysis. Figure 2.3 provides a visual comparison of the two classifications. The main changes are at the margins of some of the larger classes, where the discriminant function classification provides “smoother” boundaries with fewer outliers. The match between the two classification schemes was high and showed that the climate variables could successfully be used to predict the Baseline98 classification membership. Figure 2.4 illustrates the gradient of the bioclimatic characteristics for the classes and shows how well the classes distinguish the different climatic patterns within Britain and Ireland
Figure 2.3: Comparison of (a) the Baseline98 classification produced using cluster analysis with (b) the classification produced using discriminant function analysis.
(a) (b)
18 MONARCH 2 Report – Chapter 2 ______
Figure 2.4: The gradient of key climate variables for the 26 Baseline98 classes.
MONARCH 2 Report – Chapter 2 19 ______
2.4 Conservation characterisation
The different classes were characterised according to their nature conservation interest by examining the range of protected habitats in each class and the percentage of the class that is within a protected area. The following tables highlight the most important classes in conservation terms. Table 2.3 shows the percentage of area designated in each class by country. Tables 2.4 and 2.5 give a summary of the location, size and habitat types in each class for Britain and Ireland.
Table 2.3: Bioclimatic classes characterised by area (total km2) and percentage cover of designated sites. Bold figures highlight the largest areas/percentage coverage across the classes. Class Total area N. Ireland Eire Wales Scotland England Total % (km2) ASSI SAC SSSI SSSI SSSI designated (km2) (km2) (km2) (km2) (km2) area in class 1 21300 114.8 581.2 108 881.2 118.8 8.5 2 2000 193.8 9.7 3 11900 32.8 637.5 512.5 9.9 4 31200 21.0 2568.8 77 181.9 9.1 5 7700 9.5 68.75 371 881.2 275 20.9 6 3600 987.5 27.4 7 15700 245 350 3.8 8 18100 69.8 756.2 128 50 5.5 9 13600 404 581.2 1168.7 15.8 10 79400 24.8 31.3 153 1950 1618.7 4.8 11 7700 4.0 593.8 139 206.3 50 12.9 12 41900 543.2 912.4 88 31.3 393.8 4.7 13 4400 1343.8 30.5 14 24800 19 1681.3 6.9 15 7100 700 450 16.2 16 2000 31.3 126 156.3 15.7 17 43800 388 112.5 2950 7.9 18 5600 443.8 7.9 19 2200 106.3 4.8 20 11500 568.8 843.8 12.3 21 1000 116 87.5 68.75 27.2 22 9900 6.0 418.8 4.3 23 2000 1081.3 54.1 24 4000 368.8 9.2 25 1800 425 23.6 26 4300 400 9.3 N.B. 1. Method probably slightly overestimates area of sites due to coarseness of grid. Grid resolution varied per country due to processing times involved (Northern Ireland: 500m; Wales 1000m; England, Scotland, Eire: 2500m). 2. In Irish Republic, SAC's used as the most complete designation. 3. No designated sites in Isle of Man and Channel Islands.
For Ireland, bioclimate class 4 holds the greatest size of protected area, but class 23 holds the greatest percentage of protected area for the size of class. For Great Britain Class 10 holds the greatest amount of protected area but this is a large class anyway so the proportion of it protected is only about 5%. Class 6 holds the highest proportion of protected sites for its size and these are all in Scotland.
20 MONARCH 2 Report – Chapter 2 ______
Table 2.4: Summary of class characteristics in Great Britain. Class Mean Baseline Location Area Habitat Character (>10% of Altitude (m) no of designated total number of sites in bold of Baseline squares (%) – largest number of sites in Class italics) 1 125.72 168 West Coast of Scotland, 6.6 Machair, Coastal Salt Northern Hebrides and marsh, Blanket bog isolated upland areas in Wales, 2 405.80 20 Small class in very high 9.7 Pine woodland upland areas of W Scotland 3 42.03 64 Southern Outer Hebrides and 8.0 Machair, Coastal dune slacks 4 163.49 35 Brecon Beacons and Exmoor 7.4 Upland oak woodland 5 298.76 74 Moderately high inland areas 20.6 Tall herb ledge communities, of western Scotland, Southern alpine & boreal grasslands, Uplands, Lake District, montane heath Snowdonia and Cambrians 6 617.22 36 Cairngorms tops 27.4 Alpine & boreal grasslands, Montane heath, Tall herb ledge communities, 7 75.26 157 Coastal SW England, Southern 3.8 Coastal dune slacks Wales, Isle of Wight and Channel Islands 8 45.00 50 Gower Peninsula, Isle of Man 3.6 Coastal dune slacks 9 277.38 135 Moderately high uplands in 16.0 Upland hay meadows, Blanket NW Highlands, Southern bog Uplands, Peak District and Cambrians 10 97.39 790 Lowland areas of Eastern 4.7 Raised bog, Drought prone Scotland, England, Midlands, acid grassland, Lowland and Welsh borders calcareous grassland, Upland oak, Beech woodland, Coastal saltmarsh, Pine woodland, Coastal dune slacks, Wet heath 11 131.84 58 Moderately high areas of 6.8 Coastal saltmarsh Western coastal Highlands, Lake District, Snowdonia 12 186.08 92 Sheltered areas of western 5.6 Raised bog Pennines 13 Not Present n/a 14 36.67 248 Eastern coast of England and 6.9 Coastal dune slacks, some sheltered coastal areas in Drought prone acid Western and Northern Wales grassland and Scotland 15 421.45 71 High upland mountains of 16.2 Alpine & boreal grassland, Central Highlands, Southern Pine woodland, Upland hay Uplands and Pennines meadows, Montane heath 16 317.94 18 High upland areas of Lake 15.7 Wet heath District, Brecon Beacons, Cambrian Mountains, Exmoor and Dartmoor 17 76.33 438 Lowland plains of South 7.9 Beech woodland, Lowland central England, Cheshire, and calcareous grassland, coastal areas of Solway Firth Drought prone acid grassland, Raised bog, Coastal saltmarsh, Coastal dune slacks 18 Not Present n/a 19 116.00 22 Upland areas in Hebrides 4.8 Machair
MONARCH 2 Report – Chapter 2 21 ______
Class Mean Baseline Location Area Habitat Character (>10% of Altitude (m) no of designated total number of sites in bold of Baseline squares (%) – largest number of sites in Class italics) 20 286.78 115 Eastern fringes of Cairngorms, 12.3 Upland hay meadow Lammermuir Hills, Cheviots, and northern Pennines 21 352.10 10 Isolated, wet high mountain 27.2 Tall herb ledge communities areas of Lake District, Snowdonia and Skye 22 37.13 93 Orkneys and Shetlands 4.5 Coastal saltmarsh 24 287.82 40 Moderately high western 9.2 Pine woodland edges of Scottish Highlands 25 498.89 18 Mountain tops in Western 23.6 Pine woodland Highlands 26 433.88 43 Mountainous areas in eastern 9.3 Pine woodland, Montane Cairngorms heath, Upland hay meadows Total 2795
Table 2.5: Summary of class characteristics for the Republic of Ireland and Northern Ireland. Class Mean Baseline Location Area Habitat Character (>10% of Altitude (m) no of designated total number of sites in bold of Baseline squares (%) – largest number of sites in Class italics) 1 261.16 45 Isolated uplands of Ireland and 15.5 Wet heath, Montane Heath, NI Upland Oak, Blanket Bog 3 76.16 55 Northern NI 12.2 Machair, Coastal Dune Slacks, Coastal saltmarsh, Montane Heath 4 97.22 277 Inland Western Ireland 9.3 Lowland calcareous grassland, Raised Bog, Blanket Bog, Upland Oak , Coastal Salt Marsh, Machair, Coastal dune slacks, Wet heath, Montane Heath 5 267.67 3 Isolated upland sites in NW 26.1 Wet Heath Ireland 8 34.21 131 Coastal Eastern and Southern 6.3 Coastal saltmarsh, Coastal Ireland dune slacks, Machair, Lowland calcareous grassland 9 366.00 1 Moderately high upland site in 0.0 Wet Heath, blanket bog Sperrin Mountains 10 55.50 4 Isolated lowland sites east of 14.0 Coastal saltmarsh, raised & Sperrin Mountains blanket bog 11 203.63 19 Moderately high areas of NW 31.5 Montane Heath Ireland 12 120.50 327 Central Eastern Ireland 4.5 Raised bog, Lowland calcareous grassland, Upland oak, Wet Heath & Blanket bog 13 161.93 44 Exposed moderately high sites 30.5 Montane heath, Wet heath, of SW and Central West coast Blanket bog of Ireland 16 264.50 2 Isolated sites upland on NW 15.7 Montane heath coast of Ireland 18 76.34 56 Lowland coastal areas of SW 7.9 Coastal saltmarsh and Central Western Ireland
22 MONARCH 2 Report – Chapter 2 ______
Class Mean Baseline Location Area Habitat Character (>10% of Altitude (m) no of designated total number of sites in bold of Baseline squares (%) – largest number of sites in Class italics) 22 123.00 6 Moderately high upland areas 1.0 Coastal dune slacks North of Sperrin mountains 23 237.80 20 Moderately high areas in SW 54.1 Montane heath and Central Western Ireland Total 990
2.5 Identification of climatically sensitive areas
2.5.1 Climatically sensitive areas
The discriminant function analysis was rerun using the climate variables for the UKCIP98 2050s High scenario. The analysis assigned squares to classes based on a comparison of distance to a class centroid in relation to that of other classes. However, the nature of the analysis is such that a square will always be assigned to a class regardless of the goodness of fit of that square’s bioclimate to the class mean. Since it is expected that the future climate of Britain and Ireland may differ greatly from that experienced in the baseline, a cut-off limit was imposed to highlight squares that were outliers within a class. The cut-off was defined as being the maximum squared Mahalanobis distance (854.49) between any two classes in the baseline classification. Hence, if a future climate square was further from the centroid of its class than the maximum distance of the most dissimilar baseline classes, it was considered to be beyond the climate range of the Baseline98 classification (Figure 2.5). Of the 3785 squares in the classification, 1586 were classified as being beyond the Baseline98 classification and hence in having no suitable analogue under the future scenario. Figure 2.6 shows the pattern of classification for the cluster and discriminant function predicted memberships and for the pattern under the 2050s High scenario. The map on the bottom right of this figure also shows the pattern of squares that fall outside of the scope of the Baseline98 classification.
Figure 2.5: A schematic diagram illustrating the principle behind the method used to identify climate range beyond the Baseline98 boundary by determining the Mahalanobis distances of the squares that are particularly sensitive to climate change.
Baseline Classification
Baseline Class C
Mahalanobis distance Baseline Class A between most dissimilar Baseline Class B classes in baseline(Maximum Mahalanobis distance)
Baseline Class E
Baseline Class D
Baseline Class F
Maximum Mahalanobis distance
Altered climate grid X square
MONARCH 2 Report – Chapter 2 23 ______
Figure 2.6: Maps showing the pattern of classification for the Baseline98 classes, the discriminant function analysis of the baseline data for the UKCIP98 2050s High scenario data and those squares whose 2050s High climate places them beyond the range of the Baseline98 classification.
2.5.2 Potential new classes
Figure 2.7 shows several ways of assessing the difference between the Baseline98 and UKCIP98 2050s High classifications. The two maps based on squared Mahalanobis distance explore: (a) how the climate of an anomalous square relates to the climate of the other anomalous squares; and (b) the absolute distance of a square from the baseline classification. In addition, the bottom right map (d)
24 MONARCH 2 Report – Chapter 2 ______
shows how the 2050s High climate may itself be classified using the same PCA and clustering technique as used in the Baseline98 classification and compares this to the baseline98 classification (c). The clustering of the 2050s High climate produced a natural break at 22 classes, suggesting less distinction between climate conditions in the future, particularly across southern and central England, though the southern coast does become a distinct class in the future, which extends further inland than the coastal classes of the baseline.
Snowdonia, Sussex, Kent and the South Essex coast show the greatest difference between the baseline classification and the future 2050s High climate. In Ireland, Fermanagh, Galway and Connemara all show patterns of climate that differ greatly from that experienced across the region in the baseline. For Scotland, Lanarkshire and the western Grampians show the greatest difference from baseline conditions (Figure 2.7).
Figure 2.7: The bioclimate classes classified as sensitive mapped according to their Mahalanobis clusters (a) or Malhanobis distance (b) along with comparison of the baseline98 (c) and UKCIP98 2050s High (d) bioclimate classifications.
MONARCH 2 Report – Chapter 2 25 ______
2.5.3 Selecting the case study areas
One of the aims of MONARCH 2 was to find an objective way to select areas of Britain and Ireland for more detailed analysis using the species impact models (see Chapter 3). The areas identified as being beyond the Baseline98 classification under their 2050s High climate were examined to identify areas with both high conservation interest and apparent sensitivity to climate change. These areas were considered to be most suitable for testing the working of the impact models.
The objective selection of areas had to be constrained by more pragmatic requirements:
• In order to provide a useful test for the species impact models, the selected areas should cover a range of habitat types; • To aid the selection of species to model, the areas selected needed to represent the country-based interests of the project's funders; this meant that four case study areas would be needed. • One area should highlight cross-border aspects of any conservation policy or management responses to climate change.
2.5.4 Key conservation changes
Some of the key issues for conservation policy and management in terms of climate change impacts (Hossell et al, 2001) are:
1. Loss of climate suitable for a protected habitat 2. Shift of a habitat to a new area 3. Need for dispersal over great distances 4. Influx of non-native species into the study area
Each of these changes may be measured by a change to the future pattern of the bioclimatic classification:
1. Loss of a class or series of classes that support all or significant areas of a habitat in the whole study area 2. Local arrival of a class associated with a new habitat in one or more of the regions of the study area 3. Largest shift of northern boundary or contraction of southern boundary of a class across the study area 4. Influx of new climatic conditions into the study area.
2.5.4.1 Loss of climate class associated with a habitat
Tables 2.6 to 2.9 below show the main habitat groups by their Baseline98 class for the areas identified in the UKCIP98 2050s High analysis as being sensitive to climate change.
Table 2.6: England: Areas and their habitats most sensitive to changes in climatic conditions. Key area affected Baseline class Habitat lost Comments Pennines and Cheviot 9, 15, 20 Upland Hay Meadow Large contraction of class range and Hills loss from W of Pennines Norfolk, Lincolnshire 10, 14, 17 Drought Prone acid Some Northward shift of conditions grassland into Scotland –
26 MONARCH 2 Report – Chapter 2 ______
Table 2.7: Scotland: Areas and their habitats most sensitive to changes in climatic conditions. Key area affected Baseline class Habitat lost Comments West Coast of Scotland & 1 Blanket bog Equal largest number of protected Hebrides sites Central West Highlands of 9 Blanket bog Equal largest number of protected Scotland sites Cairngorm tops, & 5, 6, 15, 26 Montane Heath/Pine Large reduction in area of suitable Grampians, Southern woodland future conditions, possible shift of Uplands Welsh and Lake District conditions into Central West Highlands Cairngorm tops, & 5, 6 Tall herb ledge Large reduction in area of suitable Grampians communities future conditions, possible shift of Welsh and Lake District conditions into Central West Highlands Cairngorm tops, & 5, 6, 15 Alpine & boreal Large reduction in area of suitable Grampians, Southern grasslands future conditions, possible shift of Uplands Welsh and Lake District montane heath conditions into Central West Highlands
Wales
No loss of major habitats apparent, since incoming classes have a similar habitat type. However, some vegetation changes may be expected (perhaps in pattern of habitat composition), since there are a large number of climate sensitive squares in the country. But the incoming classes have similar habitat types so the changes may be restricted to variation in species composition within existing communities.
Table 2.8: Ireland: Areas and their habitats most sensitive to changes in climatic conditions. Key area affected Baseline class Habitat lost Comments South of Lough Neagh, 4, 12, 13, 23 Lowland Calcareous Replaced by coastal class conditions Cork, Central, Eastern grassland, Raised bog, Ireland Blanket bog
2.5.4.2 Potential local arrivals
Scotland and Ireland both gain “new” classes from the Baseline98 classification under the 2050s High scenario. The table below shows what those new classes are, what habitats they are associated with in Britain, the areas that they occupy and the classes they displace.
Table 2.9: Possible new arrival of habitats based on the change in bioclimatic class. Key area affected Baseline class Habitat potentially Comments arriving? South of Lough Neagh, Cork, 7, 18 Coastal dune slacks, Largely replace squares in Central Eastern Ireland coastal salt marsh classes 4, 12, 13 and 23 Inner Hebrides round Mull 18 Coastal saltmarsh New class to Scotland
2.5.4.3 Greatest shifts in class boundaries
Class 22 shows the greatest shift in its southern boundary under the 2050s High scenario. The class area contracts to Orkney and Shetland with the class’ climatic conditions no longer represented in Northern Ireland at all.
MONARCH 2 Report – Chapter 2 27 ______
2.6 Selected case study areas
On the basis of both criteria for the selection of areas and the potential changes in conservation interest for which the models need to be tested, the case study areas chosen for further analysis in MONARCH 2 were:
1. Hampshire, England (potential influx of new species from continental Europe) 2. Snowdonia, Wales (large bioclimate change in area of high conservation interest) 3. Central Highlands, Scotland (potential large reduction in sensitive habitat) 4. A cross border area around Cuilcagh/Pettigo in Ireland (representing class 12 with greatest area of ASSIs).
Figure 2.8 shows the location of these case study areas.
Figure 2.8: The location of the case study areas.
2.7 Selection of habitats and species
Habitat and species selection was undertaken by members of the Research Team in conjunction with the funders and other stakeholders. A workshop to assist this process was held in all the case study areas, apart from Central Highlands. The selection was constrained by a protocol, detailed below, and by the limited capacity to model species within a relatively short time frame.
The case study areas were selected to encompass an area of 1,000 – 2,500 km2 and thereby suitable for landscape-scale modelling of climate change impacts. This size was sufficient to embrace the habitats of conservation interest for which the region as a whole had been selected.
28 MONARCH 2 Report – Chapter 2 ______
2.7.1 Selection criteria
The aim was to focus on two habitats per case study area and to choose four or five species for each habitat according to the criteria below.
2.7.1.1 Habitats
BAP priority habitats (or EU Habitats Directive descriptors), which are:
• characteristic of the current bioclimatic zone (ideally with a restricted geographic range nationally), and, • important in defining the landscape of the case study area.
It is important to note that these selected habitats did not cover the whole case study area, which was larger to facilitate dispersal modelling and determine species that may move into the selected habitats.
2.7.1.2 Species
Within the selected habitat(s), species were identified that fulfilled the following criteria:
• Dominant species: characteristic of each habitat selected, as these will be fundamental to ecosystem functioning (the NVC was used when appropriate to inform choice). The target was to choose two or three species from this category; by nature, dominant species generally would be plants; • Rare or Flagship species: ideally chosen from other taxa (not plants) that are interactive with the dominant species identified above, as they may be affected significantly by any change in status of the latter. The target was to choose one species from this category. This category was accepted as not being necessary for each habitat selected; • Recruitment species: species able to move into the habitat from other habitats within the case study area or from elsewhere, which may have a significant impact on ecosystem functioning. The target was to choose one or two species from this category.
It is important to note that keystone species status was not selected as a criterion, as although such species are critical to ecosystem dynamics they are very difficult to identify without experimental data.
The impact of climate change on habitats was not considered directly, instead the response of species was modelled (Chapter 3) and the implications for community composition deduced (Chapter 4). A maximum of ten species per case study was modelled (due to budgetary and time constraints). The selected species and habitats are given in Table 2.10.
2.8 Bioclimatic classification of the case study areas
The climate variables in each of the study areas were updated using the UKCIP02 datasets. This process was designed to make use of the increased resolution of the UKCIP02 datasets to 5km x 5km. In recalibrating the baseline98 classification using the UKCIP02 baseline (Baseline02) climate, the pattern of the classification in the study areas has changed somewhat (results for each case study area are presented in Chapters 6 to 9). In particular, the squared Mahalanobis distances (a measure of the closeness of fit of the climate of a square to the mean for the class) for the Baseline02 classification were much higher than for the Baseline98, suggesting that there was discrepancy between the baseline climates of the two datasets. With the change in the resolution of the data, it was not unexpected for squares to move to a similar class with the UKCIP02 data. However, in some cases the UKCIP02
MONARCH 2 Report – Chapter 2 29 ______
data placed squares beyond the limits of the Baseline98 classification, indicating that the higher resolution data were markedly different from the UKCIP98 data.
Table 2.10: The habitats and species selected for study within the case study areas. D: dominant, F: Flagship R: rare, Rec: Recruitment Hampshire Beech hangers Wet heath Beech - Fagus sylvatica (D) Cross-leaved heather - Erica tetralix (D) Ash - Fraxinus excelsior (D) Purple moor grass - Molinia caerulea (D) Dog’s mercury - Mercurialis perennis (D) Heather - Calluna vulgaris (Rec) Yellow-necked mouse - Apodemus flavicollis (F) Bog bush-cricket - Metrioptera brachyptera (F) Hawfinch - Coccothraustes coccothraustes (R) Curlew - Numenius arquata (F) Central Highlands Caledonian pine woodland Upland/montane heath Scots pine - Pinus sylvestris (D) Heather - Calluna vulgaris (D) Silver birch - Betula pendula (D) Bilberry - Vaccinium myrtillus (D) Sessile oak - Quercus petraea (Rec) Cowberry - Vaccinium vitis-idaea (D) Hairy wood ant - Formica lugubris (R) Stiff sedge - Carex bigelowii (D) Willow tit - Parus montanus (Rec) Ptarmigan - Lagopus mutus (F) Snowdonia Upland oak woodland Upland/montane heath Sessile oak - Quercus petraea (D) Heather - Calluna vulgaris (D) Bluebell - Hyacinthoides non-scripta (F) Bilberry - Vaccinium myrtillus (D) Common cow wheat - Melampyrum pratense (D) Bracken - Pteridium aquilinum (Rec) Pied flycatcher - Ficendula hypoleuca (F) Western gorse - Ulex gallii (Rec) Stiff sedge - Carex bigelowii (D) Cuilcagh/Pettigo Blanket bog Crowberry - Empetrum nigrum (D) Golden plover - Pluvialis apricaria (F) Hare’s-tail cotton grass - Eriophorum vaginatum (D) Lesser twayblade - Listera cordata (R) White beak-sedge - Rhynchospora alba (D) Downy birch - Betula pubescens (Rec) Deergrass - Trichophorum cespitosum (D) Bracken - Pteridium aquilinum (Rec) Sphagnum moss - Sphagnum cuspidatum (D)
An investigation of the mean variable values for the two baseline datasets showed that sunshine hours showed the greatest difference. The level of discrepancy between the two baseline classifications was somewhat reduced by removing sunshine hours as a variable from the UKCIP02 climate variables. However, despite the removal of the sunshine hours data from the UKCIP02 dataset, there were still some differences between the Baseline98 and 02 classifications, although some of the changes were no more than squares moving between closely related classes. Some classes are closer to each other in ‘statistical space’ than others, so the probability of shifting classes is greater for some classes than for others. This has important implications for estimating the degree of change for a grid square because classes are not uniformly “distant” from each other.
In Hampshire, for example, most of the changes apparent between the Baseline98 and the Baseline02 classifications seem to be small changes reflecting the improved resolution of the UKCIP02 baseline data. However, for the Scottish and Welsh case study areas the changes between classes represented large shifts and in some cases (particularly in Scotland) the changes were sufficient that the UKCIP02 baseline data did not fit within the Baseline98 classification. This means that, for some squares, the effect of the change in baseline data was as great as the effect of applying the 2050s High scenario data. This would suggest that the Baseline98 classification is not a good representation of the pattern of climate in such areas. It is not possible, however, to say if the differences in the datasets would have affected the basis upon which the study area was selected. Such an assessment would require the creation of a Baseline02 classification at the national level and then its perturbation using the 2050s High emission scenario (essentially a repeat of the Baseline98 classification process but using the UKCIP02 dataset).
30 MONARCH 2 Report – Chapter 2 ______
More detail on the nature of the differences between the classification pattern using the 10km x 10km UKCIP98 data and that using the 5km x 5km UKCIP02 data is discussed in the individual case study area chapters. However, in order to discover why the discrepancies between the two baseline classifications may have occurred, a comparison was made of the ways in which the UKCIP98 and UKCIP02 baseline datasets were derived.
2.8.1 Relationship between 10km UKCIP98 and 5km UKCIP02 data sets
2.8.1.1 Differences in the number of stations used to generate the datasets
The UKCIP98 data were derived from a restricted number of stations, ranging from 80 for the wind speed variables to 750 for the rainfall (Hulme and Jenkins, 1998). By contrast the UKCIP02 data made use of the Meteorological Office’s more extensive archive of UK observations. Table 2.11 shows the total number of sites and the number of sites per month available for each variable type. As was shown with the UKCIP98 Irish data, the use of too few meteorological stations can have a significant effect on the quality of the interpolated data (see Hossell et al., 2001 for more details of this issue) by smoothing over the variation that does occur.
Table 2.11: Number of sites used to provide data for the UKCIP02 baseline dataset (after Hulme et al., 2002). Variable type Total no. of sites No. of sites per month Temperature 1400 550 Rainfall (no missing days allowed) 11000 4000 Sunshine 650 300 Wind (1969-90) 400 200 Cloud (used to calculate sunshine hours) 400 200
2.8.1.2 Differences in interpolation techniques
The UKCIP98 baseline dataset used a partial thin-plate splines technique to interpolate the station data (Barrow et al., 1993). The method uses elevation, latitude and longitude as independent predictor variables to improve the pattern of the gridded data. The UKCIP02 data are interpolated using an inverse cubed weighting technique, with latitude, longitude, altitude, coastal proximity and urbanisation used for different variables as independent predictors of the climate variables. This technique was preferred over less well-proven methods such as splines and kriging (Hulme et al., 2002).
2.8.1.3 Local spatial variation in climate variables
There may be significant differences in the spatial pattern of climate between the 10km x 10km and 5km x 5km scale where there are local differences in topography. Proximity to the coast may also affect spatial variation in climate at this scale.
Whilst it is difficult to assess the effects of these individual differences between the datasets, it is clear that considerable differences do exist between the 10km and 5km baseline datasets. The significance of the differences needs to be noted when undertaking analyses using both datasets or comparing impacts studies that have used the different scenario sources since the starting point for the analyses will be different, as well as the scenario data.
2.9 Uncertainty in future climate change projections
The UKCIP02 climate scenarios (Hulme et al., 2002) have been used in the MONARCH 2 study for model development and application. All the UKCIP02 scenarios are based on the high-resolution
MONARCH 2 Report – Chapter 2 31 ______
regional climate model (HadRM3) from the Hadley Centre for Climate Prediction and Research. Four scenarios are available reflecting differences in greenhouse gas emissions, labelled Low Emissions, Medium-Low Emissions, Medium-High Emissions and High Emissions. Although the UKCIP02 scenarios represent the effects of uncertainties in future emissions on UK climate, they do not cover uncertainties in how the climate system will respond to these emissions (scientific or modelling uncertainty). The reason for this is that, at the time the UKCIP02 scenarios were designed, only the Hadley Centre had a set of model experiments for Europe conducted using regional climate models at a high spatial resolution of 50km2 (Hulme et al., 2002). To address this limitation, scientific uncertainties have been explored by analysing the range of regional climates over Britain and Ireland simulated by other global climate models. It is important that MONARCH 2 research explores this uncertainty so that results can be framed within this wider risk context.
Results from four global climate models (GCMs) have been made available through a data exchange agreement with the EC 5th Framework ACCELERATES and ATEAM projects. These are the DOE PCM model from the USA, the CSIRO2 model from Australia, the CGCM2 model from Canada, and the HadCM3 model from the UK. It is important to note that results from other regional climate models were not available and hence, these additional scenarios are based on coarser scale results from global climate models with spatial resolutions over the UK ranging from 250 to 360 km. These raw climate model outputs were downscaled to a 10-minute grid for Europe within the ACCELERATES/ATEAM projects. Each GCM pattern of change has been scaled to four emissions storylines from the IPCC Special Report on Emissions Scenarios (SRES): B1, B2, A2 and A1FI. These correspond to the UKCIP02 Low, Medium-Low, Medium-High and High Emissions scenarios, respectively.
The UKCIP02 patterns and magnitudes of climate change for the 2050s time slice (2041-2070 average) are compared with those from the four global climate models over Britain and Ireland in Figures 2.9 to 2.12. The patterns for summer temperature change for the UKCIP02 scenarios are similar to the HadCM3 model, the global model used to drive the Hadley Centre high-resolution regional model, with a south-east to north-west gradient (Figures 2.9 and 2.10). However, the magnitude of change differs slightly under both the B1 (Low) and A1FI (High) emissions scenarios, with higher increases in the south-east and lower increases in the north-west for the HadCM3 model. The CGCM2 model also exhibits a south-east to north-west gradient in summer temperature change under both emissions scenarios, but the magnitude of change is less than the UKCIP02 scenarios. The other models differ from the UKCIP02 scenarios in terms of both pattern and magnitude of change. In particular, the PCM model shows significantly lower increases in summer temperature ranging from 0.3 to 1.0 under the B1 emissions scenario and from 1.0 to 2.0 under the A1FI emissions scenario. This compares with ranges of 0.3 to 2.1 and 0.5 to 3.3 for the UKCIP02 Low and High scenarios, respectively.
Changes in summer and winter precipitation for the UKCIP02 High scenario and the comparative A1FI emissions scenario for the four global climate models are shown in Figures 2.11 and 2.12. The UKCIP02 and HadCM3 models are significantly drier than the other three climate models in summer. Mean changes in summer precipitation across Britain and Ireland are –21.8 and –16.1% for the UKCIP02 and HadCM3 models, respectively, whilst the CSIRO2, CGCM2 and PCM models show mean changes of –0.9, +0.8 and +1.4%, respectively. The UKCIP02 High scenario exhibits a south to north gradient, with decreases in summer precipitation projected for virtually all of Britain and Ireland. Alternatively, the PCM, CSIRO2 and CGCM2 models project increases in summer precipitation for Scotland and Northern Ireland and either small increases or small decreases for England, Wales and the Republic of Ireland, depending on the model.
Changes in winter precipitation are more consistent between the different models. All show increases in precipitation, with the exception of the PCM model which projects slight decreases for northern England, southern Scotland and parts of northern Ireland. Changes in winter precipitation range from –2 to +24% across the different models for the A1FI (High) emissions scenario.
32 MONARCH 2 Report – Chapter 2 ______
Figure 2.9: Change in summer mean temperature (oC) for four global climate models under the B1 SRES emissions scenario compared to the UKCIP02 Low emissions scenario for the 2050s.
UKCIP02 Low CGCM2 B1 CSIRO2 B1
HadCM3 B1 PCM B1
MONARCH 2 Report – Chapter 2 33 ______
Figure 2.10: Change in summer mean temperature (oC) for four global climate models under the A1FI SRES emissions scenario compared to the UKCIP02 High emissions scenario for the 2050s.
UKCIP02 High CGCM2 A1FI CSIRO2 A1FI
HadCM3 A1FI PCM A1FI
34 MONARCH 2 Report – Chapter 2 ______
Figure 2.11: Change in summer precipitation (%) for four global climate models under the A1FI SRES emissions scenario compared to the UKCIP02 High emissions scenario.
UKCIP02 High CGCM2 A1FI CSIRO2 A1FI
HadCM3 A1FI PCM2 A1FI
MONARCH 2 Report – Chapter 2 35 ______
Figure 2.12: Change in winter precipitation (%) for four global climate models under the A1FI SRES emissions scenario compared to the UKCIP02 High emissions scenario.
UKCIP02 High CGCM2 A1FI CSIRO2 A1FI
HadCM3 A1FI PCM2 A1FI
36 MONARCH 2 Report – Chapter 2 ______
2.10 Discussion and conclusions
This chapter has outlined the basis on which the UKCIP98 baseline data for 1961-90 were used to define a bioclimatic classification, which was then characterised using protected site and BAP habitat distribution information. At the Britain and Ireland scale of the initial MONARCH study, the 10km gridded dataset of the UKCIP98 scenarios provided an appropriate resolution to derive the Baseline98 classification. It allowed the development of a classification of sufficient detail that could be readily matched to the data on nature conservation sites and habitats. Perturbing this dataset with the UKCIP98 2050s High scenario highlighted a number of areas that appear to be particularly sensitive to climate change. Based on the conservation information and bioclimatic sensitivity data, four case study areas were selected in which to further test the species impact models. These areas were Hampshire, Central Highlands, Snowdonia and a cross border area around Cuilcagh/Pettigo in Ireland.
In focusing in more detail on the case study areas, it is clearly valuable to use climate data at a higher resolution. Hence, the bioclimatic classification was recalibrated in each area to the UKCIP02 baseline data. In doing this analysis, differences between the two baseline datasets were highlighted. These were particularly apparent in the upland case study areas and may reflect the lower number of stations used in the UKCIP98 dataset. However, since the two UKCIP baseline datasets have been developed using such different techniques and with different sources of climate data, it is impossible to determine why the datasets produce such different climate patterns over the case study areas. It also means that it is difficult to compare any analyses that have been undertaken on both datasets, since it is not possible to determine if any difference in results is due to the different scenario projections within the datasets or related to the different starting points of the baselines. This adds to the uncertainty of the climate change impact projections.
2.11 References
Barrow, E. M., Hulme, M. and Jiang, T. (1993). A 1961-90 baseline climatology and future climate change scenarios for Great Britain and Europe. Part I: A 1961-90 Great Britain baseline climatology. A report prepared for the TIGER IV3a consortium, Climatic Research Unit, Norwich. 43pp.
Hossell, J. E., Riding, A. E. and Brown, I. (2003). The creation and characterisation of a bioclimatic classification for Britain and Ireland. Journal for Nature Conservation, 11, 5-13.
Hossell, J. E., Riding, A. E., Dawson, T. P. and Harrison, P. A. (2001). Bioclimatic classification for Britain and Ireland. In: Harrison, P. A., Berry, P. M. and Dawson, P. M. (Eds.) Climate Change and Nature Conservation in Britain and Ireland: Modelling natural resource responses to Climatic Change (the MONARCH project). UKCIP Technical Report, Oxford, pp 13-22.
Hulme, M. and Jenkins, G. (1998). Climate change scenarios for the United Kingdom. Technical Report No 1, Scientific Report, UK Climate Impacts Programme, Norwich. 60pp.
Hulme, M., Jenkins, G.J., Lu, X., Turnpenny, J. R., Mitchell, T.D., Jones, R.G., Lowe, J., Murphy, J.M., Hassell, D., Boorman P., McDonald, R. & Hill, S. (2002). Climate Change Scenarios for the United Kingdom: The UKCIP02 Scientific Report. Tyndall Centre for Climate Change Research, School of Environmental Sciences, University of East Anglia, Norwich, UK, 120 pp.
MathSoft (1997). S-PLUS 4 Guide to Statistics. Seattle, Data Analysis Products Division, MathSoft.
Parmesan, C., Ryrholm, N. Stefanescu, C., Hill, J.K., Thomas, C.D., Descimon, H., Huntley, |B., Kaila, L., Kullberg, J., Tammaru, T., Tennent, J., Thomas, J.A., and Warren, M. (1999). Poleward shifts in geographical ranges of butterfly species associated with regional warming. Nature 399, 579- 583.
StatSoft, Inc (2001). STATISTICA for Windows. Tulsa, OK, StatSoft, Inc.
MONARCH 2 Report – Chapter 2 Annex 37 ______
Chapter 2 Annex Technical details of climatic analysis
A2.1 Climatic variables
Table A2.1 shows the variables used in the Baseline98 classification of MONARCH 2, with shaded rows indicating the variables not included in the classification developed under MONARCH 1.
Table A2.1: Climate variables used in the Monarch 2 bioclimatic classification. Variable type Abbreviation Unit of Number Description measurement of variables Mean Temp MeanT °C 12 Minimum temp MinT °C 12 Growing Degree Days >5°C GDD °C 1 Rainfall PPT mm/day 12 Potential evapotranspiration PET mm/day 12 Hydrologically effective HER mm/day 12 rainfall Absolute minimum Abs Min °C 1 temperature over 20 years Maximum temperature of the MaxT °C 1 warmest month Mean wind speed MeanW m/s 12 Average spring rainfall Spngppt mm/month 1 Total Sunshine hours Sun hours/month 12 Calculated from cloud cover for all timeslices Growing season length GSL Days 1 Calculated using the sign curve estimate of the daily temperature. Growing season starts on 10th day above 5°C and ends on 5th day below 5°C
A2.2 Principal Component Analysis results
The 89 variables given in Table A2.1 were used within the PCA to reduce the number of variables for the clustering analysis and to remove any correlation between the original variables. Table A2.2 shows the eigenvalues for each of the seven factors derived from the PCA. The eigenvalue is a measure of the number of the original variables explained by the new factor. These factors accounted for more than 96% of the variation within the Baseline98 climate data.
Table A2.2: Eigenvalues from PCA Factor Eigenvalue % Total Cumulative Cumulative variance Eigenvalue % 1 28.63453 32.17363 28.63453 32.17363 2 27.33975 30.71882 55.97428 62.89245 3 13.34653 14.9961 69.32081 77.88855 4 11.89566 13.36591 81.21647 91.25446 5 2.142563 2.407375 83.35903 93.66183 6 1.530370 1.719517 84.88940 95.38135 7 0.899150 1.010281 85.78855 96.39163
38 MONARCH 2 Report – Chapter 2 Annex ______
So for example, the eigenvalue for Factor 1 shows that 28.6 of the 89 starting variables are explained by this new factor.
For the UKCIP98 2050s High scenario PCA eight principal components were derived (Table A2.3), suggesting a more complex climatic pattern under this scenario.
Table A2.3: Eigenvalues for the eight principal components in the UKCIP98 2050s High PCA
Factor Eigenvalue % Total Cumulative Cumulative variance Eigenvalue % 1 28.82451 32.387 28.82451 32.387 2 26.89026 30.214 55.71477 62.601 3 13.50827 15.178 69.22304 77.779 4 11.90176 13.373 81.1248 91.152 5 1.500363 1.6858 82.62516 92.8378 6 1.563382 1.7566 84.18855 94.5944 7 1.046196 1.1755 85.23474 95.7699 8 1.089764 1.2245 86.32451 96.9944
A2.3 Hierarchical clustering
This is calculated using squared Euclidean distance as an indication of the maximum dissimilarity between members of each class (Venables and Ripley, 1994). Figure A2.1 gives the squared Euclidean distances for class solutions between two and 50.
Figure A2.1 Euclidean distances for the first 50 cluster solutions for the Baseline98 classification.
MONARCH 2 Report – Chapter 2 Annex 39 ______
A2.4 Discriminant function analysis
An alternative measure of similarity of the cluster analysis and discriminant function analysis predictions for class membership may be measured by a similarity index (S) (Carey et al., 1995):
S=2c/(a1+a2)
Where c is the number of 10km2 grid squares in common between two classes and a1 and a2 are the numbers of squares in those classes respectively. With this measure the match between the original and the discriminant function classification is 81.67%.
Table 2.4 provides the breakdown of the membership of the two classifications for Britain and Ireland.
Table 2.4: Comparison of class membership between cluster analysis and discriminant function analysis Britain Ireland
Class number Clustered Class Predicted Class Clustered Class Predicted Class numbers Numbers numbers Numbers 1 145 168 38 45 2 22 20 3 47 64 36 55 4 47 35 251 277 5 88 74 4 3 6 35 36 7 156 157 8 54 50 146 131 9 126 135 6 1 10 729 790 58 4 11 59 58 14 19 12 129 92 288 327 13 3 43 44 14 312 248 15 68 71 16 13 18 2 2 17 426 438 56 56 19 22 22 20 131 115 21 8 10 22 96 93 29 6 23 19 20 24 29 40 25 18 18 26 32 43
Table 2.5 shows some of the climatic characteristics of the classes with variables selected to represent the first three factors of the PCA.
40 MONARCH 2 Report – Chapter 2 Annex ______
Table 2.5: Mean values for key bioclimatic variables in each baseline class (bold shows greatest value, italics shows lowest value). Related 1 2 3 PCA Factor Class Summer Winter GDD >5 °C GSL Absolute Mean Jan Wind HER HER (days) minimum Speed (m/s) (mm/day) (mm/day) temperature (°C) 1 0.75 4.49 1240.11 245.30 -17.37 7.29 2 3.77 10.03 839.23 189.36 -20.87 7.59 3 0.44 3.72 1443.49 307.19 -15.87 8.13 4 0.48 4.04 1580.33 311.26 -15.71 5.87 5 1.76 6.63 1044.85 211.87 -19.19 7.39 6 0.79 5.00 605.37 160.46 -23.68 7.10 7 0.01 3.17 1980.47 352.11 -14.52 6.43 8 0.15 2.88 1789.57 339.38 -14.92 6.66 9 0.48 4.39 1097.73 210.33 -19.54 6.20 10 0.02 1.78 1518.78 249.32 -17.68 5.48 11 1.33 5.87 1300.90 249.71 -17.35 7.17 12 0.28 3.26 1420.46 258.43 -17.14 5.22 13 1.56 6.45 1471.61 305.28 -15.81 6.29 14 0.02 1.65 1618.78 270.52 -16.84 6.62 15 0.94 5.85 866.57 188.40 -21.21 6.54 16 1.75 7.35 1284.13 234.13 -18.37 6.70 17 0.01 2.30 1768.72 280.75 -16.53 5.10 18 0.79 5.36 1788.58 353.29 -14.05 6.56 19 2.46 8.24 1203.46 245.77 -17.27 8.08 20 0.19 2.80 1013.07 201.80 -20.12 6.07 21 3.85 9.79 966.88 208.25 -19.21 8.44 22 0.18 3.32 1090.51 234.32 -17.27 7.79 23 2.49 7.98 1391.14 300.05 -15.93 6.47 24 2.71 8.19 988.53 205.76 -19.45 7.66 25 2.71 9.06 737.19 178.28 -21.84 7.53 26 0.35 3.59 772.49 177.63 -22.13 6.69
A2.5 Baseline98 and Baseline02 classifications
With the removal of the sunshine hours from the Baseline98 and 02 datasets the correspondence between the original hierarchical clustering classification (Baseline98) and the discriminant function classification was reassessed to determine the effect of the change in variables. The ability of the reduced Baseline98 set to predict the original hierarchical clustering classification fell slightly from 82.77 to 81.96%.
Even following the removal of the sunshine hours data, there were some discrepancies between the classification of squares in the study areas for the UKCIP98 and UKCIP02 datasets. Some of these changes were relatively small as some classes are climatically very close. Those classes where the statistical separation is relatively small are marked with an asterisk in Table 2.6 and the squared Mahalanobis distance to the nearest class is given.
MONARCH 2 Report – Chapter 2 Annex 41 ______
Table 2.6: The three nearest neighbours to each class and identification of classes that are nearest to other classes (*) based on the squared Mahalanobis distance.
Class Nearest Classes Squared Mahalanobis distance to nearest class 1st 2nd 3rd 1* 12 11 4 23.6 2 25 21 24 83.4 3* 4 8 1 39.2 4* 12 1 8 21.1 5* 11 16 24 35.5 6 26 15 9 316.6 7 17 8 4 80.6 8* 4 3 17 32.1 9* 20 12 15 30.1 10* 14 17 12 19.5 11* 1 5 4 25.3 12* 4 10 1 21.1 13 23 11 4 67.9 14* 10 17 3 19.5 15* 9 5 20 46.5 16 5 11 13 55.1 17* 10 4 1 22.4 18 13 4 8 68.2 19 24 11 15 93.1 20* 9 10 12 30.1 21 2 19 24 88.4 22 10 3 14 70.1 23 13 16 18 67.9 24 5 19 2 64.1 25 2 24 5 83.4 26 20 6 15 66.7
A2.6 Uncertainty in climate change scenarios
Changes in mean temperature and precipitation for all climate models and all seasons are summarised in Tables 2.7 and 2.8. The values highlight the uncertainty surrounding projections of future levels of climate change by indicating the magnitude and pattern of change produced by four other Global Climate Models, which use different parameters and climate sensitivity, by comparison to the UKCIP02 scenarios. The assessment indicates that UKCIP02 scenarios present a slightly warmer and drier summer pattern of change than the other four models considered.
A2.7 References
Carey, P. D., Preston, C., Hill, M. O., Usher, M. and Wright, S. (1995). An environmentally defined biogeographical zonation of Scotland designed to reflect species distributions. Journal of Ecology, 83, 833-845.
Venables, W.N. and Ripley, B.D. (1994). Modern Applied Statistics with S-Plus. New York, Springer- Verlag.
42 MONARCH 2 Report – Chapter 2 Annex ______
Table 2.7: Mean, minimum and maximum changes in mean temperature (oC) across Britain and Ireland for the UKCIP02 Low and High scenarios and the comparative B1 and A1FI emissions scenarios for four global climate models for the 2050s. Climate B1 (Low) emissions A1FI (High) emissions model Spring Summer Autumn Winter Spring Summer Autumn Winter UKCIP02: Mean 1.10 1.46 1.46 1.02 1.75 2.32 2.33 1.62 Minimum 0.50 0.30 0.60 0.60 0.80 0.50 0.90 0.90 Maximum 1.30 2.10 1.80 1.30 2.10 3.30 2.90 2.10 HadCM3: Mean 1.35 1.39 1.38 1.46 2.01 2.30 2.23 1.88 Minimum 0.67 0.30 0.70 0.80 0.93 0.43 0.87 1.10 Maximum 1.83 2.53 1.87 2.00 2.40 4.20 3.30 2.43 CGCM2: Mean 1.00 1.37 1.27 0.96 1.49 2.30 1.90 1.68 Minimum 0.90 1.10 1.07 0.83 1.13 1.27 1.60 1.27 Maximum 1.30 1.57 1.43 1.37 2.00 2.83 2.10 2.13 CSIRO2: Mean 1.09 1.56 1.95 1.80 1.23 1.61 1.89 1.89 Minimum 0.23 0.97 1.47 1.40 0.63 0.93 1.47 1.40 Maximum 1.90 1.93 2.33 1.97 1.90 1.97 2.20 2.13 PCM: Mean 0.67 0.40 0.76 0.66 1.39 1.21 2.00 1.65 Minimum 0.50 0.33 0.70 0.43 1.10 1.00 1.80 1.23 Maximum 1.40 0.97 1.03 2.27 3.03 2.00 2.73 4.97
Table 2.8: Mean, minimum and maximum changes in precipitation (%) across Britain and Ireland for the UKCIP02 Low and High scenarios and the comparative B1 and A1FI emissions scenarios for four global climate models for the 2050s. Climate B1 (Low) emissions A1FI (High) emissions model Spring Summer Autumn Winter Spring Summer Autumn Winter UKCIP02: Mean -0.98 -13.70 -2.86 7.55 -1.55 -21.77 -4.55 11.99 Minimum -5.60 -20.20 -6.90 0.60 -8.90 -32.10 -11.00 1.00 Maximum 7.40 3.00 3.00 14.50 11.70 4.80 4.80 23.10 HadCM3: Mean 3.58 -11.50 2.74 9.25 3.21 -16.13 6.43 14.94 Minimum -2.77 -18.90 -5.03 -0.67 -4.53 -28.93 -0.77 -2.13 Maximum 9.77 3.53 10.03 18.03 9.23 0.97 15.63 23.57 CGCM2: Mean 4.19 -0.81 4.50 6.95 3.02 0.83 0.43 11.78 Minimum -0.53 -8.00 -3.27 -0.30 -7.27 -18.57 -4.43 3.07 Maximum 8.13 6.90 12.00 10.37 12.57 11.13 8.93 17.87 CSIRO2: Mean 6.86 4.00 4.97 4.45 4.10 -0.87 4.50 8.01 Minimum 1.00 -1.73 -1.43 0.77 -1.57 -6.03 2.20 3.23 Maximum 9.67 6.77 9.67 9.83 7.60 2.93 8.70 17.57 PCM: Mean 3.38 2.53 0.53 1.64 8.05 1.44 1.04 2.68 Minimum 1.87 -0.50 -2.20 0.13 3.37 -8.27 -5.30 -1.90 Maximum 8.97 6.23 5.23 10.57 21.60 12.03 14.80 18.97
MONARCH 2 Report – Chapter 3 43 ______3 Species distribution and dispersal modelling
P.A. HARRISON, R.G. PEARSON, T.P. DAWSON, S. FREEMAN, J.E. HOSSELL, H. LYONS, P. SCHOLEFIELD AND P.M. BERRY
Summary
Three models for analysing the impacts of climate and land cover change on potential species distribution are described. The SPECIES model employs an Artificial Neural Network (ANN) to characterise bioclimate envelopes based on inputs generated through a climate-hydrological process model. An important element of the model is its multi-scale approach whereby the bioclimate envelope of a species is first identified at the European scale before application at a finer resolution within Britain and Ireland. This enables climatic range margins that are currently outside Britain and Ireland, but which may move into these countries under future climate scenarios, to be identified. The model therefore does not extrapolate outside its training dataset when used to predict the suitable climate space for species in Britain and Ireland under potential future climates.
The downscaled SPECIES model integrates the European scale climate-driven simulation with fine- scale land cover data in a second ANN to generate regional scale suitability surfaces for species. This provides an insight into the roles of climate and land cover as determinants of species’ distributions and enables predictions of distributions under scenarios of changing climate and land cover type to be examined either separately or together and for interactions to be explored. A generalised additive model is used to produce consistent scenarios of changes in key vegetative land cover types under climate change. The downscaling of the SPECIES model also facilitates coupling of the modelled species’ suitability surfaces with dynamic simulations of species dispersal. Species dispersal over a specified time step is simulated within the boundaries of the climate and land cover suitability surfaces. The model is based on a spatially explicit cellular automaton, which simulates the stochastic dispersal of species in terms of two main processes: the release of a number of propagules by an existing population and the redistribution of the propagules according to a dispersal function. The three inter-related species models produce maps of the probability of occurrence for each species; based on dispersal characteristics, climatic suitability and land cover for future scenarios.
3.1 Introduction
Climate change and land cover change, leading to habitat loss, are two of the most important factors threatening ecosystems worldwide (Parmesan and Yohe, 2003, Sala et al., 2000). Considered individually, the threats posed are significant, but the interaction between the two factors could be disastrous (Travis 2003). A number of studies have modelled the potential impacts of climate change on the distribution of species, applying climate-driven simulations across a range of scales, study areas and species (for reviews see Guisan and Zimmermann, 2000; Pearson and Dawson, 2003). However, few modelling studies have explicitly addressed climate – land cover interactions. To study the combined effects of climate change and habitat fragmentation (as driven by changes in land cover), a novel modelling approach has been developed that integrates climate and land cover drivers in a scale-dependent hierarchical manner.
The spatial scale at which species distribution modelling is undertaken is of fundamental importance, with the selection of appropriate spatial extent and data resolution for a given application essential. It has been proposed that climate impacts on the distribution of species will be most apparent at macro- scales, with broad spatial extents and coarse data resolutions most appropriate for correlating climate with species distributions. This is the premise behind the development of the SPECIES model, which uses an Artificial Neural Network (ANN) to simulate the potential climate space of species at the European scale (Pearson et al., 2002). It has also been hypothesised that within the climate space defined by synoptic climate conditions other factors influence the distribution of species in a hierarchical manner, with different factors being better correlates at different scales (Collingham et
44 MONARCH 2 Report – Chapter 3 ______al., 2000; Franklin, 1995; Pearson and Dawson, 2003). Thus, it is proposed that land cover may be considered the dominant control of species presence and absence at a finer spatial resolution than climate. This is the premise behind the development of the downscaled SPECIES model, which integrates the broad-scale climate-driven simulation with finer-scale land cover data (Pearson et al., 2004). Outputs from the continental scale ANN are used as inputs to a second ANN, along with land cover data, to generate regional scale suitability surfaces for species. A suitability surface is defined as a landscape identifying areas where a species could potentially grow and reproduce, and is analogous to an approximation of the spatial manifestation of the fundamental niche. The downscaled SPECIES model has been designed to be applicable to scenarios of future climate and land cover change. Land cover change scenarios are based on the characterisation of climate envelopes for existing land cover classes and their subsequent projection under climate change. The downscaling of the SPECIES model also facilitates coupling of the modelled species’ suitability surfaces with dynamic simulations of species dispersal. Regional-scale predictions of potential future environmental impacts will be more suited to the requirements of local conservation policy planning than those of previous studies looking at the potential impacts and policy implications of broad-scale climate change, such as those investigated in MONARCH 1 (Harrison et al., 2001).
3.2 SPECIES model
The SPECIES (Spatial Estimator of the Climate Impacts on the Envelope of Species) model was used to simulate the impacts of climate change on the potential climatic suitability of individual species in the MONARCH 1 project (Berry et al., 2001). A full description of the model is given in Pearson et al. (2002). The model uses an ANN to integrate bioclimatic variables for predicting the distribution of species through the characterisation of bioclimatic envelopes. A number of integrated algorithms, including a climate-hydrological process model, are used to pre-process climate (temperature, precipitation, solar radiation, vapour pressure and wind speed) and soils (AWC – available water holding capacity) data to derive relevant bioclimatic variables for input to the neural network. Those variables found to be most successful for bird distributions (Harrison et al., 2003) and other taxa (Berry et al., 2003) are given in Table 3.1.
Table 3.1: Bioclimatic input variables used for birds and other taxa in the SPECIES model. Birds Other taxa Growing degree days > 5°C Growing degree days > 5°C Absolute minimum temperature expected over a Absolute minimum temperature expected over a 20-year period 20-year period Mean summer temperature (May, June, July) Annual maximum temperature Mean summer precipitation (May, June, July) Accumulated annual soil water deficit Mean winter precipitation (December, January, Accumulated annual soil water surplus February) Mean summer water availability (May, June, July)
The model is trained using existing empirical data on the European distributions of species to enable the full climate space of a species to be characterised and to capture their response to climatic conditions that might be expected under future scenarios. A kriging interpolation function is applied to the presence/absence distributions of each species to provide a smoothed suitability surface. The data are then randomly divided into three groups for training, validating and testing the neural network. The validation set ensures that the network does not over-train on the training data, thus losing its ability to generalise, while the test data is used to independently verify the prediction.
Two methods for assessing the predictive performance of each network have been used: Cohen’s Kappa statistic of similarity (k) and the Area Under the Receiver Operating Characteristic Curve (AUC). Kappa is a commonly used statistic that provides a measure of proportional accuracy, adjusted for chance agreement (Cohen, 1960). Kappa varies from 0, indicating no agreement between
MONARCH 2 Report – Chapter 3 45 ______observed and predicted distributions, to 1 for perfect agreement. AUC is an unbiased measure of prediction accuracy calculated from the Receiver Operating Characteristic (ROC) curve. The ROC curve describes the compromise that is made between the sensitivity (defined as the proportion of true positive predictions versus the number of actual positive sites) and false positive fraction (the proportion of false positive predictions versus the number of actual negative sites). This index is independent of both species prevalence and the decision threshold (Fielding and Bell, 1997). AUC ranges from 0.5 for models with no discrimination ability, to 1 for models with perfect discrimination.
Tables 3.2 and 3.3 show the statistical performance of the European neural network models at replicating the test dataset, which was not used for model training. The AUC statistic is greater than 0.9 for all European models, indicating very good discrimination ability. The maximum Kappa statistic is slightly lower for most species, but this is to be expected as the index can vary between 0 and 1. Here, 18 species show values greater than 0.85, indicating excellent agreement between observed and simulated distributions, 12 species show values between 0.7 and 0.85 indicating very good agreement, and 2 species show a value between 0.55 and 0.7 indicating good agreement.
Table 3.2: Statistics showing predictive performance of the SPECIES model for the bird species (see Table 2.10 for common names). Species Kappa Numenius arquata 0.68 Pluvialis apricaria 0.86 Coccothraustes coccothraustes 0.74 Ficendula hypoleuca 0.81 Lagopus mutus 0.82 Parus montanus 0.81
Table 3.3: Statistics showing predictive performance of the SPECIES model for other taxa (see Table 2.10 for common names). Species AUC Kappa Species AUC Kappa Hampshire, England: Snowdonia, Wales: Fraxinus excelsior 0.986 0.885 Quercus petraea 0.983 0.850 Fagus sylvatica 0.977 0.794 Melampyrum pratense 0.989 0.899 Mercurialis perennis 0.982 0.888 Hyacinthoides non-scripta 0.986 0.883 Apodemus flavicollis 0.960 0.714 Calluna vulgaris 0.991 0.895 Erica tetralix 0.986 0.865 Vaccinium myrtillus 0.988 0.903 Molinia caerulea 0.986 0.912 Carex bigelowii 0.953 0.697 Calluna vulgaris 0.991 0.895 Ulex gallii 0.996 0.760 Metrioptera brachyptera 0.992 0.907 Pteridium aquilinum 0.993 0.940 Central Highlands, Scotland: Cuilcagh/Pettigo Peatlands, Ireland: Pinus sylvestris 0.983 0.887 Betula pubescens 0.989 0.921 Betula pendula 0.987 0.886 Empetrum nigrum 0.972 0.796 Formica lugubris 0.949 0.706 Eriophorum vaginatum 0.995 0.944 Quercus petraea 0.983 0.850 Listera cordata 0.971 0.816 Vaccinium myrtillus 0.988 0.903 Pteridium aquilinum 0.993 0.940 Calluna vulgaris 0.991 0.895 Rhynchospora alba 0.976 0.848 Carex bigelowii 0.953 0.697 Sphagnum cuspidatum 0.963 0.747 Vaccinium vitis-idaea 0.995 0.927 Trichophorum cespitosum 0.991 0.903
Once a network is trained, validated and tested at the European scale, it can then be used to produce a map of simulated suitable climate space for baseline (1961-90) climate and to estimate the potential re-distribution of this climate envelope under alternative climate change scenarios, at a finer 5km x 5km spatial resolution in Britain and Ireland.
46 MONARCH 2 Report – Chapter 3 ______
3.3 Downscaled SPECIES model
The SPECIES model has been downscaled to identify areas suitable for a given species at a finer resolution than in the purely climate-driven simulations (Pearson et al., 2004). This refined modelling approach is presented in Figure 3.1. The top half of the schematic presents the original SPECIES model, with continental-scale ANN training driven by climate at a broad spatial resolution of 0.5o latitude/longitude. The British-scale climate suitability surface is then incorporated, along with land cover classes, as input into a second ANN that is trained against British species distributions at a finer spatial resolution of 10km. This second ANN, which aims to characterise the relationship between species distribution, climate and land cover, has then been used to simulate regional scale suitability surfaces for species at 10 km and 1 km resolutions. Within the MONARCH 2 study, the downscaled ANN has been run using output climate suitability surfaces for Great Britain for those species associated with the Hampshire, Snowdonia and Central Highlands case studies, and output climate suitability surfaces for Ireland for those species associated with the Cuilcagh/Pettigo Peatlands case study. Land cover data for Great Britain are based on the CEH Land Cover Map 2000 and for Ireland on the Corine land cover map, which are both available at a 1km spatial resolution.
Figure 3.1: Schematic of the downscaled SPECIES model (taken from Pearson et al., 2004).
Pre-processing: European scale Temperature climate database • indicators • Growing degree days Continental scale GB scale • Soil water (SPECIES model): climate database indicators climate driven (with scenarios) Training using European species’ distribution
GB climate suitability surface
GB land-cover classes (25 class system)
Regional scale: climate and land cover driven Training using GB species’ distribution
Predicted species’ distribution
Tables 3.4 and 3.5 show the statistical performance of the regional trained neural network models at replicating the observed species’ distributions. For taxa other than birds, the Area Under the Receiver Operating Characteristic Curve (AUC) was used to assess the predictive performance of each network, as described for the European networks. Kappa values were not used at the regional scale because this statistic is affected by species prevalence such that rare species tend to result in very low kappa statistics that are not necessarily indicative of model performance. Alternatively, AUC is an unbiased measure of prediction accuracy, which is independent of both species prevalence and the decision threshold (Fielding and Bell, 1997).
MONARCH 2 Report – Chapter 3 47 ______
For bird species, the comparative method of Fewster and Buckland (2001) was used which is particularly well-suited to highly mobile organisms, such as birds. The method works by calculating the ‘best attainable match’, after permitting mismatched squares to swap status with their near neighbours (within a given radius, here 30 km) until a maximum number of agreements between model and data, at the level of the single square, is produced. This coefficient, the ‘Best Attainable Match’ (BAM) thus takes account of the positions of matched and unmatched squares, as well as their number, and can then be used as an alternative means of locating an optimal threshold from the suitability surface. Calculations were carried out using the BAM software of Fewster and Buckland (2001).
Table 3.4: Statistics showing predictive performance of the downscaled SPECIES model for birds, both in raw and kriged form. BAM = Best Attainable Match, SMC = Simple Matching Coefficient, the number of squares assigned to the correct status (see Fewster and Buckland, 2001). Species Raw Atlas Data Kriged Atlas Data Threshold SMC (%) BAM (%) Threshold SMC (%) BAM (%) Numenius arquata 0.40 80.5 89.9 0.50 79.7 87.9 Pluvialis apricaria 0.50 88.0 92.7 0.45 88.0 94.4 Coccothraustes 0.25 88.4 95.0 0.20 90.2 93.8 coccothraustes Ficendula hypoleuca 0.30 81.7 92.5 0.40 83.1 92.2 Lagopus mutus 0.30 97.0 98.7 0.45 96.6 98.5 Parus montanus 0.45 84.7 94.8 0.55 82.8 93.1
Table 3.5: Statistics showing predictive performance of the downscaled SPECIES model for taxa other than birds. Species AUC Species AUC Hampshire, England: Snowdonia, Wales: Fraxinus excelsior 0.971 Quercus petraea 0.740 Fagus sylvatica 0.913 Melampyrum pratense 0.761 Mercurialis perennis 0.875 Hyacinthoides non-scripta 0.897 Apodemus flavicollis (M) 0.823 Calluna vulgaris 0.890 Erica tetralix 0.914 Vaccinium myrtillus 0.931 Molinia caerulea 0.867 Carex bigelowii 0.957 Calluna vulgaris 0.890 Ulex gallii 0.868 Metrioptera brachyptera (I) 0.883 Pteridium aquilinum 0.913 Central Highlands, Scotland: Cuilcagh/Pettigo Peatlands, Ireland: Pinus sylvestris 0.949 Betula pubescens 0.784 Betula pendula 0.896 Empetrum nigrum 0.660 Formica lugubris (I) 0.747 Eriophorum vaginatum 0.569 Quercus petraea 0.740 Listera cordata 0.775 Vaccinium myrtillus 0.931 Pteridium aquilinum 0.734 Calluna vulgaris 0.890 Rhynchospora alba 0.807 Carex bigelowii 0.957 Sphagnum cuspidatum 0.513 Vaccinium vitis-idaea 0.921 Trichophorum cespitosum 0.826 M = Mammal I = Insect
The results for bird species show occasional variation in the level of the optimal threshold, but remarkable consistency between the quality of the model fit at those thresholds, between the raw and kriged analyses. The Best Attainable Match ranges between 87.9% for Numenius arquata (curlew) and 98.5% for Lagopus mutus (ptarmigan) among the six species, compared to the simple proportions of correctly predicted cell statuses, which ranged from 79.7% to 96.6%. All subsequent analyses are based on the kriged data. Predicted distributions based on these data and a threshold as determined by the BAM coefficient in Table 3.4 reveal a strikingly accurate depiction of the British distributions for all six species.
48 MONARCH 2 Report – Chapter 3 ______
Results for taxa other than birds show that 8 of the species modelled have AUC values greater than 0.9 indicating very good discrimination ability, whilst 16 species have values between 0.7 and 0.9 indicating reasonable discrimination ability and 3 species have values less than 0.7 indicating poor discrimination ability. The discrimination ability of the models is generally less at the regional scale than at the European scale, but this reflects the greater fragmentation, and sometimes greater rarity, of observed species’ distributions at finer spatial resolutions.
Once a network is trained, validated and tested at the regional scale (on a 10km grid for Great Britain or Ireland), it can then be used to produce a map of the simulated distribution for baseline (1961-90) climate and land cover and to estimate the potential re-distribution of a species under alternative climate and land cover change scenarios at a finer 1km x 1km spatial resolution in each case study region.
3.3.1 Identifying decision thresholds
To aid the interpretation and presentation of model results it is useful to identify a threshold value above which model outputs are considered to represent species presence. The choice of threshold value is important because model outputs, when mapped as presence/absence, may look quite different dependent on the threshold applied. A threshold is commonly identified by maximising the agreement between observed and simulated distributions. One approach is to use the threshold value that maximises the Kappa statistic of agreement, as described in the previous section (and applied in Pearson et al., 2002). An alternative approach, based on the ROC procedure, is to plot sensitivity against specificity (defined as the proportion of true negative predictions versus the number of actual negative sites) at a series of thresholds and to apply the threshold value at which these two curves cross (Figure 3.2, threshold a; Thuiller et al., 2003). In this way, the cost arising from an incorrect decision is balanced against the benefit gained from a correct prediction (Manel et al., 2001).
Figure 3.2: Sensitivity and specificity plotted against threshold for defining decision thresholds. Threshold a is assigned at the point where the two curves cross. Thresholds b and c are defined by sensitivities of 90 and 95 % respectively (taken from Pearson et al., 2004).
The above threshold values, based on maximising agreement between observed and simulated distributions, have been calculated for the European scale ANN results in the MONARCH 2 study. However, it may be argued that these approaches do not represent the most appropriate threshold for identifying those sites where a species could exist (i.e. the potential distribution). Given the many factors that influence the actual distribution of species, maximising the fit between simulated
MONARCH 2 Report – Chapter 3 49 ______suitabilities and the observed distribution is likely to result in an underestimation of the extent of the potential distribution. In fact, we should be more concerned with minimising the number of sites with observed presences that are predicted to be unsuitable, than with minimising the number of sites without actual presences that are simulated as suitable. More formally, the primary concern when identifying a decision threshold should be to minimise the false negative fraction (the proportion of false negative predictions versus the number of actual positive sites). Minimising the false negative fraction is analogous to maximising sensitivity (since sensitivity = 1 – false negative fraction). It is therefore appropriate to define thresholds by assigning cut-off sensitivities as outlined in Figure 3.2. Cut-off values defined by sensitivity values of 90 and 95% (thresholds b and c in Figure 3.2) have been applied in MONARCH 2 for the regional scale ANN results. These thresholds are presented along with those delimited by balancing sensitivity and specificity, giving three levels of confidence in model simulations.
3.4 Land cover change scenarios
A land cover model has been developed in order to create scenarios of land cover change that are consistent with the UKCIP02 climate change scenarios for application to the downscaled SPECIES model. The modelling procedure involved the development of a series of relationships between climate and soils variables (listed in Table A3.2) and the presence/absence of a land cover class. The contribution of the variables to the prediction of presence/absence of a land cover class was explored through the development of a generalised additive model using a logistic link function within Statistica 6.1 (Statsoft Inc, 2000). The parameters that were accepted and retained by the regression procedure were then used to generate predictive maps. The models are based on land cover data from the CEH Land Cover Map 2000 (LCM2000) for Great Britain and the Corine land cover map for Ireland. Baseline climate data covering the 30-year period from 1971 to 2001 were used to match the timescale over which the land cover data were collated. The modelling also made use of the soils datasets assembled for the project (see Chapter 1.3.1.3). Since the Scottish soils data were derived from a separate dataset and contained different variables, the land cover modelling was developed for England and Wales, Scotland and Ireland separately.
The LCM2000 data used were the 27 subclasses (level 2 of the dataset), whilst for the Corine data 44 classes were available. Both datasets provide information on the percentage of a grid cell occupied by each land cover class and are available at a 1km resolution. However, the climate data for the modelling was only available at a 5km grid scale size, so all other data were aggregated to this level. The land cover class percentage coverage data were converted to presence/absence information using a cut-off threshold of 5%, above which a habitat was said to be present, to reduce the potential for errors resulting from the interpretation of the original satellite imagery. The base presence/absence data, identified in this manner, was then used to create the 5 x 5 km surfaces on which the models were based. The model was applied to all land cover classes that were considered to be impacted by climate change, but excluded land covers that were ubiquitous or where the baseline pattern of land cover data was considered to be a poor representation of reality. Table 3.6 lists the LCM2000 classes modelled for England and Wales and Scotland and the Corine classes modelled for Ireland. Classes excluded from the model for any of the reasons given above were still used as input to the downscaled SPECIES model, but their distributions under the climate change scenarios were not adjusted from their baseline values.
The climate and land cover class data were randomly sampled and partitioned into training (70% of dataset) and validation (30% of dataset) subsets for use in the model development and validation analyses. Probability values were obtained for each respective land cover class and the threshold probability for presence of a land cover class was determined by maximising the agreement between the training and validation datasets. Interpretation of the predictive power of the models and the level of agreement between the training and validation sets was made using Cohen’s Kappa statistic of similarity (k) and the Area Under the Receiver Operating Characteristic Curve (AUC), which are described in Section 3.2.
50 MONARCH 2 Report – Chapter 3 ______
Table 3.6: Land cover classes used in the modelling study from LCM2000 for England & Wales (E&W) and Scotland (S), and from Corine for Ireland.
LCM2000 classes Corine classes Code Description Code Description 8 Bog, deep peat (E&W; S) 12 Non-irrigated arable land 9 Dense dwarf shrub (E&W; S) 18 Pastures 10 Open dwarf shrub (E&W; S) 20 Complex cultivation patterns 11 Montane habitats (S) 21 Land principally occupied by agriculture 12 Broad leaved / mixed woodland (S) 23 Broad-leaved woodland 15 Neutral grass (E&W; S) 24 Coniferous forest 17 Bracken (E&W; S) 26 Natural grasslands 18 Calcareous grass (E&W; S) 27 Moors and heathland 19 Acid grassland (E&W; S) 29 Transitional woodland-shrub 20 Fen, marsh, swamp (E&W) 33 Burnt areas 35 Inland marshes 36 Peat bog
Figure 3.3 shows the pattern of representation of baseline land cover classes across each of the model areas. For every class the maximised kappa values, AUC values and percentage of squares correctly predicted by the model have been calculated (Tables 3.7 to 3.9) and an interpretation of the predictive power of the model made based on the Kappa statistic (Manel et al, 2001; Landis and Koch, 1977). The Receiver Operating Characteristic curves (Figures A3.4- A3.6 in the Annex to this Chapter) show that all three models perform well for most land cover classes. The models for Scotland have consistently lowest predictive power, whilst the England and Wales models are generally best. The lower predictive power of the Scottish datasets is probably a function both of the poorer soils datasets available and the patchiness of the baseline distributions of the land cover types. For Scotland, the soil information was related to depth and water content, whereas for England and Wales the variables also included information on clay content and humus content. There are also difficulties associated with modelling a finely mosaiced land cover distribution as occurs for most of the land cover types in Scotland, since the presence/absence pattern occurs at a finer resolution than was used for modelling.
For Ireland there is wide variation in the reliability of the models. Results from the Irish models for classes 21 (predominantly occupied by agriculture), 23 (broad leaved forest), 29 (transitional woodland shrub) and 35 (inland marshes) should be viewed with caution. Similarly the bracken class (17) and neutral grassland class (15) models have very low predictive power in Scotland.
Table 3.7: Validation statistics for land cover models constructed for Scotland. Class Description Kappa AUC % Match Interpretation C8 Bog, deep peat 0.440 0.82 87.9 Moderate C9 Dense dwarf shrub 0.421 0.77 74.7 Moderate C10 Open dwarf shrub 0.545 0.85 55.4 Moderate C11 Montane habitats 0.494 0.99 47.0 Moderate C12 Broad leaved / mixed woodland 0.387 0.80 66.4 Slight to fair C15 Neutral grass 0.229 0.71 66.2 Slight to fair C17 Bracken 0.181 0.79 43.4 Little or no agreement C18 Calcareous grass 0.404 0.88 82.6 Moderate C19 Acid grassland 0.346 0.73 66.9 Slight to fair
MONARCH 2 Report – Chapter 3 51 ______
Figure 3.3: Comparison of observed and predicted presence/absence data for acid grassland (England & Wales and Scotland) and moors and heathland (Ireland). Green = present. Purple = absent.
##### ########### ############# #################### ############# ######################### ############# ######################### ##################### ################################### ######################### ########################################## ############################## ############################################## ################# ################################## ###################################### ######################################################################## ############################################## ####################################################################### ####################################################### ################################################################################################ ####################################################### ################################################################################################################ ############################################################## ############################################################################################################################# ####################################################################### ################################################################################################################### ############################################################################### ####################################################################################################################################################### ############################################################################### ########################################################################################################### ######################################################################### #################################################################################################################################### ################################################################################ ################################################################################################################################## ###### # # ########################################################### ###### # ################################################################################################################################### ###################################################################### ############################################################################################################################################## ########################################################################### ############################################################################################################################################# ############################################################################# ########################################################################################################################################################### ###################################################################### ###################################################################################################################################### #################################### #### ######################################################################## #### ############################################################################# #### ############################################################################################################################################### #### ##################################################################################### ########################################################################################################################################################### #### ###################################################################################### #### #################################################################################################################################################################### ############# ## ############################################################################################# ###################### # ### ################################################################################################################################################################################## ###################################### ########################################################################################### #################################################### ########################## ######################################################################################################################################################################## ################################################################################################################################# ################################################################################################################################################################################################################################################## ###################################################################################################################### ################################################################################################################################################################################################################################################# ############################################################################################################################# ######## # #### ############################################################################################################################################################################################################################################# #################### # #### ######################## #################################################################################################### ############################ ################################################## ####################################################################################################################################################################################################### ######################################################### #### ########################################################################################################## ################################### ################## ################################################################################################################################################################### ################################################################# ####################################################################################################################################################### ######################################################################################################################################################################################################################################################################################################### ############################################################################################################################################ ################################################################################################################################################################################################################################################################################################### ######################################################################################################################################################### ###################################################################################################################################################################################################################################################################################################### ############################################################################### ######################################################################################################################################################## ########################################################################################################################################## ############################################################################################################################################################################################################################################################################################ ############################################################################################################################################################### ####################################################################################################################################################################################################################################################################################################################### ################################################################################################################################################################# ###################################################################################################################################################################################################################################################################### ############################################################################################################################################################################ ################################################################################################################################################################################################################################################################################################################################## ######################################################################################################################################################################## ######################################################################################################################################################################################################################################################################################################################################################### ################################################################################################################################################################################# ###################################################################################################################################################################################################################################################################################################################################################################### # ############################################################################################################################################# # ############################################################################################################################################################################################################################################################################################################################## ########################################################################################################################################################################### # ###################################################################################################################################################################################################################################################################################################################################################### # ###### #################### ############################################################################################################################# ###### ############################################ ###################################################################################################################################################################################################################################################### ###### ################## ###################################################################################################################### # ############################################# ################################################################################################################################################################################################################################ ####################### ######################### ################################################################ ############################################## ########################################################################## ############################################################################################################### ############# ########################################################################################################## ###### # ########################### ### ############################################################################################################################################################################################################## ###### # ########################################################################################################################## ###################################################################################################################################################################################################################### ####################################################################################################################################### #################################################################################################################################################################################################################################################################### ############################################################################################################################################################ ################################################################################################################################################################################################################################################################################################################### ########################################################################## #################################################################################################################################### ###################### ################################################################################################################################# ######################################## ############################################################################################################################################################################################################################################################# ################################################################################################################################ ############################################################################################################################################################################################################################################################# ################################################################################################################################## ####### ############################################################################################################################################################################################################################################################ ############## ######################################################################################################## # #### ############################################################################################################################################################################### ####### ####### # #### ############################################## ### ################## #### ###### ############################################################################################ ##### ############################# ############ ###### ############################################# # ### ### ############################################################################################# ###### ##### ############################################ ################################################################## ################################################ ############################################################################################### ################## ############ ############################ #################### ###################### #### ############################################ ######## ##### ##### ### ##### ##### ###### ##### ##### Original LCM2000 data Baseline predicted presence/absence
# # # # #### ##### ### ## ###### ## ### ### ### ## ### ## # #
# #### # #### ################ ###### ############## ########### # ######################## ### ###################### ### ###################### ########## ##################################### ########## ############################################### ############### ######################################## ############### ############################################# ################## ########################################### ################## ########################################### ################## #################################### ################# ##################################### ############ ################################# ############# ################################## ### ############## # ### ############## # #### # ################################ ##### ############################# # ######################################### ########################################## ####### # ######################################## # ######## # ######################################## # ###### ## ################################# ############################ ##### ##### #################################### ###################### #### ## ####### ######## #################################### #### ############################################## ## ####### # #### ##################################### ########################### # ######################## ## ###### ################## ############################# ## ###### ################## ############################# #### ####### ########################################################################################## ### # ####### #### ######################################################################################### #### ########## ######################################################################################## ## ######### ######## ################################################################################## ############################################################################################## ######### # ################################################################################### ######################################################################################### ######### ################################################################################ ## ######################################################################################## ## # ####################################################################################### ############################################ ############################################ ######################################################################################### ################################################################################# ################################################################################ ###################################################################################### ########################################################################################## ########################### ######## ####################################################### ###### ## ############## ##################################################################### #### ############# #### ####### ############################################################ # ######## ####################################################################### ########## ## ###### ################################################################ #### ################################## #### ################################## ####################################################################### ## # ######## ########################################################### ## # # ################################################################# ##################################################################### ######################################################################### ############################################################# # ############################################################## # ############################################################## ############### ############################################# ## #### ###################################################### ## ## ## ##################### ########## ############# ## ############ # ###### ############ ##################### ###### ###### ########### # ## ##### # #### ################# ############# ######## #### ########### ## ### ########## ########################################## ############## ### ######### ################################################### # ## ### ################# ########## ####################################### ##### ####### ########### ######### ########################################### ## ## #### ############################################################## #### # ### ############################################################ ## # ####### ############################################### ####### # ####################################################### ################################ ################################ ############################################################ ############################################################### ## ########################################################## ## ############################################################# ####################################################### ########################################## ####################################################### ########################################################## ######################################################## ######################################################## ########################### ########################### ################################# # ###################################### # ### ######## ######### ### ####### ######## #### ### ###
Original LCM2000 data Baseline predicted presence/absence # # ###### ###### ############## ####### ############## ####### ############### ########### ############### ########### ######### #### ############ ######### #### ############ ############################## ############################## # ################################ # ################################ ################################# ################################# ################################## ################################## #################################### #################################### ##################################### ##################################### ##################################### #### ##################################### #### #### ################################### #### ################################### ################################### ################################### ################################# # ################################# # ##################################### # ##################################### # ######################################## ######################################## # # ###################################### # # ###################################### ########### ## #################################### ########### ## #################################### ######### ########################################## ######### ########################################## # ##################################################### # ##################################################### ################################################ ### ################################################ ### ################################################ ################################################ ### ############################################## ## ### ############################################## ## # ############################################## # ############################################## ################################################# ################################################# # ############################################# # ############################################# ################################################ ################################################ ################################################# ################################################# ################################################# ################################################# #################################################### #################################################### ###################################################### ###################################################### ##################################################### ##################################################### ## ################################################ ## ################################################ ### ############################################# ### ############################################# # ################################################ # ################################################ # ####################################### # ####################################### # ######################################## # ######################################## ### ############################################ ### ############################################ ############################################# ############################################# ############################################# ############################################# ############################################## ############################################## ############################################# ############################################# ############################################# ############################################# ############################################ ############################################ ######## ###################################### ######## ###################################### ######## ###################################### ######## ###################################### ## # ########################################### ## # ########################################### ############################################## ############################################## ############################################### ############################################### ################################################# ################################################# ############################################### ############################################### # ############################################### # ############################################### ###### ################################################ ###### ################################################ ############################################### # ## # ############################################### # ## # # ######################################### #### # # ######################################### #### # ################################# ################################# ################################### ################################### ################################## ################################## ############################ #### ############################ #### # ###### ################### ### # ###### ################### ### ### ##################### ### ##################### ##################### ##################### ###### ############## ###### ############## ## # ############ ## # ############ ########## ########## ## # # ## # # Original Corine data Baseline predicted presence/absence
52 MONARCH 2 Report – Chapter 3 ______
Table 3.8: Validation statistics for land cover models constructed for England and Wales. Class Description Kappa AUC % Match Interpretation C8 Bog, deep peat 0.589 0.90 93.8 Moderate C9 Dense dwarf shrub 0.590 0.87 73.9 Moderate C10 Open dwarf shrub 0.664 0.89 86.1 Substantial C15 Neutral grass 0.314 0.71 20.6 Slight to fair C17 Bracken 0.576 0.91 58.0 Moderate C18 Calcareous grass 0.449 0.84 15.9 Moderate C19 Acid grassland 0.598 0.87 38.9 Moderate C20 Fen, marsh, swamp 0.418 0.85 48.1 Moderate
Table 3.9: Validation statistics for land cover models constructed for Ireland. Class Description Kappa AUC % Match Interpretation 12 Non-irrigated arable land 0.741 0.94 77.0 Substantial 18 Pastures 0.500 0.93 83.8 Moderate 20 Complex cultivation patterns 0.488 0.83 66.6 Moderate 21 Land principally occupied by agriculture 0.280 0.68 56.1 Slight to fair 23 Broad-leaved forest 0.225 0.67 67.3 Slight to fair 24 Coniferous forest 0.370 0.75 58.3 Slight to fair 26 Natural grasslands 0.462 0.80 63.6 Moderate 27 Moors and heathland 0.509 0.83 70.2 Moderate 29 Transitional woodland-shrub 0.247 0.65 54.7 Slight to fair 33 Burnt areas 0.499 0.99 87.1 Moderate 35 Inland marshes 0.268 0.78 79.8 Slight to fair 36 Peat bogs 0.592 0.88 68.3 Moderate
3.5 Dispersal model
The ability of species to track changes in the regional suitability surfaces simulated by the downscaled SPECIES model under the climate and land cover change scenarios will be dependent on the dispersal mechanisms by which migrations occur. The potential for species to migrate rapidly over large distances is a question of critical conservation importance as populations are increasingly threatened by climate change and the fragmentation of habitats (Sala et al., 2000; Hannah et al., 2002; Parmesan and Yohe, 2003). A spatially explicit cellular automata model has been developed to investigate the ability of species to migrate rapidly through fragmented landscapes (Pearson and Dawson, 2004).
The model operates in discrete time and space and simulates stochastic dispersal at the landscape scale, with cell sizes of 10002 metres. As such, the dispersal kernel is designed to be flexible and to incorporate rare long-distance dispersal events, focusing on those few seeds that are expected to travel at least several hundred metres and thus drive migration at coarse spatial resolutions (Nathan et al., 2002). The model does not aim to simulate the fate of individual seeds, since this would be computationally impractical at the scale of analysis, but rather describes the dispersal of ‘propagules’, defined as the minimum number of individuals of a species capable of successfully colonising a new cell (Higgins et al., 2003a). The basic steps used in the model are set out in Figure 3.4. Having assigned species’ parameters and initialised the model with landscape suitabilities and initial populations, the model describes three basic steps: (i) survival, (ii) within-cell population dynamics, and (iii) dispersal. The survival step is analogous to mortality since any population falling on an unsuitable cell fails to survive. Cell suitability is defined as binary suitable or unsuitable in the current study, and suitabilities are changed across time steps according to the regional suitability surfaces produced by the downscaled SPECIES model for the UKCIP02 climate change scenarios and the combined climate and land cover change scenarios. Suitability is defined using the 95% cut-off threshold from the ROC curve described in Section 3.3.1.
MONARCH 2 Report – Chapter 3 53 ______
Within-cell population dynamics are incorporated to determine the number of propagules released by a populated cell in each time step. The model flow places dispersal properly within an organism’s life cycle, with the test for survival carried out before populations can release propagules (Hassell et al. 1995). Population growth is not initiated until the population in a cell has been established for a set period, defined by the number of time steps required for an individual plant to reach reproductive maturity. The number of propagules then released by a populated cell in each time step is determined by a combination of the inherent fecundity of the species (how many seeds does an individual plant produce?) and the size of the population. The likelihood of a population releasing propagules is assumed to increase with population size and population size is assumed to increase through time (except for extinction). A population density growth function is therefore incorporated within each cell, such that older populations have a higher probability of releasing propagules. Population growth is defined as following a Sigmoidal curve as the population rises from low density and saturates at the highest number permitted by the environmental resources. Such sigmoidal growth curves assume that population density is only affected by intraspecific competition (between individuals of the same species) and take no account of interspecific competition (between individuals of different species), yet have been observed in many natural situations (e.g. Alliende and Harper 1989).
Figure 3.4: Flow diagram detailing the steps undertaken in the dispersal simulation (taken from Pearson and Dawson, 2004).
54 MONARCH 2 Report – Chapter 3 ______
The model simulates the redistribution of the propagules according to a dispersal kernel. A dispersal kernel is a function describing the probability of a propagule dispersing to distance x from its source. The dispersal kernel included in the model is based on Clark et al. (1998) and can describe Gaussian, Exponential and ‘fat-tailed’ distributions (Figure 3.5). The ‘fat-tailed’ dispersal function enables low- probability long-distance dispersal events to be simulated. Long-distance dispersal (LDD) concerns that small percentage of seeds that travel significantly beyond the expected dispersal distance. Mechanisms by which LDD may be achieved are diverse and include the catching of seeds in updrafts, dispersal by birds in nest material and movement of seeds whilst attached to the fur of mammals (Higgins et al., 2003b). Evidence for LDD has been drawn from the palaeoecological record (Huntley and Birks, 1983; Davis and Shaw, 2001) and from contemporary observations, particularly of island colonisation and alien plant spread (Pitelka and Group, 1997; Clark, 1998; Higgins and Richardson, 1999; Cain et al., 2000; Horn, 2001; Gomez, 2003). The function is controlled by two parameters: a 'distance' parameter, which controls the mean distance a propagule can travel, and a 'shape' parameter (kurtosis, c) which controls the fatness of the tail.
Figure 3.5: Dispersal kernels used in the model, after Clark et al. (1998). Each curve has the same mean and maximum dispersal distance, but a different amount of kurtosis. The high kurtosis of the fat-tailed kernel means that propagules will have a low, but not insignificant, probability of dispersing a long distance from the source. Plotting the probabilities on a log scale (inset) clarifies this potential for long-distance dispersal (taken from Pearson and Dawson, 2004).
C = 0.5 C = 1.0
}C = 2.0 p.d.f.
Gaussi an: c = 2.0 (l ow kurtosis)
Exponential: c = 1.0
‘Fat-tailed’: c = 0.5 (high kurtosis)
051015 distance
The probabilistic nature of the dispersal process demands that the direction and distance of each dispersal event be selected through the generation of random numbers. Since we are simulating dispersal as a non-deterministic process, the model is run using a Monte Carlo approach. Hereby, the dispersal process is run many times so as to build up a probability surface identifying those cells more/less likely to be populated under certain dispersal assumptions.
MONARCH 2 Report – Chapter 3 55 ______
3.4.1 Application of the dispersal model to artificially fragmented landscapes
An investigation of how different formulations of the dispersal kernel affect the ability of species to disperse through fragmented landscapes is reported in Pearson and Dawson (2004) based on artificial landscapes with differing amounts of suitable habitat cells and different patterns in the distribution of these cells. Illustrative results are shown in Figure 3.6 for a fat-tailed dispersal kernel with a mean and maximum dispersal distance of 1 and 15, respectively. Dispersal occurs across a two-dimensional grid of 100 x 100 cells with an artificially fragmented landscape containing 20% suitable habitat. Populations were initialised on suitable habitat in the centre four cells of the grid at time t = 0 and the model was run for 40 time steps and 10,000 Monte Carlo iterations. The probability of species dispersing within the patch of suitable habitat surrounding the initial populations is very high, whilst the probability that the species could disperse to the isolated regions of suitable habitat at the edges of the grid is much lower. These results demonstrate how the incorporation of long distance dispersal events within the model enable species to jump across patches of unsuitable habitat.
Figure 3.6: Example of species dispersal through an artificially fragmented landscape with 20% suitable habitat showing the probability of dispersal on a log scale.
Log scale:
1.01.00000
le a c s g o L
0.00.00002
3.4.2 Application of the dispersal model within the MONARCH 2 study
The aim of the dispersal modelling work in MONARCH 2 was to ascertain whether species can track the changes in the suitability surface predicted by the downscaled SPECIES model for the UKCIP02 climate change scenarios for the 2020s and 2050s. The UKCIP02 scenarios for the 2020s are based on the average of the time period 2011 to 2040, whilst scenarios for the 2050s are based on the period 2041 to 2070. The mean point of these 30-year periods is 2025 and 2055. Hence, the dispersal model was run for 25 and 55 time steps for the 2020s and 2050s scenarios, respectively, assuming a base year of 2000. To allow the species time to react to the new climate and land cover suitability surfaces (for example by moving into new areas of suitable habitat), the scenario suitability surfaces are switched before 2025 and 2055. Specifically, the model was run with baseline suitability surfaces for 10 time steps then this was switched to the 2020s suitability surface and the model was run for a
56 MONARCH 2 Report – Chapter 3 ______further 15 time steps. At this point the dispersal probability map for the 2020s was saved. The model was then run for a further 30 time steps before the 2020s suitability surface was switched to the 2050s suitability surface and finally the model was run for 15 time steps before saving the dispersal probability results for the 2050s scenario.
The model required parameterisation for six species-dependent variables before it could be applied within the four case study areas. These are maximum and mean dispersal distance, the shape parameter for the dispersal kernel, net reproductive rate, years to reach reproductive maturity and fecundity. Information on each variable was gathered from an extensive search of the ecological literature, supplemented by expert opinion. As specific information was rarely available, categories were defined to assist with the parameterisation of the model based on sensitivity analyses showing the implications of independent and combined variations in the main parameters.
These categories were: • Shape of the dispersal kernel (see Figure 3.5): Fat-tailed distribution (0.5) - birds, insects and herbaceous with light, wind-dispersed seeds Exponential distribution (1.0) - trees, very heavy seeds Gaussian distribution (2.0) - mammals
• Net reproductive rate: Slow growth (1.5) - Perennials, trees and less than univoltine (one generation a year) organisms Medium growth (2.0) - Annuals, univoltine organisms Rapid growth (3.0) - Multi-voltine plants and organisms
• Fecundity: 5 categories, ranging from low (parameter value = 1) for species that produce few seeds, to high (value = 5) for species that produce many seeds.
Dispersal grids were created for the case study areas consisting of all 1km2 grid cells that fell into the area plus a buffer zone surrounding the case study of approximately 50% of its area in order to minimise effects caused by propagules dispersing off the edge of the grid.
3.5 Discussion and conclusions
The complexity of the natural system presents fundamental limits to predictive modeling. The bioclimate envelope approach used in the SPECIES model can provide a useful first approximation as to the potentially dramatic impact of climate change on biodiversity. However, it is stressed that the spatial scale at which these models are applied is of primary importance, and that model results should not be interpreted without due consideration of the limitations involved. There are important limitations to the predictive capacity of bioclimatic models, regardless of the methodology used to characterize the bioclimate envelope. Three of the main criticisms of the bioclimatic approach are biotic interactions, evolutionary change and species dispersal (Pearson and Dawson, 2003). The latter criticism has been addressed in the MONARCH 2 project by coupling the modelled species’ suitability surfaces resulting from the downscaled SPECIES model with dynamic simulations of species dispersal. The importance of biotic interactions between species, such as competition, predation and symbiosis with other species, have been shown to have important impacts on species distributions. Changes to the distribution of a single species could have significant knock-on impacts on the distributions of many other species. It is thus apparent that modelling strategies based on bioclimate envelopes alone may in some cases lead to predicted distributions that are, in fact, wildly incorrect. However, it is argued that applying bioclimatic models at macro-scales, where climatic influences on species distributions are shown to be dominant, can minimize the impact of biotic interactions (Pearson and Dawson, 2003). The implications of rapid evolutionary change for bioclimate envelope modelling are important since the assumption of niche conservatism, whereby rates of adaptation are slower than extinction rates, will be wrong for species experiencing sufficiently
MONARCH 2 Report – Chapter 3 57 ______rapid adaptation. Predicting adaptive changes to species in response to climate change presents a huge challenge to vegetation modellers and has not been accounted for within the MONARCH2 modelling framework. It is thus apparent that applications of bioclimate envelope models for predicting distribution changes over the next century are most appropriate for species not expected to be able to undergo rapid evolutionary change over this timescale. This is most likely to be the case for long-lived species and poor dispersers, since intergenerational selection and/or selection at expanding range margins is required for evolutionary processes to take effect (Pearson and Dawson, 2003).
The bioclimate envelope approach used in the SPECIES and downscaled SPECIES models is based on Artificial Neural Networks (ANNs). These have increasingly been employed in ecological studies as an alternative to more traditional statistical techniques (Lek and Guegan, 1999). The advantages and disadvantages of using ANNs for characterising species distributions have been discussed in detail by Hilbert and Ostendorf (2001) and Pearson et al. (2002). Of particular note here is the ability of ANNs to identify non-linear responses to environmental variables and to incorporate multiple types of input variables, including categorical (e.g. land cover classes) and non-categorical (e.g. climate suitability) data. A notable disadvantage is that the relative contribution of different input variables is not immediately identified in an ANN, though further analysis of the network can increase the explanatory power of the approach (Gevrey et al., 2003).
The interaction between climate and habitat availability plays an important role in determining the biogeography of species. The downscaled SPECIES model was developed to enable the combined effects of climate and land cover change on individual species to be studied and to help uncouple effects of climate and habitat change in the interpretation and prediction of species’ distribution. The model defines the relationship between climate, land cover and species’ distributions at a 10km2 spatial resolution before applying these relationships at a 1km2 resolution. The presence/absence of land cover types at a 10km2 resolution does not always provide a good correlate with species’ 2 distributions. This is due to the fact that at this resolution nearly all 10km cells incorporate at least a small patch of suitable land cover (i.e. a ‘presence’), leading to blanket coverage throughout the study region. In order to better identify correlations between land cover type and species’ distributions it would be necessary to adopt a finer resolution of analysis at which patterns in the distribution of suitable land cover are apparent in the dataset. This was not possible in the MONARCH 2 study because 10km2 was the finest resolution at which observed species distributions were available for Britain and Ireland. Whilst the inclusion of land cover in the downscaled SPECIES model was able to improve the simulation of current species’ distributions at the national scale in many cases, the statistics describing the discrimination ability of the models were lower than those based on climate alone, which are derived at the European scale. Furthermore, the testing of the downscaled SPECIES model under the climate and land cover change scenarios (see Chapters 6 to 9) showed that the signal from the impact of climate change (i.e. gains and losses in climate space) was suppressed in some predictions when compared with outputs from the original SPECIES model. This is likely to be a result of the architecture of the ANN, which consisted of 22 input nodes related to land cover classes for Great Britain (and 38 for Ireland) but only 1 input node related to the climate suitability surface. A less complex approach to the integration of land cover constraints to species’ distributions, utilising simple land-cover ‘masks’, may have been more informative and is recommended for future research.
It is important to note that the land cover model results only represent projections in the location of the climate envelopes that are currently occupied by these land cover classes. It does not consider any economic or social drivers that may affect the location or management of such land cover types. As illustrated in projects such as RegIS, for certain land cover types, such as those dominated by agricultural or forestry production, it is clear that land use and agricultural policy will have as great an influence on the future location of such land cover types as climate (Holman and Loveland, 2002). The model proved most successful where the climate and soil conditions provided strong restrictions on the locations of land cover types (e.g. for peat or calcareous based land covers). All results for classes showing a poor agreement between the training and validation datasets should be treated with caution. There is a need to extend the range of climatic conditions over which the land cover model
58 MONARCH 2 Report – Chapter 3 ______was developed using European land cover data in order to capture the full range of class types that may exist under future climatic conditions. This would require the modelling to rely solely on the Corine dataset, since it is the only one that extends beyond Britain and Ireland. Ideally this wider modelling exercise would also include some measure of socio-economic indicators as covariates within the model in order to capture the non-climatic elements of the baseline distributions.
A number of limitations to the modelling approach have been noted, including biotic interactions, evolutionary change, the restricted explanatory power of ANNs and the reliance on correlations between observed distributions and environmental variables. Further limitations are inherent in the availability and accuracy of datasets. Data can rarely be generated for all resolutions and for all spatial extents, but rather tends to be available for large extents at coarse resolutions, or small extents at fine o resolutions. Thus, in the MONARCH 2 study species distributions were obtained at 0.5 resolution for Europe, 10 km2 resolution for Britain and Ireland, and 1 km2 resolution for the local case studies. It has been necessary to design the modelling framework to take best advantage of the available data. Questions regarding the accuracy of the data also arise, in particular regarding the assumption that observed species absences are true absences, and not a result of insufficient sampling (Griffiths et al., 1999). The use of species records spanning many years is also potentially problematic, since distributions are dynamic over relatively short time-scales. The use of mean 1961-90 climate data aims to reduce this effect, though the single year (1998) ‘snapshot’ of land cover will add an element of error to the modelling. Base errors arising from data limitations are unavoidable. However, the level of success that has been achieved in modelling species distributions has demonstrated that biogeographical trends can be identified regardless of the imperfect data that is so often all that is available in ecological studies.
Differences between decision thresholds must be considered when interpreting presence/absence maps generated from such models. It has been argued here that, rather than maximising agreement between observed and simulated distributions, a more appropriate approach to identifying decision thresholds is to minimise the number of observed presences falling outside the simulated distribution. The three-level approach to presenting model output applied in this study makes the interpretation of results less dependent on the choice of a single threshold and facilitates the identification of broader potential distributions. These broader potential distributions based on the 95% cut-off threshold provided the regional scale suitability surfaces for coupling with dynamic simulations of species dispersal.
The incorporation of long distance dispersal (LDD) within the dispersal model enables investigation of the potential for species to migrate rapidly under future climate change. However, identifying which plants are most likely to disperse via LDD in the future is problematic, particularly given that we can expect long-distance events to be caused by non-standard means of dispersal (Higgins et al., 2003b), so making the categorisation of species as more or less likely to disperse difficult. We can be certain, however, that different dispersal mechanisms will result in very different abilities of species to keep track of changing climate regimes (Collingham et al., 1996; Collingham and Huntley, 2000), which will have important consequences for the future composition and functioning of ecological communities (Berry et al., 2002; Pearson and Dawson, 2003).
The great complexity of natural systems suggests that there are fundamental limits to the prediction of future species’ distributions. Combining the complexities arising from biotic interactions, evolutionary change, modelling approach and data accuracy, along with the uncertainties all too evident in predictions of future climate and land cover change, it is apparent that accurate predictions of future species distributions are not currently possible (Pearson and Dawson, 2003). The development of dynamic global vegetation models (DGVMs), which include mechanistic representations of physiological, biophysical and biogeochemical processes, has demonstrated significant progress in the modelling of vegetation–climate interactions at the global scale (Woodward and Beerling, 1997; Cramer et al., 2001). Recent development of these techniques for application at regional scales, including the breaking down of ecosystem processes into key components with characteristic spatial and temporal scales shows much promise (Sykes et al., 2001). However, the
MONARCH 2 Report – Chapter 3 59 ______complexity of DGVMs makes their parameterisation and validation problematic, and does not currently allow their widespread application to specific species and regions (Pearson and Dawson, 2003). Alternatively, the relatively simple SPECIES bioclimate envelope model linked in a scale- dependent hierarchical manner with land cover data and a dynamic model of species dispersal can provide a useful starting point when applied to suitable species and at appropriate spatial scales. The importance of the model predictions undertaken in the MONARCH 2 project should not be underestimated, though model predictions should be interpreted with due caution and should be viewed as first approximations indicating the potential magnitude and broad pattern of future impacts, rather than as accurate simulations of future species distributions.
3.6 References
Alliende, M.C. and Harper, J.L. (1989). Demographic studies of a dioecious tree. I. Colonisation, sex and age-structure of a population of Salix cinerea. Journal of Ecology, 77, 1029-1047.
Berry, P.M., Dawson, T.P., Harrison, P.A, Pearson, R.G. and Butt, N. (2003). The sensitivity and vulnerability of terrestrial habitats and species in Britain and Ireland to climate change. Journal for Nature Conservation, 11, 15-23.
Berry P.M., Dawson T.P., Harrison P.A. and Pearson R.G. (2002). Modelling potential impacts of climate change on the bioclimatic envelope of species in Britain and Ireland. Global Ecology and Biogeography, 11, 453-462.
Berry, P.M., Vanhinsbergh, D., Viles, H.A., Harrison, P.A., Pearson, R.G., Fuller, R., Butt, N. and Miller, F. (2001). Impacts on terrestrial environments. In: Harrison, P.A., Berry, P.M. and Dawson, T.P. (Eds.) Climate Change and Nature Conservation in the Britain and Ireland: Modelling natural resource responses to climate change (the MONARCH project). UKCIP Technical Report, Oxford.
Cain M.L., Milligan B.G. and Strand A.E. (2000). Long-distance dispersal in plant populations. American Journal of Botany, 87, 1217-1227.
Clark J.S. (1998). Why trees migrate so fast: confronting theory with dispersal biology and the Paleorecord. The American Naturalist, 152, 204-224.
Clark J.S., Fastie C., Hurtt G., Jackson S.T., Johnson C., King G.A., Lewis M., Lynch J., Pacala S., Prentice C., Schupp E.W., Webb III T. and Wyckoff P. (1998). Reid's paradox of rapid plant migration: dispersal theory and interpretation of paleoecological records. BioScience, 48, 13-24.
Cohen, J. (1960). A coefficient of agreement for nominal scales. Educational and Psychological Measurement, 20, 37-46.
Collingham Y.C., Hill M.O. and Huntley B. (1996). The migration of sessile organisms: a simulation model with measurable parameters. Journal of Vegetation Science, 7, 831-846.
Collingham Y.C. and Huntley B. (2000). Impacts of habitat fragmentation and patch size upon migration rates. Ecological Applications, 10, 131-144.
Collingham, Y.C., Wadsworth, R.A., Huntley, B. and Hulme, P.E. (2000). Predicting the spatial distribution of non-indigenous riparian weeds: issues of spatial scale and extent. Journal of Applied Ecology, 37, 13-27.
Cramer, W., Bondeau, A., Woodward, F.I., Prentice, I.C., Betts, R.A., Brovkin, V., Cox, P.M., Fisher, V., Foley, J.A., Friend, A.D., Kucharik, C., Lomas, M.R., Ramankutty, N., Sitch, S., Smith, B., White, A. and Young-Molling, C. (2001). Global response of terrestrial ecosystem structure and
60 MONARCH 2 Report – Chapter 3 ______
function to CO2 and climate change: results from six dynamic global vegetation models. Global Change Biology, 7, 357–373.
Davis M.B. and Shaw R.G. (2001). Range shifts and adaptive responses to Quaternary climate change. Science, 292, 673-679.
Fewster, R.M. and Buckland, S.T. (2001). Similarity indices for spatial ecological data. Biometrics, 57, 495-501.
Fielding, A. H. and Bell, J. F. (1997). A review of methods for the assessment of prediction errors in conservation presence/absence models. Environmental Conservation, 24, 38-49. Franklin, J. (1995). Predictive vegetation mapping: geographic modelling of biospatial patterns in relation to environmental gradients. Progress in Physical Geography, 19, 474-499.
Gevrey, M., Ioannis, D. and Lek, S. 2003. Review and comparison of methods to study the contribution of variables in artificial neural network models. Ecological Modelling, 160, 249-264.
Gomez J.M. (2003). Spatial patterns in long-distance dispersal of Quercus ilex acorns by jays in a heterogeneous landscape. Ecography, 26, 573-584.
Griffiths, G. H., Eversham, B. C. and Roy, D. B. (1999). Integrating species and habitat data for nature conservation in Great Britain: data sources and methods. Global Ecology and Biogeography, 8, 329-345.
Guisan, A. and Zimmermann, N. E. (2000). Predictive habitat distribution models in ecology. Ecological Modelling, 135, 147-186.
Hannah L., Midgley G.F. and Millar D. (2002). Climate change-integrated conservation strategies. Global Ecology and Biogeography, 11, 485-495.
Harrison, P.A., Vanhinsbergh, D.P., Fuller, R.J. and Berry, P.M. (2003). Modelling climate change impacts on the distribution of breeding birds in Britain and Ireland. Journal for Nature Conservation, 11, 31-42.
Harrison, P.A., Berry, P.M. and Dawson, T.P. (Eds.) (2001) Climate Change and Nature Conservation in the Britain and Ireland: Modelling natural resource responses to climate change (the MONARCH project). UKCIP Technical Report, Oxford.
Hassell, M.P., Miramontes, O., Rohani, P. and May, R.M. (1995). Appropriate formulations for dispersal ability in spatially structured models: comments on Bascompte and Sole. Journal of Animal Ecology, 64, 662-664.
Higgins, S.I., Lavorel, S. and Tackenberg, O. (2003a). Plant dispersal and habitat loss synergies. In: Climate Change and Biodiversity: synergistic impacts (eds. Hannah, L. and Lovejoy, T.E.), pp. 71-76. Conservation International, Washington.
Higgins K. and Richardson D.M. (1999). Predicting plant migration rates in a changing world: the role of long-distance dispersal. The American Naturalist, 153, 464-475.
Higgins, S.I., Nathan, R. and Cain, M.L. (2003b). Are long-distance dispersal events in plants usually caused by non-standard means of dispersal? Ecology, 84, 1945-1956.
Hilbert, D. W. and Ostendorf, B. (2001). The utility of artificial neural networks for modelling the distribution of vegetation in past, present and future climates. Ecological Modelling, 146, 311-327.
MONARCH 2 Report – Chapter 3 61 ______
Holman, I and Loveland, P (2002). Regional climate change impact and response studies in East Anglia and North West England (RegIS). CC0337, Final report to MAFF, Soil survey and Land Research Centre, Cranfield.
Horn H.S. (2001). Long-distance dispersal of tree seeds by wind. Ecological Research, 16, 877-885.
Huntley B. and Birks H.J.B. (1983). An atlas of past and present pollen maps for Europe: 0-13,000 B.P. Cambridge University Press, Cambridge. Landis, J.R, and Koch, G.G. (1997). The measurements of observer agreement for categorical data. Biometrics, 33(1), 159-174.
Lek, S. and Guegan, J. F. (1999). Artificial neural networks as a tool in ecological modelling: an introduction. Ecological Modelling, 120, 65-73.
Manel, S., Williams, H. C. and Ormerod, S. J. (2001). Evaluating presences-absence models in ecology: the need to account for prevalence. Journal of Applied Ecology, 38, 921-931.
Nathan R., Katul G.G., Horn H.S., Thomas S.M., Oren R., Avissar R., Pacala S.W. and Levin S. (2002). Mechanisms of long-distance dispersal of seeds by wind. Nature, 418, 409-413.
Parmesan, C. and Yohe, G. (2003). A globally coherent fingerprint of climate change impacts across natural systems. Nature, 421, 37-42.
Pearson, R.G. and Dawson, T.P. (2004). Long-distance plant dispersal and habitat fragmentation: identifying conservation targets for landscape planning under climate change. Biological Conservation, in review.
Pearson, R. G. and Dawson, T. P. (2003). Predicting the impacts of climate change on the distribution of species: are bioclimate envelope models useful? Global Ecology and Biogeography, 12, 361-371.
Pearson, R.G., Dawson, T.P. and Lui, C. (2004). Modelling species distributions in Britain: a hierarchical integration of climate and land-cover data. Ecography, 27, 285-298..
Pearson, R.G., Dawson, T.P., Berry, P.M. and Harrison, P.A. (2002). SPECIES: a spatial evaluation of climate impact on the envelope of species. Ecological Modelling, 154, 289–300.
Pitelka L.F. and Group P.M.W. (1997). Plant migration and climate change. American Scientist, 85, 464-473.
Sala, O. E., Chapin III, F. S., Armesto, J. J., Berlow, E., Bloomfiled, J., Dirzo, R., Huber-Sanwald, E., Huenneke, L. F., Jackson, R. B., Kinzig, A., Leemans, R., Lodge, D. M., Mooney, H. A., Oesterheld, M., Poff, N. L., Sykes, M. T., Walker, B. H., Walker, M. and Wall, D. H. (2000). Global biodiversity scenarios for the year 2100. Science, 287, 1770-1774.
StatSoft, Inc (2001). STATISTICA for Windows. Tulsa, OK, StatSoft, Inc.
Sykes, M.T., Prentice, I.C., Smith, B., Cramer, W. and Venevsky, S. (2001). An introduction to the European Terrestrial Ecosystem Modelling Activity. Global Ecology and Biogeography, 10, 581–593.
Thuiller, W., Vaydera, J., Pino, J., Sabate, S., Lavorel, S. and Garcia, C. (2003). Large-scale environmental correlates of forest tree distributions in Catalonia (NE Spain). Global Ecology and Biogeography, in press.
Travis, J. M. J. (2003). Climate change and habitat destruction: a deadly anthropogenic cocktail. Proceedings of the Royal Society of London, Series B, 270, 467-473.
62 MONARCH 2 Report – Chapter 3 ______
Woodward, F.I. and Beerling, D.J. (1997) The dynamics of vegetation change: health warnings for equilibrium ‘dodo’ models. Global Ecology Biogeography Letters, 6, 413–418.
MONARCH 2 Report – Chapter 3 Annex 63 ______
Chapter 3 Annex Technical details of land cover modelling
A3.1 Model derivation
Table A3.1 indicates the range of additional climate variables that were calculated from the UKCIP02 data. Some of these variables were also used in the bioclimatic classification (see Chapter 2). Table A3.2 lists the combined climate and soils datasets that were initially entered into the modelling phase for each country/country group.
Table A3.1: Calculated parameters for the UKCIP02 data. Parameter Description GDD Sum of Growing degree days above 5.0oC (used in UKCIP98 – 10km) GDD04 Sum of Growing degree days above 4.0 oC (used in UKCIP02 baseline – 5km) Abs Tmin Calculated using Prentice et al. (1992) as the absolute minimum temperature of the coldest month over a 20 year period StartGS Starting date of growing season assuming the start is the 10th consecutive day with temperatures above 5.0oC (UKCIP98 10km method) StartGS02 Starting date of growing season assuming the start is the 5th consecutive day with temperatures above 5.0 oC (UKCIP02 5km method) EndGS End date of growing season assuming the end is the 5th consecutive day with temps below 5.0 oC (used in both UKCIP98 and UKCIP02 data) LengthGS Number of days starting on the 10th consecutive day with temperatures above 5.0oC and ending on the 5th consecutive day with temps below 5.0 oC (used in UKCIP98 data) LengthGS02 Number of days starting on the 5th consecutive day with temperatures above 5.0 oC and ending on the 5th consecutive day with temps below 5.0 oC (used in UKCIP02 baseline)
The contribution of the variables to the prediction of presence/absence of a land cover class was explored through the development of a generalised additive model using a logistic link function (Equation 1) within Statistica 6.1 (Statsoft Inc, 2000). Parameters were entered in a backward stepwise manner, with a maximum number of iterations set to 100, significance of entry and removal set to 0.05, sweep delta set to 1E-7, convergence at 1E-7 with sigma restricted estimated. The parameters that were accepted and retained by the regression procedure were then used to generate predictive maps using equation 2.
Equation 1 p 27 log)(logit p = log)(logit o += ∑ 1 xbb ii 1− p i=1 Equation 2 + • + + • xm)bm...x1b1exp(b0 y = •++•++ xm)bm...x1b1exp(b01
64 MONARCH 2 Report – Chapter 3 Annex ______
Table A3.2: Modelled parameters included in the land cover class model. Country England and Wales Scotland Ireland Soil CALCPERC TAWC1 SoilAWC Parameters PEATPERC SS_TAWC HUMOSEPERC DEPTH1 SBPEATPERC GLEYGWPERC GLEYSWPERC Climatic Annual Rainfall Total Annual Rainfall Total Annual Rainfall Total Parameters Summer Rainfall Total Summer Rainfall Total Summer Rainfall Total Winter Rainfall Total Winter Rainfall Total Winter Rainfall Total Annual Tmean Annual Tmean Annual Tmean Summer Tmean Summer Tmean Summer Tmean Summer Tmin Summer Tmin Summer Tmin Winter Tmean Winter Tmean Winter Tmean Winter Tmin Winter Tmin Winter Tmin Summer Tmax Summer Tmax Summer Tmax Winter Tmax Winter Tmax Winter Tmax Winter PET Annual PET Winter Radiation Summer PET Winter PET Summer Radiation GDD>5°C Summer PET Annual Radiation GDD>4°C GDD>5°C Annual Wind Abs Tmin GDD>4°C Winter Wind StartGS Abs Tmin Summer Wind StartGS02 StartGS GDD>5°C EndGS StartGS02 GDD>4°C EndGS Abs Tmin LengthGS StartGS LengthGS02; StartGS02 LengthGS02; EndGS LengthGS LengthGS02;
A3.2 Results
Figures A3.1 to A3.3 show an interpretation of the predictive power of the models based on the kappa threshold (after Manel et al., 2001).
The interpretation from these land cover class graphs shows that for some land classes there is only slight agreement between the predicted and raw data. However, analysis of agreement using ROC curves (Figures A3.4 to A3.6), indicate that the model has performed well across most of the land classes. It was concluded, therefore, that although the kappa values provide an indication of a good threshold, a better threshold might be obtained by further analysis of the ROC curves.
MONARCH 2 Report – Chapter 3 Annex 65 ______
Figure A3.1: Kappa statistics for Scotland.
C20 C19 Kappa interpretation C18 0-0.2 Little or no agreement 0.2-0.4 Slight to fair C17 0.4-0.6 Moderate C15 0.6-0.8 Substantial
Land class C10 0.8-1.0 Almost total agreement C9 C8
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Ka ppa va l u e
Figure A3.2: Kappa statistics for England and Wales
C19 C18 Kappa interpretation C17 0-0.2 Little or no agreement C15 0.2-0.4 Slight to fair C12 0.4-0.6 Moderate C11 0.6-0.8 Substantial Land class C10 0.8-1.0 Almost total agreement C9 C8
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Ka ppa va l u e
Figure A3.3: Kappa statistics for Ireland
36 35 33 32 31 Kappa interpretation 30 0-0.2 Little or no agreement 29 0.2-0.4 Slight to fair 27 0.4-0.6 Moderate 26 0.6-0.8 Substantial Land class Land 24 0.8-1.0 Almost total agreement 23 21 20 18 12
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Kappa value
66 MONARCH 2 Report – Chapter 3 Annex ______
Figure A3.4: Receiver operating characteristic curves for the Scottish land cover classes. Scotland 1.0
0.8
0.6
0.4
0.2
0.0 0.0 0.2 0.4 0.6 0.8 1.0 0.00.20.40.60.81.00.00.20.40.60.81.0
Landclass: C8 Landclass: C9 Landclass: C10 1.0
0.8
0.6
0.4
0.2 Sensitivity 0.0 0.0 0.2 0.4 0.6 0.8 1.0 0.00.20.40.60.81.00.00.20.40.60.81.0
Landclass: C11 Landclass: C12 Landclass: C15 1.0
0.8
0.6
0.4
0.2
0.0 0.0 0.2 0.4 0.6 0.8 1.0 0.00.20.40.60.81.00.00.20.40.60.81.0
Landclass: C17 Landclass: C18 Landclass: C19 1-Specificity
Figure A3.5: Receiver operating characteristic curves for the England and Wales land cover classes.
1.0
0.8
0.6
0.4
0.2
0.0 0.0 0.2 0.4 0.6 0.8 1.0 0.00.20.40.60.81.00.0 0.2 0.4 0.6 0.8 1.0
Landclass: C8 Landclass: C9 Landclass: C10 1.0
0.8
0.6
0.4
0.2 Sensitivity 0.0 0.0 0.2 0.4 0.6 0.8 1.0 0.00.20.40.60.81.00.00.20.40.60.81.0
Landclass: C15 Landclass: C17 Landclass: C18 1.0
0.8
0.6
0.4
0.2
0.0 0.0 0.2 0.4 0.6 0.8 1.0 0.00.20.40.60.81.0
Landclass: C19 Landclass: C20 1-Specificity
MONARCH 2 Report – Chapter 3 Annex 67 ______
Figure A3.6: Receiver operating characteristic curves for the Irish land cover classes.
Ireland 1.0
0.8
0.6
0.4
0.2
0.0 0.00.20.40.60.81.00.00.20.40.60.81.00.0 0.2 0. 4 0.6 0. 8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
Landclass: C12 Landclass: C18 Landclass: C20 Landclass: C21 1.0
0.8
0.6
0.4
0.2
0.0 0.00.20.40.60.81.00.00.20.40.60.81.00.0 0.2 0. 4 0.6 0. 8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
Landclass: C23 Landclass: C24 Landclass: C26 Landclass: C27 1.0
0.8
Sensitivity 0.6
0.4
0.2
0.0 0.00.20.40.60.81.00.00.20.40.60.81.00.0 0.2 0. 4 0.6 0. 8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
Landclass: C29 Landclass: C30 Landclass: C31 Landclass: C32 1.0
0.8
0.6
0.4
0.2
0.0 0.00.20.40.60.81.00.00.20.40.60.81.00.0 0.2 0. 4 0.6 0. 8 1.0
Landclass: C33 Landclass: C35 Landclass: C36 1-Specificity
68 MONARCH 2 Report – Chapter 4 ______
4 Implications for the Composition of Species Communities
G.J. MASTERS AND N.L. WARD
Summary
One of the limitations of the climate envelop, land cover and dispersal models developed for MONARCH 2 was their inability to take into consideration inter-specific competition and interactions (see Chapter 3). In an attempt to address this issue an assessment of species interactions (type and strength) was undertaken to estimate habitat level impacts of simulated distribution changes. Of particular concern were the consequences of species arriving (through dispersal) or leaving (through local extinction) due to climate change within the selected habitats of each case study area. An interaction matrix was developed based on ecological knowledge of all the selected species and those with which they are understood to interact. This informed the conceptual models by enabling an assessment of the likely ecological impacts of the loss or arrival of the species. The predictions from the downscaled SPECIES and dispersal modelling were examined using the Leaver conceptual model if a species was predicted to be lost within the case study area or using the Arriver model if it was predicted to disperse within the case study area. The conceptual modelling provided an assessment of how communities/habitats may respond to a particular species’ response to climate change, particularly on community composition. The conceptual models could not be applied if no change in species distribution was predicted.
4.1 Introduction
Changes in species composition, abundance, species richness, diversity and/or functional type affect the paths and efficiency with which resources are processed within an ecosystem and thus affect function. However, the conceptual framework developed in this chapter was based on species and their interactions by creating a descriptive species interaction matrix and Arriver and Leaver models. Classically, this approach can be considered as community composition and dynamics or by the relatively new term of metacommunities (Leibold et al., 2004).
Extinction is often a global phenomenon but can also refer to loss from a country, region, habitat or community. Throughout this chapter, local extinction is used to describe the natural loss of a species from a habitat or community. With climate change, species are expected to move from or become extinct from existing habitats/communities when the changing environmental conditions (or their consequences) begin to become intolerable for co-existence. However, as is shown for the extension in range northwards from southern refugia after the last ice age (during a period of warming), all the organisms within a community/ecosystem do not move together as a unit. Instead, species showed a large degree of individualism in their timing, rate and direction of response (Huntley, 1991). Communities will therefore undergo change with the loss of species and the addition of new species so that new species assemblages will be created. This implies that there will be some degree of community assembly, disassembly and reassembly. The impact of species loss and addition in terms of changes in species richness, functional richness and community composition are of crucial importance when attempting to model and predict the effects of climate change on a habitat or community.
4.2 The species Arriver and Leaver conceptual models
These models were used to examine the community level impact of a species arriving or a species leaving. There are obviously other alterations to the habitat that were not included in MONARCH 2 models but are recognised in the discussions within each case study chapter.
MONARCH 2 Report – Chapter 4 69 ______
A species interaction matrix informed these two conceptual models. Such an approach was based on the following assumptions (supported by the literature): • Changes in species composition, abundance, richness, and/or functional type are inter-related; • A community’s response to changes in biodiversity may depend on its composition, i.e. which functional types and species are lost and which remain; • Some species matter more than others: loss of a species can have a disproportionate impact on the community (e.g., loss of a keystone or dominant species); • At least one species per functional group is essential for the persistence (stability) of a species community, more than one species per functional group may insure against community collapse in times of disturbance or environmental change.
4.2.1 The species interaction matrix
To inform the modelling process an extensive literature review was conducted to identify information regarding the effects of species arrival or departure from communities or habitats. Additionally, an extensive literature review was conducted to identify the species relations from which the species interaction matrix was developed. This was obviously different and developed separately for each of the selected species and habitats. A generic species interaction matrix is shown in Figure 4.1 to show the general format and specific examples can be found in Chapters 6-9.
Figure 4.1: Interactive matrix of a community. The matrix consists of a generalised food web combined with species interaction direction and strength. All species interact directly or indirectly with each other. Only direct interactions are illustrated for clarity.
GPr P = plant/ producer H = herbivore SPr SPr Pr = predator C = communitors SO = soil organisms S prefix = specialist SH GH SH G prefix = generalist Circle size represent relative abundance Arrow thickness represents interaction strength
C C P1 P2
GPr GPr SSO GSO SSO
Nutrient availability
70 MONARCH 2 Report – Chapter 4 ______
The species interaction model (Figure 4.1) informs two conceptual models detailing community response pathways for two situations:
(i) the Leaver model, which examines the consequences of a species leaving a community (Figure 4.2) and; (ii) the Arriver model, which examines the consequences of a species arriving in a community (Figure 4.3).
In both models species are classified as being dominant, sub-dominant or rare (defined below). These species categories were also, in part, a component of the species selection criteria (see Chapter 2.7.1.2). Keystone species (defined below) are also considered for completeness but, in relation to the case study areas, it is difficult to identify true keystone species (as opposed to simply identifying a top predator, for example) without experimental manipulation of the communities being considered so they were not identified within the case studies. Keystone status is a functional attribute distinct from the others that are related to species abundance.
Dominant species: those species that are characteristic of each habitat and so are generally the more abundant or frequent species within the habitat, to a large extent governing the type and abundance of other species in the community (after Greig-Smith, 1983; Tilman, 1982).
Sub-dominant species: those species that are common but not as frequent/abundant as dominant species. These species reflect the middle ground between scarce/rare species and the dominants. Generally, these species are often important for community persistence and stability. Their position within the community is probably a result of interactions with other species, such as the dominant(s) (after Tilman, 1982).
Rare species: although these species can often be characteristic of a particular species community they are of low abundance. This does not mean that they are of low importance, as in aggregate they contribute (if not determine) the diversity and species richness of a community as a whole (after Odum, 1989).
Keystone species: those species that have a major influence on the structure and composition of an ecosystem or community. Its presence impacts many other members of the community, disproportionate to its abundance within the community, and if it becomes extinct from the community, there can be far-reaching consequences for the habitat, generally through initiating changes in community structure and composition (often a loss of diversity) (after Paine, 1966; 1969).
The models recognise that communities change with time, so a once rare species, can, with a disturbance or simply with enough time, become a dominant species (Figures 4.2 and 4.3).
4.2.1.1 The Leaver model
Figure 4.2 shows that for a species to have a significant impact on community composition, through causing community reassembly, then the departing species needs to be a dominant species (a driver of the ecosystem), a keystone species (if one can be identified) or come from a functional group where there is no functional type redundancy (the species is the sole representative of that particular functional type). Functional type redundancy occurs when there is more than one species as members of a single functional group, i.e. replication of species of the same functional type (after Odum, 1989). If the species that leaves does not meet any of these requirements (with conservative estimates for keystone species) then the effect of its departure on the ecosystem will be negligible or slight. Even so, with time a very small immediate effect on species composition can lead to a new community forming by altering the long-term natural successional trajectory.
MONARCH 2 Report – Chapter 4 71 ______
Figure 4.2: The Leaver model. This conceptual model explores the consequences of a species leaving a community due to climate change. The model recognises that communities are dynamic entities changing over time (e.g. succession) and that climate change is going to be a major disturbance on community structure, function and dynamics.
Community collapse New Community Dominant Species ct ffe • strong links with other species r e ajo • driver of belowground processes M Reassembly
Yes Keystone species? , nge ha No c ion al ss Sub-dominant tur ce Leaver a uc Species N . s e.g No Functional Type Redundancy? Yes
t Rare Species fec ef • weak interaction strength ible Existing Community glig • often dependent on biotic interactions Ne
4.2.1.2 The Arriver model
The consequences of a species arriving in a new community or habitat are more complex than for a species leaving (Figure 4.3). The propagule pressure (number of propagules, e.g. seeds, of a particular species entering a habitat or community) for the arriver will determine the probability of recruitment and establishment within the ecosystem. Local extinction of the arriving species can occur during either recruitment or establishment, but the probability of extinction is greater with less propagule pressure. Once the species has established it can be considered a coloniser. The consequences for community composition depend on what the coloniser’s status becomes. If the coloniser invades and expands to become a dominant species or represents a new functional type/keystone species, then there will a large impact on community composition, with probably some form of community collapse and reassembly leading to a highly modified community.
72 MONARCH 2 Report – Chapter 4 ______
Figure 4.3: The Arriver model. This conceptual model explores the consequences of a species arriving in a community due to climate change. The model recognises that communities are dynamic entities changing over time (e.g. succession) and that climate change is going to be a major disturbance on community structure, function and dynamics.
t Community t n n e e m Dominant Species Large impact collapse m sh it li Highly u r b c ta modified e s R E community Expansion Reassembly
New Functional Type Sub-dominant Keystone Species species Coloniser , ge
Arriver n ha n Existing l c io a ss Functional ur ce at c Type N su g. e.
ct ffe Rare Species e e ibl glig Ne
Existing community Extinct No impact
4.2.2 Model application and links with SPECIES and dispersal modelling
If species within the case study areas were predicted to redistribute outside that area then the Leaver conceptual model was applied while the Arriver model was applied if species were predicted to spread into the case study area. The interaction matrix was used to assess possible impacts of each species' movement on the community. The models could not be applied if there was no change in species' distribution predicted per test area, but inferences on possible community impacts due to climate change could still be made using the interaction matrix and the literature.
4.3 The application of the Arriver and Leaver conceptual models
For application to the case study areas, an interactive matrix for each of the species selected for each of the habitats within the case study area was constructed from primarily the scientific literature, and if required, from the grey literature. One criterion for building the species relations shown in the interactive matrices was a supporting reference. The downscaled SPECIES and dispersal modelling results (Chapter 3) informed the species interaction matrices if a selected species was predicted to leave a selected habitat or if a selected recruitment species would arrive in a selected habitat. The relevant conceptual model (Arriver or Leaver) was applied to the relevant species interaction matrix and the consequences for the selected habitat (community) of a species arriving or leaving predicted.
Four scenarios were developed to predict the effects on the ecosystem of a species leaving. These involved applying the Leaver model (Figure 4.2) to the interactive matrix for the following scenarios: loss of a dominant species, loss of a rare species, loss of a functional type (with no functional
MONARCH 2 Report – Chapter 4 73 ______redundancy) and loss of a functional type (with functional redundancy). Similar approaches can be applied to the Arriver model. The results for the selected species are presented within Chapters 6-8.
4.4 Discussion
The development of the Arriver and Leaver conceptual models and the interactive species matrix approach that underpins the models goes some way to understanding community level impacts. However, the conceptual models and the interactive matrices can only explore scenarios, they are very difficult to quantify, even if the data existed to enable quantification. Assembling a full interactive matrix for each case study was limited by lack of data on all species present; some groups are covered relatively well while others are missing from local datasets. All the information gathered on species had to be supported; hence the emphasis on the scientific literature as a source, but data was limited by the scope of publishing. Nevertheless, changes in community composition and complexity can be predicted from the application of these models.
4.5 References
Greig-Smith, P. (1983). Quantitative Plant Ecology (3rd edn.). Blackwell Scientific Publications, Oxford.
Huntley, B. (1991). How plants respond to climate change: migration rates, individualism and the consequences for plant communities. Annals of Botany (Supplement 1), 67, 15-22.
Leibold, M.A., Holyoak, M., Mouquet, N., Amarasekare, P., Chase, J.M., Hoopes, M.F., Holt, R.D., Shurin, J.B., Law, R., Tilman, D., Loreau, M. and Gonzalez, A. (2004). The metacommunity concept: a framework for multi-scale community ecology. Ecology Letters, 7, 601-613.
Odum, E.P. (1989). Ecology and our endangered life-support systems. Sinauer Associates, Massachusetts.
Paine, R.T. (1966). Food web complexity and species diversity. American Naturalist, 100, 65-75.
Paine, R.T. (1969). A note on trophic complexity and community stability. American Naturalist, 103, 91-93.
Tilman, D. (1982). Resource competition and community structure. Monographs in population biology, 17. Princeton University Press, New Jersey.
74 MONARCH 2 Report – Chapter 5 ______
5 Impacts on coastal environments
G.E. AUSTIN AND M.M. REHFISCH
Summary
This chapter describes methods developed to predict the effects of climate change on the distribution and numbers of wintering waterbirds in Britain and Ireland. Models that predict the effect of habitat change resulting from sea-level rise (previously developed under MONARCH 1) were integrated with models that predict the redistribution of waterbirds with increasingly mild winters. This redistribution has been detected in the majority of waterbirds in which it has been investigated for the UK. The integrated modelling approach was applied and assessed for seven species (oystercatcher Haematopus ostralegus, ringed plover Chraradrius hiaticula, knot Calidris canutus, sanderling C. alba, dunlin C. alpina, curlew Numenius arquata and redshank Tringa tetanus) using the estuaries of the Suffolk coastline as a case study and subsequently applied within the Hampshire case study.
This analysis has identified strong relationships between weather and waterbird winter distributions, and shown that already these species are redistributing in response to the changing climate. In particular, average minimum temperatures on the muddy estuaries of the east coast help explain a proportion of the variation in the distribution of six of the seven species considered. Although the distribution of four of the species considered is significantly associated with weather, in no species is this expected to result in a large increase in numbers on the estuaries of the Suffolk coast under the UKCIP02 scenarios. In most cases, consideration of the baseline predictions for the capacity of these estuaries suggests that there is currently surplus capacity and this situation is unlikely to change under the various predictions for sea level rise.
Consequently, when considering those aspects of the birds' response to climate change that we have been able to model, there is probably little cause for concern that the estuaries of the Suffolk coast will not be able to hold the expected numbers of waders under the various UKCIP02 scenarios. There are still many other factors affecting waterbird distributions that remain unquantified and further work will be required before a qualitative tool upon which to base management targets for these natural resources is developed. The largest unknown amongst these is likely to be the effect of climate change on their Arctic breeding grounds and the availability of stop-over and wintering sites along their migration routes. While developing the models, initial work was done towards quantifying coastal climatic zones, used here to classify estuaries, based upon their winter weather. It is suggested that this aspect could be further developed to allow a bioclimatic modelling approach similar to that used within MONARCH for other taxonomic groups to be applied to coastal waterbirds.
5.1 Introduction
In international terms, Britain and Ireland are ornithologically important, partly for the vast numbers of waterbirds that winter on its estuaries (Moser, 1987; Cayford and Waters, 1996; Rehfisch et al., 2003). The waterbirds are attracted by a combination of productive wetlands and relatively mild winters. Many estuaries have been designated SSSI or ASSI, SPA, or Ramsar (often all three) on the strength of the over-wintering waterbirds that they support. Monitoring waterbirds is relatively straightforward, and being near the top of the food chain the state of their populations provides a useful proxy for the "health" of the estuarine habitat more generally. Climate change has the potential to affect over-wintering waterbirds in two ways. Firstly, rising sea levels will directly affect the availability of the habitats and prey favoured by these birds, especially within an estuarine context, and, secondly; there will be the direct effect of changed meteorological conditions on the birds, their habitats and their food. Within MONARCH 1, as part of research into the impacts of climate change on coastal environments, both these aspects were explored. Within MONARCH 2 these relationships were further explored with the aim of developing modelling protocols that could be used to predict the
MONARCH 2 Report – Chapter 5 75 ______impact of climate change at a local scale. Such models would make it possible to predict with a certain degree of confidence how many waterbirds would be present in an area according to the total numbers present in Britain and Ireland. As part of MONARCH 2, models were developed and then tested on the estuaries of the Suffolk coast and subsequently on those adjacent to the New Forest in the Hampshire case study area (see Chapter 6).
5.2 Methods
5.2.1 Modelling changes in waterbird distributions in relation to weather
The research undertaken in MONARCH 2 considered the effects of changing winter weather on estuarine waterbirds. It has been established that the winter distribution of many wader species, as monitored by the Wetland Bird Survey (WeBS), has been changing over the past three decades, with a general pattern of shifts towards the north and east during this period (Austin et al., 2000; Austin et al., 2001; Rehfisch and Crick, 2003). Generalized Linear Models (GLMs) were used to investigate whether the numbers of estuarine waders counted at a site was related to local winter weather. This analysis showed that for four out of 10 wader species certain aspects of winter weather (mean wind speed or mean minimum temperature) significantly explained part of the variation in numbers. However, although statistically significant, the proportion of variation in bird numbers at a given site that was explained by the local weather was small and thus the predictive capabilities of these models were small.
MONARCH 2 also reported on research undertaken by the BTO on behalf of the Wetland Bird Survey (WeBS) that had taken a broader view, in that it considered the proportion of the national population of each species wintering in one particular region and related this to the winter weather averaged around the British coast. This research offered a more promising method of predicting changes in distributions of waterbirds due to climate change. A GLM was formulated to specifically test the hypothesis that the proportion of birds wintering in Wales and southwest England, a region of Britain in which numbers of wintering waders had fallen steadily since the mid-1980s (Austin et al., 2000), was related to average coastal weather conditions elsewhere. The results suggested that increasingly mild winter weather in the east of England allowed an increasingly large proportion of the national population of eight of the nine common and widespread species considered to take advantage of the rich food supplies found on the relatively muddy east coast estuaries (Austin and Rehfisch, in press). It follows that as weather patterns continue to change with global warming, the distributions of wading birds are likely to continue to change in response.
It was not known whether this approach would be transferable to other regions or scaleable to smaller regions or possibly individual sites. If transferable, it would make it possible to model how the distributions of wintering waders in Britain and Ireland might respond to climate change. As part of MONARCH 2 the WeBS modelling protocol was developed further to determine whether it would be transferable to other regions and scales within Britain and Ireland.
5.2.2 Modelling changes in waterbird densities in relation to sea level rise
MONARCH 1 considered the effects that sea level rise can be expected to have on the extent and quality of estuarine habitats that support the majority of Britain and Ireland’s coastal waterbirds (Rehfisch et al., 2003). The BTO and Centre for Ecology and Hydrology (CEH), developed techniques to allow estuarine waterbird densities on British estuaries to be predicted from estuary sediments, morphology and geographical location. This technique was predisposed to making predictions of how waterbird densities could be affected by changes in the shape of estuaries with sea level rise. Accordingly, these methods were further developed as part of MONARCH 1 to produce a modelling protocol for predicting the densities of waterbirds that would be expected following changes in estuary morphology due to sea level rise and any managed response to that sea level rise (Austin et al., 2001, Austin et al., submitted). This was done for a range of waterbird species, and
76 MONARCH 2 Report – Chapter 5 ______those models for Haematopus ostralegus (oystercatcher), Calidris canutus (knot), C. alpina (dunlin), Numenius arquata (curlew) and Tringa totanus (redshank) proved robust to rigorous testing. These five species account for over 90% of the waders that winter on British estuaries (Rehfisch et al., 2003) and about 60% of those wintering on Irish estuaries (Colhoun, 2001). These models were applied to the two case study areas in MONARCH 2 (Austin et al., 2001; Austin and Rehfisch, 2003). For each estuary, expected changes in estuary morphology due to sea level rise were determined using detailed topographic data gathered using an airborne Light Detection and Ranging (LIDAR) remote sensing system and potential sea defence management plans, these data being provided by the Environment Agency. From these two case studies, a general pattern of change to estuarine habitat with sea level rise emerged. Where sea defences are maintained, the estuary morphology is expected to be relatively stable and the MONARCH models predicted no substantial changes in bird densities. Where a policy of managed realignment of sea defences is to be adopted, many estuaries can be expected to become wider as land is claimed or reclaimed by the sea, and in turn sediments would be expected to become increasingly sandy (Yates et al., 1995). The MONARCH models predict that in such cases the estuaries will support lower densities of species such as redshank, dunlin and curlew, which are found at greatest densities on muddier sediments (Austin et al., 1996). In contrast they would be capable of supporting higher densities of species such as oystercatcher that tend to favour sandier sediments.
On estuaries where land claim for agriculture has been a historic feature, a substantial increase in estuarine area would be possible if the current sea defences were to be breached or if schemes for managed realignment were to be implemented, thus allowing agricultural land to revert to intertidal habitats. Where areas of land-claim have been urbanised or industrialised there is probably little scope for managed realignment. Where estuaries are bounded by hard natural features there is likely to be little substantial change in estuary shape and, particularly, area. The impact of sea-level rise will, therefore, differ between regions of Britain and Ireland depending on the regional differences in geology, isostatic realignment and historical land-claim. In England, the most vulnerable estuaries to sea-level rise are those in East Anglia and the southeast. These regions will experience relatively high mean sea-level rise due to the effects of isostatic adjustment and because the general topography has led to a history of land-claim (Austin et al., 2001), making the future of large areas of agricultural land in the region dependent on the maintenance and improvement of existing sea defences. Many of the estuaries in the area hold internationally important numbers of waterbirds (Musgrove et al., 2001). In contrast, in northern England, Wales and Scotland no estuaries have been identified as being susceptible to significant changes in shape due to rising sea levels (Frazier, 1999). In Ireland, estuaries with a history of significant land-claim are concentrated in the southeast.
MONARCH, then, had established a preliminary modelling protocol for assessing the possible effects of sea level rise on estuarine waterbirds. As part of MONARCH 2 the aim was to integrate such predictions with those from the weather related models (section 5.2.1) to give an overall assessment of the consequences of global climate change on estuarine habitats and to test the modelling protocol on four estuaries in Hampshire.
5.2.3 Developing the regional waterbird and weather modelling protocol
The WeBS modelling approach was developed to test the specific hypothesis that the downward trend in the numbers of many species of wader in Wales and southwest England was associated with increasingly mild winter weather. This association was found to be significant for a range of wader species. It was suggested that this could be due to growing numbers of their winter populations taking advantage of the increasingly mild winter weather, which made it energetically less expensive to winter on the relatively food-rich, muddy estuaries of the east coast, despite them remaining colder than the west (Austin and Rehfisch, in press).
5.2.3.1 The Wetland Bird Survey Weather Model
Research undertaken by WeBS sought the existence of relationships between the proportion of the UK population of each species over-wintering on estuaries in the region of interest and the climatic
MONARCH 2 Report – Chapter 5 77 ______conditions in the UK using a Generalised Linear Model. The proportion, i.e., the total population in the region of interest (Totalreg) divided by the total UK population (TotalUK) was modelled by logistic regression. The models were binomial and specified a logit link. Population values for each species were calculated as the sum of birds at all sites having first used the Underhill indexing algorithm (Underhill and Prŷs-Jones, 1994) to estimate missing counts.
UK wader populations can vary between years and so differences in distribution could result from less attractive sites only supporting high numbers in years when UK populations are high. The WeBS models addressed this by including the UK national index (NATINDEX = TotalUK ÷ TotalUK for January on base year) as an explanatory variable. The principal circumstance under which the UK national index might significantly improve a species model would be where movement into or out of the UK, either due to annual fluctuation or a long-term trend, has been disproportional between regions. The principal explanatory variables of interest for the WeBS research were MINTEMP, RAIN and WIND, the monthly averages of daily minimum temperature, rainfall and windspeed, respectively. Those values were derived from Meteorological data from the British Atmospheric Data Centre (BADC). Models were fitted to the data using a stepwise approach. Consequently the final model for each species would include those parameters from the equation:
Logit(Totalreg÷TotalUK) = μ + β(MINTEMP) + γ(RAIN) + δ(WIND) + ε(NATINDEX) [Equation 1] that were significant in explaining the variation in the proportion of the UK population over-wintering in the target region. A positive parameter estimate for minimum temperature would indicate that during warmer winters a larger proportion of the species wintered in the target region. Similarly, positive parameter estimates for rain or wind would indicate that a higher proportion of the population was found in the target region during wetter or windier winters respectively.
5.2.3.2 Developments for MONARCH 2
Weather data
The WeBS model had used weather variables derived by averaging Meteorological Office data for weather stations associated with estuaries. For MONARCH 2, it was desirable to make the baseline weather data as compatible as possible with the predicted scenarios. Thus, the UKCIP02 baseline data were used to derive weather variables for Britain. Similar data for Ireland were only available for Northern Ireland. Irish Meteorological Office data, therefore, were obtained to characterise weather in Ireland.
The WeBS model had considered three weather variables (Equation 1), chosen because they could be expected to influence the energy budgets of estuarine waders. Although weather data from Ireland were to be incorporated into the MONARCH 2 analysis it was not apparent how raw Irish Meteorological Station data and interpolated UKCIP02 data could be combined into a single average variable for each weather parameter. This necessitated each weather parameter being averaged to give separate variables for, at least, Britain and Ireland. For compatibility across the whole of Ireland, Irish Meteorological Station data were combined with similar data for Northern Ireland previously obtained for MONARCH rather than use the UKCIP02 data for Northern Ireland.
Extending the approach of calculating separate weather variables for each weather parameter for Britain and Ireland, the possibility of further subdividing weather variables to take account of the longitudinal clines in weather across Britain which lead to different conditions on the east and west coasts was explored. This would be advantageous as the WeBS analysis had suggested that it was east coast weather rather than coastal weather generally that was driving the observed shifts in wader distributions. Accordingly, a coastal climatic zonation of UKCIP02 data was derived for Britain (described in the following section).
78 MONARCH 2 Report – Chapter 5 ______
Wader numbers on British estuaries vary markedly through the winter, normally reaching peaks during midwinter, although the exact phenology and patterns may differ between species and sites. The WeBS model had addressed this seasonal variation by considering month as an additional class variable to those given above in Equation 1. Within MONARCH 2, analyses that included data from Ireland were restricted to January because, firstly, bird data from other winter months were generally not available for many Irish sites and, secondly, the data obtained from the Irish Meteorological Office were restricted to January for budgetary reasons.
Coastal Climatic Zones
Coastal climatic zones were developed to determine appropriate geographical locations upon which to base a west - east split of the UKCIP02 baseline data for Britain for modelling purposes rather than to describe in detail a coastal zonation similar to the MONARCH bioclimatic zonation. The subset of UKCIP02 data derived from 5 km grid squares bisected by the British coastline were selected for each of the three weather parameters for December, January and February using a Geographic Information System (GIS, ESRI, 2002). These data were then subjected to a Ward’s minimum variance cluster analysis (SAS Inc., 2002) to produce classifications based on 3, 5, 10, 15 and 20 clusters. The results from these analyses were then plotted in the GIS. None of these classifications allowed the purely west / east split desired but comparing classifications allowed the geographical stability of boundaries between clusters to be assessed. Those boundaries that were geographically stable over a range of classification being those between coastlines with major climatic differences.
With 10 clusters or more the general pattern was similar. In the case of the 10-cluster classification (Figure 5.1a) two clusters (2 and 5) accounted for all of the east coast of Britain between Beachy Head in the south and the Moray Firth in the North. Cluster 5 was also found concentrated in the upper reaches of a number of large west coast estuaries (The Severn, Mersey and Solway). The south coast of England between Beachy Head and Exmouth fell largely into cluster 4, which, together with clusters 1 and 3 accounted for most of the remaining coast of England and Wales. The remaining clusters were concentrated on the west coast of Scotland and the Western Isles, with clusters 6 and 7 tending to be associated with coastline exposed to westerly weather systems and clusters 8 to 10 tending to be associated with more sheltered coastline. The 15- and 20-cluster classifications tended to differ from the 10-cluster classification principally in that western Scotland and the Western Isles were further divided. The classification of the east coast of Britain was particularly stable between 20, 15 and 10 cluster classifications.
The sought after division between east and west coast Britain became even more clear-cut with the 5- cluster classification (Figure 5.1b). The east coast of Britain between Beachy Head and the Moray Firth fell completely within cluster 3. The remainder of England, southern Scotland and the remainder of the north and east Scotland coasts fell largely into cluster 1. The final three clusters were principally confined to the west coast of Scotland and the Western Isles although cluster 2, which occurred on the more exposed coastline in Scotland, also occurred to a small degree in the shelter of estuaries in southwest England and Wales. Clusters 4 and 5 were associated with more sheltered coastline.
The three cluster classification lost the sought after division between east and west coast Britain (Figure 5.1c). Clusters 1 and 3 from the 5-cluster classification merged to give a single cluster covering most of Britain other than the west of Scotland and the Western Isles.
The persistence of zone boundaries associated with the east coast of Britain across all but the most restrictive classification suggested that weather variables representing the averages of the weather parameters for east coast Britain should be based on data from coastal grid cells between Moray Firth clockwise to Beachy Head. It also suggested that the remainder of Britain should be divided into two areas, broadly speaking the remainder of England and Wales and the west of Scotland and Western Isles. Weather variables for the south and west of Britain were, therefore, based on the averages of weather parameters for coastal grid cells from Beachy Head clockwise to Firth of Clyde. The west of
MONARCH 2 Report – Chapter 5 79 ______
Scotland and Western Isles do not contribute significantly to the bird data set being analysed here as there are no estuaries north of the Firth of Clyde supporting significant numbers of the principal "estuarine" wader species. Accordingly, weather variables for that region were not included in the models that follow because of the necessity of keeping the number of potential explanatory variables to a reasonable level.
Figure 5.1a: Distribution of coastal weather zonation classes based on 10-cluster classification. Classes are numbered by decreasing frequency of occurence.
Cluster 1 of 10 Cluster 2 of 10 Cluster 3 of 10 Cluster 4 of 10 Cluster 5 of 10
Cluster 6 of 10 Cluster 7 of 10 Cluster 8 of 10 Cluster 9 of 10 Cluster 10 of 10
Figure 5.1b: Distribution of coastal weather zonation classes based on 5-cluster classification. Classes are numbered by decreasing frequency of occurence.
Cluster 1 of 5 Cluster 2 of 5 Cluster 3 of 5 Cluster 4 of 5 Cluster 5 of 5
80 MONARCH 2 Report – Chapter 5 ______
Figure 5.1c: Distribution of coastal weather zonation classes based on 3-cluster classification. Classes are numbered by decreasing frequency of occurence.
Cluster 1 of 3 Cluster 2 of 3 Cluster 3 of 3
The MONARCH 2 models
The new MONARCH 2 models were extensions of the WeBS model described above (Equation 1, section 5.2.3.1). The dependent variable was unchanged. The potential explanatory weather variables, which in the WeBS model had represented the UK average, were each replaced by up to three coastal zone averages, one each for the east coast Britain, the south and west coast Britain, and for Ireland. For these analyses, the national index, which in the WeBS model is referred to as the UK Index, was replaced by the British Index and, additionally, the Irish Index was considered for some models.
For Britain, annual values for the weather variables were derived for each of the two retained coastal zones by averaging the January value for each of the three weather parameters, mean minimum temperature, mean rainfall and mean wind speed from those 5km grid cells in the UKCIP02 baseline data that are bisected by coastline. For Ireland, annual values for comparable weather variables were obtained by averaging the January value for each of the same three weather parameters from six Irish Meteorological Office and eight UK Meteorological Office weather stations from Northern Ireland, chosen for their proximity to the coast (Table 5.1).
Table 5.1: Meteorological stations from which data were used to derive coastal weather parameters for Ireland. Meteorological Station Location Balliwatticock 54° 32′ N 5° 40′ W Mourne Grange 54° 4′ N 6° 2′ W Jordontown 54° 41′ N 5° 53′ W Bann 55° 10′ N 6° 45′ W Lough Foyle 55° 4′ N 7° 4′ W Belmullet 54° 14′ N 10° 0′ W Casement Aerodrome 53° 18′ N 6° 26′ W Cork Airport 51° 51′ N 8° 29′ W Dublin Airport 53° 26′ N 6° 14′ W Malin Head 55° 22′ N 7° 20′ W Roche’s Point 51° 48′ N 8° 15′ W Rosslare 52° 15′ N 6° 20′ W Shannon Airport 52° 42′ N 8° 55′ W Valentia 51° 56′ N 10° 15′ W
For Britain, annual indices for each wader species were derived using standard WeBS methodology based on the Underhill indexing algorithm (Underhill and Prŷs-Jones, 1994). Thus indices were based
MONARCH 2 Report – Chapter 5 81 ______on all sites where at least 50% of potential counts had been made, values for missing counts being imputed. In practice this includes all but a few estuaries holding very few birds. A similar approach was used to derive indices for Ireland, however, the coverage of estuarine sites was less complete. Thus Irish annual indices were based on data from eight sites in Northern Ireland that are covered by WeBS and, following consultation with the organisers of the Irish Wetland Bird Survey (I-WeBS), a representative sample of 12 estuarine sites from the remainder of Ireland (Table 5.2). The latter were chosen for the quality and quantity of their data and in order to obtain as wide a geographical coverage as possible.
Table 5.2: Estuaries used to derive Irish wader indices. Site Location Clonakilty Bay (Cork) 51º 35' N 8º 52' W Cork Harbour (Cork) 51º 51' N 8º 17' W Ballymacoda (Cork) 51º 54' N 7º 55' W Dungarvan Harbour (Waterford) 52º 5' N 7º 37' W Bannow Bay (Wexford) 52º 14' N 6º 48' W Inner Galway Bay (Galway) 53º 12' N 9º 1' W Baldoyle (Dublin) 53º 25' N 6º 8' W Broadmeadow (Malahide) Estuary (Dublin) 53º 28' N 6º 10' W Rogerstown Estuary (Dublin) 53º 30' N 6º 0' W Boyne Estuary (Louth) 53º 44' N 6º 15' W Dundalk Bay (Louth) 53º 56' N 6º 19' W Carlingford Lough 54º 2' N 6º 10' W Dandrum Bay 54º 15' N 5º 49' W Outer Airds 54º 28' N 5º 32' W Strangford Lough 54º 28' N 5º 37' W Belfast Lough 54º 41' N 5º 50' W Larne Lough 54º 50' N 5º 47' W Lough Swilly (Donegal) 54º 58' N 7º 39' W Lough Foyle 55º 4' N 7º 4' W Bann Estuary 55º 10' N 6º 45' W
While including separate variables in the analysis for different parts of Britain and Ireland might be expected to improve the resulting models, differences between the completeness of bird data and the quality of the weather data between Ireland and Britain caused a number of problems. When bird data from Ireland were excluded from the analysis all 30 years of bird data for Britain could be included. When Irish bird data were included in the models those models could only be based on nine winters (1984/85 to 1986/87 and 1993/94 to 1998/99), partly because no data were available for most other years for most of the Irish sites, and partly because the inclusion of those data for additional years from the remaining sites led to the overall level of imputed counts exceeding the accepted limits for the Underhill indexing algorithm. Although weather data were available for both Britain and Ireland for the entire period for which bird data were available, weather variables for Ireland were derived from a small number of weather stations. These data were thus more susceptible to being influenced by a few anomalous values and local conditions than were the equivalent UKCIP02 baseline (already smoothed to remove anomalies) based variables. To address these problems, three basic GLMs were considered for each species using, in each case, a stepwise approach for variable selection.
Model BI1: Applicable to Britain and Ireland. These models considered as explanatory variables all three weather parameters for January only, for each of the east coast of Britain, the south and west coasts Britain and the Irish coast together with both British and Irish bird indices.
Model GB1: Applicable to Britain only. These models considered as explanatory variables all three weather parameters, for both the east coast of Britain and the south and west coasts of Britain together
82 MONARCH 2 Report – Chapter 5 ______with the British bird index. These models are unaffected by any limitations of the Irish data. They do not allow for bird movements between Ireland and Britain or weather conditions in Ireland affecting the distribution of birds in Britain. As no Irish data were included, alternative versions based on January data only and based on December to February inclusive were possible.
Model GB2: Applicable to Britain only. These models considered as explanatory variables all three weather parameters, for each of the east coast of Britain, the south and west coasts of Britain and the Irish coast together with the British bird index. These models, based on January data only, are unaffected by limitations of the Irish bird data but are only applicable to regions within Britain. Whereas they allow for weather conditions in Ireland affecting the distribution of birds within Britain, they do not allow for bird movements between Ireland and Britain.
Determining the geographical extents over which the modelling protocol is reliable
As stated earlier, it was not known whether this modelling protocol, although based on one that had successfully explained part of the variation in numbers of waders in southwest Britain, would be transferable to other regions or scaleable to smaller regions. Although numbers of waders have decreased in some areas and increased in others, numbers in some areas have changed little. In the latter cases it may be that weather conditions have not changed or that emigration has been balanced by immigration. Consequently underlying processes which may be driving waterbird redistributions, and which may include climate change, would not yet be evident in changing bird numbers (the dependent variable) and thus no significant associations would be expected. In order to determine the wider applicability of the modelling protocol, the models were run in an automated manner for a range of species, and for a range of incrementally sized regions (all single estuaries, all possible groupings of 3, 5, 7, 9, 11, 13 and 15 adjacent estuaries and estuaries grouped by EA or SEPA regions). These analyses were run using British sites only because bird data from groups of adjoining sites were not available for Ireland and thus the proportion of a population in a given region of Ireland (the dependent variable) could not be derived.
Rather than develop the modelling protocol based on a single species as originally intended a range of species was used as it was felt that this would give a better assessment of the general applicability of the approach.
Geographical areas where models might be expected to be obtained for each of seven species (Oystercatcher Haematopus ostralegus, Ringed Plover Chraradrius hiaticula, Knot Calidris canutus, Sanderling C. alba, Dunlin C. alpina, Curlew Numenius arquata and Redshank Tringa totanus) were assessed, within a GIS (ESRI, 2002), by plotting for each model and for each incremental increase in the number of adjacent sites being considered, the central site of each group of sites. Sites were displayed according to whether weather variables explained, significantly, part of the variation in the proportion of each species’ national population wintering on each group of sites. The results for all seven species are summarised below (Table 5.3) and maps for Dunlin are given by way of illustration (Figure 5.2).
For each species models were obtained for up to 934 unique groups of adjacent sites for each of the three model forms (BI1, GB1 and GB2). With such a large number of models it was not feasible to critically assess each. Furthermore, particularly for the Britain and Ireland models (BI1), there was a high variable to sample size ratio. It was, therefore, necessary to assess the probability that models to be used for a particular study area could have been obtained by chance. This was done using randomisation techniques (Manly, 1991). Using 9999 repetitions, the dependent variable (over-winter average bird numbers) was matched randomly to the independent variables rather than matched to the relevant winter. The independent variables were not randomly sorted with respect to each other thus maintaining any relationships between the co-variates themselves. The proportion of cases where the model was deemed significant represents the probability of having obtained that model by chance. Consequently, it would be reasonable to accept a model with a particular set of explanatory variables if it was obtained less than 500 times from the 9999 repetitions (equivalent to P < 0.05).
MONARCH 2 Report – Chapter 5 83 ______
Table 5.3: Summarised results of geographical locations where models explaining bird distribution in terms of weather may be expected. In order to provide a geographical context, groups of adjacent sites were classified according to which of the three SEPA areas or six coastal EA regions the centre site lies within. Species are listed in taxonomic order (oc=oystercatcher, rp=ringed plover, kn=knot, ss=sanderling, dn=dunlin, cu=curlew, rk=redshank). Species codes are in uppercase when more than 50% of their regional models included weather (P<0.05) and lowercase when 25%-50% of their regional models included weather (P<0.05). Models were assessed at three different geographical scales: Local, one to five adjacent sites; Intermediate, seven to 11 adjacent sites, and; Extensive, 13 to 15 adjacent sites. Note that the true geographical extent of the three categories will differ between different areas of the country depending on the linear density of estuaries. Additionally models were considered for all estuaries contained within each of the SEPA or EA regions.
Local scale (1 to 5 adjacent sites) Model BI1 Model GB1 Model GB2 SEPA - Highland, Grampian & Western Isles dn,cu dn,CU dn,cu SEPA – Southwest Area OC,dn,cu oc,cu oc,cu,rk SEPA – Southeast Area oc,kn,cu rp oc,rp EA – North East Region OC,rp,dn,cu,rk kn,dn oc,KN,dn EA – North West Region oc, kn, rk kn kn EA – Wales oc,rp,kn,ss,DN,rk kn,ss,DN DN EA – Anglian Region OC,rp,KN,dn,cu,rk dn dn,cu EA – South West Region oc,rp,kn,dn oc,rp,dn,cu oc,rp,dn,cu,rk EA – Southern Region kn,dn,rk dn,rk oc,rp,dn,RK
Intermediate scale (7 to 11 adjacent sites) Model BI1 Model GB1 Model GB2 SEPA - Highland, Grampian & Western Isles rp,DN,CU CU CU SEPA – Southwest Area oc,dn oc,rp,cu oc,cu SEPA – Southeast Area OC,dn oc,rp OC,rp EA – North East Region OC kn,cu oc,kn,cu EA – North West Region KN,ss,dn rp,KN,ss EA – Wales oc,rp,DN,rk ss,DN DN EA – Anglian Region OC,dn,cu cu oc,cu EA – South West Region oc,dn,cu,rk dn,CU CU EA – Southern Region oc,kn,dn,cu,RK rp,DN,rk oc,rp,DN,RK
Extensive scale (13 to 15 adjacent sites) Model BI1 Model GB1 Model GB2 SEPA - Highland, Grampian & Western Isles rp,DN,CU rp,CU,rk oc,CU SEPA – Southwest Area OC,RK cu,RK cu,RK SEPA – Southeast Area OC,dn,cu OC,rp OC,rp EA – North East Region OC,cu cu OC EA – North West Region oc,dn KN,dn,cu KN EA – Wales DN rp,dn dn EA – Anglian Region OC,dn CU oc,CU EA – South West Region oc,rk oc,rp,dn,CU oc,rp,dn,CU,rk EA – Southern Region oc,cu,RK dn,rk oc,dn,rk
Regions (all sites within SEPA or EA boundary) Model BI1 Model GB1 Model GB2 SEPA - Highland, Grampian & Western Isles RP,SS,DN CUjw oc,rp,kn,ss,dn,cu,rk SEPA – Southwest Area SSw oc,rp,kn,ss,dn,cu,rk SEPA – Southeast Area DN OCw,CUjw,RKjw oc,rp,kn,ss,dn,cu,rk EA – North East Region SS DNjw oc,rp,kn,ss,dn,cu,rk EA – North West Region RPw,KNjw,SSjw,CUjw oc,rp,kn,ss,dn,cu,rk EA – Wales OC,RP,DN RPw,KNw,SSjw,DNw,CUw oc,rp,kn,ss,dn,cu,rk EA – Anglian Region OC oc,rp,kn,ss,dn,cu,rk oc,rp,kn,ss,dn,cu,rk EA – South West Region OC,RP,KN oc,rp,kn,ss,dn,cu,rk oc,rp,kn,ss,dn,cu,rk EA – Southern Region oc,rp,kn,ss,dn,cu,rk oc,rp,kn,ss,dn,cu,rk Ireland as part of Britain &Ireland SS N/A N/A Northern Ireland as part of Ireland RP N/A N/A Republic of Ireland as part of Ireland RP N/A N/A
84 MONARCH 2 Report – Chapter 5 ______
Figure 5.2: Assessment of the geographical areas where models were obtained in which weather explained, significantly, part of the variation in dunlin numbers between winters. Solid circle = central site of adjacent groups for which at least one weather variable was included in the model; shaded circles = central site of adjacent group of sites for which no models containing weather variable were obtained. These data are summarised for all species considered in Table 5.3.
Number of adjacent sites in group Dunlin Model BI1 Dunlin Model GB1 Dunlin Model GB2
3
7
11
15
MONARCH 2 Report – Chapter 5 85 ______
5.3 Suffolk coastline case study
The Suffolk coastline was chosen for the first case study because its estuaries are surrounded by low- lying land previously claimed for agriculture and thus offering considerable scope for managed re- alignment in response to sea-level rise. Estuaries along this part of the coast have particularly muddy sediments which support high densities of invertebrates and thus of the wading birds that prey upon them. WeBS research suggests that increasingly mild winters in this part of Britain are driving many of the observed shifts in wader distribution as birds adjust their trade-off between the risk of cold weather mortality on this relatively cold coastline and the benefit of richer feeding conditions (Austin and Rehfisch, in press).
5.3.1 Impact of climate change
Using the protocol detailed above (5.1), statistically valid models based on weather variables were obtained for four of the six species considered (Table 5.4). The models for oystercatcher and redshank predict the proportion of the British population of each to be found on the Suffolk estuaries excluding the Stour and Orwell. The Stour and Orwell estuaries were excluded from the analysis because bird numbers on these adjoining sites have been decreasing across a range of species in contrast to the remainder of the EA Anglian Region, probably due to local habitat degradation (Armitage et al., 2002). When these two sites were included no predictive models were obtained for any species. While no models for the Suffolk coast were obtained for either ringed plover or dunlin, models for these species were obtained for the whole of the EA Anglian Region. No model was derived for sanderling because in most winters none have been recorded by WeBS on these estuaries.
Table 5.4: Details of the Generalised Linear Models relating the proportions of the British population of wader species over-wintering in EA Anglian Region or on the estuaries of the Suffolk coastline to weather. Applicability – the region was chosen to provide the most reliable models for each species. Parameters – independent variables retained by the model and the associated parameter estimates (β,γ,δ in equation 1). Partial t values and probability – indicates the significance of each parameter included in the model. Partial R2% – indicates the percentage of the variation in the dependent variable explained by each parameter. Adjusted model R2% – indicates the percentage of the variation in the dependent variable explained by the full model (adjusted for multiple parameters). Probability of obtaining the model by chance – as assessed by the randomisation testing using 9999 repetitions (see end of section 5.2.3.2).
Species Applicability Parameters Parameter Partial t value Partial R2% Probability estimate of obtaining (adj. model the model by R2%) chance
Oystercatcher Suffolk Intercept -7.775 Coastline East-coast Minimum 0.277 t1=3.09, P=0.0044 24.8% P=0.0303 Temperature (22.2%) Ringed Plover EA Anglian Intercept 0.132 Region East-coast Minimum 0.179 t1= 2.14, P=0.0414 11.9% Temperature -0.181 t1= -2.42, P=0.0225 14.9% P=0.0025 West-coast Wind (21.6%) Dunlin EA Anglian Intercept -0.742 Region GB Index -0.229 t1= -4.30, P=0.0002 27.3% East-coast Minimum 0.147 t1= 3.41, P=0.0020 21.3% P=0.0015 Temperature (45.0%) Redshank Suffolk Intercept -3.853 Coastline East-coast Minimum 0.119 t1= -2.62, P=0.0137 19.2% P=0.0254 Temperature (16.4%)
86 MONARCH 2 Report – Chapter 5 ______
The average minimum East coast temperature explained, significantly, part of the variation in the proportion of the British population wintering on the Suffolk or Anglian coast of all four species. In all cases the positive parameter estimate for this variable indicates that the higher the average minimum east coast temperature, the higher the proportions of the species over-wintering in Britain that do so on the estuaries of the Suffolk coast. This gives strong support for the hypothesis that temperatures on the east coast are driving the observed shifts in the distributions of waders wintering in Britain. However, the proportions of the variation explained by the weather variables in these models are comparatively low at between 12% and 25%. This indicates that factors other than the weather variables considered have an influence on wader distributions. Amongst such factors, water quality, human disturbance, predation pressure, and habitat loss are likely to be particularly important and are the subjects of ongoing research by BTO. Consequently, these weather based models, although being built upon highly significant associations, have relatively weak predictive power as can be seen from the wide confidence limits of the resulting predictions when the models are used to produce predictions for the proportions of these species over-wintering in EA Anglian Region or the estuaries of the Suffolk coast under the Low and High UKCIP02 scenarios (Figures 5.4a to 5.4d). For any given species the predictions are neither statistically different from the predictions made under the alternative scenarios nor the values recorded over the past three decades.
Figure 5.4a: Observed numbers of oystercatcher over-wintering on the Suffolk estuaries between the winters of 1969/70 and 1999/2000 inclusive and predicted numbers (with 95% confidence intervals) made under the UKCIP02 Low and High scenarios for 2020s, 2050s and 2080s. The models used predicted the proportion of birds over-wintering in Britain that do so on the estuaries of the Suffolk Coast. These predictions have been converted to numbers using both minimum and maximum recorded values for birds over-wintering in Britain. Because the proportion is not dependent on country-wide numbers this gives an indication of the extremes in numbers that might be expected to over-winter on these estuaries.
Observed Predictions: Observed Predictions: based on minimum British index based on maximum British index 3000 3000
2500 2500
2000 2000
1500 1500
1000 1000
Over-winter average numbers 500 Over-winter average numbers 500
0 0 w w w w w w h 5 o o gh o gh 0 gh g 70 75 85 90 9 70 8 85 95 o o 9 9 L L Hi L Hi 9 L Lo Hi L Hi 19 1 1980 19 1 19 0 High 0 0 19 1975 19 1 1990 19 0 High 0 0 20 2 5 80 20 2 5 80 80 20 20 205 20 20 2080 20 20 205 20 20 20
Oystercatcher - Numbers predicted under all UKCIP02 scenarios considered are well within the recorded range over the past three decades although below the average over that period (Figure 5.4a). For any given time-frame the differences between the predictions of numbers made under the High and Low scenarios are no larger than the observed between winter variation over the past three decades.
MONARCH 2 Report – Chapter 5 87 ______
Figure 5.4b: Observed numbers of ringed plover over-wintering on the Suffolk estuaries between the winters of 1969/70 and 1999/2000 inclusive and predicted numbers (with 95% confidence intervals) made under the UKCIP02 Low and High scenarios for 2020s, 2050s and 2080s. The models used predicted the proportion of birds over-wintering in Britain that do so on the estuaries of EA Anglian region. These predictions have been converted to numbers using both minimum and maximum recorded values for birds over-wintering in Britain. This gives an indication of the extremes in numbers that might be expected to over-winter on the estuaries within this region. The resulting predictions have been further adjusted to give those numbers expected on Suffolk estuaries while using both minimum and maximum values for the proportion of birds over-wintering in EA Anglian region that do so on these estuaries. This latter proportion is not dependent on country-wide numbers (otherwise a model based on the estuaries of the Suffolk coast alone would have been obtained). This gives a indication of the extremes in numbers that might be expected to over-winter on these estuaries.
Observed Predictions: Observed Predictions: based on minimum British index based on maximum British index 160 160
140 140
120 120
100 100
80 80
60 60
40 40 Over-winter average numbers average Over-winter numbers average Over-winter 20 20
0 0 w h w h h 0 5 5 0 5 0 5 0 5 0 5 h 7 8 o 7 9 9 o 97 99 Hig 98 L 19 1 1980 19 1 199 0 Low 0 L 197 19 1 198 19 19 Hig 0 Hig 0 Hig 2 8 20 Low20 50 5 8 0 0 0 20 2020 High 2050 Low2050 High 20 2080 2 2 2 20 2080 20Low
Figure 5.4c: Observed numbers of dunlin over-wintering on the Suffolk estuaries between the winters of 1969/70 and 1999/2000 inclusive and predicted numbers (with 95% confidence intervals) made under the UKCIP02 Low and High scenarios for 2020s, 2050s and 2080s. The models used predicted the proportion of birds over-wintering in Britain that do so on the estuaries of EA Anglian region. The Dunlin national index is an explanatory variable in the predictive model used. In order to obtain an indication of how the size of the population affects the estimates, separate predictions have been made while using minimum and maximum recorded values of the Dunlin national index from the past three decades. The resulting predictions have been further adjusted to give those numbers expected on the estuaries of the Suffolk coast while using the mean value for the proportion of birds over-wintering in EA Anglian the extremes in numbers that might be expected to over-winter on these estuaries.
Observed Predictions: Observed Predictions: based on minimum British index based on maximum British index 8000 8000
6000 6000
4000 4000
2000 2000 Over-winter average numbers Over-winter average numbers
0 0 h h w h 0 h g 0 0 h 70 80 85 9 95 ig 70 8 85 9 95 ig 9 9 9 Low Low Hi 9 9 Lo Low 1 1975 19 1 19 1 0 H Hig 19 1975 19 1 19 1 0 H Hig 0 0 80 0 0 050 05 050 080 High 202 202 2 2 2080 20Low 202 202 2 205 2080 2Low
88 MONARCH 2 Report – Chapter 5 ______
Ringed plover - Values predicted under all UKCIP02 scenarios considered are well within the recorded range over the past three decades and close to the average over that period (Figure 5.4b). For any given time-frame the difference between predictions made under the High and Low scenarios are small compared to the observed between winter variation over the past three decades.
Dunlin - Numbers predicted under all UKCIP02 scenarios considered are generally within the recorded range over the past three decades (Figure 5.4c). However, the model for this species shows that in years when the number of birds over-wintering in Britain has been relatively low and during this period the proportion of birds wintering on the estuaries of the East Anglian coast has been relatively high. If this trend continues then the proportion of birds (although not the number) over- wintering there may increase beyond recent levels as dunlin forsake estuaries to the west of Britain and probably Ireland. Also, if the numbers of dunlin over-wintering in Britain and Ireland were to decline, although numbers on the estuaries of the East Anglian coast may remain relatively stable, the proportion of the British and Irish population supported by these estuaries would increase.
Figure 5.4d: Observed numbers of redshank over-wintering on Suffolk estuaries between the winters of 1969/70 and 1999/2000 inclusive and predicted values (with 95% confidence intervals) made under the UKCIP02 Low and High scenarios for 2020s, 2050s and 2080s. The models used predicted the proportion of birds over-wintering in Britain that do so on the estuaries of the Suffolk Coast. These predictions have been converted to numbers using both minimum and maximum values for birds over- wintering in Britain. Because the proportion is not dependent on country-wide numbers this gives a indication of the extremes in numbers that might be expected to over-winter on these estuaries.
Observed Predictions: Observed Predictions: based on minimum British index based on maximum British index 7000
6000 6000
5000
4000 4000
3000
2000 2000 Over-winter average numbers Over-winter average numbers 1000
0 0 h w h w w ig o g o o igh 75 85 95 Low L L 70 80 90 L High 9 0 H Hi H 1970 19 1980 19 1990 1 0 50 0 80 19 1975 19 1985 19 1995 0 5 020 2 080 202 202 20 20 20 2080 High 2 20 2050 Low2050 High 2080 2Low
Redshank - Numbers predicted under all UKCIP02 scenarios considered are well within the recorded range over the past three decades and close to the average over that period (Figure 5.4d). For any given time-frame the differences between the predictions of redshank numbers made under the High and Low scenarios are no larger than the observed between winter variation over the past three decades.
5.3.2 Impact of sea level rise
Models have been developed for predicting the density of birds held on estuaries from aspects of estuary morphology for five species of wader – oystercatcher, knot, dunlin, curlew and redshank (section 5.2.2). The application of these models for predicting the change in capacity of estuaries to hold wader densities in response to estuary realignment has been demonstrated previously for the Deben estuary in Suffolk and the Duddon estuary in Cumbria (Austin et al. 2001).
MONARCH 2 Report – Chapter 5 89 ______
Following the protocol developed in MONARCH 1, management plans from the Suffolk Estuarine Strategies (EA, 1999) and predicted sea level rise under the UKCIP 2020s and 2050s High scenarios for climate change were incorporated into the ArcView Geographic Information System (ESRI, 2002). The necessary management plans were available for three Suffolk estuaries, the Deben, the Alde / Ore and the Blyth. Similar data were not available for the Stour and Orwell estuaries. The predictions of sea level rise used were those previously made for 2020s and 2050s under UKCIP Low and High scenarios (Austin et al., 2001). LIDAR data were used to assess whether management compartments defined in the management plans would be subject to inundation with sea level rise and thus predict the future shape of each estuary. Under the existing management strategies there is very little scope for significant changes to the shapes of the three estuaries considered (Figure 5.5). There would be a small increase in the overall area of both the Blyth and the Alde / Ore while the area of the Deben effectively remains at its current size.
The changes to the values of the variables used by the MONARCH models were, therefore, small for all estuaries considered and consequently the nature of the sediments would not be expected to change sufficiently to have a substantial effect on the densities of birds that they are likely to support. This is because the basic shape of all the estuaries will remain "long and narrow" and thus sediments are likely to remain muddy. Accordingly, predictions for each estuary were only made under those management scenarios that would allow the greatest scope for realignment, i.e., allowing all management compartments, other than those designated as hold the line, to be included in the tidal regime where the LIDAR data indicated that they would be inundated by spring high tides with predicted sea level rise (Figures 5.6a-e). Furthermore, most of the area of those compartments that would no longer be defended under the management plan would be inundated under our predictions of sea level rise associated with the UKCIP Low scenario for the 2020s, thus leaving little scope for further change. Consequently the predictions under the various scenarios and time-frames would be expected to be similar.
Oystercatcher – Relatively small numbers of oystercatcher winter on the estuaries of the Suffolk coast, a consequence of their preference for sandier sediments. The Blyth and Deben baseline predictions (model predictions under current estuary shape) are three-fold higher than, but not significantly different from, those observed over the past three decades (Figure 5.6a). This suggests that other factors not accounted for by the estuary morphology models such as the impacts of low shellfish stocks in the region (Atkinson et al., 2003) have held numbers of oystercatchers on these estuaries below those that they would otherwise have been capable of supporting. The baseline predictions for the Alde / Ore are within the range of values observed over the past three decades. As expected, there is little change in the predicted capacity of the estuaries to hold oystercatchers under the various scenarios for sea level rise. Proportionally small changes in capacity over their baseline values may be expected on the Blyth and Deben. The proportional increase in capacity of the Alde / Ore over its baseline value could be more substantial. Overall there could be an increased capacity for the estuaries of the Suffolk coast to support oystercatchers, however, the predictions under the various scenarios are not significantly different from those observed over the past three decades.
90 MONARCH 2 Report – Chapter 5 ______
Figure 5.5: Suffolk Estuarine Strategies management compartments (left) and the maximum expected change (right) to the extent of the three estuaries considered under these management strategies and predicted sea level rise under the UKCIP 2020s and 2050s High scenarios for climate change.
Blyth
Alde / Ore
Deben
Management Predicted Extent Do nothing Possible defence Predicted extent by 2020 Managed realignment Predicted additional extent by 2050 Hold the line Seawall CHaMPs dependent Current extent Seawall 012345678910Km Current extent
MONARCH 2 Report – Chapter 5 91 ______
Figure 5.6a: Density and numbers of oystercatcher over-wintering on the Blyth, Alde / Ore and Deben. Observed values between the winters of 1969/70 and 1999/2000 inclusive and predicted estuary capacity to hold oystercatchers (with 95% confidence intervals) made under predictions of sea level rise for 2020s and 2050s.
1200 3.0 Blyth 1000 2.5
2.0 800
1.5 600
1.0 400 Average number Average Density (birds/ha) Density 0.5 200
0.0 0 h w h e w h 0 5 0 ig o 0 5 0 o ig 7 7 8 7 7 8 85 lin 9 9 9 985 L 9 9 9 9 e L High H 1 1 1 1 1990 1995 0 1 1 1 1 1990 1995 s 0 0 0 20 Low20 H a 5 Baseline 0 0 05 050 Hig B 0 2 2 2 2 202 202 2050 2Low
2.5 2500 Deben 2.0 2000
1.5 1500
1.0 1000 Average number Average Density (birds/ha) Density 0.5 500
0.0 0 e w h 0 5 0 igh 0 5 0 o ig ow igh Low 7 7 8 85 lin L 7 7 8 H 9 9 9 9 e L H 9 9 9 985 0 1 1 1 1 1990 1995 0 0 0 1 1 1 1 1990 1995 5 2 2 5 50 H Baseline Bas 0 0 0 0 2020 2020Low High2050 20 2 2 2 2
2.5 Alde / Ore 5000
2.0 4000
1.5 3000
1.0 2000 Average number Average Density (birds/ha) Density 0.5 1000
0.0 0 e w h w 0 5 0 o 5 0 5 o gh 7 7 8 lin ig 9 9 9 9 9 985 e L H 9 9 L Hi 1 1 1 1 1990 1995 s 0 0 1970 1975 1980 198 1 1 0 0 a 2 2 B 0 050 Low Baseline 020 Low 05 20 2 2 2050 High 2 2020 High205 2
Knot – Historically numbers of knot on the estuaries of the Suffolk coast have been very low (Figure 5.6b). Knot tend to favour larger sites with wide intertidal areas such as The Wash and the Thames that are capable of supporting the large flocks formed by this species. Knot are also known to move between sites both within and between winters to a greater degree than most wader species (Rehfisch et al., 1996, Rehfisch et al., 2003). The patterns of occurrence on the estuaries of the Suffolk coast over the past three decades are typical of smaller sites with very few or no birds being recorded in most years, punctuated by years when substantial flocks are recorded. Because the models are based on average values the sporadic occurrence of knot on these sites leads to baseline predictions that are higher than in a typical winter but within the range of historic values. The predictions under the sea level rise scenarios do not differ substantially from those of the baseline and none of these differences
92 MONARCH 2 Report – Chapter 5 ______are statistically significant. Thus overall there is little to suggest that the capacity of the estuaries of the Suffolk coast to support Knot will change.
Figure 5.6b: Density and numbers of knot over-wintering on the Blyth, Alde / Ore and Deben. Observed values between the winters of 1969/70 and 1999/2000 inclusive and predicted estuary capacity to hold knot (with 95% confidence intervals) made under predictions of sea level rise for 2020s and 2050s.
800 2.0 1.8 Blyth
1.6 600 1.4 1.2 1.0 400 0.8 0.6
Average number Average 200
Density (birds/ha) Density 0.4 0.2 0.0 0 0 5 5 ine igh gh 0 5 ine igh gh 80 l Low i l Low i 97 9 98 99 H 970 98 985 99 H 1 1975 1 1 1990 1 se 0 1 1975 1 1 1990 1 se 0 a 20 5 50 H a 20 5 50 H B 0 B 0 2020 20Low 20 2 2020 20Low 20 2
2.0 2000 1.8 Deben 1.6 1500 1.4 1.2 1.0 1000 0.8 0.6 500 Average number Average Density (birds/ha) 0.4 0.2 0.0 0 w w 0 5 gh 0 5 igh 70 75 8 85 9 i 70 75 8 85 9 9 9 eline Low H 9 Low H 19 1 19 1 1990 19 s 0 19 1 19 19 1990 19 0 50 50 Ba 020 High 05 Baseline 020 High 2020 2Lo 2 20 2020 2Lo 205 20
2.0 4000 1.8 Alde / Ore 1.6 3000 1.4 1.2 1.0 2000 0.8 0.6 1000 Average number Average Density (birds/ha) 0.4 0.2 0.0 0 5 0 0 5 0 0 igh 8 Low Low High 8 Low Low H 97 99 0 97 99 0 1970 1 19 1985 1 1995 20 1970 1 19 1985 1 1995 20 Baseline 020 High Baseline 020 High 20 2 205 2050 20 2 205 2050
Dunlin – Dunlin are found at their highest concentrations on muddy estuaries and consequently the estuaries of the Suffolk coast support relatively high densities of this species (Figure 5.6c). The model for dunlin is particularly good (explaining 84% of the variation in bird density - Austin et al., 2001) and the baseline predictions suggest that potentially these densities could be even higher. As expected, the predicted values under the various scenarios for sea level rise on any one estuary are all very similar. Those for the Blyth and Alde / Ore are higher than for the baseline while those for the Deben are lower - although these differences are not statistically significant. While overall there could be an
MONARCH 2 Report – Chapter 5 93 ______increased capacity for the estuaries of the Suffolk coast to support dunlin the comparison between the baseline predictions and historic values suggest that there is already surplus capacity. Habitat availability in this area is therefore only likely to become limiting if the number of birds arriving on the Suffolk coast were to increase several-fold.
Figure 5.6c: Density and numbers of dunlin over-wintering on the Blyth, Alde / Ore and Deben. Observed values between the winters of 1969/70 and 1999/2000 inclusive and predicted estuary capacity to hold dunlin (with 95% confidence intervals) made under predictions of sea level rise for 2020s and 2050s.
8000 20 Blyth
6000 15
10 4000
5 number Average 2000 Density (birds/ha) Density
0 0 e w h w h 0 5 5 in o ow g 5 5 5 ine o ow g 7 8 9 l L L 7 8 9 l L L 9 9 9 e Hi 9 9 9 e Hi 1 1975 1980 1 1990 1 s 0 0 1970 1 1980 1 1990 1 s 0 0 5 5 5 5 Ba 0 0 Ba 0 0 20202020 High2 2 20202020 High2 2
18 20000 16 Deben 18000 14 16000 14000 12 12000 10 10000 8 8000 6 6000 Average number Average Density (birds/ha) Density 4 4000 2 2000 0 0 e w h w h e w w h 5 0 in o ig 5 5 0 o ig 7 9 l Lo 7 8 9 lin Lo 9 9 e H 9 9 9 e H s 0 0 1970 1 1980 1 1 1995 0 0 1970 1 1980 1985 1 1995 a 2 5 2 5 B 0 0 Bas 0 0 2 2020 Hig 2050 2L 2 2020 High2050 2L
20 40000 18 Alde / Ore 16 30000 14 12 10 20000 8 6 10000 Average number Average Density (birds/ha) Density 4 2 0 0 e w h e w h 0 5 5 n g 0 5 5 n g 7 8 9 li 7 8 9 li 9 9 9 e Lo Low Hi 9 9 9 e Lo Low Hi 1 1975 1980 1 1990 1 s 0 0 1 1975 1980 1 1990 1 s 0 0 0 5 5 2 5 5 Ba 020 High 0 0 Ba 0 020 High 0 0 20202 2 2 2 2 2 2
Curlew – The number of curlew over-wintering on the estuaries of the Suffolk coast has increased steadily over the past three decades (Figure 5.6d). Although the predictive power of the model for this species is the least powerful being used (explaining 33% of the variation in bird density - Austin et al., 2001), comparison of the historic values with the baseline predictions suggest that while numbers
94 MONARCH 2 Report – Chapter 5 ______on the Deben may be close to capacity there is currently surplus capacity on the Blyth and Alde / Ore. While mindful of the wide confidence limits of the predictions for this species, those made under the various scenarios for sea level rise suggest that the capacity of the estuaries of the Suffolk coast to support curlew will remain sufficient to absorb a several-fold increase over current numbers. The only substantial change in curlew density is likely to occur on the Alde / Ore complex but this will be largely offset by the increase in area.
Figure 5.6d: Density and numbers of curlew over-wintering on the Blyth, Alde / Ore and Deben. Observed values between the winters of 1969/70 and 1999/2000 inclusive and predicted estuary capacity to hold curlew (with 95% confidence intervals) made under predictions of sea level rise for 2020s and 2050s.
600 1.6 Blyth 1.4 500
1.2 400 1.0
0.8 300
0.6 200
0.4 number Average Density (birds/ha) 100 0.2
0.0 0 w h h 0 5 0 o 0 5 0 5 ine ow ow 8 8 9 95 line Low 70 L 9 9 9 e L Hig 9 el L High 1970 1975 19 1 1 1 0 1 1975 198 198 199 199 s 0 as 20 2 50 High B 0 Ba 0 0 2 2020 Hig 2050205 20202 2050 2
1.6 1600 Deben 1.4 1400
1.2 1200
1.0 1000
0.8 800
0.6 600
400 0.4 number Average Density (birds/ha) Density 0.2 200
0.0 0 w h w h 5 o 5 0 5 line ow 70 9 Lo 75 80 e L Low 9 0 L 0 Hig 9 9 98 s High 1 1975 1980 1985 1990 19 2 5 0 1970 1 1 1 199 199 0 Baseline 0 020 Hig 0 Ba 2 2 2 205 2020202 20502050 High
3.5 Alde / Ore 4000 3.0
2.5 3000
2.0 2000 1.5
1.0 1000 Average number Average Density (birds/ha)Density 0.5
0.0 0 e w w h 0 5 5 0 0 5 0 7 7 lin 7 8 9 o ig 985 990 99 975 98 eline L Low 19 19 1980 1 1 1 se 19 1 1 19 19 1995 s 0 0 H 0 a 50 Lo 2 B 020 High 0 050 High Ba 2020 2Low 2 2 202 20 205 2050 High
MONARCH 2 Report – Chapter 5 95 ______
Figure 5.6e: Density and numbers of redshank over-wintering on the Blyth, Alde / Ore and Deben. Observed values between the winters of 1969/70 and 1999/2000 inclusive and predicted estuary capacity to hold redshank (with 95% confidence intervals) made under predictions of sea level rise for 2020s and 2050s.
1600 5 Blyth 1400
4 1200
1000 3 800
2 600
Average numberAverage 400
Density (birds/ha) Density 1 200
0 0 e w h h 0 5 o ow 0 5 0 5 0 ow 7 8 90 line 9 9 9 elin Hig L 97 e L Hig 1 1975 1980 1 1 1995 s 0 L 0 1 197 198 198 199 1995 0 Low0 2 50 High 5 Ba 0 0 Bas 05 2 202 20502 20202020 High20 2
4 5000 Deben
4000 3
3000 2 2000
1 number Average Density (birds/ha) Density 1000
0 0 h e w 0 5 0 5 ow 0 5 5 0 o 7 7 80 9 9 7 7 8 9 lin 9 9 9 Hig L 9 9 e High 19 1 1 1985 19 1 seline 0 0 19 1 1980 19 1 1995 0 0 Low 5 50 High Ba 02 05 Bas 020 L 0 2020 2Low 2 2050 High 2 202 20 2
7 Alde / Ore 10000 6 8000 5
4 6000
3 4000 2 Average number Average Density (birds/ha) Density 2000 1
0 0 e w h w 0 5 0 5 0 5 5 5 7 9 o ow 7 8 line o 9 980 9 Hig L 9 9 990 e L High 197 1 1 1985 199 1 0 L 0 197 1 1980 1 1 199 s 0 2 50 High 20 High 50 Baselin 0 0 Ba 0 0 2 202 20502 2020 2Low 2 205
Redshank – Redshank are found at their highest concentrations on muddy estuaries and consequently the estuaries of the Suffolk coast support relatively high densities of this species (Figure 5.6e). The model for redshank is particularly good (explaining 87% of the variation in bird density - Austin et al., 2001) and the baseline predictions suggest that potentially these densities could be even higher on the Deben and Alde / Ore complex. The model suggests that while densities on the Blyth and Deben are unlikely to change substantially with sea level rise, those on the Alde / Ore may decrease. However this decrease is not statistically significant and would be offset by the increase in area. Overall the capacity of the estuaries of the Suffolk coast to support redshank is unlikely to change much with sea level rise and is probably capable of absorbing a substantial increase in birds. Habitat
96 MONARCH 2 Report – Chapter 5 ______availability in this area is therefore only likely to become limiting if the number of birds arriving on the Suffolk coast were to increase several-fold.
5.3.3 Overall Impact
The modelling has considered both how changes in climate might affect the distribution of waders within Britain and Ireland and the implications this may have for the numbers of these birds over- wintering on the estuaries of Suffolk coast; and how rising sea levels might affect the availability and nature of the intertidal habitat required by these birds. In order to assess the overall impact of climate change, predictions from these two processes must be considered together. Predictions have been made for both aspects in the case of oystercatcher, dunlin and redshank, climate models only for ringed plover, and sea level rise models only for knot and curlew.
While other climate change driven factors will also impact on these birds it has not been possible to address these here. The largest unknown amongst these is likely to be the effect of climate change on their Arctic breeding grounds and the availability of stop-over and wintering sites along their migration routes (Lindström and Agrell, 1999; Rehfisch and Crick, 2003). It is possible that alternative sites nearer to the breeding grounds may become increasingly suitable for over-wintering. This may lead to the phenomenon of "short-stopping" whereby birds over-winter further to the north or east in areas previously unsuitable because of severe climate, and result in a reduction in the number of waders over-wintering in Britain and Ireland. Although the distribution of waders in Britain has changed (Austin and Rehfisch, 2003; Rehfisch et al., 2003) and over-wintering numbers of eight species have declined through the 1990s (Rehfisch et al., 2003) there is as yet no direct evidence for this effect in waders. Ongoing work being carried out under the WeBS partnership at a wider European scale may help determine whether the recent decline in wader numbers overwintering in Britain and Ireland can be explained by an increase in waders numbers over-wintering on continental Europe. It has already been recorded that some species of wildfowl are increasingly over-wintering in Eastern Europe on water-bodies that no longer remain frozen for long periods throughout the winter.
The MONARCH 2 models provide predictions of how the number of waders over-wintering in Suffolk may change and whether with managed realignment there is likely to be sufficient habitat of suitable quality to support these numbers. Although the distribution of four of the species considered is significantly associated with weather (the proportion of waders in Britain that over-winter on the Suffolk or East Anglia coast being related to coastal minimum temperatures), in no case is this expected to result in a large increase in numbers on the estuaries of the Suffolk coast under the UKCIP02 scenarios if the present relationships remain true. In most cases, consideration of the baseline predictions for the capacity of these estuaries suggests that there is already surplus capacity and this situation is unlikely to change under the various predictions for sea level rise (Table 5.5). Consequently, when considering those aspects of the birds' response to climate change that we have been able to model, there is probably little cause for concern that the estuaries of the Suffolk coast will not be able to hold the expected numbers of waders under the various UKCIP02 scenarios.
5.4 Discussion and conclusions
The coastal work tackled under MONARCH 2 has been successful in providing further support for the hypothesis that changes in weather patterns over the past three decades have resulted in a broad scale redistribution of waders over-wintering in Britain. In particular, average minimum temperatures on the muddy estuaries of the east coast help explain a proportion of the variation in the distribution of six of the seven species considered. Winter temperatures are thus likely to be at least one of the causal factors leading to the redistribution of over-wintering waders in Britain. This is an important result because relatively few studies (Parmesan and Yohe, 2003; Root et al., 2003) have been able to demonstrate a link between changes in faunal or floral distributions and changes in climatic conditions and yet such a mechanism is an inescapable assumption of all models that seek to predict future changes in distributions based upon contemporary associations including MONARCH.
MONARCH 2 Report – Chapter 5 97 ______
Table 5.5: Comparison of the numbers of wader predicted for the estuaries of the Suffolk coast against predicted capacity. The range given for baseline capacity are the values obtained by summing respectively the lower and upper 95% confidence limits, for the number of each species as predicted by the sea level rise models, based on current estuary morphology, across all the Suffolk estuaries. The range given for baseline numbers is that obtained from recorded values, summed across those estuaries for the winters 1969/20 to 1999/2000. The range given for capacity under UKCIP scenarios are the values obtained by summing respectively the lower and upper 95% confidence limits based on future predictions of estuary morphology across the Suffolk estuaries. The range given for numbers under the UKCIP scenarios are the values obtained by summing, respectively, the lower 95% confidence limit of numbers estimated under minimum recorded GB index and upper 95% confidence limit of numbers estimated under maximum recorded GB index, across these estuaries.
Prediction Oystercatcher Ringed plover Knot Dunlin Curlew Redshank
Baseline Capacity 0 - 4744 N/A 0 - 4264 16396 - 42438 848 - 5229 6165 - 12007 Numbers 11 - 605 24 - 102 0 - 438 2475 - 5184 307 - 771 537 - 3485
2020 Low Capacity 0 - 6369 N/A 0 - 5210 19461 - 50983 385 - 4945 6570 - 12218 Numbers 23 - 900 18 - 109 N/A 2242 - 5149 227 - 1036 494 - 4304
2020 High Capacity 0 - 7463 N/A 0 - 6046 21590 - 57287 348 - 5402 7199 - 13527 Numbers 24 - 931 18 - 110 N/A 2273 - 5211 226 - 1036 501 - 4363
2050 Low Capacity 0 - 6691 N/A 0 - 5444 20266 - 53188 400 - 5161 6860 - 12756 Numbers 26 - 1035 19 - 114 N/A 2371 - 5417 226 - 1033 521 - 4558
2050 High Capacity 0 - 7776 N/A 0 - 6269 22288 - 59281 357 - 5527 7341 - 13907 Numbers 31 - 1256 21 - 122 N/A 2539 - 5819 224 - 1029 556 - 4937
The coastal work has been less successful in its second aim of developing these associations into universal tools that could be used for conservation management at the local scale. However, the approach may work in some regions. The "estuary morphology models" provide a robust and reliable method for predicting densities of a variety of wader species from estuary morphology (Austin et al., 1996; Austin and Rehfisch, 2003) and thus the capacity of estuaries to hold these birds. However, it is only appropriate to apply these models to estuaries where the managed response to sea level rise allows a fairly substantial re-alignment of the estuary extent and consequentially a change to the values of the morphological variables used by those models. This condition was met only for the Alde / Ore estuaries on the Suffolk coast and not for the estuaries of the New Forest coast (see Chapter 6). Although based on highly significant associations, the proportion of variation in wader distributions explained by winter weather alone is relatively low because of the many other factors such as land- claim, water quality and recreational disturbance that also influence the distributions. However, from those aspects of the birds' response to climate change that we have been able to model, the Suffolk estuaries should have the capacity to hold numbers of waders similar to those at present under the various UKCIP scenarios.
Thus the overall approach taken has successfully captured those aspects of change in wader distributions that can be expected in response to changes in estuary morphology with sea level rise and to continued direct response to changes in winter weather. However, a substantial part of the variation in wader distributions still remains unquantified. This overall approach provides a useful qualitative tool for assessing the direction in which these factors may drive future wader distributions, but, given that many other factors unrelated to climate change appear to have influenced past distributions, the approach has proved to be less useful as a quantitative tool upon which to base targets for the management of these natural resources. However, the importance of developing models for predicting waterbird distributions cannot be over-stated, as waterbirds are the principal feature of most coastal SPAs designated under the Habitats Directive (Stroud et al., 2001). Consequently, further research towards this goal is warranted.
98 MONARCH 2 Report – Chapter 5 ______
The modelling approach adopted for other species within the MONARCH project is adaptable to coastal waterbirds. This would require the development of coastal winter bioclimatic zones and we have gone some way towards creating them. The models would need to be adapted to work with densities rather than simple presence / absence data and to incorporate annual weather data rather than a baseline average given that distributions have been changing over the past three decades (Austin and Rehfisch, in press; Rehfisch et al., 2003). Such an approach may prove more successful at integrating changes due to sea level rise with changes due to weather patterns. However, work towards incorporating the effect of other factors such as land-claim, water quality and disturbance into the models is also recommended. This is because whatever the modelling approach adopted, the likelihood of predicted distributions of any species, waders or otherwise, being realised will be related to the amount of observed variation that can be explained by the factors considered by the models and to the availability of realistic and accurate future scenarios of these factors.
A cautionary note is needed, resulting from the approach that has been taken for predicting the likely effect of climate change on wintering waterbirds. The available waterbird data come from a survey that has collected data at relatively fine spatial and temporal scales for longer and in a more complete manner than that available for the large majority of surveys of other faunal or floral groups in the world (Musgrove et al., 2001). The observed changes in the British wader distributions have been clearly associated with changes in winter weather (Rehfisch et al., 2003; Austin and Rehfisch, in press). The habitat associations of waders are better known than for most other biological groups (e.g., Goss-Custard, 1977a; Goss-Custard, 1977b; Quammen, 1982; Goss-Custard and Yates, 1992; Yates et al., 1993; Goss-Custard et al., 1994; Goss-Custard 1995; Yates and Goss-Custard 1997; Rehfisch et al., 1999; Rehfisch et al., 2000). Many other aspects of wader biology have been extensively studied. However, even with this wealth of data and interpretative information, and ignoring the likely considerable impacts of climate change on the Arctic breeding grounds of waders (Lindström and Agrell, 1999; Rehfisch and Crick, 2003), it is still proving difficult to predict clear effects of climate change on the wader distributions due to the amount of unexplained variation in the sound models used to describe wader habitat usage and distributional change with weather. It is important for all predictive work on faunal distributions, whether based on this approach or bioclimatic relationships, to at least attempt to determine the size of the likely error in the predictions; otherwise the value of the predictions may be questioned. Even if these errors prove to be large, the MONARCH predictive assessment should still provide scenarios of likely change that can help guide national conservation priorities towards wildlife, and to help assess the approach that should be taken towards the anthropogenic factors that are leading to the climatic change.
5.5 References
Atkinson, P.W., Clark, N.A., Bell, M.C., Dare, P.J., Clark, J.A. and Ireland, P.L. (2003). Changes in commercially fished shellfish stocks and shorebird populations in the Wash, England. Biological Conservation, 114, 127-141.
Armitage, M.J.S., Burton, N.H.K., Atkinson, P.W., Austin, G.E., Clark, N.A., Mellan, H.J. and Rehfisch, M.M. (2002). Reviewing the Impact of Agency Permissions and Activities on Bird Populations in Special Protection Areas: Level 1 Interpretation. BTO Research Report No. 296 for the Environment Agency. British Trust for Ornithology, Thetford, Norfolk, UK.
Austin, G. and Rehfisch, M.M. (in press). Shifting non-breeding distributions of migratory fauna in relation to climatic change. Global Change Biology
Austin, G. and Rehfisch, M.M. (2003). The likely impact of sea level rise on waders (Charadrii) wintering on estuaries. Journal for Nature Conservation, 11, 43-58.
Austin, G.E., Rehfisch, M.M., Viles, H.A., and Berry, P.M. (2001). Impacts on coastal environments. In: Harrison, P.A., Berry, P.M. and Dawson, T.P. (Eds.) Climate Change and Nature Conservation in
MONARCH 2 Report – Chapter 5 99 ______
Britain and Ireland – modelling natural resource responses to climate change (the MONARCH project). UKCIP Technical Report, Oxford, pp177-228.
Austin, G.E., Peachel, I. and Rehfisch, M.M. (2000). Regional trends in coastal wintering waders in Britain. Bird Study, 47, 352-371.
Austin, G.E., Rehfisch. M.M., Holloway, S.J., Clark, N.A., Balmer, D.E., Yates M.G., Clarke, R.T., Swetnam, R.D., Eastwood, J.A., le V. dit Durell, S.E.A., West, J.R. and Goss-Custard, J.D. (1996). Estuaries, Sediments and Shorebirds III: Predicting Wader Densities on Intertidal Areas. BTO Research Report No. 160 to ETSU (T/04/00207/REP). British Trust for Ornithology, Thetford, Norfolk.
Boyd, H. and Madsen, J. (1997). Impacts of global change on arctic-breeding bird populations and migration. In: Oechel, W.C., Callaghan, T.Gilmanov, T., Holten, J.I., Maxwell, B., Molau, U. and Sveinbjornsson, B. (Eds). Global change and Arctic terrestrial ecosystems. Springer-Verlag, New York, USA, pp. 201-217.
Cayford, J. and Waters, R. (1996). Population estimates for waders (Charadrii) wintering in Great Britain, 1987/88-1991/92. Biological Conservation, 77, 1-17.
Colhoun, K. (2001). Irish Wetland Bird Survey 1998-1999: Results of the fifth season of the Irish Wetland Bird Survey; Including summarised results from Northern Ireland. Bird Watch Ireland.
Environment Agency (1999). Suffolk Estuarine Strategies Phase II – Report C Deben Estuary. Environment Agency, Anglian Region, Peterborough.
ESRI (2001). Environmental Systems Research Institute, Inc. USA.
Frazier, S. (Ed.) (1999). A Directory of Wetlands of International Importance. Wetlands International and Ramsar Convention Bureau. CD.
Goss-Custard, J. (1977a). The ecology of the Wash 3. Density related behaviour and the possible effects of the loss of feeding grounds on wading birds (Charadrii). Journal of Applied Ecology, 14, 721-739.
Goss-Custard, J.D. (1977b). Predator responses and prey mortality in redshank, Tringa totanus (L) and a preferred prey, Corophium volutator. Journal of Animal Ecology, 46, 21-35.
Goss-Custard, J.D. and Yates, M.G. (1992). Towards predicting the effect of salt-marsh reclamation on feeding bird numbers on the Wash. Journal of Applied Ecology, 29, 330-340.
Goss-Custard, J.D., Caldow, R.W.G., Clarke, R.T., Durell, S.E.A le V.dit, Urfi, J. and West, A.D. (1994). Consequences of habitat loss and change to populations of wintering migratory birds: predicting the local and global effects from studies of individuals. Conservation: the Science and the Action (eds. Coulson, J and Crockford, N.J.). Ibis, 137, S56-66
Goss-Custard, J.D. (1995). Effects of habitat loss and habitat change on estuarine shorebird populations. Coastal Zone Topics: Process, Ecology & Management 1: 61-67
Lindström, Å. and Agrell, J. (1999). Global change and possible effects on the migration and reproduction of arctic-breeding waders. Ecological Bulletins. 47, 145-159.
Manly, B.F.J. (1991). Randomization and Monte Carlo methods in biology. Chapman and Hall, London, UK.
100 MONARCH 2 Report – Chapter 5 ______
Moser, M.E. (1987). A revision of population estimates for waders (Charadrii) wintering on the coastline of Britain. Biological Conservation, 39, 153-164.
Musgrove, A.J., Pollitt, M.S., Hall, C., Hearn, R.D., Holloway, S.J., Marshall, P.E., Robinson, J.A. and Cranswick, P.A. (2001). The Wetland Bird Survey 1999-2000: Wildfowl and Wader Counts. BTO/WWT/RSPB/JNCC, Slimbridge.
Parmesan, C. and Yohe, G. (2003). A globally coherent fingerprint of climate change impacts across natural systems. Nature 421, 37-42.
Quammen, M.L. (1982). Influence of subtle substrate differences on feeding by shorebirds on intertidal mudflats. Marine Biology, 71, 339-343.
Rehfisch, M.M., Austin, G.E., Freeman, S.N., Armitage, M.J.S. and Burton, N.H.K. (2004). The possible impact of climate change on the future distributions and numbers of waders on Britain's non- estuarine coast. In: Rehfisch, M.M., Feare, C., Jones, N.V. and Spray, C. (eds), Coastal Birds and Climate Change. Ibis, 146 (suppl 1), 70-81
Rehfisch, M.M. and Crick, H.Q.P. (2003). Predicting the impact of climate change on Arctic-breeding waders. Wader Study Group Bulletin, 100, 86-95.
Rehfisch, M.M., Austin, G.E., Armitage, M.J.S., Atkinson, P.W., Holloway, S.J., Musgrove, A.J. and Pollitt, M.S. (2003). Numbers of wintering Waterbirds in Great Britain and the Isle of Man (1994/1995-1998/1999): II. Coastal waders (Charadrii). Biol. Conservation, 112, 329-341.
Rehfisch, M.M., Austin, G.E., Clark, N.A., Clarke R.T., Holloway, S.J., Yates, M.G., Durrel, S.E.A. le V. dit, Eastwood, J.A., Goss-Custard, J.D., Swetnam, R.D. and West, J.R. (2000). Predicting densities of wintering Redshank Tringa totanus from estuary characteristics: a method for assessing the likely impact of habitat change. Acta Ornithologica, 35, 25-32.
Rehfisch, M.M., Holloway, S.I., Yates, M.G., Clarke, R.T., Austin, G., Clark, N.A., Durell, S.E.A., le V. dit, Eastwood, J.A., Goss-Custard, J.D., Swetnam,R.D. and West, J.R. (1999). Predicting the effect of habitat change on waterfowl communities: a novel empirical approach. In: Goss-Custard, J. Rufino R.& Luis A. (eds.) Predicting Habitat Loss, pp. 116-126. HMSO, London.
Rehfisch, M.M., Austin, G.E., Freeman, S.N., Armitage, M.J.S., and Burton, N.H.K. (2004). The possible impact of climate change on the future distributions and numbers of waders on Britain’s non- estuarine coast. In: Rehfisch, M.M., Feare, C., Jones, N.V. and Spray, C. (eds), Coastal Birds and Climate Change. Ibis.
Root, T.L., Price, J.T. Hall, K.R. Schneider, S.H., Rosenzweig, C. and Pounds, J.A. (2003). Fingerprints of global warming on wild animals and plants. Nature 421, 57-60.
SAS (2001). The SAS Institute Inc. Cary, NC, USA.
Stroud, D.A., Chambers, D., Cook, S., Buxton, N., Fraser, B., Clement, P., Lewis, P., McLean, I., Baker, H. and Whitehead, S. (2001). The UK SPA network: its scope and content. Volume 1: Rationale for the selection of sites. JNCC, Peterborough, UK.
Underhill, L.G. and Prŷs-Jones, R. (1994). Index numbers for waterbird populations. I. Review and methodology. Journal of Applied Ecology, 31, 463-480.
Yates, M.G., Clarke, R.T., Swetnam, R.D., Eastwood, J.A., Durell, S.E.A. le V. dit, West, J.R., Goss- Custard, J.D., Clark, N.A., Holloway, S.J. and Rehfisch, M.M. (1996). Estuary, Sediments and
MONARCH 2 Report – Chapter 5 101 ______
Shorebirds I. Determinants of the Intertidal Sediments of Estuaries. A report by the Institute of Terrestrial Ecology under contract to ETSU (ETSU Project T/04/00201/REP).
Yates, M.G., Goss-Custard, J.D., McGrorty, S., Lakhani, K.H., Durell, S., E.A. le V. dit, Clarke, R.T., Rispin, W.E., Moy, I., Yates, T., Plant, R.A., Frost, A.E. (1993). Sediment characteristics, invertebrate densities and shorebird densities on the inner banks of the Wash. Journal of Applied Ecology, 30, 599- 614.
Yates, M.G. and Goss-Custard, J.D. (1997). The development of a correlative approach relating bird distribution and remotely-sensed sediment distribution to predict the consequences to shorebirds of habitat change and loss. In: J.D. Goss-Custard, R. Rufino & A Luis (Eds.) Predicting and detecting the effect of habitat loss and change on wetland bird populations, ITE Symposium no. 30; Wetlands International publication no. 42. HMSO, London, pp. 138-144.
102 MONARCH 2 Report – Chapter 6 ______
6 Impacts for the Hampshire case study area
J.E. HOSSELL, G.E. AUSTIN, P.M. BERRY, N. BUTT, H.Q.P. CRICK, S. FREEMAN, P.A. HARRISON, G.J. MASTERS, A. MORRISON, M.M. REHFISCH, P. SCHOLEFIELD, N. WARD AND I. WILDE
Summary Hampshire was selected as a case study area because of its high sensitivity to climate change. Both habitats (beech hangers and wet heathland) were chosen because of their possible sensitivity to climate change and their conservation importance in the area. Beech hangers were represented by beech (Fagus sylvatica), ash (Fraxinus excelsior) and dog’s mercury (Mercurialis perennis) as dominants; hawfinch (Coccothraustes coccothraustes) as a rare species and yellow-necked mouse (Apodemus flavicollis) as a flagship species. Cross-leaved heath (Erica tetralix) and purple moor- grass (Molinia caerulea) were chosen as dominants for the wet heathland; heather (Calluna vulgaris) as a recruitment species; and both bog bush cricket (Metrioptera brachyptera) and curlew (Numenius arquata) as flagship species. The work on estuarine waterbirds was applied to oystercatcher (Haematopus ostralegus) and redshank (Tringa totanus).
The research showed that: 1. Hampshire may experience climatic conditions well beyond those currently seen in Britain and Ireland by the 2050s. Under the 2050s High scenario, Hampshire experiences climatic conditions more akin to the climate of continental Europe. However, wet heath and beech hanger species are predicted to show relatively little effect from climate change according to the MONARCH models.
2. Under future climate change scenarios, grassland shows a large reduction in extent across southern England, including Hampshire. A conclusion is that a wider European baseline dataset would be needed to capture the type of climate potentially to be experienced in this part of the country.
3. The dispersal model shows a limited extension in range for F. sylvatica and F. excelsior, due to a combination of the relatively short timescale considered and with respect to the long time to reproductive maturity of these species.
4. There is a potential for suitable climate for M. perennis to extend throughout much of the eastern part of Hampshire by the 2020s and the 2050s. However, water stress, canopy development and grazing could not be assessed here but they could limit dispersal.
5. The models predict that A. flavicollis will spread such that a large part of the county could be colonised by the 2050s, although dispersal could be affected by factors such as competition with other rodents. A. flavicollis is expected to become more widespread within beech hangers. Some loss of suitable climate space for C. coccothraustes is predicted.
6. E. tetralix has the potential to disperse from sites in the New Forest and the east of the county until under the 2050s scenario all suitable space has been filled. M. caerulea and C. vulgaris have more limited potential to extend their ranges due to a combination of limited dispersal capability and/or habitat availability. The predicted distribution of N. arquata might be more restricted in Hampshire under both the 2020s and 2050s scenarios than it is today although it is already rare.
7. The modelling has only examined the impacts of change with respect to average climatic conditions. For some species, the combination of wetter winters and drought in summer may lead to stress, while for long-lived species, extreme events such as droughts or storms may have significant effects on individuals already stressed by changes under these climatic changes.
MONARCH 2 Report – Chapter 6 103 ______
8. Based on climate change and mean sea-level rise predictions, alterations to the morphology of the coastal estuaries should not significantly affect bird numbers on intertidal flats. Modelling suggests that H. ostralegus will show no real change in numbers relative to those seen over the last 30 years on the Hampshire estuaries, and likewise T. totanus if total numbers wintering in the UK remain low. An increase in total numbers of T. totanus in the UK would increase Hampshire estuary populations by about 50% and hence increase competition.
6.1 Introduction
This chapter explores the results of all the modelling described in chapters 2-5 for the Hampshire case study area. The character of the study area is described in terms of the habitats and species selected for further study and with relation to the bioclimatic classification using the UKCIP98 and UKCIP02 datasets. The sensitivity of the area to the 2050s Low and High scenarios is explored in terms of the effect on the bioclimatic pattern, the species distributions and the land cover and agricultural land use. The effect of these factors on the dispersal capability of selected species within beech hanger and wet heath habitats is also discussed. An assessment is made of the effects of climate change on an area of the New Forest coastline for wading bird species.
6.2 Bioclimatic classification
Table 6.1 shows the bioclimatic classification of the county under the Baseline98 classification defined in MONARCH 2 using the UKCIP98 data and the percentage of grid squares assigned to each class under the UKCIP02 (Baseline02) data and 2050s Low and High emission scenarios. It is clear that the reclassification of the Baseline02 dataset into the Baseline98 classification has caused a shift in the pattern of bioclimatic classes across the county.
Table 6.1: Percentage of grid squares within each of the classes of the bioclimatic classification for the Baseline98 and Baseline02 datasets and for the 5km 2050s High and Low emission scenarios. Class 3 4 7 8 10 12 14 17 26 Baseline98 6.9 93.1 Baseline02 13 44.3 41.7 0.5 2050 Low 51.6 30.2 6.8 8.3 3.1 2050 High 68.3 31.8
6.2.1 Baseline classification
The distribution of the classes and the relationship between the Baseline98 and Baseline02 classifications is shown in Figure 6.1. In the Baseline98 classification, climatic variation across Hampshire was defined by just two classes, 17 and 7. At the 5km Baseline02 scale, surprisingly neither of these classes is represented. Instead there is a shift to classes 8, 10, 12 and 14. The four classes that were represented at the 5km scale were not present in the original 10km bioclimatic classification for this area. Examination of the pattern of the classes shows that to some degree the Baseline02 data is picking up local spatial variation such as influence from the coast. The shift from class 17 to class 8 around the coastal area represents definition of slightly warmer, wetter and windier conditions. From Table A2.6, it can be seen that the classes predicted with the Baseline02 data are similar in statistical terms: 17 is close to 10; and 7 is close to 8.
104 MONARCH 2 Report – Chapter 6 ______
Figure 6.1: Pattern of baseline 10km and 5km bioclimatic classes across Hampshire. Symbols within the 5km grid squares relate to the Baseline02 classification and symbols at the intersection of grid squares relate to the Baseline98 classification.
However, looking further at the difference between the UKCIP98 and UKCIP02 datasets (Figure 6.2 and Table 6.2) it is clear how greatly the climate of the two baselines differs. For PET in particular, (one of the clearest determinants of the 4th factor of the PCA) the lowest Baseline98 values are greater than the highest Baseline02 values across the study region. Similarly for Growing Degree Days the Baseline98 values show much higher baseline conditions than under the Baseline02 dataset. These differences may in part be related to the manner of deriving the two datasets. In particular, the 5km UKICP02 dataset incorporates a variable that accounts for the proximity of a grid cell to the sea, which in turn is likely to reduce the temperature of coastal cells and to more accurately define the coastal areas. Unlike the Scottish and Welsh case study areas, however, the Baseline02 data do still fit within the climatic limits of the Baseline98 classification.
Table 6.2: Comparison of mean values of key variables averaged over the Hampshire case study area from each of the data sets. The variables selected represent those related to the first four components of the PCA in the derivation of the bioclimatic classification Spring Growing Degree January wind speed July Potential precipitation total Days >5°C (m/s) Evapotranspiration (mm) (mm/day) Baseline98 112.71 1816.12 4.88 3.87 Baseline02 113.29 1808.12 4.84 3.59 2050 Low 99.87 2327.09 4.96 3.91 2050 High 91.97 2661.57 5.04 4.28
MONARCH 2 Report – Chapter 6 105 ______
Figure 6.2: Pattern of climate at 10km and 5km resolution for baseline conditions across Hampshire. Symbols within the 5km grid squares relate to the Baseline02 classification and symbols at the intersection of grid squares relate to the Baseline98 classification.
106 MONARCH 2 Report – Chapter 6 ______
6.2.2 Climate change scenarios – 2050s Low and High
Figures 6.3 and 6.4 show the pattern of bioclimatic classes for the future climate change scenarios. The pattern indicates a distancing of the climate of the area from the Baseline98 conditions (Tables 6.1 and 6.2). For the 2050s Low emissions scenario, the squared Mahalanobis distance (Chapter 2.5.1) of a number (18 out of 192) of the coastal squares indicates that the climate of the squares is beyond the limit of the Baseline98 classification. The cut-off level used to indicate squares with climate conditions outside of the Baseline classification is relatively generous (>1347) and it represents the greatest distance between the centres of any two classes, plus the distance of the furthest square from its class in the Baseline98 classification. The remaining squares change class but appear to retain a climate that is represented within the Baseline98 classification. The revised pattern of classes shows a shift inland of the coastal class 8 particularly along the eastern edge of the New Forest. Many of class 12 squares change to class 4, which are statistically very close, with class 4 being slightly warmer but slightly wetter in winter and having a longer growing season. Class 12 squares also change to class 14 in more sheltered areas (especially to the north of the South Downs), representing isolated squares with a much warmer and drier conditions. Most of the squares to the north west of the case study area and across the New Forest, which were classified into class 10 under the Baseline02 climate conditions, also change to class 4 under the 2050s Low emission scenarios. This represents a shift to slightly warmer conditions but a longer growing season and much wetter winter conditions.
Figure 6.3: Pattern of bioclimatic classes under the UKCIP02 2050s Low scenario. Symbols within the 5km grid squares relate to the Baseline02 classification and symbols at the intersection of grid squares relate to the Baseline98 classification.
MONARCH 2 Report – Chapter 6 107 ______
Figure 6.4: Pattern of bioclimatic classes under the UKCIP02 2050s high scenario.
Under the 2050s High emission scenario all of the squares appear to have climatic conditions beyond what is experienced in the Baseline98 classification more akin to the climate of continental Europe.
6.3 Land cover changes
Figure 6.5 shows the LCM 2000 percent coverage, the 5% cut-off presence and absence and the model predicted presence and absences for neutral grassland in Hampshire. It is clear the 5% cut-off threshold does not affect the pattern of the land cover across the area. The model predictions are relatively good for the coastal areas of the county but poorer for the northern and south-western areas. The Kappa statistic for the baseline is 0.314, which indicates poor agreement between the validation and the training dataset. By comparison Figure 6.6 shows the same information for land cover class 19, acid grassland. Here the kappa statistic 0.598 provides a moderate match between modelled and predicted datasets.
108 MONARCH 2 Report – Chapter 6 ______
Figure 6.5: LCM 2000 percent coverage, the 5% cut-off presence/absence and the model predicted presence/absence for neutral grassland in Hampshire.
MONARCH 2 Report – Chapter 6 109 ______
Figure 6.6: LCM 2000 percent coverage, the 5% cut-off presence/absence and the model predicted presence/absence for acid grassland in Hampshire.
Under the future climate change scenarios the grassland classes show a large reduction in their presence across southern England, including Hampshire. Figure 6.7 shows the pattern of change for neutral grassland. The difference between the 2020s and the 2050s pattern for Low and High scenarios is not great, though the 2020s Low shows slightly greater presence of these land cover types than the 2050s High scenarios. The loss of the grassland classes may in part be due to the limited climatic range of the baseline on which the model was derived. A wider (European) baseline dataset is needed in order to capture the influx of new climate and their associated land covers into this part of the country.
110 MONARCH 2 Report – Chapter 6 ______
Figure 6.7: Neutral grassland under future climate scenarios.
6.4 Beech hangers
Evidence from old maps suggest that woodland along the chalk escarpment, from Petersfield to near Selborne, was very limited until well into the 19th century, but some such as Selborne Hanger were well established in the 18th century (Brewis et al., 1996). Some may have been planted for stock or game husbandry. The hangers contain a diversity of vascular plants and bryophytes and are notable for the number of colonies of long-leaved helleborine (Cephalanthera longifolia), more than in any
MONARCH 2 Report – Chapter 6 111 ______other area of Britain of similar size (Brewis et al., 1996). The East Hampshire hangers are a SAC, partly because of the mix of woodland Asperulo-Fagetum beech and Tilio-Acerion forests they contain. They are vulnerable to nutrient runoff from surrounding agricultural land and poor management.
6.4.1 Species modelling
This case study area was chosen because of its high sensitivity to climate change and, therefore, it would be expected that species’ response would be marked. None of the species were directly chosen for their sensitivity, but from previous modelling with the UKCIP98 scenarios, Fagus sylvatica has shown some sensitivity in southern England (Berry, et al., 2002).
The SPECIES model trained well at the European scale, with the independent AUC statistic greater than 0.9 for all plant species, indicating very good discrimination ability (Table 3.3). The maximum Kappa statistic is slightly lower for most species, with two species showing values greater than 0.85, indicating excellent agreement between observed distributions and simulated climate space, and the other three values (for F. sylvatica, A. flavicollis and C. coccothraustes) between 0.7 and 0.85 indicating very good agreement. In the case of the former, there is not a particularly close match on the eastern range margin (Figure 6.8). For the latter, it is probably due to its slightly patchy distribution (Figure 6.9) and the climate space simulation misses out the western half its distribution in England at the European scale, which will affect the training of the downscaled SPECIES model.
Figure 6.8: The European observed distribution (a) and simulated climate space (b) for Fagus sylvatica.
(a) (b)
Figure 6.9: The European observed distribution (a) and simulated climate space (b) for Coccothraustes coccothraustes.
(a) (b)
112 MONARCH 2 Report – Chapter 6 ______
The three plant species and A. flavicollis show some loss of suitable climate space, in East Anglia under the 2050s High scenario and for F. sylvatica and A. flavicollis this is more widespread, extending into Central England. C. coccothraustes showed a slightly different pattern, with more climate space being lost in southern England (Figure 6.10), but this may be a function of the different version of SPECIES being used for birds (Chapter 3).
Figure 6.10: SPECIES model outputs for Coccothraustes coccothraustes: (a) observed 10km distribution, (b) climate suitability surface from the European-trained network, and suitable climate space under the UKCIP02 scenarios (c) 2020s Low (d) 2020s High (e) 2050s Low (f) 2050s High.
(a) (b)
(c) (d)
(e) (f)
MONARCH 2 Report – Chapter 6 113 ______
The downscaled model training statistics are lower (Table 3.5), for reasons discussed in Chapter 3, although F. sylvatica now does better than some other species, due to its close association with broad- leaved woodland. The downscaled SPECIES model uses the presence/absence data provided by the Biological Records Centre, which does not distinguish between the native and planted status of F. sylvatica, thus it shows almost the whole of the country as representing suitable space and only some of the higher parts of the Grampians and North West Highlands of Scotland as unsuitable (Figure 6.11). This picture is maintained throughout the climate change scenarios. This shows that F. sylvatica is able to grow in a wide variety of circumstances and it is only altitudinal related variables that constrain it. This is shown more clearly when the outputs are shown on the 1 km grid where suitability in highland areas is confined to the lower elevations (Figure 6.12).
Figure 6.11: Model outputs for Fagus sylvatica: (a) climate suitability surface from the European- trained network, (b) observed 10km distribution, (c) climate and land cover suitability surface from the downscaled model, and (d) presence/absence surface from the downscaled model.
(a) (b)
(c) (d)
114 MONARCH 2 Report – Chapter 6 ______
Figure 6.12: 1km suitability surface for Fagus sylvatica under baseline climate.
Unsuitable 95% AUC threshold 90% AUC threshold Optimum AUC threshold
The overall pattern for A. flavicollis is better with the inclusion of land cover and the low kappa is because much of southern England is simulated as being suitable, whereas it has a much more restricted fragmented distribution and only for this species is some of the loss of climate space picked up by the downscaled model (Figure 6.13). A better match would be achieved with a higher cut off. In contrast, for F. sylvatica, but not Fraxinus excelsior, many coastal squares have a slightly lower suitability, so would be included if a lower threshold was used. This could be important as it particularly affects the south coast including Hampshire (Figures 6.11d and 6.12).
F. excelsior is shown to be absent from the higher land in northern Scotland and the downscaled SPECIES model suggests that a slightly wider area, based on climate and land cover is unsuitable (Figure 6.14). At the 1 km grid scale, as for F. sylvatica, it is the higher parts of the country and eastern England, which are shown as unsuitable. The same is true for Mercurialis perennis, where climate limits the simulated suitable habitat for the species in the Central and North West Highlands of Scotland, compared with actual distribution (Figure 6.15). This could be a function of average climate data being used for grid squares and M. perennis being found in valleys at lower altitudes within the Highlands. The downscaled network for C. coccothraustes, begins to capture the amount of fragmentation present in Britain and a good Best Attainable Match is achieved (Figure 6.15) and, under the climate scenarios and land cover, it would lose a large amount of its suitable area (Figure 6.16).
MONARCH 2 Report – Chapter 6 115 ______
Figure 6.13: Model outputs for Apodemus flavicollis: (a) climate suitability surface from the European-trained network, (b) observed 10km distribution, (c) climate and land cover suitability surface from the downscaled model, and (d) presence/absence surface from the downscaled model.
(a) (b)
(c) (d)
116 MONARCH 2 Report – Chapter 6 ______
Figure 6.14: Model outputs for Fraxinus excelsior: (a) climate suitability surface from the European- trained network, (b) observed 10km distribution, (c) climate and land cover suitability surface from the downscaled model, and (d) presence/absence surface from the downscaled model.
(a) (b)
(c) (d)
MONARCH 2 Report – Chapter 6 117 ______
Figure 6.15: Model outputs for Mercurialis perennis: (a) climate suitability surface from the European-trained network, (b) observed 10km distribution, (c) climate and land cover suitability surface from the downscaled model, and (d) presence/absence surface from the downscaled model.
(a) (b)
(c) (d)
118 MONARCH 2 Report – Chapter 6 ______
Figure 6.16: 1 km downscaled modelling outputs for Coccothraustes coccothraustes: (a) current predicted distribution, (b) 2020s Low, (c) 2020s High, (d) 2050s Low, and (e) 2050s High predicted distributions, on the basis of climate space and land cover.
(a)
(b) (c)
(e) (f)
The dispersal model shows a limited range expansion for F. sylvatica and F. excelsior, due to a combination of the time scale being considered and the long time to reproductive maturity of the species (Figures 6.17 and 6.18). Thus they have the potential to disperse into other suitable areas in Hampshire, but in reality are probably constrained from so doing by the presence of other land uses. M. perennis, however, has the potential to spread throughout much of the eastern part of Hampshire under the 2020s scenarios and that under the 2050s scenarios large parts of the New Forest could be
MONARCH 2 Report – Chapter 6 119 ______colonised (Figure 6.19). A number of factors, however, could affect the realisation of this, including water stress (Grimme, 1984), canopy development (Graves, 1990; Klotzli and Walther, 2000) and grazing (Crampton et al., 1998; Kirby and Thomas, 2000).
The dispersal model shows that A. flavicollis could spread from its main centre of distribution, as well as from the two populations in the north of the county (Figure 6.20). Under the 2020s scenarios it could start to spread into the New Forest and then consolidate its position by the 2050s. A large part of the county could be colonised by this time period, due to the dispersal capabilities of this species and the availability of suitable habitat, although if this scenario was realised many of the populations in the east of the county would be small and scattered, thus questioning their long-term viability. It is shown as having the potential to disperse into Southampton and A. flavicollis can be found in urban areas, but as a study in Prague showed, it does not have a high ability to penetrate urban areas (Frynta et al., 1994). The realisation of this dispersal would be affected by factors such as competition with other rodents, including, wood mouse (A. sylvaticus), bank vole (Clethrionomys glareolus) and striped field mouse (Apodemus agrarius).
Figure 6.17: Fagus sylvatica dispersal model outputs.
2020s Low scenario 2020s High scenario
2050s Low scenario 2050s High scenario
120 MONARCH 2 Report – Chapter 6 ______
Figure 6.18: Fraxinus excelsior dispersal model outputs.
2020s Low 2020s High
2050s Low 2050s High
For the beech hangers, the projected changes in species location and extent are only slight. This reflects the relatively short timescale that is under consideration in relation to the long-lived tree species. This is not to say that other species may not be more severely affected by climate change. Moreover, changes in abundance and phenology may have significant effects on community composition, structure and dynamics. The modelling has only examined the impacts of changes to average climatic conditions. For long-lived species extreme events such as droughts or storms may have large effects on individuals already stressed by changes in average conditions.
MONARCH 2 Report – Chapter 6 121 ______
Figure 6.19: Mercurialis perennis dispersal model outputs.
2020s Low scenario 2020s High scenario
2050s Low scenario 2050s High scenario
6.4.2 Implications for the composition of species communities
6.4.2.1 The dominant plant species and climate change
The dispersal model suggests that the beech hanger ecosystems will remain within the climatic tolerances of the dominant species that have been modelled (Table 6.3). On this basis very little change in community composition is expected. However MONARCH 2 does not take into account extreme events that may become more frequent under future climate scenarios (Hulme et al., 2002). Storms can fell healthy trees as well as trees in decline creating gaps, similar to those produced by the death of mature trees, which are generally filled by F. sylvatica crown enlargement, but storms can also produce larger gaps, which will create opportunities for regeneration. Where storms are rare each storm will create larger gaps, which may be filled by even aged trees, but as storms become more frequent the gaps created will become smaller and the patchiness of the woodland may increase (Pontailler et al., 1997).
122 MONARCH 2 Report – Chapter 6 ______
Figure 6.20: Apodemus flavicollis dispersal model outputs.
2020s Low scenario 2020s High scenario
2050s Low scenario 2050s High scenario
Table 6.3: A summary of the dispersal model predictions for the modelled species within the beech hangers. Species Category Predicted space change Fagus sylvatica beech Dominant species No change Fraxinus excelsior ash Dominant species No change Mercurialis perennis dog’s mercury Dominant species No change Coccothraustes coccothraustes hawfinch Rare species Decrease Apodemus flavicollis yellow necked mouse Flagship species Increase
In Wealden Edge hanger, areas on the steeper thinner soils usually occupied by NVC 12 Fagus sylvatica-Mercurialis perennis woodland have already been found to be particularly prone to storm damage. Past storm damage at Selborne Hill has produced areas of open canopy with dense regenerating scrub below (Jonathan Cox Associates, 1997).
Beech hanger communities often grow on thin free draining soils and may, therefore, be impacted by a rise in the frequency of drought, due to an increase in very dry summers (Hulme et al., 2002). In southern England, F. sylvatica is already showing effects of droughts, which include reduced gain in stem diameter, a high root mortality partly compensated for by greater fine root growth (Leuschner et
MONARCH 2 Report – Chapter 6 123 ______al., 2001), decreased crown density and reduced apical growth in post drought years (Stribley and Ashmore, 2002). Small leaves, dieback and mast years are also known to occur during or directly after a drought year (Cannell and Sparks, 1999). However, late spring frosts, may kill flowers on the trees and extend the period between seed crops in F. sylvatica (Matthews, 1955). Both drought and late frosts may affect recruitment and seed predator dynamics. At Noar Hill NNR, the collective impact of the 1976 drought and the gales of the late 1980’s have caused the loss of high canopy mature beech forest, with less than 10% of the canopy now being representative of large areas of High Wood hanger (Jonathan Cox Associates, 1997). This indicates the importance of extreme events for the beech hangers.
6.4.2.2 Coccothraustes coccothraustes (Hawfinch)
The dispersal model suggests that there may be some loss of suitable space from the beech hangers, for C. coccothraustes (Table 6.3). As the loss of a species will cause an obvious change in the composition of species communities and may have further knock-on indirect effects on community composition and structure, it is important to examine any possible impacts that may occur as a result of this species leaving (Chapter 4).
The interaction web (Figure 6.21), which illustrates the major biotic interactions associated with this species, indicates that it has a very broad diet. Generally, kernels and seeds are eaten in the autumn and winter, whereas in spring the main food is buds and shoots, and in the summer insects are eaten. The chicks are fed on insects, and successful breeding is dependent on a high abundance of defoliating larvae especially caterpillars (Mountford, 1956). For all these diet items C. coccothraustes faces competition from other species within the beech hanger ecosystem (Figure 6.21), suggesting that it is not the sole representative of a functional group. In addition, the predators of C. coccothraustes, are quite generalist and few.
On this basis, it is suggested that the departure of C. coccothraustes, from the ecosystem, is only likely to have a negligible impact on community species composition within the beech hangers (Figure 6.22). This conclusion is further supported by the fact that C. coccothraustes, is not particularly site constant, today. County histories and annual ornithological records indicate that there have been substantial fluctuations of breeding and wintering populations at localities over time, which cannot be attributed to any specific factor (Mountford, 1956). Furthermore, winter flocks of C. coccothraustes are know to wander widely in search of food, often totally decimating one food supply before moving on to the next (Mountford, 1956).
6.4.2.3 Apodemus flavicollis (Yellow necked mouse)
Indications from the dispersal model suggest that A. flavicollis will become more widespread within the beech hangers than it is today (Table 6.3), and may thus colonise new woodlands. The interaction web (Figure 6.23) suggests that one of the significant functions of A. flavicollis is the predation of seeds. In this role, wood mouse (Apodemus sylvaticus), bank vole (Clethrionomys glareolus) and grey squirrel (Sciurus carolinensis) join it, as well as many birds (e.g. C. coccothraustes). Seeds make up approximately 80% of the diet of A. flavicollis with animal food and green plants each accounting for another 10% or so of the diet. Fungi that are eaten in the summer and autumn make up a maximum of 5% of the diet (Hansson, 1985). The diet of A. sylvaticus has also been studied and the proportions of the different food types eaten are very similar (approximately 15% animals, 70% seeds, 5-10% forbs) (Hansson, 1985) whilst C. gareolus eats less seeds and a higher proportion of plant material. There is also overlap between the species in feeding sites, with all species feeding under the ground, on the ground and in trees (although A. flavicollis and C. glareolus are thought to forage here to a greater extent than A. sylvaticus). Furthermore A. flavicollis has a more restricted diet than A. sylvaticus, feeding on only a sub-set of the food sources that A. sylvaticus feeds on.
124 MONARCH 2 Report – Chapter 6 ______
Figure 6.21: Figure 6.21: An interaction web for C. coccothraustes, showing the main interactions within the beech hangers. Blue lines represent competitive interactions; red lines predator-prey interactions; brown lines host-parasite interactions and black lines all other interactions. The thickness of the arrows give an indication of interaction strength.
Other interactions Protocalliphora Accipiter nisus competitors azurea Columba palumbus (Aves) (esp on fledglings) (Diptera) Sciurus carolinensis Apodemus sylvaticus (Aves) use old (Mammalia) (Mammalia) hawfinch nests Dasypsyllus gallinulae Garrulus glandarius (Siphonaptera) (Aves) (on nests) Clethrionomys Apodemus flavicollis Nest materials glareolus (Mammalia) Sciurus carolinensis Lonicera Menacanthus setosus (Mammalia) (Mammalia) (on nests) Fringilla periclymenum (Mallophaga) montifringilla Fringilla coelebs (Aves) (Aves) Clematis vitalba Brueelia juno (Mallophaga) Parus caeruleas Pyrrhula pyrrhula (Aves) (Aves) roots predators Columba palumbus Sitta europaea (Aves) (Aves) Betula sp. Parus major Parus palustris (Aves) (Aves) Grasses
Coccothraustes
Tortix viridana coccothraustes Taxus baccata (Lepidoptera) Carpinus betulus Earthworms Aphids Ilex aquifolium Phyllobius spp. (Coleoptera) Prunus spinosa Biston betularia (Lepidoptera) Acer pseudoplatanus Operophtera brumata (Lepidoptera) Crataegus mongyna Elasmostethus Viburnum Sambucus nigra Pupae interstinctus opulus (Heteroptera) Corylus Ulmus glabra Viburnum Grasses Rumex sp. avellana Fraxinus excelsior Larvae Spiders lantana Euonymus Troilus luridus Hedera Rosa europaeus Prunus avium Fagus sylvatica Acer campestre Moths Beetles (Heteroptera) helix canina agg.
MONARCH 2 Report – Chapter 6 125 ______
Figure 6.22: The pathway (marked in orange) through the Leaver model predicted for C. coccothraustes.
Community collapse t New Community Dominant Species fec ef jor on • Strong links with other species Ma cti fun • Driver of belowground processes on Reassembly
Yes
Keystone species? e, ng ha n No c io al ss Sub-dominant ur e Leaver at cc N su Species g. e. No Functional Type Redundancy? Yes
ct Rare Species ffe le e • weak interaction strength igib egl ion Existing Community N nct • often dependent on biotic interactions fu on
A. sylvaticus and C. glareolus share similar predators and parasites, suggesting that A. flavicollis cannot be viewed as a novel functional group within the beech hanger communities that it is predicted to colonise under the modelled scenarios. Using these conclusions to navigate through our Arriver conceptual model suggests that this species is unlikely to have any great impact on the species composition of beech hanger communities, as shown in Figure 6.24.
A. flavicollis is already present within some beech hanger woodlands, for example Ashford Hangers NNR, and given the good dispersal capabilities (as shown by the parameters for the dispersal model), it may be interesting to consider why the yellow-necked mouse is currently absent from some of the hangers. One explanation may be that the coverage of the fine scale distribution maps for the area in question is not complete and that A. flavicollis is already present in the hangers. Alternatively, the woodlands from which it has not been recorded, may not contain suitable habitat, and may thus not be colonised under climate change either. Unlike A sylvaticus, which has a wide habitat breadth, A. flavicollis, has been shown to favour woodlands with a high level of canopy cover, a greater amount of fallen timber and woodlands which contain good seed/ fruit producing tree species (Marsh and Harris, 2000).
126 MONARCH 2 Report – Chapter 6 ______
Figure 6.23: The interaction web for A. flavicollis, displaying the main interactions. Grey shading is used to indicate species that do not currently occur within the beech hangers of the test study area. Blue lines represent competitive interactions; red lines predator-prey interactions; brown lines host- parasite interactions and black lines all other interactions. The thickness of the arrows gives an indication of interaction strength.
Mercuralis perennis Geum urbanum Food items Hard tree seed/fruit Buds Centipedes Insects seeds fruit
Fagus sylvatica Fagus sylvatica Snails seeds Fraxinus excelsior flowers seeds Fungi Corylus avellana Forbs Worms seeds
Predators
Vulpes vulpes (Mammalia) Ctenophthalamus nobilis Apodemus (Siphonaptera) Neotrombiculla autumnalis Mustela nivalis (Acari) (Mammalia) flavicollis Ixodes ricinus (Acarina) Mustela erminea (Mammalia) Ixodes trianguliceps Laelaps spp. (Acarina) Meles meles (Acari) (Mammalia) Parasites Martes martes Apodemus sylvaticus (Mammalia) (Mammalia) Clethrionomys glareolus Buteo buteo (Mammalia) (Aves) Sciurus carolinensis Falco tinnunculus (Mammalia) Fraxinus excelsior (Aves) Plant material saplings for nests Muscardinus avellanarius and other saplings Strix aluco Tyto alba (Mammalia) (Aves) (Aves) Competitors Other interactions
However, there are further considerations with regards to the composition of species communities, as competition within the ‘granivorous’ functional group may lead to changes in relative abundance of the species, and the addition of a further seed-eating species may lead to recruitment limitation in plants within the ecosystem.
Leaf bud burst in F. excelsior is dependent on cumulative spring temperatures, and furthermore flower bud production is dependent on the time of leaf bud burst in the preceding year. The earlier the leaf bud burst was in the preceding year the greater the number of trees fruiting and the larger the crop per tree (Tapper, 1996). Individual trees vary in the temperature threshold required for leafing and therefore they also vary in the frequency and quantity of fruit production. Ash, unlike beech, is not limited to fruiting in alternate years by a limit in resources; fruit set in one year does not affect the probability of fruit set in the following year. Fruits are relatively ‘cheap’ to produce and in a long- term Swedish study the majority of trees produced fruits in two successive years, with a proportion producing good crops in both years (Tapper, 1996). It is therefore conceivable that climate change will lead to both an increased seed crop and seed crops being produced with greater frequency.
MONARCH 2 Report – Chapter 6 127 ______
Figure 6.24: The pathway through the Arriver model predicted to be followed by A. flavicollis (shown in orange).
Large impact t n on ecosystem Community t e n e m Dominant Species collapse h function m s it li Highly ru b c ta modified e s R E community Reassembly Expansion New Functional Type Sub-dominant Species Keystone Coloniser species e, ng Arriver ha n Existing l c io ra ss Functional tu ce a uc Type N . s .g e
t fec ef Rare Species ble igi n gl tio Ne nc fu on
Existing community Extinct No impact on ecosystem function
Fruit production in F. sylvatica is also influenced by climatic factors, the most important of which appear to be summer (July) temperature and sunshine in the year preceding the mast year. High temperature, high amounts of sunshine and low precipitation are required for the initiation of flower bud development. However in the fruiting year, late spring frosts can be highly detrimental, killing flowers, and leading to a failure in seed crop (Matthews, 1955) and an extended period between crops. In this instance, both higher summer temperatures and an enhanced frequency of extreme events could be of relevance in determining seed production in F. sylvatica.
Both F. sylvatica and F. excelsior have a monophagous Lepidopteran seed predator, as well as a wide range of more generalist invertebrate and vertebrate predators, such as A. flavicollis. Masting in F. excelsior, currently regulates its seed predator, Pseudargrotoza conwagana, and satiation in mast years has been demonstrated in Sweden. However, if fruit were produced on an annual basis in the future (see discussion above), this could lead to an expansion of this seed predator. In contrast to F. excelsior, under future climatic conditions, biennial F. sylvatica masting will still be maintained, and therefore its monophagous seed predator, Cydia fagiglandana, will not have the opportunity to undergo a similar population expansion. The excess of seeds in mast years will therefore be available to other seed predators, such as rodents, and for recruitment (Nilsson and Waestjung, 1987).
Seed predators also will have greatly reduced impacts on regeneration if masting by trees leads to their satiation. Under climate change this is likely to be the case in F. sylvatica, but the effects of this may be reduced if other tree species in the community play a role in maintaining strong seed predator populations by producing fruit annually, especially if the fruit is less preferred than that of F.
128 MONARCH 2 Report – Chapter 6 ______sylvatica. Rodents may cache surplus seeds and although this could lead to seedling dispersal, seed stores are often too deep underground, to allow successful germination and recruitment.
S. carolinensis populations are currently limited by both food supply and cold winter weather (Gurnell, 1996) suggesting the higher populations may occur with climate change. As well as predating seeds S. carolinensis may also cause severe damage to trees, through bark stripping. A recent study in the beech woods of the Chilterns indicated that 75% of young trees had squirrel damage and this can seriously affect the further growth and form of affected trees (Rayden and Savill, 2004). Furthermore, S. carolinensis preferentially strip the bark of F. sylvatica and Acer pseudoplatanus (sycamore) trees and tend to avoid F. excelsior, Betula pendula (silver birch) and Prunus trees, and may, therefore, influence the future species composition of the beech hangers.
Seed predators can also have marked effects on the composition of tree species, through selective foraging. For example, wood mice and bank voles select Ulmus glabra (wych elm) seeds in preference to the seeds of F. excelsior, even where Fraxinus seeds occur in substantially higher densities. Thus rodents can cause local extinctions of tree species, such as U. glabra (Hulme and Hunt, 1999).
However, climate related changes in the disturbance regime and the ground flora, as well as impacts on seed germination, will also all affect community composition and structure. For example, mild winters have been demonstrated to have a negative impact on the germination success of beech nuts (Pontailler et al., 1997).
6.5 Wet heathland
Annex 1 of the European Habitats Directive 1992 lists both Northern Atlantic wet heaths with Erica tetralix and Temperate Atlantic wet heaths with Erica ciliaris and Erica tetralix, which are priority habitats and lowland heathland is UKBAP Priority habitat. In Europe, they are threatened by direct loss through conversion to other land uses, as well as indirect changes from atmospheric deposition. In the United Kingdom these are also concerns, along with encroachment by scrub and trees. In Hampshire, this lack of management, fragmentation and disturbance from development all pose threats (Hampshire Biodiversity Partnership, 1998).
In MONARCH 1, wet heaths were found to be associated with all the bioclimate classes (Harrison et al., 2001), indicating its widespread occurrence under a range of climatic conditions. Lowland wet heath often occurs on gleyed and peaty soils with impeded drainage, although if it is too wet then mire may develop with Eriophorum spp. and Sphagnum spp. In Hampshire, there are 16 SSSIs with lowland heath, bog and acid grassland covering 4.4% of the county and it is a key area for this habitat as it contains about 13% of the lowland heathland left in Europe and 30% of UK total (Hampshire Biodiversity Partnership, 1998). The wet heathland is found in the New Forest and in the eastern part of the Hampshire Basin (Figure 6.25).
MONARCH 2 Report – Chapter 6 129 ______
Figure 6.25: Map of the distribution of wet heathland in Hampshire.
Three plant species were chosen to represent the wet heaths: Erica tetralix (cross-leaved heath), and Molinia caerulea (purple moor grass), are dominants in M16, while Calluna vulgaris (heather) is a constant, but potential recruitment species if the heaths dry. The insect, Metrioptera brachyptera (bog bush cricket) is a flagship species and the bird, Numenius arquata (curlew) is a typical species.
6.5.1 Species modelling
Four of the wet heath species show good agreement at the European scale between the observed distribution and simulated climate space, using both the AUC and the Kappa statistics (Table 3.3), although for the bird, Numenius arquata, the Kappa statistic is 0.68 and this may be a function of its more fragmented European distribution, especially in the northern part of its range (Figure 6.26). All the plants show no alteration in suitable climate space under the climate change scenarios, but M. brachyptera loses a considerable amount of suitable climate space in Central England and East Anglia.
Figure 6.26: The European observed distribution (a) and simulated climate space (b) for Numenius arquata.
(a) (b)
130 MONARCH 2 Report – Chapter 6 ______
As with the beech hanger species, the downscaled model statistics were lower, especially for the Kappa statistic (Table 3.5), as the range margins within England are not particularly well captured, but it does lead to a more restricted suitability, compared with the climate only suitability surface (Figure 6.27). This could partly be a function of other factors, such as management or habitat destruction, affecting the distribution of the associated heathland land cover type. The loss of suitable climate space for M. brachyptera is only partly reflected in the downscaled model outputs, owing to the patchy availability of suitable space. For N. arquata, there is a high Best Attainable Match, and Hampshire provides something of a local stronghold for the curlew in the breeding season, with the species largely absent from neighbouring regions. The 1 km downscaled network identified both these local occurrences in Hampshire and the regions devoid of the species around it (Figure 6.28).
Figure 6.27: Model outputs for Molinia caerulea: (a) climate suitability surface from the European- trained network, (b) observed 10km distribution, (c) climate and land cover suitability surface from the downscaled model, and (d) presence/absence surface from the downscaled model.
(a) (b)
(c) (d)
MONARCH 2 Report – Chapter 6 131 ______
Figures 6.28: 1 km downscaled modelling outputs for Numenius arquata: (a) current predicted distribution, (b) Low 2020, (c) High 2020, (d) Low 2050, and (e) High 2050 predicted distributions, on the basis of climate space and land cover.
(a)
(b) (c)
(d) (e)
132 MONARCH 2 Report – Chapter 6 ______
The dispersal model shows that E. tetralix has the potential to disperse from sites in the New Forest and in the east of the county, until under the 2050s scenario all suitable space has been filled (Figure 6.29), but in reality it may be constrained by development and geology. The areas surrounding the populations to the east of Southampton Water are considered unsuitable, and so dispersal does not occur from these locations. Metrioptera brachyptera also could spread out from a similar area and become widespread in the New Forest and move into Dorset, under the 2020s scenarios (Figure 6.30). This continues under the 2050s Low scenario, but under the 2050s High scenario some locations, particularly in the New Forest start to become unsuitable as the climate becomes hotter and drier. Molinia caerulea and Calluna vulgaris have a more limited potential for spread from their existing foci due to a combination of limited dispersal and/or habitat availability (Figures 6.31 and 6.32).
Figure 6.29: Erica tetralix dispersal model outputs.
2020s Low scenario 2020s High scenario
2050s Low scenario 2050s High scenario
MONARCH 2 Report – Chapter 6 133 ______
Figure 6.30: Metrioptera brachyptera dispersal model outputs.
2020s Low scenario 2020s High scenario
2050s Low scenario 2050s High scenario
134 MONARCH 2 Report – Chapter 6 ______
Figure 6.31: Molinia caerulea dispersal model outputs.
2020s Low scenario 2020s High scenario
2050s Low scenario 2050s High scenario
MONARCH 2 Report – Chapter 6 135 ______
Figure 6.32: Calluna vulgaris dispersal model outputs.
2020s Low scenario 2020s High scenario
2050s Low scenario 2050s High scenario
6.5.2 Implications for the composition of species communities
Wet heaths are dependent on water for their persistence. Any increase in the period of high water levels would allow species typical of valley bogs to out-compete typical wet heath species, whereas an increase in the length of dry periods could allow dry heath or acid grassland species to invade. The constant species of wet heath are Calluna vulgaris, Erica tetralix, Molinia caurulea, Sphagnum compactum and Sphagnum tenellum and of these, the first three were modelled within this study. Although these all presently occur within wet heaths, C. vulgaris is generally less abundant than the other two species.
As shown in Table 6.4, the dispersal model indicates that there will generally be no loss of the modelled dominant plant species from the main wet heath areas in the New Forest and in the north- east of the county.
136 MONARCH 2 Report – Chapter 6 ______
Table 6.4: A summary of the outputs of the dispersal model for the wet heath areas Species Category Predicted space Calluna vulgaris Ling Dominant species No change Erica tetralix Cross-leaved heath Dominant species No change Molinia caurulea Purple moor grass Dominant species No change Metrioptera brachyptera Bog bush-cricket Flagship species Increase in 2020s Decrease in 2050s Numenius arquata Curlew Bird species Decrease
6.5.2.1 The dominant species and climate change
Climate change is predicted to produce changes in the seasonal patterns of rainfall (generally wetter winters and drier summers) as well as an increase in the number of extremely dry summers (Hulme et al., 2002). Increased frequency and severity of summer droughts will have a significant impact on wet heath communities. Changes in the water table, as indirect impacts of climate change such as increased water abstraction, also may influence the occurrence and composition of wet heath communities.
Comprehensive comparisons of the rooting systems and physiology have been undertaken by a number of authors in an attempt to explain differences in dominance between common heathland species (E. tetralix, C. vulgaris and M. caerulea). These are of interest when considering the impacts of water deficit on the composition of the heathland vegetation. E. tetralix was observed to occur at higher abundance in areas where the water table was high, on a wet heath in Bramshill Forest, Hampshire (Rutter, 1955), and other authors have also confirmed that it is commonly found in wet, often waterlogged soils (e.g. Bannister, 1964). This species has a shallow rooting system, with about a third of all roots located in the organic matter at the soil surface, and a maximum root depth of 32- 40cm (Sheikh and Rutter, 1969). This shallow rooting habit means that it evades high concentrations of carbon dioxide and hydrogen sulphide found in wetter soils, to which it is sensitive, but makes it highly susceptible to drought. Furthermore, E. tetralix appears unable to reduce transpiration rates in response to soil drying and this results in water deficits, leaf abscission and death of the plant (Bannister, 1966).
In contrast to E. tetralix, high M. caerulea abundance at Bramshill Forest, Hampshire was observed in areas with a low water table, and large fluctuations in summer water table depth (Rutter, 1955). This species has a deep root system, probably reaching depths of 1m (although 95% of roots are in the top 25cm), and extensive arenchyma means that roots can survive in waterlogged soils (Sheikh, 1970). This deeper rooting system may allow it to cope more effectively than E. tetralix, with low soil moisture levels and a low water table during the summer months, which may occur as a result of climate change.
The abundance of C. vulgaris appears to be less influenced by soil moisture, as Rutter (1955) could find no relationship with water table depth and Bannister (1964) noted that pure stands of C. vulgaris can grow on quite wet soils. However, in really wet areas, C. vulgaris is dependent on M. caerulea tussocks as germination sites (Rutter, 1955). C. vulgaris is also better able to withstand soil drying, than E. tetralix, as it is able to reduce its transpiration rate, although some loss in turgidity is still apparent (Bannister, 1966).
Although drought will determine the distribution and abundance of the dominant heathland species in this habitat, other differences between them, for example in nutrient availability and resistance to burning, may also affect competition and hence abundance of these important species. If M. caerulea becomes more abundant relative to C. vulgaris and E tetralix, this may have a significant impact on
MONARCH 2 Report – Chapter 6 137 ______the habitat, as M caerulea is a tussocky grass which provides a different habitat structure to that of dwarf shrubs.
Finally, as wet heaths dry out due to the predicted increased frequency of extreme droughts there is a greater susceptibility to fire. Hence, there could be a multiple stresses upon wet habitats, leading to potentially extreme community collapse and re-assembly with a different suite of species. The area of wet heath almost sandwiched between dry heath and wetter areas (bogs/marshes) could be reduced dramatically under such a scenario.
6.5.2.2 Metrioptera brachyptera (Bog bush-cricket)
Although some of the animal species on heathlands are present because they are associated with a particular host plant (e.g. the majority of species in the interaction webs), many are present because heathland sites present suitable physical conditions (especially in terms of microclimate) (Webb, 1986). The bog bush-cricket Metrioptera brachyptera is likely to fall into the latter group (Detzel, 1998). The spread of this species within Hampshire as predicted by the dispersal models, may be linked to changes in the availability of suitable micro-climatic conditions. The interaction web for this species suggests that although interactions with wet heath species do occur, these could be replaced by similar interactions within other habitats. In Baden-Wurttemberg, this species is found in all types of heathland, disturbed bogs and mires, unproductive wet meadows and open pine forests with a Vaccinium ground flora. Although this species does require a damp place for egg development (eggs are laid in damp earth, peat, moss, and the stems of higher plants), dense vegetation in warm and dry habitats, probably meets this requirement (Detzel, 1998). This species is often recorded from tall vertically structured and grassy vegetation patches.
Similarly for the wet heath, the interaction web (Figure 6.33) suggests that the predicted arrival and departure of M. brachyptera will have little impact on the communities affected. M. brachyptera is an omnivore, which feeds on a wide selection of foods. It is also eaten by a wide variety of birds and spiders, and may sometimes get caught by sundews. None of these interactions are specialised, and none of the species featured in the web (Figure 6.33) are totally dependent on one another.
6.5.2.3 Numenius arquata (Curlew)
Climate change will not only affect plant and invertebrate species but also bird species such as the Numenius arquata. The dispersal models indicate that the predicted distribution of the curlew will be smaller in Hampshire under both the 2020s and the 2050s scenarios, than it is today. Curlews use wet heath habitats during the breeding season, both as a feeding site and nesting site although grassland also is used. Hence, as wet heaths and the wetter habitats (bogs) dry due to climate change, as previously discussed, then the curlew will lose valuable habitat and potentially this will lead to increased pressure in terms of space, food and niches, on other habitats within the larger region.
However, although curlews use wet heath habitats, they are not common within these habitats. There were a maximum of 100 pairs recorded in the New Forest in the 1990s (Conservation Progress Report, 2004), and no N. arquata were recorded during surveys of Hazeley Heath, DERA Farnborough, and Eelmoor marsh, three important wet heath sites. Therefore, this species should be considered as a rare species within the community and the Leaver model (Figure 6.34), suggests that its departure will have little impact on the community of the wet heath ecosystem.
This conclusion is further strengthened by recent evidence that suggest that N. arquata may have an even smaller impact on wet heaths. Curlews breeding in the New Forest are dependent on the coastal habitats of Southampton Water and the west Solent for feeding early in the season, when food is scarce at the breeding site. Later, when the young are free flying, the coastal sites are important as refuges (Conservation Progress Report, 2004).
138 MONARCH 2 Report – Chapter 6 ______
Figure 6.33: The interaction web for Metrioptera brachyptera, showing the main interactions. Red lines show predator-prey interactions, brown lines host-parasite interactions, and black lines all other interactions. Species in grey boxes are not currently present within the Hampshire wet heaths.
Coleoptera Erica tetralix Lépidoptera ‘prey’ Heteroptera Diptera Nymphal Orthoptera larvae flowers & seeds caterpillars
Leaves and stems Potentilla erecta Calluná vulgaris Narthecium ossifragum Hypochoeris radicata Galium uliginosum of herbs flowers & seeds flowers & seeds flowers & seeds
Metrioptera grasses brachyptera
predators Drosera intermedia Drosera anglica Eggs laid in Molinia tussocks Metrioptera brachyptera Drosera rotundifolia Sylvia undata ( Aves ) Eggs laid in mosses Anthus pratensis Saxicola torquata ( Aves ) ( Aves ) Eggs laid in stems of higher plants Decticus Spiders verrucivorus Other interactions Metrioptera Staphylinidae brachyptera Nematodes (Coleoptera ) ( Orthoptera ) e.g. Gordis species Mites e.g Emberiza schoeniclus Erythraeus phalangii Asilidae ( Aves ) ( Diptera ) Parasites Argiope brunnichi
6.5.2.4 Implications for the composition of species communities
Under climate change, wet heath communities may not occupy the same patch of land as at present, but changes in the relative abundances of the three modelled species may occur. With drying out there may be a general shift of the boundaries of the communities, with valley bog and mire, possibly losing habitat area, although this will obviously depend overwhelmingly on the hydrology of the specific site in question. The locations that currently contain wet heath may in future support a community representative of drier soil conditions. Such changes are unlikely to be visible in the current study, as a consequence of the species that were selected. In the New Forest, wet heath grades into an Ulex minor - Agrostis heath which grades into a Calluna – Ulex minor heath with increasing dryness of sites (Rodwell et al., 1991). However because all these communities contain the species modelled, the possible change in community type with climate change (especially on sites containing wet heath, but not bogs) will not be evident from the models. It may be interesting and informative to model species, such as bristle bent (Agrostis curtisii) and bell heather (Erica cinerea), which may enter the community and species, such as Sphagnum compactum and S. tenellum, which may be lost with decreasing wetness.
MONARCH 2 Report – Chapter 6 139 ______
Figure 6.34: The pathway through the Leaver model predicted for Numenius arquata, shown in orange.
Community collapse New Community Dominant Species Major effect on function • Strong links with other species • Driver of belowground processes Reassembly
Yes Keystone species? No Natural change, Sub-dominant e.g. succession Leaver Species No Functional Type Redundancy? Yes
Rare Species Negligible • weak interaction strength Existing Community • often dependent on biotic interactions effect on function
The interaction web and conceptual model approach developed under MONARCH works well for the two species in this habitat predicted to undergo changes in suitable space, namely bog bush cricket and curlew. Indeed, the robustness of the approach is further illustrated by considering what would happen if either C. vulgaris or E. tetralix had been predicted to depart from the community (both of these species are predicted to undergo no change). Under such circumstances, the conceptual Leaver model alone would have predicted that this would have a large impact on community composition due to the departure of dominant species. However, this would have been a spurious prediction, as, for example, the interaction web for E. tetralix (Figure 6.35) shows a large number of the interactions within the descriptive community are not species specific to E. tetralix. In fact, they are actually specific to the ericoid species of heathland (C. vulgaris, E. cinerea and E. tetralix) (shown in pink) or even less specific than that (shown in blue). Similar conclusions have been reached regarding the fauna of Calluna and Erica (Richards, 1924; Webb, 1989). Therefore, the majority of the interactions are general and so any within community change in relative dominance (as would occur if Erica or Calluna had dispersed out of the community) would not result in a change in functional group (Chapter 4) and hence would continue to operate as before, even though there has been a noticeable change in the species composition of the community.
140 MONARCH 2 Report – Chapter 6 ______
Figure 6.35: The interaction web for E. tetralix, showing the main interactions. Yellow boxes indicate species that have specific interactions with E. tetralix; pink boxes indicate species that also interact with other heaths and C. vulgaris; blue boxes represent species which interact more widely and grey boxes indicate that the species is not currently present on wet heath in the test study area.
Flower visitors Bombus terrestris Andrena fuscipes Lassioglossum prasinum Andrena argentata (Hymenoptera) (Hymenoptera) (Hymenoptera) (Hymenoptera) Bombus ruderarius pathogens Other interactions Bombus campestris (Hymenoptera) Hoplitis claviventris (Hymenoptera) (Hymenoptera) Bombus sylvarum Gibbera salisburgensis Molinia tussocks Bombus lucorum (Hymenoptera) form rooting Anthophora furcata (Hymenoptera) medium for (Hymenoptera) Erica tetralix Colletes floralis Bombus vestalis Colletes (Hymenoptera) Bombus bohemicus (Hymenoptera) competitors (Hymenoptera) succinctus (Hymenoptera) Nomada flavopicta Vollucella bombylans (Hymenoptera) Molinia caerulea (Diptera) Mycorrhizae Bombus jonellus Calluna vulgaris Bombus hortorum (Hymenoptera) Hymenoscyphus ericae (Hymenoptera) (Ascomycete) Frakliniella intonsa (Thysanoptera) Ceratothrips ericae (Thysanoptera)
Strophosoma lateralis Erica tetralix (Coleoptera) Ematurga atomaria Dyscia fagaria (Lepidoptera) (Lepidoptera) Coleophora pyrrhulipennellla sheep (Lepidoptera)
Parvonia Coleophora juncicolella Pempelia palumbella cattle parvonia (Lepidoptera) (Lepidoptera)4+ (Lepidoptera) Ischnocoris Ceratothrips ericae (Thysanoptera) grouse angustulus (Heteroptera) Scolopostethus decoratus Pachythelia (Heteroptera) Macrothylacia rubi Pyla fusca villosella (Lepidoptera) (Lepidoptera) (Lepidoptera) Heliothis maritima Altica ericeti Eupithecia (Lepidoptera) Tetralicia ericae (Coleoptera) goosensiata Acanthococus devoniensis (Diptera) (Sternorrhyncha) (Lepidoptera) Lycophotia Amblyptilla acanthadactyla porphyrea (Lepidoptera) Thrips tabaci Cleora cinctaria Macroderma Xenolechia aethiops (Lepidoptera) (Thysanoptera) (Lepidoptera) micropterum (Lepidoptera) Scythris empetrella (Heteroptera) Frakliniella (Lepidoptera) intonsa Perconia strigillaria Diacrisia sannio Acleris hyemana Aeolothrips ericae (Lepidoptera) (Thysanoptera) (Thysanoptera) (Lepidoptera) (Lepidoptera) Ancylis uncella Metrioptera (Lepidoptera) Plebejus argus Argyrotaenia ljungiana Lycophotia porphyrea Neofaculta ericetella Pleurota bicostella Micrelus ericae brachyptera (Lepidoptera) (Lepidoptera) (Lepidoptera) (Lepidoptera) (Lepidoptera) (Coleoptera) (Orthoptera) ‘predators’
6.6 Estuarine waterbirds
The New Forest coastline was chosen as the second waterbird case study because it is adjacent to the Hampshire case study area chosen for wider consideration by the MONARCH 2 project. This stretch of coastline includes four estuaries – Southampton Water, Beaulieu, Lymington and Chichester Harbour. In contrast to the Suffolk coastline study area the estuaries of the New Forest coastline are characterised by sea defences protecting land reclaimed for industry and with substantial urbanisation. Furthermore, the topography of the land adjacent to these estuaries rises steeply compared to that along the Suffolk coast. Its estuaries are surrounded by low-lying land previously claimed for agriculture and thus offering considerable scope for managed re-alignment in response to sea-level rise. Unlike the east coast of Britain where numbers of many wader species have been increasing, and the south-west coast of Britain where numbers of many species of waders have been increasing, no marked general trend in wader numbers has been identified along the south coast (Austin et al., 2000). This would be as expected, as losses due to birds moving east could be balanced by gains due to birds moving in from the west.
MONARCH 2 Report – Chapter 6 141 ______
6.6.1 Impact of climate change
Using the protocol detailed in Chapter 5, weather did not help to explain the variation in the proportion of the population of any species over-wintering on British estuaries that do so on the four estuaries of the Hampshire case study area. An alternative analysis, that considered the proportion of the population over-wintering on British estuaries that do so along the entire south coast between The Dart estuary and Pegwell Bay, found that weather could explain the variation in this proportion for two of the seven species considered (H. ostralegus and T. tetanus). Both models proved to be statistically valid following the randomisation tests (Table 6.5).
Table 6.5: Details of the Generalized Linear Models relating the proportions of the British population of two wader species over-wintering on the estuaries of the south coast of Britain to weather. Parameters – independent variables retained by the model and the associated parameter estimates (β,γ,δ in equation 1). Partial t values and probability – indicates the significance of each parameter included in the model. Partial R2% – indicates the percentage of the variation in the dependent variable explained by each parameter. Adjusted model R2% – indicates the percentage of the variation in the dependent variable explained by the full model (adjusted for multiple parameters). Probability of obtaining the model by chance – as assessed by the randomisation testing using 9999 repetitions (see Chapter 5).
Species Applicability Parameters Parameter Partial t value Partial R2% Probability of Estimate obtaining the (Adjusted model by chance model R2%)
H. ostralegus South Coast Intercept -3.293 East-coast Minimum 0.043 t1=2.05, P=0.0497 12.6% P=0.0203 Temperature (9.6%)
T. totanus South Coast Intercept -2.957 GB Index 0.056 t1= 4.10, P=0.0003 28.2% P=0.0016 East-coast Mean -0.715 t1= -4.56, P<0.0001 24.8% Windspeed (49.6%)
6.6.1.1 Haematopus ostralegus (Oystercatcher )
The average minimum East coast temperature explained, significantly, part of the variation in the proportion of the H. ostralegus over-wintering in Britain that do so on the south coast. The positive parameter estimate for this variable indicates that the higher the average minimum east coast temperature the higher the proportion of the British population over-wintering on the south coast. However, the proportion of the variation explained by this association is comparatively low at 9.6%. This indicates that the cumulative effect of factors other than the weather variables considered have a strong influence on H. ostralegus numbers on the estuaries of the south coast, and consequently, although being built upon a significant association, the model has weak predictive power and the resulting predictions have wide confidence limits. This is apparent when the model is used to produce predictions of the number of H. ostralegus over-wintering on the estuaries of the Hampshire coastline under the UKCIP02 Low and High scenarios (Figure 6.36).
The numbers of H. ostralegus over-wintering on the estuaries of the Hampshire coast has increased from the 1970s to the late 1980s, since when numbers have fluctuated between about 800 and 1000 birds (Figure 6.36). The numbers predicted under the UKCIP02 scenarios are similar to the historic range.
142 MONARCH 2 Report – Chapter 6 ______
Figure 6.36: Observed numbers of H. ostralegus over-wintering on the New Forest coast estuaries between the winters of 1969/70 and 1999/2000 inclusive and predicted values (with 95% confidence intervals) made under the UKCIP02 Low and High scenarios for 2020s, 2050s and 2080s. The model used predicted the proportion of birds over-wintering in Britain that do so on the estuaries of the south coast. These predictions have been converted to numbers using both minimum and maximum recorded values for birds over-wintering in Britain. This gives an indication of the extremes in numbers that might be expected to over-winter on the estuaries within this region. The latter have been further adjusted to give those numbers expected on the estuaries of the New Forest coast while using both minimum and maximum values for the proportion of birds over-wintering on the estuaries of the south coast that do so on these estuaries. This latter proportion is not dependent on countrywide numbers (otherwise a model based on the estuaries of the New Forest coast alone would have been obtained). This gives an indication of the extremes in numbers that might be expected to over-winter on these estuaries.
Observed Predictions: Observed Predictions: based on minimumrecorded British index based on maximum recorded British index 1600 1600
1400 1400
1200 1200
1000 1000
800 800
600 600
400 400 Over-winter average numbers average Over-winter numbers average Over-winter 200 200
0 0
igh igh igh 0 970 Low 985 995 Low 1 1975 1980 1985 1990 1995 0 Low H H 197 1975 1980 1 1990 1 High 080 High 202 2020 2050 2050 H 2080 2080Low 20202020 2050 Low2050 High 2080 2Low
6.6.1.2 Tringa totanus (Redshank)
The British index for T. totanus and the average east coast wind speed explained, significantly, part of the variation in the proportion of the T. totanus over-wintering in Britain that do so on the south coast. The positive parameter estimate for average east coast wind speed indicates that the higher the average wind speed on the east coast the higher the proportion of the British population over- wintering on the south coast. The overall proportion of the variation explained by these associations is 49.6%, which indicates that the cumulative effect of factors, other than those considered, also have an influence on T. totanus numbers on the south coast. The predicted numbers of T. totanus that may over-winter on the estuaries of the Hampshire coast under the Low and High UKCIP02 scenarios are given below (Figure 6.37).
Predictions of the number of T. totanus that may over-winter on the estuaries of the Hampshire coast under the UKCIP02 scenarios suggests that at low values for the British index, numbers are unlikely to increase beyond those recorded over the past three decades (Figure 6.37). However, if the numbers of T. totanus over-wintering in Britain remain at levels that are as high as or higher than those of recent years then competition for space on these estuaries might increase as potentially numbers could increase by some 50% over contemporary values if they were to approach the upper confidence limit of the prediction.
MONARCH 2 Report – Chapter 6 143 ______
Figure 6.37: Observed numbers of T. totanus over-wintering on the New Forest coast estuaries between the winters of 1969/70 and 1999/2000 inclusive and predicted values (with 95% confidence intervals) made under the UKCIP02 Low and High scenarios for 2020s, 2050s and 2080s. The T. totanus national index is an explanatory variable in the predictive model used. In order to obtain an indication of how the size of the population affects the estimates, separate predictions have been made while using minimum and maximum recorded values of the Dunlin national index from the past three decades. This gives an indication of the extremes in numbers that might be expected to over-winter on these estuaries. The latter have been further adjusted to give those numbers expected on the estuaries of the New Forest coast while using both minimum and maximum values for the proportion of birds over-wintering on the estuaries of the south coast that do so on these estuaries. This latter proportion is not dependent on countrywide numbers (otherwise a model based on the estuaries of the Suffolk coast alone would have been obtained). This gives an indication of the extremes in numbers that might be expected to over-winter on these estuaries.
Observed Predictions: Observed Predictions: based on minimumrecorded British index based on maximum recorded British index 2000 2000
1500 1500
1000 1000
500 500 Over-winter average numbers average Over-winter numbers average Over-winter
0 0
igh igh 0 igh 970 Low H 985 995 Low 1 1975 1980 1985 1990 1995 20 Low 197 1975 1980 1 1990 1 0 050 080 Low 020 050 Low 080 Low80 High 2 2020 High 2 2050 H 2 2080 2 2020 H 2 2050 High 2 20
6.6.2 Impact of sea level rise
Detailed estuarine management plans of the type available for the estuaries of the Suffolk coast (see Chapter 5) do not exist for the estuaries on the Hampshire coast. Existing coastline management plans do not provide the detail required to assess how realignment of sea defences will affect the morphology of the estuaries themselves and they are not available in a readily accessible format. Consequently a summary of the coastline management strategies was considered in conjunction with Ordnance Survey maps (1:25,000 scale) in order to determine whether the stretches of coastline for which re-alignment of defences was being considered was likely to affect the morphology of the estuaries within the study area. From this exercise we have concluded that there is no latitude for morphological changes on the estuaries being considered sufficiently large to significantly affect the values of the parameters used by the MONARCH models. This is largely because of the nature of the land use adjacent to the these estuaries which is characterised by urbanisation and industrial development with hard defences that, unlike those defending agricultural land, will need to be maintained under all realistic socio-economic scenarios. In the few areas where this is not the case the land rises too steeply for sea level rise to lead to significant areas being inundated by high tides. Consequently, we predict that the future capacity of the estuaries of the New Forest coast will not change as a result of changes to the morphology of those estuaries.
This does not imply that sea-level rise will not affect habitats on these estuaries and coastal squeeze with loss of coastal-marshes may become an important conservation issue. Storms or extreme events could lead to more significant impacts. However, these complex issues are not ones that can be addressed by the MONARCH sea level rise models that relate to the extent of intertidal flats.
Although we predict that the future capacity of the estuaries of the Hampshire coast is unlikely to change as a result of changes in estuary morphology, it is possible to speculate whether there is
144 MONARCH 2 Report – Chapter 6 ______currently any surplus capacity within these estuaries, which would be capable of supporting any future increase in numbers of waders. H. ostralegus numbers over-wintering on these estuaries have increased steadily from the 1970s to the late 1980s and it may be that their numbers have now stabilised at between about 800 and 1000. Apart from two winters when particularly low numbers were recorded, T. totanus numbers have fluctuated at between 800 and 1200 birds with no distinct trend through time. This indicates that if H. ostralegus and T. totanus numbers were to increase above 1000 and 1200 birds respectively then competition for space between individuals would be greater than that encountered over the past three decades. The weather based models suggest that this will not be the case for H. ostralegus but that it might be for T. totanus under the UKCIP02 scenarios if their over-wintering numbers remain near to or higher than the higher values recorded in recent winters.
6.7 Discussion and conclusions
The bioclimatic classification has illustrated how sensitive the area is to changing climate. Whilst the effects of the 2020s Low scenario are largely within the bounds of climate experienced under the Baseline98 classification, the changes projected under the 2050s High scenario are significantly different from those experienced in the area at present.
The land cover modelling shows good prediction of baseline land cover for the coastal areas but poorer match for the northern and south-western areas. There is a complete loss of bracken (only a small percentage of land cover in baseline) by 2050s High. For the grassland habitats there is some redistribution of and loss of neutral grasslands but a complete loss of calcareous grassland land cover by 2050s High. Acid grasslands show a severe contraction of land cover. This may be due to the climate conditions exceeding those on which the model was trained and may not be a reflection of the fact that grassland land cover cannot exist in these areas.
The sensitivity of the area to changes under the 2050s High scenario, as shown in the bioclimate work, is not so strongly reflected in the species modelling results. This is partly due to the choice of long-lived tree species within the beech hangers habitat.
For the species modelling the following conclusions were drawn:
• Generally, the predicted climate change is not shown to be significant for the modelled species. • The European SPECIES model performed well, but fragmented distributions at both this scale and finer resolutions can lead to a lower agreement between the actual distributions and simulated climate space. • The downscaled model slightly improved the simulated distribution for F. sylvatica, but there was little difference for other non-bird species in terms of the statistics, although often the visual pattern is improved, as instead of blanket suitable climate space the simulated distribution is more refined e.g. C. vulgaris and the other dominant wet heath species. The climate response of species was masked by the inclusion of land cover, although not totally for A. flavicollis and M. brachyptera. • Dispersal is limited for trees, but A. flavicollis with a high maximum dispersal parameter has potential for spread but ecological factors may constrain it. Similarly M. perennis may be indirectly affected. The dispersal model parameterisation is important in influencing the apparent dispersal capabilities of species. • None of the plants appear sensitive to climate change, despite the predicted high sensitivity of this area, although both the birds could experience large losses of suitable space and M. brachyptera is sensitive to the level of change under the 2050s High scenario.
MONARCH 2 Report – Chapter 6 145 ______
In terms of species composition:
• Very little impact on species community composition is foreseen, based on the outputs of the dispersal model, as the dominant species remain within the community. • Extreme events such as storms and droughts may have the largest impact on species community composition. They have not been assessed in MONARCH 2. • Wet heath communities may shift in response to climate driven changes in water regime and may change the relative abundance of the modelled dominant species.
For estuarine birds the modelling projected the following effects:
• Based on climate change and mean sea-level rise predictions, alterations to the morphology of the coastal estuaries should not significantly affect bird numbers on intertidal flats. • Modelling suggests that H. ostralegus will show no real change in numbers relative to those seen over the last 30 years on the Hampshire estuaries, and likewise T. totanus if total numbers wintering in the UK remain low. An increase in total numbers of T. totanus in the UK would increase Hampshire estuary populations by about 50% and hence increase competition.
6.8 References
Austin, G.E., Peachel, I. and Rehfisch, M.M. (2000). Regional trends in coastal wintering waders in Britain. Bird Study, 47, 352-371.
Bannister, P. (1966). Erica tetralix L. Biological flora of the British Isles. Journal of Ecology, 54(3), 795-813.
Bannister, P. (1964). The water relations of certain heath plants with reference to their ecological amplitude: II. Field studies. Journal of Ecology, 52(3), 481-497.
Berry, P.M., Dawson, T.P., Harrison, P.A., and Pearson, R.G. (2002). Impacts on native woodland dynamics and distribution. In: Broadmeadow, M. S. J. (Ed.) Climate change and UK forests. Forestry Commission Bulletin 124. Forestry Commission. Edinburgh, pp169-180.
Berry, P.M., Vanhinsbergh, D., Viles, H.A., Harrison, P.A., Pearson, R.G., Fuller, R., Butt, N. and Miller, F. (2001). Impacts on terrestrial environments. In: Harrison, P.A., Berry, P.M. and Dawson, T.P. (eds.) Climate Change and Nature Conservation in the UK and Ireland: Modelling natural resource responses to climate change (the MONARCH project). UKCIP Technical Report, Oxford, pp43-150.
Brewis, A., Bowman, P. and Rose, F. (1996). The Flora of Hampshire. Harley Books, Colchester, in association with the Hampshire and Isle of Wight Wildlife Trust.
Cannell, M.G.R. and Sparks, T.H. (1999) Health of beech trees in Britain. In: Cannell, M.G.R., Palutikof, J.P. and Sparks, T.H. (Eds.) Indicators of Climate Change in the UK, Department for Environment, Transport and the Regions, London.
Conservation Progress Report: (2004) The Hampshire Ornithological Society. http://www.hants.org.uk/hos/conservation/conservationrev17.pdf
Crampton, A.B., Stutter, O., Kirby, K.J. and Welch, R. (1998). Changes in the composition of Monks Wood National Nature Reserve (Cambridgeshire, UK) 1964-1996. Arboricultural Journal, 22(3), 229-245.
146 MONARCH 2 Report – Chapter 6 ______
Detzel, P. (1998). Die Heuschrecken Baden-Wurttembergs. Verlag Eugen Ulmer, pp580.
Frynta, D., Vohralik, V. and Reznicek, J. (1994). Small mammals (Insectivora, Rodentia) in the city of Prague: Distributional patterns. Acta Societatis Zoologicae Bohemicae, 58(3-4), 151-176.
Graves, J.D. (1990). A model of the seasonal pattern of carbon acquisition in two woodland herbs Mercurialis perennis L. and Geum urbanum L. Oecologia, 83, 479-484.
Grimme. K. (1984). Water relations of Mercurialis perennis and Asarum europaeum in their natural habitat. Flora Jena, 175(4), 249-256.
Gurnell, J. (1996). The effects of food availability and winter weather on the dynamics of a grey squirrel population in Southern England. Journal of Applied Ecology, 33, 325-338.
Hampshire Biodiversity Partnership (1998). Biodiversity Action Plan for Hampshire, Volume 1. pp10.
Hansson, L. (1985). The food of bank voles, wood mice and yellow-necked mice. In J.R. Flowerdew, J. Gurnell and J.H.W. Gipps (eds.) The ecology of woodland rodents: bank voles and wood mice. Symposium of the Zoological Society of London, 55, 141-168.
Harrison, P.A., Berry, P.M. and Dawson, T.P. (2001). Climate Change and Nature Conservation in the UK and Ireland: Modelling natural resource responses to climate change (the MONARCH project). UKCIP Technical Report, Oxford.
Hulme, M., Jenkins, G.J., Lu, X., Turnpenny, J. R., Mitchell, T.D., Jones, R.G., Lowe, J., Murphy, J.M., Hassell, D., Boorman P., McDonald, R. & Hill, S. (2002). Climate Change Scenarios for the United Kingdom: The UKCIP02 Scientific Report. Tyndall Centre for Climate Change Research, School of Environmental Sciences, University of East Anglia, Norwich, UK, 120 pp.
Hulme, P.E. and Hunt, M.K. (1999) Rodent post-dispersal seed predation in deciduous woodland: predator response to absolute and relative abundance of prey. Journal of Animal Ecology, 68(2), 417- 428.
Jonathan Cox Associates (1997). East Hampshire Hangers NVC Survey. A report to English Nature. Unpublished report.
Kirby, K.J. and Thomas, R.C. (2000). Changes in the ground flora in Wytham Woods, southern England from 1974 to 1991: Implications for nature conservation. Journal of Vegetation Science, 11(6), 871-880.
Klotzli, F. and Walther, G.R. (2000). The behavior and dynamics of some dominant herbaceous plants of Swiss deciduous forests. Fragmenta Floristica et Geobotanica, 45(1-2), 111-121.
Leuschner, C., Backes, K., Hertel, D., Schipka, F., Schmitt, U., Terborg, O and Runge, M. (2001). Drought responses at leaf, stem and fine root levels of competitive Fagus sylvatica L. and Quercus petraea (Matt.) Liebl. Trees in dry and wet years. Forest Ecology and Management, 149, 33-46. Marigo, G., Peltier, J. P., Girel, J. and Pautou G. (2000). Success in the demographic expansion of Fraxinus excelsior L. Trees, 15(1), 1-13.
Marsh, A.C.W. and Harris, S. (2000). Partitioning of woodland habitat resources by two sympatric species of Apodemus: Lessons for the conservation of the yellow-necked mouse (A. flavicollis) in Britain. Biological Conservation, 92 (3), 275-283.
Marsh, A.C.W., Poulton, S and Harris, S. (2001). The Yellow-necked mouse Apodemus flavicollis in Britain: its status and analysis of factors affecting distribution. Mammal Review, 31, 203-227.
MONARCH 2 Report – Chapter 6 147 ______
Matthews, J.D. (1955). The influence of weather on the frequency of beech mast years in England. Forestry, 28, 107-116.
Mitchell-Jones, A.J., Amori, G., Bogdanowicz, W., Krystufek, B., Reijndres, P. J. H., Penuelas, J. and Boada, M. (2003). A global change-induced biome shift in the Montseny mountains (NE Spain). Global Change Biology, 9(2), 131-140.
Mountford, G. (1957). The Hawfinch. The New Naturalist. Collins. 176pp.
Nilsson, S.G. and Wastljung, U. (1987). Seed predation and cross-pollination in mast-seeding beech (Fagus sylvatica) patches. Ecology, 68(2), 260-265.
Pontailler, J-Y., Faille, A. and Lemee, G. (1997). Storms drive successional dynamics in natural forests: a case study in Fontainebleau forest (France). Forest Ecology and Management, 98, 1-15.
Preston, C.D., Pearman, D.A. and Dines, T.D. (2002). New Atlas of the British Flora. Oxford University Press, Oxford, 910pp.
Rayden, T.J. and Savill, P.S. (2004). Damage to beech woodlands in the Chilterns by the grey squirrel. Forestry, 77, 249-253.
Richards, O.W. (1924). Studies on the ecology of English heaths: III. Animal communities of the felling and burn successions at Oxshott Heath, Surrey. Journal of Ecology, 14(2), 244-281.
Rodwell, J.S. (Ed.), Pigott, C.D., Ratcliffe, D.A., Malloch, A.J.C., Birks, H.J.B., Proctor, M.C.F., Shimwell, D.W., Huntley, J.P., Radford, E., Wigginton,M.J. and Wilkins, P. (1991). British plant communities. Volume 2. Mires and heaths. Cambridge University Press, pp628.
Rutter, A.J. (1955). The composition of wet-heath vegetation in relation to the water-table. Journal of Ecology, 43(2), 507-543.
Sheikh, K.H. (1970). The responses of Molinia caerulea and Erica tetralix to soil aeration and related factors: III. Effects of different gas concentrations on growth in solution culture; And general conclusions. Journal of Ecology, 58, 141-154.
Sheikh, K.H. and Rutter, A.J. (1969). The responses of Molinia caerulea and Erica tetralix to soil aeration and related factors: I. Root distribution in relation to soil porosity. Journal of Ecology, 57(3), 713-726.
Stribley, G.H. and Ashmore, M.R. (2002). Quantitative changes in twig growth pattern of young woodland beech (Fagus sylvatica L.) in relation to climate and ozone pollution over 10 years. Forest Ecology and Management, 157, 191-204.
Webb, N.R. (1989). The invertebrates of heather and heathland. Botanical Journal of the Linnean Society, 101: 307-312.
Webb, N.R. (1986). Heathlands. A natural history of Britain’s lowland heaths. The New Naturalist. Collins, pp 223.
148 MONARCH 2 Final Report – Chapter 7 ______
7 Impacts for the Central Highlands case study area
P.M. BERRY, N. BUTT, H.P.Q. CRICK, S. FREEMAN, P.A. HARRISON, J.E. HOSSELL, G. MASTERS, P. SCHOLEFIELD AND N. WARD
Summary
The Central Highlands were selected because of their high potential for change, which could lead to the altitudinal upward movement of habitats and species, as well as gains or losses. The former was a reason for choosing upland and montane heath. These habitats were represented by bilberry (Vaccinium myrtillus) and heather (Calluna vulgaris) as dominants in the upland heath, cowberry (Vaccinium vitis-idaea), which is common to both, stiff sedge (Carex bigelowii), as typical of montane heath and ptarmigan (Lagopus mutus) as a flagship species. Caledonian pine woodland was selected as the other habitat because of its historical and high conservation interest. Scots pine (Pinus sylvestris) and silver birch (Betula pendula) were chosen as dominants, sessile oak (Quercus petraea) and willow tit (Parus montanus) as recruitment species and hairy wood ant (Formica lugubris) as a rare species.
The research showed that:
1. Central Highlands may experience a loss of montane habitat with a corresponding increase in the extent of areas classified as upland.
2. The land cover model reflected this well with a significant decline in suitable climate space for montane habitats, but also a complete loss of dwarf shrub, whereas neutral grassland expands in range. This is contrary to the SPECIES and dispersal modelling which indicated no change in extent for P. sylvestris, B. pendula, C. vulgaris, V. myrtillus and V. vitis-idaea. Possible explanations include lack of model sensitivity and the lack of land cover data in these models.
3. The dominant species are expected to remain within the Black Wood of Rannoch. P. sylvestris is predicted to remain a dominant species in Caledonian pinewoods. However climate and disturbance induced changes affecting the canopy and ground-flora will have feedbacks that may alter the composition and structure of the species community.
4. The tree species showed limited dispersal, although B. pendula showed the potential to disperse greater distances (under 50km by the 2050s) due to its shorter time to reproductive maturity. F. lugubris is currently rare in the case study area (and possibly under-recorded) and shows considerable potential for dispersal under future climatic scenarios.
5. The dominant and characteristic upland heath species: V. myrtillus and C. vulgaris and V. vitis-idaea, show little change, as they are widespread throughout the area. The small changes in the dominant species modelled for upland/montane heaths means that there are few direct implications for the species communities.
6. Suitable climate space for stiff sedge (C. bigelowii) and ptarmigan (L. mutus), species characteristic of montane heaths, remains extensive across the case study area, even under the 2050s High scenario. This may be due to the coarse scale of modelling (climatic data at 5km2) lacking sensitivity to indicate changes in an area of varied altitude.
7. When land cover change was incorporated for stiff sedge, then loss of montane habitat led to a decrease in suitable area.
MONARCH 2 Final Report – Chapter 7 149 ______
7.1 Introduction
The Central Highlands were selected as a case study because of their high change potential (Chapter 2 and Section 7.2), high conservation interest, range of habitat types, altitudinal variation and potential for species loss and gain. The boundaries of the areas were chosen to include the Creag Meagaidh National Nature Reserve (NNR) and the Black Wood of Rannoch, which is an important Caledonian pine woodland (Figure 7.1).
Creag Meagaidh was designated as a SSSI in 1964 because of its geological interest and subsequently a larger area covering almost 7,000 ha was designated as a SSSI due to its botanical interest. It is also an SAC due to the presence of Annex 1 habitats: Sub-Arctic Salix spp. scrub, siliceous and alpine boreal grasslands and hydrophilous tall herb and fringe communities of montane to alpine level. The alpine and sub-alpine grasslands cover 32.5% of the NNR and Carex bigelowii, one of the modelled species, is an important component of many of these communities. Part of it is also designated as a SPA (2872.64 ha.), because of its assemblage of montane birds. It contains an important breeding population of 23 pairs of dotterel (Charadrius morinellus), which represents at least 2.7% of the breeding population in Great Britain (8 year mean, 1987-1994) (SPA description, 2001). Many other parts of the case study areas, such as Ben Alder, the Drumochter Hills and Aonach Beag, also are designated as SAC, SPA and SSSI, mostly because of their montane habitats and species and hence the selection of upland heath and montane as one of the habitats for modelling.
7.2 Bioclimatic classification
The Central Highlands case study extends from Creag Meagaidh in the north southwards to incorporate the Black Wood of Rannoch. There are 72 5-km squares within the case study area, which was represented in the UKCIP98 data by 18 10-km squares. Table 7.1 shows the bioclimatic classification of the case study area under the Baseline98 classification defined in MONARCH 2.1 and the percentage of grid squares in each class under the Baseline02 data and 2050s Low and High emission scenarios. It is clear that the reclassification of the UKCIP02 baseline data into the Baseline98 classification has caused a shift in the pattern of climate classes across the study area.
Table 7.1: Percentage of grid squares within each of the classes of the bioclimatic classification at the 10km and 5km baseline and for the 5km 2050s Low and High scenarios. Class Scenario 3 6 910111215202526 Baseline98 10km 50.00 38.89 11.11 Baseline02 5km 66.67 1.39 1.39 22.22 6.94 1.39 2050 Low 44.4 16.67 8.33 4.17 25.00 1.39 2050 High 58.33 8.33 2.78 1.39 1.39 11.11 16.67
7.2.1 Baseline classification
The distribution of the classes and the relationship between the Baseline98 and Baseline02 classifications is shown in Figure 7.2. In the Baseline98 classification, climatic variation across the area was defined by just three classes (6, 15 and 26). These classes represent three of the four coldest classes in the Baseline98 classification in terms of absolute minimum temperature. Figure 2.5 shows the gradient of the key climate variables across the 26 classes for the Baseline 98 classification.
150 MONARCH 2 Final Report – Chapter 7 ______
Figure 7.1: The main features of the case study area.
A) Dry heath (Calluna - Vaccinium (some Montane heath/ rocks fields. bracken)). B)Lowland and upland acid grassland Regenerating birch wood. (Festuca spp. (including Festuca alpina), Deschampsia flexuosa, Agrostis spp.). C) Wet heath (e.g. Erica tetralix, Mixed deciduous woodland. Myrtillus, Calluna). Native pine woodland. Very species rich grassland on steep slopes. Includes all sorts of rarities.. Forestry
Upland moor (Molinia, Agrostis, Nardus, Blanket bog (Trichophorum spp., Calluna, ptarmigan, hares, dotterel, etc). Eriophorum spp., etc).
MONARCH 2 Final Report – Chapter 7 151 ______
Figure 7.2: The relationship between the Baseline98 and Baseline02 classifications. Symbols within the 5km grid squares relate to the Baseline02 classification and symbols at the intersection of grid squares relate to the Baseline98 classification.
At the 5-km scale of the Baseline02 classification, two of these classes are still represented but the increased data resolution also enables four further classes to be differentiated and the proportions of the classes represented changes greatly. The majority of the classes in the Baseline 98 dataset belonged to the high mountain class 6, which has both the shortest growing season and the lowest growing degree-day total of all the 26 classes. Class 6 represents an extreme bioclimatic type that is significantly different from the other 25 classes (see Table 2.15). Under the Baseline02 classification, the proportion of the area in this class increases as the classification successfully picks out the high mountain tops of Creag Meagiadh, Ben Alder and the Aonach and Grey Corries to the east of Ben Nevis (on the western edge of the case study area). Figure 7.2 shows the pattern of Baseline02 class membership within the pattern provided by the Baseline98 classification across the study area.
The more moderate climate of the valley edges and bottoms through Glen Spean and along Loch Laggan is picked up by the classification of these squares into classes 9, 15 and 20. The low-lying area to the south of the case study area is Loch Rannoch, which runs west to east and is classified into classes 12, 15 and 20. By comparison, the UKCIP98 data seem to overestimate the altitude of the area, and this is reflected in the low temperatures and short growing season of the UKCIP98 climate data for this area (Table 7.2 and Figure 7.3).
Examination of the squared Mahalanobis distances for the Baseline02 dataset (which are an indication of the distance of a grid cell’s climate from the centre of its class) shows 38 of the 72 squares may be considered to be beyond the bounds of the Baseline98 classification. This would indicate that the UKCIP02 baseline climate is sufficiently different from the classes defined under the Baseline98 classification as to warrant the creation of new classes. The inability to correctly assign these squares to a class under the Baseline02 climate means that it is also not possible to predict their sensitivity to the UKCIP02 scenarios, as the starting position against which they are being measured is incorrect.
152 MONARCH 2 Final Report – Chapter 7 ______
Table 7.2: Comparison of mean values of key variables averaged over the Central Highlands case study area from each of the data sets. The variables selected represent those related to the first four components of the PCA in the derivation of the bioclimatic classification. Spring Growing Degree January wind speed July Potential precipitation Days >5°C (m/s) Evapotranspiration Total (mm) (mm/day) Baseline98 10km 194.63 712.41 7.00 2.62 Baseline02 5km 166.28 724.45 9.28 3.15 2050 Low 154.67 985.80 9.70 3.30 2050 High 147.83 1153.74 9.71 3.53
Figure 7.3: Comparison of bioclimatic variables between the UKCIP98 and UKCIP02 baseline climate data. Symbols within the 5km grid squares relate to the Baseline02 classification and symbols at the intersection of grid squares relate to the Baseline98 classification.
MONARCH 2 Final Report – Chapter 7 153 ______
The cut off for the squared Mahalanobis distance taken is relatively generous (1347), since it represents the greatest distance between the centres of any two classes plus the distance of the furthest square from its class in the Baseline98 classification. Figure 7.4 shows the pattern of squared Mahalanobis distances for the Baseline02 data. All squares with a squared Mahalanobis distance greater than 1347 should be considered to lie beyond the limits of the Baseline98 classification. The greatest distances are for those squares in the highest mountain regions of the study area. The average climatic values for these squares, when compared to the national or local average for their class at the 10km classification, indicates a much cooler and more exposed climate (Table 7.3).
Table 7.3: Average climate conditions for squares in each class for key variables under the Baseline98 (local =average of squares in the study area, national=average for the class across Britain and Ireland), Baseline02 and 2050s Low and High emission scenarios. Spring Mean Jan Mean July Number of Class Scenario Rainfall Total GDD >5°C wind speed PET (mm/day) Squares (mm) (m/s) Baseline98 National 141.03 1393.63 8.07 2.72 113 3 2050 High 160.14 1139.02 10.08 3.57 42 Baseline98 National 187.58 599.84 7.11 2.63 33 Baseline98 Local 192.08 663.86 7.04 2.59 9 6 Baseline02 166.75 622.88 10.55 3.21 48 2050 Low 163.42 795.38 12.19 3.42 32 2050 High 151.92 722.07 15.51 3.77 6 Baseline98 National 169.34 1125.69 6.10 3.04 139 Baseline02 143.28 1067.66 4.74 2.78 1 9 2050 Low 167.52 1173.70 7.69 3.15 12 2050 High 140.64 1315.12 7.16 3.20 2 Baseline98 National 112.50 1514.99 5.45 3.50 794 10 2050 High 116.01 1293.20 7.12 3.62 1 Baseline98 National 213.84 1300.55 7.26 2.91 82 11 2050 High 185.48 1364.78 7.38 3.49 1 Baseline98 National 144.46 1424.56 5.24 2.64 425 Baseline02 140.85 1119.19 4.86 3.13 1 12 2050 Low 128.10 1384.33 5.59 3.24 6 2050 High 130.69 1518.92 5.92 3.44 8 Baseline98 National 197.77 861.92 6.53 2.75 76 Baseline98 Local 206.07 748.95 7.06 2.61 7 15 Baseline02 174.01 916.82 6.96 3.05 16 2050 Low 207.78 1012.90 9.29 3.09 3 Baseline98 National 138.77 1025.85 6.04 3.00 116 Baseline02 131.93 912.78 6.56 3.07 5 20 2050 Low 130.25 1068.63 7.87 3.23 18 2050 High 114.83 1121.55 8.90 3.42 12 Baseline98 National 281.40 730.90 7.48 2.51 16 25 Baseline02 240.51 842.59 8.24 3.09 1 Baseline98 National 157.70 801.95 6.61 2.88 41 26 Baseline98 Local 166.05 803.02 6.61 2.80 2 2050 Low 160.14 861.01 12.93 3.05 1
7.2.2 Climate change scenarios – 2050s Low and High
Figures 7.5 and 7.6 show the pattern of bioclimatic classes for the future climate change scenarios. The pattern indicates a further distancing of the climate of the area from the Baseline98 conditions
154 MONARCH 2 Final Report – Chapter 7 ______
(Tables 7.1 and 7.2). All the squares under future scenario conditions lie beyond the Baseline98 classification based on the squared Malhanobis distance measure. This is not surprising for the 2050s High emissions scenario, given the fact that this area was chosen for its sensitivity to this scenario using the UKCIP98 data. But even under the 2050s Low emission conditions the classification shows that the future pattern of climate is considerably different from either the Baseline02 or the Baseline98 conditions.
The classification of squares using the 2050s Low emission scenario data results in a shift towards a number of the classes associated with upland rather than mountain top climatic conditions. So for example, squares classified as 15 or 20 in the Baseline02 are reclassified to class 9 with the 2050s Low emissions data. Class 9 is associated with moderately high areas of the north-west Highlands under the Baseline98 classification.
Figure 7.4: Pattern of squared Mahalanobis distances for the Baseline02 data.
MONARCH 2 Final Report – Chapter 7 155 ______
Figure 7.5: The pattern of bioclimatic classes for the 2050s Low scenario.
156 MONARCH 2 Final Report – Chapter 7 ______
Figure 7.6: The pattern of bioclimatic classes for the 2050s High scenario. Symbols within the 5km grid squares relate to the Baseline02 classification and symbols at the intersection of grid squares relate to the Baseline98 classification.
The map of the Mahalanobis distance differences (Figure 7.7) between the Baseline98 and the 2050s High emissions scenario shows that the squares whose climate moves furthest from the Baseline98 values are at higher altitudes, particularly the square containing Creag Meagaidh itself. These squares are classified as class 6 in the Baseline02. The persistence of several squares within class 6 under all scenarios is an indication of the extreme climate conditions within these squares even given climate change. The squares have very low growing degree-days, absolute minimum and low July PET values, such that even with warming of 1-1.5°C under the 2050s Low and 2.5°C under the 2050s High scenarios the values remain well below the class mean. However, the squared Mahalanobis distance of these squares from the class centre is extremely high due to the difference between the Baseline98 class mean wind speed and that Baseline02 and 2050s emission scenario data (see Table 7.3).
The assignment of squares to classes under the 2050s High emissions scenario becomes increasingly unreliable as the level of climate change increases. For example most squares fall into Class 3 under the 2050s High emissions scenario, which may be seen as a “catch-all” class for this area, with relatively high wind speed but moderate levels of the other variables being the primary discriminating features. Its main location under the Baseline98 classification is on the west coast of Scotland and the northern tip of Ireland.
MONARCH 2 Final Report – Chapter 7 157 ______
Figure 7.7: The pattern of bioclimatic classes for the 2050s High scenario and the Mahalanobis distances.
Figure 7.8 shows the mean canonical score of the first seven of the 25 discriminant functions or canonical roots against each of the classes. This shows how the functions relate to each class. Table 7.4 shows which climate variables are related most strongly to each of the first 7 roots. The functions for class 3 all cluster around zero, with the roots associated with wind speed, 5 and 6, showing greatest difference from the others and having negative scores, and root 7 having a positive score. The wind speeds in the Baseline02 are higher across the study area than the in the Baseline98 dataset. Although they do not increase for the future scenario data, as the low temperatures are moderated and become less of a distinctive feature of the climate of these squares, so the wind speed becomes more significant in assigning the squares to classes. The high wind speeds provided by the UKCIP02 dataset may be more realistic than those within the UKCIP98 data. In the latter dataset, the average altitude for the area is 522m, whereas in reality much of the area is above 800m, with some summits
158 MONARCH 2 Final Report – Chapter 7 ______above 1000m. Though it should also be noted that microclimatic variations do also occur e.g. due to sheltering effects of vegetation and topography but such variation is beyond the resolution of the bioclimatic classification to detect.
Figure 7.8: The mean canonical score of the first seven of the 25 discriminant functions against each of the classes.
Table 7.4: Association of climatic variables with the first 7 discriminant functions. (Whilst 25 roots are created in the discriminant function analysis, the interpretation of each of these in relation to the climate variables becomes increasingly difficult as the power of the root declines. Hence only the first 7 were examined for the analysis.) Discriminant Positive canonical scores (>0.2) Negative canonical scores (<-0.2) Functions 1 Rainfall, HER, Spring rain total Mean & min T, sun, GDD, GSL, Abs min, Max T 2 Winter Mean T, Min T, GSL, Abs Min 3 Nov, Dec, Jan mean T & min T, GSL, Apl sunshine 4 Rainfall, HER (except May & Jun), May-Aug Mean T, Jun-Aug Min T, PET, June Spring Rainfall HER, Jun-Aug Sun, GDD, Max T 5 Jun, Jly Mean T, Apl rain, Nov, Dec sun, Wind speed Max T 6 March-Sep Mean T, June HER, Nov, Dec, Jan, Mch, Sep-Dec rain, Feb, Sep-Dec HER, Jan Sun, Max T, GDD Wind, Mch Sunshine 7 May, Aug-Oct Min T, Jan-Jun, Sep-Oct May HER PET, Jan, Feb, Jun, Aug-Nov Wind, Feb, May Sunshine
MONARCH 2 Final Report – Chapter 7 159 ______
7.3 Land cover changes
Figure 7.9 shows the Land Cover Map 2000 percentage coverage, the presence and absence using a 5% cutoff and the modelled presence and absence for montane habitats across the case study area. The habitat type is widespread across the area and the model prediction of its suitable climate space shows a moderate agreement with its location across Scotland as a whole and moderate to good across the case study area. Under the future change scenarios, there is a slight redistribution of the habitat within the study areas under the 2020s scenarios (Figure 7.10) and a slight loss in the south and east of the area. By the 2050s low scenario further habitat area is lost in the south. But, the greatest change is under the 2050s High scenario when all but 14 of the 72 squares lose the climate space suitable for this land cover type. The remaining squares are around the Grey Corries and Creag Meagaidh.
Several other land cover types show significant change between the baseline and the 2050s High emissions scenarios. Figure 7.11a shows the modelled baseline presence/absence of dwarf shrub land cover. The match between the modelled and Land Cover Map data is moderate for this class in Scotland and the land cover occurs in around 1/3rd of the study area, largely in the east. By the 2050s High emissions scenario there is no suitable climate space identified for this land cover type within the study area.
In contrast, neutral grassland occurs in few squares within the study area under the modelled baseline climate, but by the 2050s, there are large areas of the region that have a climate suitable for this land cover type (Figure 7.11b). The areas identified are those lower altitude squares that represent the valleys running through the Creag Meagaidh region around Glen Spean, Glen More and Loch Rannoch. However the results of the neutral grassland model should be treated with caution as its predictive power is low (see Chapter 3.4).
7.4 Caledonian pine woodland
The area of native woodland in the Scottish Highlands is 210,754 hectares, composed of equal proportions of native and planted native woodland, with Betula spp. being the most common in the latter, followed by P. sylvestris, then Quercus spp. (MacKenzie and Callender, 1995). A review of the state of P. sylvestris woods between 1975 and 1994 suggests that their decline has been halted, but that there has been little of the desired improvement in their extent and condition (Callender, 1994).
The Biodiversity Action Plan (Biodiversity: UK Steering Group Report, 1995) for native pine woodlands aims: • to maintain the current “core areas” of pinewoods and improve their condition; • to expand their area by 5600 ha by 2005 predominantly through natural regeneration; • to create conditions for a further 5600 ha to be naturally regenerated over the next 20 years; • to establish 25,000 ha of new native pinewoods on suitable sites within the natural range of pinewoods.
This, however, needs to take the potential impacts of climate change into account, especially when thinking about the future range of the species. This is discussed in Section 7.4.2.1.
160 MONARCH 2 Final Report – Chapter 7 ______
Figure 7.9: Land Cover Map 2000 percentage coverage, the presence and absence using a 5% presence cutoff and the modelled presence and absence for montane habitats.
MONARCH 2 Final Report – Chapter 7 161 ______
Figure 7.10: Montane habitat under the UKCIP02 climate change scenarios.
162 MONARCH 2 Final Report – Chapter 7 ______
Figure 7.11: The baseline and future modelled presence/absence of (a) dwarf shrub, and (b) neutral grassland.
(a)
(b)
MONARCH 2 Final Report – Chapter 7 163 ______
7.4.1 Species modelling
In the Caledonian pine woodland, none of the species were chosen for their sensitivity, but Formica lugubris was selected as a rare species. The SPECIES model trained well at the European scale for the Caledonian pine woodland species, with the independent AUC statistic greater than 0.9 for all species, indicating very good discrimination ability (Table 4.3). The plants have a maximum Kappa statistic greater than 0.85, indicating excellent agreement between observed distribution and simulated climate space, and for F. lugubris it is just above 0.7, indicating very good agreement. The lower agreement for this species is because in Europe it is found in montane (Alps, Pyrenees, Carpathians) and northern parts of Europe (Figure 7.12) and while the model picked up the general pattern, it tends to over-predict occurrences (false positives). The simulated climate space also includes the coast of northern Spain and France, suggesting that moisture may be a factor in its distribution.
Figure 7.12: The European observed distribution (a) and, simulated climate space (b) for Formica lugubris.
(a) (b)
In the downscaled modelling, the model continues to show very good discrimination ability for Pinus sylvestris showing that by the 2050s its distribution in the area may hardly alter in response to climatic change (Table 3.5). However, the AUC for the other three species is lower. This may partly be due to the presence of broad-leaved woodland in most grid squares, thus making it difficult for the model to discriminate on the basis of land cover; climate suitability being widespread for the two trees (Figures 7.13 and 7.14). For F. lugubris it is more a function of its low prevalence and very scattered observed distribution (Figure 7.15). The model outputs show that there is little change in the future suitable space, although for Q. petraea there is a loss of climate space in East Anglia, but some of this loss is masked in the downscaled model results by the land cover (Figure 7.16).
Nevertheless, unlike many of the species in other habitats, there is a poorer match between the simulated current suitability surface and the observed distribution when land cover is included. In the case of Betula pendula, which is found almost throughout Britain, there is little change and some Highland areas are simulated as unsuitable. For Q. petraea, the addition of land cover restricts the climate suitability surface, such that areas where it does occur in southern England and northern Scotland are simulated as unsuitable. In the case of the latter area, it could partly be a function of slightly lower climate suitability (Figure 7.13). The suitability surface for F. lugubris is still far too widespread compared with its actual distribution (Figure 7.15).
164 MONARCH 2 Final Report – Chapter 7 ______
Figure 7.13: Model outputs for Quercus petraea: (a) climate suitability surface from the European- trained network, (b) observed 10km distribution, (c) climate and land cover suitability surface from the downscaled model, and (d) presence/absence surface from the downscaled model.
(a) (b)
(c) (d)
MONARCH 2 Final Report – Chapter 7 165 ______
Figure 7.14: Model outputs for Betula pendula: (a) climate suitability surface from the European- trained network, (b) observed 10km distribution, (c) climate and land cover suitability surface from the downscaled model, and (d) presence/absence surface from the downscaled model.
(a) (b)
(c) (d)
166 MONARCH 2 Report – Chapter 7 ______
Figure 7.15: A comparison of the observed distribution with current simulated climate suitability surface and simulated climate and land cover suitability surface for Formica lugubris.
Observed distribution Current simulated climate Current simulated climate and land suitability surface cover suitability surface
Figure 7.16: Presence/absence suitability surfaces for Quercus petraea: (a) climate suitability surface for 2050s High, (b) climate and land cover suitability surface for current climate, and (c) climate and land cover suitability surface for 2050s High.
(a) (b) (c)
MONARCH 2 Final Report – Chapter 7 167 ______
The 1-km distributions for the trees were derived from the 5-km distributions supplied by Scottish Natural Heritage and it was assumed that the species were present in every 1-km grid square of each 5-km square with “presence”, and, which was not completely covered by water. This will give an over-estimate of the presence of the tree species, especially native Pinus sylvestris. F. lugubris is recorded as present in two 10-km squares, one of which lies at the southern edge of Black Wood of Rannoch. Scottish Natural Heritage suggested that the 1-km square closest to the centre of this 10-km square was used, together with the 12 (1km) squares in the other 10km square where broad-leaved woodland occurred, to provide the 1km distribution. The trees show a very limited dispersal, due to their longer time to reproductive maturity, than herbaceous plants. The dispersal model shows a slightly wider potential area for colonisation for B. pendula than for the other trees, because of the shorter time to reproductive maturity and longer dispersal distance (Figure 7.17).
F. lugubris, however, shows considerable potential for dispersal in future in this area (Figure 7.18), although it must be remembered that its precise current distribution is unknown. Also a study in southern Finland found that the monogynous (colonies only have one queen and inhabit one nest) F. lugubris was more common in young forests and in small old-forest fragments, while polygynous F. aquilonia was more common in old forests and in larger old-forest fragments (Punttila, 1996). Competition, therefore, allied to the structure of the landscape, may be a factor affecting the future abundance of F. lugubris.
In MONARCH 1, another species associated with pine woodland, twinflower (Linnaea borealis), was modelled, using the UKCIP98 scenarios. In Scotland, suitable climate space all but disappeared in the Southern Uplands by the 2050s High, when it is also very much more restricted to the higher elevations in the north-west Highlands and Grampians (Berry et al., 2001). Similar modelling work, also using the UKCIP98 scenarios showed that other important pine woodland species: dwarf birch (Betula nana), creeping lady’s-tresses (Goodyera repens), one-flowered wintergreen (Moneses uniflora), and Scottish crossbill (Loxia scotica) lose suitable climate space dramatically, while pine marten (Martes martes) and red squirrel (Sciurus vulgaris) showed little change (Ogawa, 2002).
The results for the Caledonian pine forest suggest that climate change will only have a minor impact in the next 50 years or so. The dominant species modelled are predicted to remain within the community, although if Q. petraea is able to establish itself, then it could have significant implications for Caledonian pine forest relict species. Other modelling work suggests that some of the associated ground flora species could lose suitable climate space in the future, although not necessarily in this case study area (Ogawa, 2002).
Another issue that has become apparent is the possible sensitivity of the modelling results to the use of different scenarios. For example, suitable climate space was predicted as becoming available for P. montanus in the Central Highlands under the UKCIP98 climate change scenarios and thus the species was selected as a possible recruit to Caledonian pine woodland. Under the UKCIP02 scenarios the climate space does not quite reach this area, due to the different patterns of temperature and precipitation change (Figure 7.19).
168 MONARCH 2 Final Report – Chapter 7 ______
Figure 7.17: Betula pendula dispersal model outputs.
Observed 1km distribution
2020s Low scenario 2020s High scenario
2050s Low scenario 2050s High scenario
MONARCH 2 Final Report – Chapter 7 169 ______
Figure 7.18: Formica lugubris dispersal model outputs.
Observed 1km distribution
2020s Low scenario 2020s High scenario
2050s Low scenario 2050s High scenario
170 MONARCH 2 Final Report – Chapter 7 ______
Figure 7.19: Suitable climate space for Parus montanus: (a) 2020s Low, (b) 2020s High, (c) 2050s Low, and (d) 2050s High.
(a) (b)
(c) (d)
7.4.2. Implications for the composition of species communities
Caledonian pine forest is represented in this case study area, by the Black Wood of Rannoch, in the southeastern section. This wood contains both Pinus sylvestris and Betula species (Betula pendula is modelled in this study) as dominants, and has a ground flora, dominated by ericoid sub-shrubs including Vaccinium myrtillus, Vaccinium vitis-idaea and Calluna vulgaris (Table 7.5). The ground flora is also moss dominated, bracken dominated or grass dominated in places.
MONARCH 2 Final Report – Chapter 7 171 ______
Table 7.5: Summary of the dispersal model results. Species Category Predicted space change Pinus sylvestris Scots pine Dominant No change Betula pendula silver birch Dominant No change Calluna vulgaris ling Dominant No change Vaccinium myrtillus bilberry Dominant No change Vaccinium vitis-idaea cowberry Dominant No change Formica lugubris hairy wood ant Flagship Increase Quercus petraea sessile oak Recruitment No change Parus montanus willow tit Recruitment Absent
7.4.2.1 Climate change and the pinewood species
7.4.2.1.1 Quercus petraea Quercus (oak) is present in the vicinity of the Black Wood of Rannoch, but not yet in the actual pinewoods themselves (Steven and Carlisle, 1959). However, the dispersal model indicates that Quercus petraea will continue to encounter suitable climate space in this area, but that expansion of range is likely to be slow (as shown in Figure 7.16). If Q. petraea arrives in the habitat, it is probable that some of the associated species will also enter the community, including fungi, mycorrhizae, mosses, lichens and invertebrates, The exact nature of this Q. petraea associated community will be dependent on numerous factors including local climate, soil type, and proximity to ponds and streams as well as tree age. Young trees, such as those which may establish at the Black Wood of Rannoch, will support a smaller number of associated species, as much of the structural diversity, and therefore niches, will not have had time to develop.
However, whether Q. petraea will establish from near-by stands will be dependent on whether it is able to regenerate in the pine wood community. In contrast to P. sylvestris and Betula, Q. petraea is able to tolerate both high levels of shading (roughly to the same extent as V. myrtillus) and better able to compete with the ground flora, allowing it to establish in closed vegetation (Jones, 1959). For example, seedlings of Q. petraea are sometimes able to grow and persist surrounded by P. aquilinum, but again smothering by dead fronds in the autumn is a problem (Jones, 1959). Grazing has been identified as a major factor limiting the regeneration of native pinewoods in Scotland (Palmer and Truscott, 2003a). In the Black Wood of Rannoch densities of deer have been estimated to be about 15 km-2 (Baines et al., 1994) and this is too high for successful regeneration of P. sylvestris to occur (Palmer and Truscott, 2003b). However, for Q. petraea, where the sub-shrub canopy is vigorous, seedlings often have yellow leaves and do not increase in height, although they are still able to persist under such conditions (Jones, 1959). In addition, grazing may affect the survival and growth of seedlings and saplings, especially as browsing of the tip may remove all foliage, as this is usually present as a single whorl at the top of the stem. Furthermore, a study of five Atlantic oakwoods, found that Q. petraea suffered the worst effects of browsing when compared to Betula, Corylus avellana (hazel), and Sorbus aucuparia (rowan) (Palmer et al., 2004). Therefore, it seems as if Q. petraea may have difficulty in regenerating within the pine woodland. Furthermore, even if it does enter the community it will, over the time scale of interest in this study, remain a relatively small tree, devoid of a large proportion of its associated flora and fauna.
7.4.2.1.2 Formica lugubris The dispersal model predicts that F. lugubris, will gain rather than lose suitable climate space, within the Black Wood of Rannoch and its vicinity (Table 7.5). Wood-ants can have far-reaching effects on the pinewood community in general, through selective foraging, competition, as a food source and through the creation of habitats. It has been suggested that wood-ants, including F. lugubris act as keystone species (Fowler and Macgarvin, 1985), and therefore on the basis of the Arriver model,
172 MONARCH 2 Final Report – Chapter 7 ______colonisation by F. lugubris into new areas, may have large implications for the structure of the species community (as shown in Figure 7.20).
However, Formica aquilonia, is also present in the Black Wood of Rannoch (Edwards and Telfar, 2001). Therefore, invasion of F. lugubris into areas already occupied by this second wood ant species which is of the same functional type is likely to have little effect on the species composition and structure of the community (Figure 7.20).
Figure 7.20: A possible outcome on the composition of the species community of the colonisation of Formica lugubris, based on Fowler and Macgarvin’s (1985) suggestion that this species is a keystone species.
Large impact t n on ecosystem Community t e n e m Dominant Species collapse h function m s it li Highly ru b c ta modified e s R E community Reassembly Expansion New Functional Type Sub-dominant Species Keystone Coloniser species e, ng Arriver ha n Existing l c io a ss Functional ur ce at c Type N su g. e.
t fec ef Rare Species ble igi n gl tio Ne nc fu on Existing community Extinct No impact on ecosystem function
MONARCH 2 Final Report – Chapter 7 173 ______
7.5 Montane/upland heath
Upland heath is a BAP priority habitat and is also listed under the EU Habitats Directive. It often grades into montane heath, and so while two of the plants chosen (Vaccinium myrtillus and Calluna vulgaris) can be considered dominants in the former, they could form recruitment species in the latter habitat, while V. vitis-idaea is found widely in all heathland types within its geographical areas. Montane heath is found on the higher parts of mountains, along with grassland and Carex bigelowii is a typical component species. It is currently affected by grazing, atmospheric pollution and fires, but under climate change its composition and diversity could be affected by species dispersing to higher latitudes.
7.5.1 Species modelling
The SPECIES model trained well at the European scale for the upland heath and montane species, with the independent AUC statistic greater than 0.9 for all species, indicating very good discrimination ability (Table 3.3). C. vulgaris, V. myrtillus and V. vitis-idaea are widespread in Europe and have a maximum Kappa statistic greater than 0.85, indicating excellent agreement between observed distributions and simulated climate space, C. bigelowii and Lagopus mutus (ptarmigan) have very good agreement (e.g. C. bigelowii has a 0.7 threshold). In the case of C. bigelowii, the simulated climate space is too wide in Scandinavia, Britain and Ireland, and the Alps are also simulated as having suitable climate space (Figure 7.21).
The downscaled SPECIES modelling results are particularly good for all the plant species (Table 3.5), with the AUC statistic dropping below 0.9 only for C. vulgaris and this is due to an over-restriction of suitable space in southern England, compared with its observed distribution (Figure 7.22). The Best Available Match for L. mutus is high at 98.7% (Table 3.4), but as it has the most restricted range of any bird species considered in this study, being a strictly montane species restricted in Europe to the high Arctic, the Alps and Pyrenees and the Scottish highlands, it is not surprising that networks on neither a European or British scale prove entirely successful in reproducing this (Figure 7.23). The addition of land cover does improve the simulated suitability surface for all the plants when compared with the actual distribution by restricting it in lowland England, although, as noted above, this is too severe for C. vulgaris (Figure 7.24). The climate change scenarios showed that both V. myrtillus and V. vitis-idaea could lose suitable climate space; the former in East Anglia and the latter in parts of southern and central England. The results from the downscaled SPECIES model does not pick up all of this loss, thus giving a wider suitability surface than would be suggested by climate alone.
Figure 7.21: The European observed distribution (a) and simulated climate space (b) for Carex bigelowii.
(a) (b)
174 MONARCH 2 Final Report – Chapter 7 ______
Figure 7.22: Model outputs for Calluna vulgaris: (a) climate suitability surface from the European- trained network, (b) observed 10km distribution, (c) climate and land cover suitability surface from the downscaled model, and (d) presence/absence surface from the downscaled model.
(a) (b)
(c) (d)
MONARCH 2 Final Report – Chapter 7 175 ______
Figure 7.23: The European observed distribution (a) and simulated climate space (b) for Lagopus mutus.
(a) (b)
Figure 7.24: A comparison of the observed distribution with current simulated climate suitability surface and simulated climate and land cover suitability surface for Calluna vulgaris.
Actual distribution Current simulated climate suitability Current simulated climate and surface land cover suitability surface
The fact that none of the species selected showed major changes in terms of availability of suitable space in the case study area could be a consequence of several factors. Firstly, they may not be sufficiently sensitive to the climate changes predicted for the area, this is particularly likely to be the case for species with a wide distribution, such as C. vulgaris and V. myrtillus. Nevertheless, they were important choices as it is essential to ensure that habitat dominants are not going to be directly adversely affected and potential changes in their abundance. Secondly, little change could be a result of not using land cover change scenarios in the downscaled SPECIES model, which would have fed through to the dispersal model. This was shown to have an influence in the case of C. bigelowii
176 MONARCH 2 Final Report – Chapter 7 ______
(Figure 7.25) and could have been important for L. mutus, which is particularly associated with the montane habitat.
Figure 7.25: A comparison of the suitability surfaces for Carex bigelowii, with and without land cover change.
CLIMATE AND LAND COVER CLIMATE CHANGE ONLY CHANGE SCENARIOS: SCENARIOS:
Baseline Baseline
2020s Low scenario 2020s Low scenario
MONARCH 2 Final Report – Chapter 7 177 ______
2020s High scenario 2020s High scenario
2050s Low scenario 2050s Low scenario
2050s High scenario 2050s High scenario
178 MONARCH 2 Final Report – Chapter 7 ______
The 1-km distributions of all the plants were derived from the 5-km distribution with all 1-km squares in a 5-km square recording a presence also deemed to contain the species, except the 1-km squares exclusively covered with water. In the case of C. bigelowii, squares of a mean altitude lower than 600m also were excluded. Most of the case study area is predicted as suitable for C. bigelowii, so this altitudinal restriction on current distribution leads to most lower altitudes being colonized under the future scenarios (Figure 7.26). This then is an artefact of the method used to derive the current distribution, if the 600m restriction is applied then there is little change in its distribution, although Beinn Eilde is progressively colonised under the dispersal model (Figure 7.27).
C. vulgaris and V. myrtillus occur throughout the case study area but the higher parts are simulated as unsuitable both now and in future and so there is no change when the dispersal model is run (Figure 7.28). V. vitis-idaea occurs and also has suitable space throughout the case study area, but under the 2050s High scenario it starts to be lost from some squares in the southern part of the area (Figure 7.29). Although the dispersal model results do not indicate any changes in the distribution in the species within montane areas of the case study area, it is suggested that in all areas there will be a reduction in late-snowbed vegetation.
The modelling work has shown that none of the upland heath or montane species investigated are likely to lose suitable space in this area, but the impacts of climate change allied to other environmental factors, such as nitrogen deposition, mean that their inter-relationships may well change and this will be explored in Section 7.5.2.
Some data limitations have been identified in that 1-km distributions were not available and thus surrogates had to be developed, based on presence in all the 1-km squares not covered by water of the 10-km distribution in the case of C. vulgaris, V. myrtillus and V. vitis-idaea or by the imposition of an artificial altitudinal limit in the case of C. bigelowii. The former led to a discrepancy between the supposed distribution and the simulated distribution, with a number of the squares at higher altitudes being predicted as unsuitable by the downscaled SPECIES model. The latter resulted in dispersal below the artificially imposed cut off altitude. In any future fine-scale modelling work distributional data at an appropriate resolution is a key requirement.
7.5.2 Implications for the composition of species communities
7.5.2.1 Upland heath
The dispersal model does not predict any change in the distribution of the upland heath species, C. vulgaris and V. myrtillus, and the generally more arctic alpine species V. vitis-idaea. No obvious visual change in upland heath would therefore be expected before the 2050s. However, the model is solely predicting the presence or absence of a species in each particular kilometre square and does not give any indication of abundance. In addition, many different communities may be represented in a single square of the model, especially in the Central Highlands where substantial changes in altitude and hence climatic conditions and probably plant abundances will occur over the area of a single square.
The SSSIs within the study area, support a number of heaths ranging from dry heaths containing C. vulgaris and bell heather (Erica cinerea) as the dominant species through to wet heaths, where C. vulgaris, cross-leaved heath (Erica tetralix) and purple moor grass (Molinia caerulea) may all be abundant. Therefore, any changes in the soil moisture regime due to climate change may result in shifts in these species. Future climate scenarios predict a slight decrease in summer precipitation of up to 20% by the 2050s, although winter rainfall is expected to be similar to the present day (Hulme et al., 2002). In addition, it is predicted that summer cloud cover will be reduced and with wind speeds remaining similar, plants may suffer from increased rates of evapo-transpiration, during the summer months.
MONARCH 2 Final Report – Chapter 7 179 ______
Figure 7.26: Carex bigelowii dispersal outputs, without an elevation mask.
Observed 1km distribution
2020s Low scenario 2020s High scenario
2050s Low scenario 2050s High scenario
180 MONARCH 2 Final Report – Chapter 7 ______
Figure 7.27: Carex bigelowii dispersal outputs, with an elevation mask.
Observed 1km distribution
2020s Low scenario 2020s High scenario
2050s Low scenario 2050s High scenario
MONARCH 2 Final Report – Chapter 7 181 ______
Figure 7.28: Calluna vulgaris dispersal model outputs.
Observed 1km distribution
2020s Low scenario 2020s High scenario
2050s Low scenario 2050s High scenario
182 MONARCH 2 Final Report – Chapter 7 ______
Figure 7.29: Vaccinium vitis-idaea dispersal model outputs.
Observed 1km distribution
2020s Low scenario 2020s High scenario
2050s Low scenario 2050s High scenario
MONARCH 2 Final Report – Chapter 7 183 ______
Studies on the flora and fauna of heathlands have suggested that a gradient in soil moisture is important in determining species assemblages. However the species assemblage of wet heaths contains many species not present on drier heaths, such as E. tetralix, the ground beetles Agonum ericeti and Carabus nitens and the spiders Pirata piraticus, Antistea elegans and Silometopus elegans, together with species present on less wet sites (Usher, 1992). This suggests that drying out of a site will lead to a net loss, of at least the invertebrates such as spiders and ground beetles, although some species less suited to the wetter conditions may become more abundant.
7.5.2.2 Montane heath
The species modelled as representatives of montane habitats are: Carex bigelowii, Vaccinium myrtillus, Vaccinium vitis-idaea, Calluna vulgaris and Lagopus mutus. However, as with the upland heath, very little change in distribution of these species is predicted to occur under the future climate scenarios, although there may be changes in abundance of these species.
Predictions of climate change suggest that the case study area will become slightly warmer by about 1 to 2.5ºC by the 2050s, and that additionally summers will become marginally drier (maximum of 20% drier under the 2050s High scenario). The amount of rain falling as snow may also decrease, but wind speeds are not expected to change (Hulme et al., 2002).
As a result, the bioclimate in the montane areas, is, for the most part, predicted to shift towards one more similar to that of upland areas, rather than of mountain tops today (Section 7.2.2). However the bioclimatic classification for the case study area also suggests that some areas located on high summits will continue to experience extreme conditions, similar to those described in Box 7.A.
Box 7.A: Current environmental conditions in montane habitats in the Central Highlands case study area.
Extreme conditions:
• Steep temperature lapse rates lead to low temperatures on the summits reducing growing season length and physiological activity of organisms.
• High cloud cover reduces ground temperature through reduction in insulation.
• Exposed areas are subject to severe frosts. In such areas only hardy plants can survive.
• Snow insulates plants from extreme temperature fluctuations, but where it is slow to melt it reduces growing season length.
• Winter drought conditions occur as water is frozen (snow and ice).
• Strong winds 1) redistribute snow and 2) increase evapo-transpiration and 3) cause mechanical damage to plants in exposed areas.
Therefore, changes in the topography, slope and microclimate can all have critical effects on the composition of the plant community, and as a result of these features, large gradients in key environmental factors can occur over fairly short spatial scales, and these can sometimes produce vegetation mosaics. The picture is further complicated where disruptions or variations have occurred. These may include a year with a shorter or longer snow-lie than usual, and although such disruptions often happen (Rodwell et al., 1992), climate change may increase their frequency still further.
184 MONARCH 2 Final Report – Chapter 7 ______
Carex bigelowii
The predicted distribution of C. bigelowii fails to give an indication of possible changes in community composition within the montane grass and moss heath communities of the case study area. This species appears to be able to tolerate the full range of conditions experienced by montane siliceous grassland communities today. It is present in communities typical of exposed situations such as the Juncus trifidus – Racomitrium lanuginosum rush heath (U9), although in the harshest of such communities, characterised by open areas of shifting gravel (Ingram, 1958) it may only occur as a casual component that has dispersed from other communities. It is also a constant species in the Nardus stricta – Carex bigelowii grass-heaths (U10), which often surround snowbeds, and in the Carex bigelowii – Polytrichum alpinum sedge heath (U11), characteristic of late snowbeds with about five months of snow-lie, which finally melt in June. C. bigelowii, may also be present in the very late snowbeds which support the Salix herbacea-Racomitrium heterostichum (U12) and Polytrichum sexangulare-Kiaeria starkei (U11) communities, albeit at a reduced density and frequency (Rodwell et al., 1992 ). Thus whilst there are substantial changes in the associated species with environmental gradients, C. bigelowii maintains its presence, and, therefore, changes in community composition as a result of climate change are not adequately portrayed. Although changes in the relative abundance of C. bigelowii may occur over this gradient, these are not picked up by the dispersal model, which solely predicts future distributions on a presence or absence basis.
Although C. bigelowii is tolerant of a wide range of exposures, soil moisture conditions, snow-lie and moderate levels of disturbance such as frost heave, changes in tiller production and growth can be influenced by shelter and temperature (Brooker et al., 2001). Increases in air temperature, such as those predicted for the case study area, have been shown to increase flowering in northern Sweden (Brooker et al., 2001). Flower initiation occurs in the summer prior to flowering and is strongly positively influenced by temperature. Once a tiller has flowered, apical dominance is removed due to the death of the main apical meristem, and this leads to the development of a series of fresh tillers from the main tiller (Brooker et al., 2001). In the year after a warm growing season, many tillers will flower and this will increase the density of tillers within a clone and increase the relative contribution of this species to the community. However, it is unclear, whether C. bigelowii flowering in the case study area is indeed temperature limited.
Vaccinium myrtillus
V. myrtillus is present within the siliceous alpine and boreal grasslands under discussion in this section but is never dominant. It attains dominance at slightly lower altitudes where conditions are somewhat less extreme. V. myrtillus is limited by exposure on the one hand (Burges, 1951) and the shortness of the growing season on the other hand (Rodwell et al., 1992). In areas of high exposure and gales such as those found in the present day Carex-Racomitrium heath, it is often dependent on shelter provided by boulders, other plants, and moss clumps (Burges, 1951).
V. myrtillus is entirely absent from the very late snowbed vegetation and occurs only infrequently in the snowbeds which melt slightly earlier (i.e. the Carex-Polytrichum sedge-heath). It is limited in vigour due to the extreme shortness of the growing season mediated through the long cover of snow, and the high moisture content of the soils in the summer as a result of impeded drainage and the occurrence of summer melt water (Rodwell et al., 1992).
On the summits of some mountains in the case study area, the conditions are expected to remain similarly harsh, although a lengthening of the growing season is predicted and this may be favourable to this species. In addition, it is possible that there may be a reduction in the quantity of snow and the duration for which it is present, which will also lead to a lengthening of the growing season so that these areas may become favourable for Vaccinium growth.
Other areas currently supporting montane communities within the case study area are predicted to become less extreme, mirroring conditions presently associated with the uplands as opposed to the
MONARCH 2 Final Report – Chapter 7 185 ______mountain tops. Under such conditions, especially in the more sheltered areas, favoured by V. myrtillus, it is possible that communities similar to those present at lower altitudes, may find suitable space, and that a general shift of communities to higher elevations will take place.
Calluna vulgaris
C. vulgaris dominated communities can often occur in mosaics with Vaccinium dominated ones, below the siliceous boreal and alpine grasslands, although they are sometimes characteristic of slightly lower altitudes. C. vulgaris can withstand higher levels of exposure than V. myrtillus and Empetrum hermaphroditum (but still less so than R. lanuginosum). However, it is probably the length of snow-lie that determines whether Calluna or Vaccinium is dominant (Metcalfe, 1950). The evergreen C. vulgaris is much less tolerant of prolonged snow-lie than V. myrtillus. Thus, with a decrease in the duration of snow-lie overall, there may be a shift in dominance in favour of C. vulgaris.
In addition, with the improving climatic conditions in at least some montane areas in this case study area, C. vulgaris may be able to penetrate further up towards the summits, with the change from C. vulgaris dominance to either Vaccinium or R. lanuginosum dominance occurring at progressively higher altitudes. Its ability to withstand the increased wind speeds present at greater altitudes, may mean that this species responds to climate change more favourably than V. myrtillus. However, C. vulgaris is vulnerable to damage from both frosts and winter drought (Gimingham, 1960) and it is these factors that may limit its upward expansion. Summer drought on exposed sites may also occur and in C. vulgaris this accelerates bud break in the following spring (Gordon et al., 1999), which may make the plant more susceptible to dieback.
Lagopus mutus
Although this bird is exclusive to the montane zone, being most abundant at approximately 900 to 1150m, it is extremely sparse in areas dominated by grassy or mossy vegetation (Watson, 1965), such as those under discussion here. Observations have suggested that L. mutus prefers areas of stunted heath containing a high abundance of either Vaccinium or Empetrum (Watson, 1965). These areas, in contrast to the grassy or mossy heaths, contain high proportions of its main food plants, V. myrtillus, crowberry (Empetrum nigrum) and C. vulgaris. Young chicks also eat invertebrates and other plant species are eaten in small quantities (Watson et al., 1998). Hence, if with climate change, the cover of these ericoid shrubs increases at the expense of grasses, mosses and sedges, this will clearly be beneficial to L. mutus.
7.6 Discussion and conclusions
The bioclimatic classification has shown that the case study area is characterised by upland and high mountain climates under the Baseline98 classification. There appear to be significant differences between the UKCIP98 and 02 baseline data in this area, which results in a number of the Baseline02 squares no longer fitting into the Baseline98 classification. The difference seems to be partly due to the differences in the wind speed data between the baseline datasets.
The distance of some of the squares in the Baseline02 classification from the Baseline98 classes is a reflection of the increased detail that the higher resolution baseline data provides. For areas of great variation in topography, such as the Central Highlands study area, the 10-km UKCIP98 dataset smoothes over these disparities. Hence whilst the 10-km bioclimatic classification is sufficient at a national scale to distinguish this mountainous region from other lower lying areas of the country it cannot provide the detail necessary to identify the wide range of climatic conditions experienced. The greater extremes of climate detailed at the 5-km resolution mean that even within the baseline data there are some squares that lie beyond the “experience” of the Baseline98 classification, so that it does not provide an accurate pattern of bioclimate across all parts of the study region. It is therefore difficult to say how sensitive parts of this region are to the climate change scenarios. However, for
186 MONARCH 2 Final Report – Chapter 7 ______those squares that are within the Baseline 98 classification, both the 2050s Low and High scenarios result in a significant shift in pattern of climate across the region, so that more squares are associated with upland instead of montane conditions. This is mirrored by the land cover change scenarios, which show a significant decrease in suitable climate space for the montane habitat under the 2050s High scenario and the dwarf shrub land cover (from LCM2000) loses all suitable climate space in the study area under this scenario. In contrast, neutral grassland shows a large increase in suitable area.
The species and dispersal modelling for the Caledonian pine woodland concluded that:
• The European trained SPECIES model performed very well, although the patchy, upland distribution of F. lugubris led to a lower agreement statistic. • The downscaled SPECIES model performed slightly less well, due to many of the species being associated with broad-leaved woodland, which is found in most squares. This meant that the addition of land cover did not always improve the match between simulated and actual distribution. It also tended to mask the responses to climate change. • The trees showed limited dispersal, although B. pendula was able to disperse further due its shorter time to reproductive maturity. F. lugubris also showed considerable dispersal potential. • The differences between the UKCIP98 and 02 scenarios were highlighted for P. montanus, whose suitable climate space extended into the case study area under the UKCIP98 scenarios, but did not under the UKCIP02 scenarios.
The implications for the species communities within the Caledonian pine woodland concluded that:
• The dominant species are expected to remain within the Black Wood of Rannoch. However climate and disturbance induced changes affecting the canopy and ground-flora will have feedbacks that may alter the composition and structure of the species community.
The species and dispersal modelling for the upland and montane heath concluded that:
• The SPECIES model applied at the European scale led to a good match between the actual and simulated distribution for all species. • The downscaled SPECIES modelling also produced good results, as the addition of land cover restricted the upland heath species’ simulated suitable space in lowland England, although it was too restrictive for C vulgaris. For V. vitis-idaea in particular, it masked some of the response to the climate change scenarios. • The lack of apparent response of species may be a function and/or the method of deriving the finer scale resolution data.
The implications for the species communities for the upland heath concluded that:
• There is little change predicted for the upland heath species communities as the dispersal model does not predict any change in the distribution of their dominant and characterisitic species (V. myrtillus and C. vulgaris and V. vitis-idaea). • However the resolution of the modelling and the small number of species modelled means that more subtle change within the community has not been picked up.
The implications for the species communities for montane heath concluded that:
• As none of the species modelled was predicted to arrive into or leave the montane community, this test area does not provide a good test for the model describing implications for the species communities developed. • However, as communities form mosaics and as large ranges in altitude and thus environmental conditions will occur within 1-km squares, the scale of the modelling, is too coarse to pick up these fine scale changes in community composition.
MONARCH 2 Final Report – Chapter 7 187 ______
• In addition, the species modelled currently grow in a wide range of communities, and this also makes the task of discerning changes more difficult.
The Central Highlands case study area has highlighted the high degree of climate change from baseline as shown by the bioclimate classification and the difficulty of classifying the climate of the 2050s High into the Baseline98 classification. This is also reflected in the high potential for loss in the montane and dwarf shrub land cover classes, but gain in neutral grassland. These changes are not reflected so clearly in the species modelling, partly as a function of few sensitive species confined to these habitats being chosen and also because land cover changes were not incorporated into the modelling process. Nevertheless, there are a number of potential direct and indirect effects of not just climate, but also changes in other environmental variables and management that may impact on the future of the species in the selected habitats.
7.7 References
Baines, D., Sage, R.B. and Baines, M.M. (1994). The implications of red deer grazing to ground vegetation and invertebrate communities of Scottish native pinewoods. Journal of Applied Ecology, 31, 776-783.
Berry, P.M., Vanhinsbergh, D., Viles, H.A., Harrison, P.A., Pearson, R.G., Fuller, R., Butt, N. and Miller, F. (2001). Impacts on terrestrial environments. In Harrison, P.A., Berry, P.M. and Dawson, T.P. (eds.) Climate Change and Nature Conservation in the UK and Ireland: Modelling natural resource responses to climate change (the MONARCH project). UKCIP Technical Report, Oxford, pp43-150.
Biodiversity: UK Steering Group Report (1995). Volume II: Action Plans. H.M.S.O., London. 259pp.
Brooker, R.W., Carlsson, B.Å. and Callaghan, T.V. (2001). Carex bigelowii Torrey ex Schweinitz (C. rigida Good., non Schrank; C. hyperborea Drejer). Biological Flora of the British Isles. Journal of Ecology, 89, 1072-1095.
Burges, A. (1951) The ecology of the Cairngorms. III. The Empetrum-Vaccinium zone. Journal of Ecology, 39, 271-284.
Callender, R.F. (1994). Native pinewoods - the last twenty years (1975-94). In Our pinewood heritage. Proc. conference, Inverness, Aldous, J. R. (ed), Forestry Commission/RSPB/SNH, 40-51.
Edwards, R. and Telfar, M. (2001) Provisional atlas of the aculeate Hymenoptera of Britain and Ireland. Part 3. Bees, Wasps and Ants Recording Society. Huntingdon: Biological Records Centre.
Fowler, S.V. and Macgarvin, M. (1985). The impact of hairy wood ants, Formica lugubris, on the guild structure of herbivorous insects on birch, Betula pubescens. Journal of Animal Ecology, 54, 847-855.
Gimingham, C.H. (1960). Calluna Salisb. Biological Flora of the British Isles. Journal of Ecology, 48, 455-483.
Gordon, C., Woodin, S.J., Alexander, I.J. and Mullins, C.E. (1999). Effects of increased temperature, drought and nitrogen supply on two upland perennials of contrasting functional type: Calluna vulgaris and Pteridium aquilinum. New Phytologist, 142, 243-258.
Hulme, M., Jenkins, G.J., Lu, X., Turnpenny, J. R., Mitchell, T.D., Jones, R.G., Lowe, J., Murphy, J.M., Hassell, D., Boorman P., McDonald, R. & Hill, S. (2002). Climate Change Scenarios for the United Kingdom: The UKCIP02 Scientific Report. Tyndall Centre for Climate Change Research, School of Environmental Sciences, University of East Anglia, Norwich, UK, 120 pp.
188 MONARCH 2 Final Report – Chapter 7 ______
Jones, E.W. (1959). Quercus L. Biological flora of the British Isles. Journal of Ecology, 47, 169-222.
Kinnaird, J.W. (1974). Effect of site conditions on the regeneration of birch (Betula pendula Roth and B. pubescens Ehrh.). Journal of Ecology, 62, 467-472.
Ingram, M. (1958). The ecology of the Cairngorms: IV. The Juncus zone: Juncus trifidus communities. Journal of Ecology, 46, 707-737.
MacKenzie, N.A. and Callander, R.F. (1995). The native woodland resource in the Scottish Highlands: a review of current statistics. Forestry Commission Technical Paper 12, 28 pp.
Metcalfe, G. (1950). The ecology of the Cairngorms: Part II. The mountain Callunetum. Journal of Ecology, 38, 46-74.
Ogawa, Y. (2002). Modelling Species’ Potential Climate Space in Pine Woodland in Response to Climate Change. MSc Dissertation, Environmental Change Institute, University of Oxford.
Palmer, S.C.F., Mitchell, R.J., Truscott, A.-M. and Welch, D. (2004). Regeneration failure in Atlantic oakwoods: the roles of ungulate grazing and invertebrates. Forest Ecology and Management, 192, 251-265.
Palmer, S.C.F. and Truscott, A.-M. (2003a) Browsing by deer on naturally regenerating Scots pine (Pinus sylvestris L.) and its effects on sapling growth. Forest Ecology and Management, 182, 31-47
Palmer, S.C.F. and Truscott, A.-M. (2003b). Seasonal habitat use and browsing by deer in Caledonian pinewoods. Forest Ecology and Management, 174, 149-166.
Punttila, P. (1996). Succession, forest fragmentation, and the distribution of wood ants. Oikos, 75, 291-298.
Rodwell, J.S., Pigott, C.D., Ratcliffe, D.A., Malloch, A.J.C., Birks, H.J.B., Proctor, M.C.F., Shimwell, D.W., Huntley, J.P., Radford, E., Wigginton, M.J and Wilkins, P. (1992), British Plant Communities. Volume 3. Grasslands and montane communities. Cambridge University Press, Cambridge, 540pp
Rodwell, J.S., Pigott, C.D., Ratcliffe, D.A., Malloch,, A.J.C., Birks, H.J.B., Proctor, M.C.F., Shimwell, D.W., Huntley, J.P., Radford, E., Wigginton, M.J. and Wilkins, P. (1991) British plant communities. Volume I. Woodlands and scrub. Cambridge University Press.
Usher, M.B. (1992). Management and diversity of arthropods in Calluna heathland. Biodiversity and Conservation, 1, 63-79.
Watson, A. (1965). A population study of Ptarmigan (Lagopus mutus) in Scotland. Journal of Animal Ecology, 34, 135-172.
Watson, A., Moss, R. and Rae, S. (1998). Population dynamics of Scottish rock ptarmigan cycles. Ecology, 79, 1174-1192.
MONARCH 2 Report – Chapter 8 189 ______
8 Impacts for the Snowdonia case study area
G.J. MASTERS, P.M. BERRY, J.E. HOSSELL, N.L. WARD, S.N. FREEMAN, A.N. BANKS, N. BUTT, H.Q.P. CRICK, P.A. HARRISON AND A. MORRISON
Summary
Snowdonia has a wide range of habitats from low-lying grasslands to high montane communities, which reflect the large altitudinal range within the area. Snowdonia is internationally recognised for its biodiversity and much of its area has been given conservation status, e.g. SSSI or SAC. The large range of Snowdonia’s habitats that are listed as of high conservation concern include upland oak woodland and upland/montane heath; the two selected habitats for this case study. For the upland oak woodland habitat, sessile oak (Quercus petraea) was selected as a dominant tree; common cow-wheat (Melampyrum pratense) as a dominant ground flora plant; and bluebell (Hyacinthoides non-scripta) and pied flycatcher (Ficedula hypoleuca) as flagship species. The species selected for the upland/montane heath habitat were heather (Calluna vulgaris) and bilberry (Vaccinium myrtillus) as dominant upland heath plants, stiff sedge (Carex bigelowii), as a dominant montane plant, and both western gorse (Ulex gallii) and bracken (Pteridium aquilinum) as recruitment species. These species were selected to reflect a continuum of habitat, from upland heath into the montane communities.
The research showed that:
1. Snowdonia is sensitive to climate change, both bioclimatically and in terms of the importance of its habitats and species. However, these sensitivities are not reflected, as a whole, by the results for the species modelled in MONARCH 2.
2. The bioclimatic classification identified Snowdonia as one of the areas where the climate is predicted to show greatest difference between the baseline bioclimate and the future 2050s High climate. The area is predicted to become warmer with slightly wetter winters and significantly drier springs/summers according to the UKCIP98 and 02 scenarios.
3. The downscaled model improved the simulation of suitable space, but tended to lead to its under-prediction, especially in southern England (Q. petraea) and eastern England (M. pratense). None of the selected dominant species in upland and montane heaths will loose suitable climate space.
4. The dispersal model showed that while Q. petraea had some potential to disperse, this was limited by its long time to reproductive maturity. Other factors that constrained species’ dispersal were the availability of suitable new areas (H. non-scripta and M. pratense) and altitude (M. pratense). All the upland/montane species showed a high potential to disperse into the simulated suitable areas and in the case of the recruitment species the issue of invasion needs considering.
5. The modelled loss of F. hypoleuca from some woodland due to restricted climate space is not predicted to significantly impact upon the composition of species communities, according to the Leaver model.
6. The Arriver model suggests that colonisation by P. aquilinum and/or U. gallii will cause large changes in community composition and have far-reaching effects on community composition and structure. The order of colonisation will influence the exact nature of any changes to community composition. If P. aquilinum is first to colonise upland heath areas, then it is possible that strong competition from this species will prevent colonisation by U. gallii.
190 MONARCH 2 Report – Chapter 8 ______
8.1 Introduction
The Snowdonia case study area is centred on the National Park which was established in 1951 and covers 213,191 ha (2,171km2 or 849 miles2) of northwest Wales (Figure 8.1). The wide range of natural and semi-natural habitats are the product of natural and anthropogenic forces and are recognised nationally and internationally through a wide range of designations. This area has a high potential for bioclimatic change (see Chapter 2 and Section 8.3), for example, showing a greater difference between the baseline classification and future climates than much of Britain and Ireland (Chapter 2).
Within the National Park, farming is predominantly pastoral with only 100 ha of arable land. Semi- natural broadleaved woodland covers 8,100 ha (10% of the Welsh resource) of which 6,000-7,000 ha are upland oakwood, with about 1,000 ha of wet woodland dominated by alder, willows and birch and only about 560 ha of upland mixed ashwood. Lowland mixed deciduous woodland covers about 160 ha. Upland heathland covers some 27,000 ha and contributes 34% of the Welsh resource. It includes 16,900 ha of dry acid heath, 3,700 ha of wet heath and 6,400 ha of upland grassland/heathland mosaic (Jones et al., 2003). Snowdonia is of major nature conservation and landscape interest with approximately 90 SSSIs and 17 National Nature Reserves and a range of habitats, including alpine/ sub-alpine heaths, acidic screes and boreal grasslands, that are recognised to be internationally important (Rhind and Jones, 2003).
Although Snowdonia has a rich variety and abundance of wildlife, national trends of reduced species abundance and habitat degradation are also evident here. With such complicated geology and habitat variety, it is not surprising to learn that more National Nature Reserves have been designated in Snowdonia than in any other comparable area in Britain. Approximately 20% of the Park is specially designated or declared under UK and European law, to protect characteristic features of its wildlife. Some of the oakwoods are internationally important because of their unusual communities of mosses and liverworts.
Recent concerns in terms of conserving Snowdonia’s biodiversity tend to have focused on overgrazing, land use change and the potential threat of acidification and climate change to some of the more vulnerable habitats (particularly upland and montane communities). This chapter reports on
Figure 8.1: Boundary and land cover of the Snowdonia Test Area. © CCW.
MONARCH 2 Report – Chapter 8 191 ______the consequences of climate change for conservation within Snowdonia National Park. The integrated modelling approach (see Chapter 1, Figure 1.1) adopted by MONARCH 2 was tested on two habitats, upland oak woodland and upland heathland/ montane, both of which are of high conservation concern within the Park.
8.2 Bioclimatic classification
The Snowdonia case study area covers 115 5-km grid squares centred on the Snowdonia National Park and adjacent coastline. In the 10-km bioclimatic classification (Baseline98), 11 classes were represented. These classes represented a mix of high mountainous and upland areas with cool and wet climatic conditions (See Chapter 2, Tables 2.7 and 2.9, for details of the Baseline98 bioclimatic and conservation characteristics).
8.2.1 Baseline classification
Under the UKCIP02 5-km baseline classification (Baseline02), ten of these 11 classes are still present but a further four are also defined (classes 12, 15, 22 and 23). Class 15 represents a cold, high mountain class most typically associated with the Central Highlands. In Snowdonia, it is found around the Cambrian Mountains to the south east of the study area. The other additional classes are lower lying and warmer classes with moderate winter rainfall but relatively dry summer conditions. These classes are found in the shadow of the hills on the west of the study area (especially class 22) and towards the coast in the south and east (especially class 12). Figure 8.2 shows the relationship of the Baseline02 classes to those of the Baseline98 classification.
The balance of the classes has changed under the different climate datasets (Table 8.1). Under the coarser resolution Baseline98 data the greatest proportion of the study area was represented by class 5, a moderately high altitude class (~300 m) with a cool short growing season, high winter wind speeds, damp summer and wet winter conditions. The Baseline02 data provides a more detailed pattern of climatic conditions, which shows an expansion of classes 11 (milder conditions at lower altitudes but still with relatively high rainfall totals) and 16 (mild still but with higher altitudes of ~300 m and high rainfall conditions). Squares within Class 9 under the Baseline98 are reassigned to class 22, which has very similar conditions under the Baseline98 dataset (see Table 2.15, Chapter 2).
Table 8.2 shows the average climatic conditions of all squares within the study area. The Baseline98 indicates a slightly milder climate than the Baseline02, with a longer but less intense growing season, higher absolute minimum temperatures and slightly lower winter wind speeds. As with Hampshire (Chapter 6), this may be a reflection of the fact that the interpolation method used to derive the Baseline02 dataset takes account of distance from the coast and hence produces a milder, wetter, maritime climate.
As with the Central Highlands case study (Chapter 7), the climatic conditions of the Baseline02 squares do not sit wholly within the statistical boundaries defined by the Baseline98 classification. Figure 8.3 shows the two baseline classifications and indicates those squares that fall outside of the Baseline98 classification according to their squared Mahalanobis distance. The cut off for the squared Mahalanobis distance taken is relatively generous (>1347), since it represents the greatest distance between the centres of any two classes plus the distance of the furthest square from its class in the Baseline98 classification. In the Baseline02, eight of the 115 squares fall beyond this limit. All but one, are in the far north of the study area, along the line of the Carnedd mountains, north-east of Snowdon. The eighth square is over Arenig Fawr in the central east of the study region. A number of these squares had been allocated to class 3 but they show a short growing season and intensity and a very high winter wind speed. They appear to have been allocated to class 3 due to high wind speed but have relatively neutral temperature and rainfall characteristics.
192 MONARCH 2 Report – Chapter 8 ______
Figure 8.2: Distribution of bioclimate classes. Figure 8.3: Distribution of bioclimate classes as UKCIP Baseline02 (in the centre of a 5 km per Figure 8.2, showing Baseline02 bioclimate grid cell) is compared with Baseline98 (at the grid cells that fall outside of the Baseline98 intersection of grid cells). classification according to the squared Mahalanobis distance.
Table 8.1: Percentage of grid squares within each of the classes of the bioclimatic classification for the Baseline98 and Baseline02 and for the 2050s Low and High emission scenario data. Scenario Class 1 3 4 5 8 9 11 12 13 14 15 16 18 19 21 22 23 Baseline98 7.9 2.6 5.3 26.3 7.9 13.2 15.8 7.9 2.6 5.3 5.3 Baseline02 2.6 3.5 2.6 4.4 7.0 2.6 29.6 6.1 4.4 17.4 1.7 4.4 13.0 0.9 2050s Low 3.5 0.9 12.2 8.7 2.6 38.3 1.7 7.0 3.5 1.7 3.5 6.1 2.6 7.8 2050s High 6.09 10.4 19.1 26.1 5.2 10.4 6.1 15.7 0.9
Table 8.2: Comparison of mean values of key variables averaged over the Snowdonia case study area from each of the data sets. The variables selected represent those related to the first four components of the PCA in the derivation of the bioclimatic classification (See Chapter 2 for more details). Spring Growing degree January wind July Potential precipitation days >5°C speed Evapotranspiration Total (mm) Mean (m/s) Mean (mm/day) Baseline98 213.77 1324.11 7.83 3.18 Baseline02 219.16 1320.86 7.18 3.10 2050s Low 195.99 1703.29 7.54 3.29 2050s High 182.36 1949.60 7.63 3.42
MONARCH 2 Report – Chapter 8 193 ______
Table 8.3: Average climate conditions for squares in each class for key variables under the Baseline98 (local: average of squares in the study area; national: average for the class across Great Britain and Ireland), Baseline02 and 2050s Low and High emissions scenarios. Spring rainfall January wind speed July PET Number of Class Scenario GDD>5°C total (mm) mean (m/s) mean (mm/day) squares Baseline98 National 172.1 1241.1 7.3 2.9 212 Baseline98 Local 190.6 1440.8 7.7 3.3 3 1 Baseline02 195.0 1208.9 9.1 3.1 3 2050s Low 212.8 1360.4 10.8 3.5 4 Baseline98 National 141.0 1393.6 8.1 2.7 113 Baseline02 217.7 1050.2 13.1 3.2 4 3 2050s Low 225.7 1479.3 13.1 3.7 1 2050s High 194.8 1444.2 13.1 3.6 7 Baseline98 National 157.3 1590.6 5.8 2.5 300 Baseline98 Local 164.1 1508.7 7.7 3.4 1 4 Baseline02 166.4 1757.4 5.8 3.0 3 2050s Low 175.1 1907.0 7.2 3.2 14 2050s High 181.4 2036.5 6.9 3.4 12 Baseline98 National 230.3 1034.8 7.4 2.8 77 5 Baseline98 Local 233.1 1139.4 7.8 3 8 Baseline02 246.7 1017.1 8.8 3.3 5 Baseline98 National 125.2 1773.6 6.8 2.9 173 Baseline98 Local 155 1651 7.9 3.5 4 8 Baseline02 131.7 1861.2 6.2 2.9 8 2050s Low 127.3 2239.4 6.6 3.2 10 2050s High 138 2344.6 7.2 3.3 22 Baseline98 National 169.3 1125.7 6.1 3 139 Baseline98 Local 177.4 1153.5 7.5 3 6 9 Baseline02 178.9 1068 7.4 3.3 3 2050s Low 165.2 1323.9 8.7 3.4 3 Baseline98 National 213.8 1300.6 7.3 2.9 82 Baseline98 Local 228.1 1367.2 7.8 3.2 8 11 Baseline02 214.8 1397.6 6.1 3.1 34 2050s Low 205.8 1637.4 6.9 3.3 44 2050s High 191.2 1845.8 7.1 3.5 30 Baseline98 National 144.5 1424.6 5.2 2.6 425 12 Baseline02 214.8 1397.6 6.1 3.1 34 2050s Low 205.8 1637.4 6.9 3.3 44 Baseline98 National 222.7 1467.2 6.3 2.4 47 13 2050s Low 269 1386.6 10.4 3.2 8 2050s High 247.8 1642 11.2 3.4 6 Baseline98 National 99.7 1694.2 6.4 3.7 251 14 Baseline98 Local 130.6 1662.2 7.7 3.5 3 2050s Low 164.6 1431.3 11.9 3.2 4 Baseline98 National 197.8 861.9 6.5 2.7 76 15 Baseline02 223.8 872.8 9.7 3.1 5 2050s Low 196.8 1109.2 11.1 3.4 2 Baseline98 National 245.9 1295.2 6.8 3.2 20 Baseline98 Local 250.3 1280.2 7.7 3.1 1 16 Baseline02 270.7 1209.6 7.8 3.1 20 2050s Low 290.2 1794.2 6 3.2 4 2050s High 268.4 1806.1 7.5 3.4 12 Baseline98 National 178.9 1817.6 6.7 2.5 57 18 2050s Low 144.3 2269.5 5.1 3 7 2050s High 135.7 2531.2 5 3.2 7 19 Baseline98 National 255.7 1151.4 8.3 2.8 18
194 MONARCH 2 Report – Chapter 8 ______
Spring rainfall January wind speed July PET Number of Class Scenario GDD>5°C total (mm) mean (m/s) mean (mm/day) squares Baseline02 300.7 1411.3 5.2 3.2 2 Baseline98 National 332.7 1019.9 8.1 2.8 11 Baseline98 Local 342.9 1160.2 8.5 3 4 21 Baseline02 406.8 1137.4 8.1 3.1 5 2050s Low 395.8 1335.4 9.5 3.3 3 Baseline98 National 108.9 1041.4 8.2 2.7 101 Baseline02 166.9 1322.5 6 3.1 15 22 2050s Low 147.9 1622.7 6.4 3.3 9 2050s High 144.5 1767.6 7.1 3.5 18 Baseline98 National 276.3 1366.8 6.6 2.4 20 23 Baseline02 355.8 692 14.7 3.2 1 2050s High 401.7 1641.3 10.9 3.5 1
8.2.2 Climate change scenarios – 2050s Low and High
The pattern of bioclimatic classes for the future emissions scenarios, 2050s Low and High, are shown in Figures 8.4 and 8.5. The 2050s Low emission scenario (Figure 8.4) results in much warmer conditions over the study area, with slightly wetter winters and slightly drier spring and summer conditions; wind speeds are little altered. The effect of these changes is to shift the pattern of classes present further into class 11 and away from the cooler classes of 16 and 22. The warm classes 4, 13 and 14 also appear in the lower lying areas to the south-west of Snowdon and along the western edge of the case study area.
Examining the squared Mahalanobis distances for the 2050s Low emissions classification, most of the squares actually fall beyond the limits for the Baseline98 classification; only 21 squares remaining within the boundaries (represented by the first shading category on Figure 8.4). This suggests that new classes, representing a climate not currently experienced in Britain and Ireland may be expected within this area by under this scenario. The remaining squares are generally the warmer climate squares towards the south east of the study area (classes 4, 8, and 11).
Figure 8.4: Bioclimatic classification for Figure 8.5: Bioclimatic classification for the UKCIP 2050s Low emission scenario. the UKCIP 2050s High emission scenario.
Under the 2050s High scenario (Figure 8.5), a similar pattern of classes exists and a similar number of squares fall outside of the classification. It is interesting to note that some squares, which lie beyond the Baseline98 classification under the 2050s Low emission scenario come back into the classification under the 2050s High scenario as they are classified into another class (e.g. between Rhinnog Fawr and the Mawddach estuary). This reflects a shift inland of the class 8, which is a coastal class under the Baseline98 classification.
MONARCH 2 Report – Chapter 8 195 ______
Under the 2050s High scenario (Figure 8.5), a similar pattern of classes exists and a similar number of squares fall outside of the classification. It is interesting to note that some squares, which lie beyond the Baseline98 classification under the 2050s Low emission scenario come back into the classification under the 2050s High scenario as they are classified into another class (e.g. between Rhinnog Fawr and the Mawddach estuary). This reflects a shift inland of the class 8, which is a coastal class under the Baseline98 classification.
8.3 Land cover changes
Figures 8.6 and 8.7 show the pattern of distribution of the land cover classes across the study area from the 1-km dataset using a 5% cutoff value for distribution from the statistical model for bog and acid grassland. Bog is not well represented in the study area and although at a national level the model is moderately good at representing the land cover, it does not capture the land cover pattern well within the case study area. By contrast, acid grassland is one of the principal land covers in the study area. The model provides a reasonable pattern (moderately good fit – see Chapter 3) of climate areas suitable for acid grassland, but over predicts in areas just beyond the study area, possibly reflecting the loss of this land cover to improved grassland in lowland areas. Figure 8.8 shows the change to bracken distribution as simulated by the changes to climate envelope of this land cover type. The results suggest that the wetter winter conditions of the 2050s may not favour this land cover type, which shows a reduction particularly in the south and east of the case study area.
Open and dense dwarf shrub are the main land cover types that represent the vegetation in the montane areas of the region. The montane land cover itself is not represented within the Land Cover Map data for the area at all, reflecting the relatively low altitude and significant maritime influence on the mountains in Snowdonia. The dwarf shrub land cover shows a small loss in potential climate space by the 2050s High scenario (Figure 8.9) particularly around the lower slopes of the upland areas, along the main valley in the north of the region and along the coast. Neutral grassland shows a similar pattern of loss but the suitable climate space for calcareous grassland, which is restricted in the baseline to the fringes around the main upland area, decreases greatly by the 2050s High scenario to leave just one coastal square where future conditions may be suitable. Figure 8.6: Distribution of bog and deep peat Figure 8.7: Distribution of acid grassland across the case study area from the 1-km square across the case study area from the 1 km land cover dataset. square land cover dataset.
196 MONARCH 2 Report – Chapter 8 ______
Figure 8.8: Changing bracken land cover type under UKCIP02 climate.
Acid grassland land cover shows a very small loss in potential climate space by the 2050s High (Figure 8.9). By contrast, bog and deep peat land cover and fen, marsh and swamp show an extension by the 2050s High scenario that matches the decline in bracken land cover.
8.4 Conservation monitoring in Snowdonia
Figure 8.10 shows the 48 SSSIs within the case study area, which are notified for habitats that contain the modelled species. These range in size from 1 ha to more than 4,000 ha. Figure 8.10 also shows the Mahalanobis distance for the square under the 2050s High scenario. This indicates those areas that lie furthest from the baseline climate classification. These climatically sensitive areas lie largely in the north of the study region along the Carnedd range and around Arenig Fawr in the east. The former is designated for the four key habitat types that cover the species modelled and with its relatively long history as an SSSI (first notified in 1971), it may have a reasonable record of monitoring data. It would seem to represent a useful area in which to monitor, to detect the effects of climate change on either climate variables or habitats and species.
MONARCH 2 Report – Chapter 8 197 ______
Figure 8.9: Changing land cover types under the UKCIP02 climate scenarios.
Figure 8.10: Location of the SSSIs and the squared Mahalanobis distance for the square under the 2050s High scenario.
198 MONARCH 2 Report – Chapter 8 ______
8.5 Upland oak woodland
It has been suggested that due to climate change upland oakwoods could suffer from dieback due to soil moisture stress, changes in timing of budburst and changes in species’ composition (Hossell et al., 2000). The upland oak woodland habitat was investigated under MONARCH 1, when it was found that of the three species modelled using the UKCIP98 scenarios, hard fern (Blechnum spicant), would not lose suitable climate space, while hay-scented buckler fern (Dryopteris aemula) would do so from the southern part of its range (although Snowdonia should just remain within its suitable climate space) and the third, serrated wintergreen (Orthilia secunda), would lose suitable climate space in upland areas, although it does not occur in Snowdonia (Berry et al., 2001).
Following the selection protocol, the species chosen for modelling in MONARCH 2 were sessile oak (Quercus petraea), a dominant tree, common cow-wheat (Melampyrum pratense), a dominant ground flora plant, while bluebell (Hyacinthoides non-scripta) and pied flycatcher (Ficedula hypoleuca) were chosen as flagship species.
8.5.1 Species modelling
The modelled climate space for the species selected in Snowdonia all showed very good discrimination ability at the European scale; with an independent AUC statistic greater than 0.9. The maximum Kappa statistic is slightly lower, with the plants having excellent agreement. Ficedula hypoleuca has very good agreement and this is because it has a patchy distribution in western Europe, being essentially a northern breeding bird in Europe, with some populations also in Spain and Southern France (Figure 8.11). In Britain, Wales is its stronghold, especially where there is an abundance of upland oak woodland, which it favours for nesting. It does, however, occur at lower densities in parts of northern and south-western England (Figure 8.12).
As with the other case study areas, the downscaled model training statistics were lower (Table 3.5), for reasons discussed in Chapter 3, although the AUC is effectively 0.9 for Hyacinthoides non-scripta and the Best Attainable Match for F. hypoleuca is 92%, this is a good result given its patchy British distribution (Figure 8.12). The lower values for the other two species may be partly a function of broad-leaved woodland occurring in most squares, thus making it difficult for the model to discriminate on the basis of land cover. A visual comparison of the outputs of the two stages of the SPECIES model for Q. petraea show that the introduction of land cover leads to some of the absences in southern and eastern England being picked up, although too much of the former is simulated as unsuitable. In the case of M. pratense, it is the eastern part of England where too much is simulated as unsuitable (Figure 8.13). For H. non-scripta the addition of land cover leads to some of the absences e.g. in East Anglia and Southern Uplands being picked up, and again there is a slight under-simulation compared to the observed distribution. Under the climate change and land cover scenarios, M. pratense was the only species to show an alteration in suitable space, as parts of southern England become viable (Figure 8.14), but the land cover overrides the loss of climate space in parts of eastern England (compare Figures 8.13 and 8.14). In contrast, F. hypoleuca shows a small loss under the 2050s High scenario (Figure 8.15). The downscaled model results incorporating land cover change for Snowdonia for M. pratense show that by the 2020s scenarios there is an increase in suitable area (Figure 8.16).
The unavailability of species’ distributions at the 1-km resolution meant that the distributions had to be derived. In the case of Q. petraea the Phase 1 survey data was used and it was assumed that Q. petraea was present in every woodland grid square. This would give a small overestimate of its distribution. A similar approach was adopted for M. pratense and H. non-scripta, but any 1-km squares that fell within a 10-km grid square in which it was absent in the New Atlas of the British and Irish Flora (Preston et al., 2002) were omitted. This method could lead to an overestimation of their distribution. This approach seemed to give reasonable results given the
MONARCH 2 Report – Chapter 8 199 ______
Figure 8.11: The European observed distribution (a) and simulated climate space (b) for Ficedula hypoleuca.
(a) (b)
Figure 8.12: The British observed distribution (a) and simulated climate space (b) of Ficedula hypoleuca. (a) (b)
woodland association of the species, but M. pratense can also be associated with other habitats, such as upland heath and thus overall its derived distribution may have been under-estimated.
The dispersal model for Q. petraea shows a limited opportunity for dispersal into adjacent suitable squares throughout most of the area or to disperse to higher elevations (Figure 8.17). The coastal area between Barmouth and Harlech appears to become unsuitable under the 2050s scenarios and thus the species could be lost from some squares. M. pratense shows potential to spread, but is constrained at altitude in the northern half of the Park (Figure 8.18). There is no opportunity for dispersal from the coastal populations due to low suitability of the land cover. The dispersal model shows a small spread for H. non-scripta, partly due to the limited dispersal potential of the species and partly as a consequence of there being little suitable space that is not already colonised. F. hypoleuca remains well-distributed in Snowdonia even under the 2050s high scenario (Figure 8.19). However, even here there are substantial losses of suitable space throughout the area, compared to its present-day occurrence, concurrent with the loss of suitable space further east.
200 MONARCH 2 Report – Chapter 8 ______
Figure 8.13: Model outputs for Melampyrum pratense: (a) climate suitability surface from the European-trained network, (b) observed 10km distribution, (c) climate and land cover suitability surface from the downscaled model, and (d) presence/absence surface from the downscaled model. (a) (b)
(c) (d)
MONARCH 2 Report – Chapter 8 201 ______
Figure 8.14: Presence/absence surfaces for Melampyrum pratense from the downscaled model, under the UKCIP02 climate change scenarios.
2020s Low scenario 2020s High scenario
2050s Low scenario 2050s High scenario
202 MONARCH 2 Report – Chapter 8 ______
Figure 8.15: Presence/absence surfaces for Ficedula hypoleuca from the downscaled model, under the UKCIP02 climate change scenarios.
2020s Low scenario 2020s High scenario
2050s Low scenario 2050s High scenario
MONARCH 2 Report – Chapter 8 203 ______
Figure 8.16: Suitability surfaces for Melampyrum pratense from the downscaled model, under combined UKCIP02 climate and land cover change scenarios.
Unsuitable 95% AUC threshold 90% AUC threshold Optimum AUC treshold
Baseline
Unsuitable Unsuitable 95% AUC threshold 95% AUC threshold 90% AUC threshold 90% AUC threshold Optimum AUC treshold Optimum AUC treshold
2020s Low climate and 2020s High climate and land cover change scenario land cover change scenario
Unsuitable Unsuitable 95% AUC threshold 95% AUC threshold 90% AUC threshold 90% AUC threshold Optimum AUC treshold Optimum AUC treshold
2050s Low climate and 2050s High climate and land cover change scenario land cover change scenario
204 MONARCH 2 Report – Chapter 8 ______
Figure 8.17: Quercus petraea dispersal model outputs.
Observed 1km distribution
2020s Low scenario 2020s High scenario
2050s Low scenario 2050s High scenario
MONARCH 2 Report – Chapter 8 205 ______
Figure 8.18: Melampyrum pratense dispersal model outputs.
Observed 1km distribution
2020s Low scenario 2020s High scenario
2050s Low scenario 2050s High scenario
206 MONARCH 2 Report – Chapter 8 ______
Figure 8.19: Ficedula hypoleuca presence/absence surfaces under UKCIP02 climate change scenarios and land cover.
Observed 1km distribution
2020s Low scenario 2020s High scenario
2050s Low scenario 2050s High scenario
MONARCH 2 Report – Chapter 8 207 ______
These results support those of MONARCH 1 (Berry et al., 2001; Berry et al., 2003) that the dominant, Q. petraea, is not vulnerable to climate change, except along the coast, but that some of the herb species, such as H. non-scripta, could start to become marginal in Snowdonia, especially if climate change is combined with other environmental pressures.
8.5.2 Implications for the species composition of upland oakwoods
Model outputs suggest that suitable climate space will remain for all the species selected to represent this habitat. In terms of community composition the models predict that the plant species (Q. petraea, H. non-scripta and M. pratense), will continue to exist together. In contrast, Ficedula hypoleuca may in the future have more restricted suitable space within the park in comparison to the existing distribution (Table 8.4).
Table 8.4: Summary of the dispersal model results. Species Category Predicted space change
Quercus petraea sessile oak Dominant No change Hyacinthoides non-scripta bluebell Flagship No change Melampyrum pratense common cow wheat Dominant No change Ficedula hypoleuca pied flycatcher Flagship Decrease
8.5.2.1 Ficedula hypoleuca (pied flycatcher)
Ficedula hypoleuca is a characteristic species of the upland oak woodlands of Snowdonia, and it inhabits the majority of suitable woodlands within the National Park, even where these are small in area. In addition, some of the bigger woodlands such as Coed Aber, support large breeding populations (Snowdonia National Park Authority, 1999). The model outputs for F. hypoleuca suggest that it may become locally extinct from some currently suitable woodland areas.
The conceptual Leaver model methodology (Chapter 4) was used to assess the impact of loss of F. hypoleuca on these woodlands. As a first step, an interaction web showing the main interactions between F. hypoleuca and other species in deciduous woodlands was constructed (Figure 8.20). This indicates that F. hypoleuca only has a moderate number of interactions within the upland oak woodland, suggesting that this species cannot be considered as a dominant within this habitat. However, because it is present in relatively high numbers and can be the most abundant breeding bird species in these woodlands (Mead, 2000), this species is also definitely not a rare species and thus should probably be viewed as sub-dominant, within our classification scheme.
Figure 8.20 further illustrates that most, if not all the interactions, are not specific one-to-one interactions, whereby F. hypoleuca is totally dependent on another species (e.g. for food), or another species is entirely dependent on F. hypoleuca. This is not surprising; F. hypoleuca is a migratory species and only forms a component of the upland oak woodland communities during the summer months. The main predators of F. hypoleuca chicks and adults during the nesting period are polyphagous, and have many alternative sources of potential prey and F. hypoleuca feeds on a great variety of insect prey, from many microhabitats, although caterpillars can be a very important component of the nestling diet.
In addition, there are bird species that have similar nest site and food requirements and compete with F. hypoleuca (Figure 8.20), such as Parus sp. (particularly great tit, P. major), and redstart (Phoenicurus phoenicurus). A study in a woodland in South Wales, demonstrated that interspecific
208 MONARCH 2 Report – Chapter 8 ______
Figure 8.20: The interaction web for F. hypoleuca, displaying the main interactions. Grey shading is used to indicate species that do not currently occur within the upland oak woodland of the test study area. Blue lines represent competitive interactions; red lines predator-prey interactions; and black lines all other interactions. The thickness of the arrows gives an indication of interaction strength.
Parus sp. Phoenicurus Muscicapa striata Dead leaves (for nest sites and phoenicurus Twigs (for food) bark (e.g. Betula) (often Quercus Animal hair food) (for food) (e.g Calluna) or Betula)
Sitta europaea (for nest sites) Dry grass Root fibres moss
Ficedula albicollis Nesting materials
Competitors adults ants Ficedula larvae hypoleuca Sphecid wasps Mustela erminea chicks Hymenoptera
Mustela nivalis Coleoptera
Dendrocopos sp. Martes martes Lepidoptera Predators (larvae)
Diptera Homoptera
Old woodpecker Holes in trees or tit nesting sites for nesting Aranea Other interactions Diet
Figure 8.21: The pathway through the Leaver model predicted to be followed by F. hypoleuca (shown in orange). See Chapter 4 for further details.
Community collapse New Community
Major effect Dominant Species strong links with other species Reassembly driver of belowground processes Yes Keystone species?
No Natural change, e.g. succession Sub-dominant Leaver Species No Functional Type Redundancy? Yes
Rare Species Negligible effect weak interaction strength Existing Community often dependent on biotic interactions
MONARCH 2 Report – Chapter 8 209 ______competition for territories and more importantly for feeding sites occurred between F. hypoleuca and P. phoenicurus. Both species feed from the ground, in the air and from the leaves of terminal twigs, suggesting that they have overlapping feeding niches (Edington and Edington, 1972). The nests, and in particular the material used to build them are also similar, for these two species. The diet of Parus sp. may also be very similar to that of F. hypoleuca (Lundberg and Alatalo, 1992), although during the breeding season, the tits took a much higher proportion of their diet from the leaves of trees, and a smaller proportion of airborne and ground dwelling prey in comparison to F. hypoleuca, in the study in South Wales (Edington and Edington, 1992). On the basis of the preceding discussion it must be concluded that F. hypoleuca fulfils similar roles to other species, notably P. phoenicurus, in the upland oak woodland, and cannot therefore be viewed as a novel functional type within the community.
As F. hypoleuca is being classed as a sub-dominant species, and is not a sole representative of a functional group, the Leaver model indicates that the loss of this species from some woodlands within the Snowdonia National Park will not greatly affect the species composition within those woodlands (Figure 8.21).
Although the dispersal model predicted that this species would remain widely distributed within Snowdonia under all scenarios, changes in abundance may still occur, as the insectivorous F. hypoleuca may be affected by climate driven alterations in peak chick food abundance. As the interaction web illustrates (Figure 8.20), F. hypoleuca feed nestlings on spiders, Lepidoptera (mostly caterpillars), Diptera, Hymenoptera (mostly larvae) and Coleoptera (Lundberg and Alatalo, 1992). Caterpillars make up between 15-65% of nestling diet in deciduous forests (Lundberg and Alatalo, 1992) and therefore the breeding success of F. hypoleuca is influenced by temporal variation in peak caterpillar abundance (e.g. winter moth abundance) and tree phenology. Breeding success may be adversely affected in a similar way to P. major (Buse and Good, 1996; Buse et al., 1999; Dury et al., 1998; Visser et al., 1998).
European-wide variations in F. hypoleuca laying date and clutch size have been examined in relation to changes in the winter North Atlantic Oscillation (NAO) index, which is broadly correlated with climatic fluctuations (Sanz, 2003). It confirms that following wetter and warmer winters, egg laying commenced earlier, although, contrary to expectation, smaller clutches were produced. Laying of smaller clutches reduces the period of egg laying and thus shortens the time between first laying and egg hatch (Sanz, 2003), possibly maintaining greater synchrony with food sources.
In the Snowdonia National Park, long-term data sets collected from Coed Aber NNR and Coedydd Maentwrog NNR indicate that the mean date of first egg laying has advanced, during the period 1976- 2001, by 0.31 to 0.33 days per year. Furthermore this advancement in laying is strongly correlated with temperature, such that for every one-degree rise in temperature, the date of mean egg laying is approximately 3 days earlier (Sparks et al., in press). Egg laying has also been reported to take place earlier in Germany, Russia and the Netherlands (e.g. F. hypoleuca are laying 10 days earlier than 20 years ago). However, it is also suggested that this advancement in the Netherlands, is not sufficient to track climatic changes and that a significant proportion of the population is now laying too late to benefit from the flush in insect abundance, and thus do worse than earlier laying pairs. This has resulted in increasing selection for earlier laying date (Both and Visser, 2001). Advancement in laying date has led to a shortening in the time gap between arrival at the breeding grounds and first laying, which is now reduced to only 5 days (Drent et al., 2003). This species was able to respond in the Netherlands, in this way because it used to arrive at the breeding ground significantly in advance of the optimal laying date.
The timing of arrival at the breeding area has not altered over the last 20 years, and arrival time is already constraining the ability of the Dutch population to respond adequately to the earlier occurrence of important phenological events mediated through increasing spring temperatures. However earlier arrival has been reported for some locations within Europe (Sparks et al., in press). It is unclear at present whether the F. hypoleuca population in the Snowdonia National Park is similarly
210 MONARCH 2 Report – Chapter 8 ______affected as the Dutch population. Although bird observatories record only small numbers of migrating F. hypoleuca, there is some evidence from Portland, which suggests that the Welsh populations have advanced the date of their spring migration. This advancement has been at a similar rate to the advancement in mean first egg date (Sparks et al., in press).
Although the Welsh population is expected to continue to arrive earlier and nest earlier, during the next 100 years, there are indications that this advancement in key phenological events is slower than the climate induced advancement in vegetation growth and invertebrate emergence (Sparks et al., in press). It is unclear whether F. hypoleuca may be able to respond to further climate change by increasing the speed of migration, decreasing the length of migration or migrating further northwards to areas in which key phenological events will still take place later (Coppack and Both, 2002).
Both the slight lack of synchrony between hatching and peak food abundance, and the reduction in clutch size, will reduce the population of F. hypoleuca in the Park. As a consequence of arriving relatively late in the oak woodland, this species may face increased competition from migratory species that are able to adjust to changing climates better, as well as with resident species. Population sizes of resident species may increase if milder winters enhance survival (Both and Visser, 2001).
In areas where natural cavities are in short supply, F. hypoleuca may have to compete with other species such as Sitta europaea, Parus sp. and P. phoenicurus for suitable nesting sites (Figure 8.20). Parus sp. already start breeding earlier and chose the most suitable nesting sites (Lundberg and Alatalo, 1992), and the problem may become worse if other species also advance their spring breeding relative to that of F. hypoleuca. Good nesting sites are important as they reduce predation of adults and chicks by stoat (Mustela erminea) and weasel (Mustela nivalis) (Figure 8.20) (Lundberg and Alatalo, 1992).
8.5.2.2 Flora species
The dispersal model suggests that Q. petraea will maintain suitable space in most of the area in which it is present today. This lack of response may reflect the fact that the timespan of the predictions is relatively short compared to the life span of individual trees (200-450 years; Jones, 1959). The maintenance of Q. petraea within upland oak woods is important, as it is a highly dominant species within the community (Snowdonia National Park Authority, 1999). Thus, species composition of the community is unlikely to be severely altered by future climatic conditions, although more subtle affects such as changes in abundance, or changes in interactions may occur.
The main direct effects of climate change on Q. petraea are briefly summarised in Box 8.A. The impacts that may have the greatest affect on the functioning of the upland oak woodland ecosystem are changes in canopy shading and the volume of leaf litter production. Increased shading may occur as a result of a possible increase in oak leaf biomass in response to elevated levels of carbon dioxide (Broadmeadow and Jackson, 2000), as well as a possible increase in growing season length due to earlier budbreak, which will mean that shading starts earlier. This earlier budbreak may also have implications for the numerous invertebrates that feed on this species (as discussed for F. hypoleuca), as well as on the ground flora.
MONARCH 2 Report – Chapter 8 211 ______
Box 8.A: Direct effects of climate change on Quercus petraea in Snowdonia National Park.
Carbon dioxide related: • Increased growth of saplings (increase in stem length, stem diameter and oak leaf biomass). • Increased leafiness will increase canopy resistance and may make trees more vulnerable to wind damage and windthrow (Broadmeadow and Jackson, 2000). • Greater production of fine roots and increased lateral root diameter with follow-on effects on root turnover and exudation (Broadmeadow and Jackson, 2000). Temperature related: • Possibility of increased growth as a result of a temperature-driven acceleration of the rate of photosynthesis and of decomposition especially N-mineralisation rates. • Potential increase in growing season length. • Possible changes in frequency of good acorn production. • Greater vulnerability to frosts due to incomplete hardening and accelerated dehardening (Saxe et al., 2001) Drought related: • Reduced height growth in Q. petraea saplings (Broadmeadow and Jackson, 2000). • Reduction in frost resistance in winter following drought (Thomas and Ahlers, 1999).
An increase in shading by Q. petraea may particularly affect H. non-scripta, which is generally recognised as a shade evader (Grabham and Packham, 1983), by reducing the phenological window during which necessary assimilation can take place. However this is assuming that H. non-scripta fails to show a phenological response to climate change. Experiments have suggested that both density (Blackman and Rutter, 1946) and seed production (Knight, 1965) are reduced with shading, and as seed dispersal only occurs over short distances, there may be a gradual loss of bluebell populations under complete, dense canopies.
The hemi-parasitic M. pratense may also be affected by increased shading as is suggested by results from a Norwegian forest, where loss of many vascular species, including M. pratense was related to canopy closure (Nygaard and Ogegard, 1999). Leaves of this species are used either to produce photosynthates or to affect high transpiration rates (Lehtila and Syrjanen, 1995), and increased shading may restrict both.
Low light levels under the oak canopy may influence regeneration and growth of Quercus seedlings and saplings, as they have minimum light level requirements that increase with sapling age (Jones, 1959), and may not be fulfilled under a denser canopy. Furthermore, defoliating insects and oak mildew (Microsphera alphitoides), the main leaf destroyers of saplings were recorded with greater frequency under low light conditions by Jarvis (1964).
As a consequence of a greater leaf biomass, there may be greater quantities of leaf litter on the woodland floor (Quercus petraea contributed 93% of the total litter at Coed Cymerau; Rieley et al., 1979), at least during some seasons, and may affect the ground flora and oak regeneration itself. Litter protects acorns from predation and moisture loss and provides favourable conditions for germination (Shaw, 1968), but H. non-scripta is unaffected by all but exceptionally high amounts of litter (Sydes and Grime, 1981) and M. pratense is currently characteristic of areas with deep litter accumulations (Edwards and Birks, 1986). Therefore, the abundance of these dominant species is unlikely to be detrimentally affected by increased litter.
In contrast to the predicted increase in oak leaf biomass and canopy cover, a greater frequency of gaps in the canopy may potentially occur as a result of climate change. Predictions suggest that whilst mean wind speeds are not expected to change in Snowdonia, gales and storms may become more frequent, leading to increased levels of windthrow, snapping of trees and branches and canopy
212 MONARCH 2 Report – Chapter 8 ______damage. This may be especially the case when increased leafiness due to elevated carbon dioxide levels (Box 8.A) increases canopy resistance and decreases the relative contribution of roots to the total biomass thus making Q. petraea trees more vulnerable to being blown over (Broadmeadow and Jackson, 2000). This will increase disturbance within the woodland and may allow species, other than oak, an opportunity to regenerate and thus alter the composition of the ground flora. Creation of gaps in the canopy will reduce the humidity within the woodland, and may affect the bryophyte fauna. Bryophytes currently form an important component of the woodland floor biomass (at Coed Cymerau NNR, mosses contributed 85-91% of the standing crop of the ground layer; Rieley et al., 1979) and thus play an important role in the habitat. North Wales also presently contains the second highest abundance of Atlantic bryophytes (Ratcliffe, 1968) and these are of high conservation importance. However, it is difficult to predict the impacts of changes in the canopy on the bryophytes as these are likely to be highly species specific, and as microclimatic effects over very small spatial scales play an important part in determining suitable habitat for these plants.
8.5.3 Conclusions for upland oak woodland
• The SPECIES model trained well at the European scale, although patchy distributions led to slightly lower measures of agreement between the obseved distributions and modelled climate space. • The downscaled model improved the simulation of suitable space, but tended to lead to its under- prediction, especially in southern England (Q. petraea) and eastern England (M. pratense). • The dispersal model showed that while Q. petraea had some potential to disperse, this was limited by its long time to reproductive maturity. Other factors that constrained species’ dispersal were the availability of suitable new areas (H. non-scripta and M. pratense) and altitude (M. pratense). • The models predict that the plants will not be adversely affected by climate change, but F. hypoleuca could become restricted in the future by lack of suitable climate space. • The Leaver model indicates that loss of F. hypoleuca from some woodlands is unlikely to have a significant impact on the composition of the community. • In woodlands where F. hypoleuca is predicted to continue to experience suitable climate space, changes in population density may take place, as a result of the slower response of this species to climate change than that of its invertebrate prey. • Climate change may also affect the growth and phenology of the dominant tree species Q. petraea with far reaching consequences for the remainder of the upland oak woodland community. These may be mediated through changes to the ground flora as a consequence of alterations in light intensity, humidity and litter production, as well as through changes to food availability and quality for herbivorous insects, which may also affect higher trophic levels (e.g. F. hypoleuca).
8.6 Montane/upland heath
In MONARCH 1, montane heath was the habitat most sensitive to climate change and thus it would be expected that the species representing this habitat would reflect this. The species chosen for MONARCH 2 were the dominant upland heath plants, heather (Calluna vulgaris) and bilberry (Vaccinium myrtillus); the dominant montane plant, stiff sedge (Carex bigelowii); while western gorse (Ulex gallii) and bracken (Pteridium aquilinum) were chosen as recruitment species. These species were selected to reflect a continum of habitat, from upland heath into the montane communities.
8.6.1 Species modelling
The European scale SPECIES model trained well (Table 3.3), with all species having an AUC greater than 0.9 (very good discrimination). The kappa statistic for C. bigelowii and U. gallii is lower, as it is affected by prevalence and both these species have restricted European distributions (Figures 8.22 and 8.23). Nevertheless, while the simulated climate space is greater than the observed distribution, the general pattern is correct.
MONARCH 2 Report – Chapter 8 213 ______
Figure 8.22: The European observed distribution (a) and simulated climate space (b) for Carex bigelowii.
(a) (b)
Figure 8.23: The European observed distribution (a) and simulated climate space (b) for Ulex gallii.
(a) (b)
The addition of land cover in the downscaled SPECIES model improves the visual match between the observed distribution and simulated suitability space for all the species (Figure 8.24), thus indicating the importance of land cover at this scale to constrain the area of suitability. The downscaled model training statistics again were lower (Table 3.5), although the AUC still is greater than 0.9 for all, except U. gallii. Only V. myrtillus has a Kappa value of more than 0.7 (very good agreement), and two (Calluna vulgaris, and P. aquilinum) have poor agreement. In the case of the former, this could be because the model over-restricts the suitable area in southern and eastern England. U. gallii is the only species to show a loss of suitable climate space under the future climate scenarios (Figure 8.25), indicating its potential sensitivity, but as most of this loss is in areas where currently it does not occur, it cannot be considered threatened. None of the upland heath/montane species show much change under the climate change and land cover change scenarios. Again U. gallii is the most responsive, but the addition of land cover in the downscaled SPECIES model suppresses some of the climate response (Figure 8.26).
The unavailability of species’ distributions at the 1-km resolution meant that the distributions had to be derived, mostly using the Phase 1 survey. The distribution of heathland was used as a surrogate for C. vulgaris, whilst V. myrtillus was assumed to be present in broad habitat classes 1, 6, 8, 10, and 12 and U. gallii in classes 8 and 10. The distribution map for P. aquilinum was derived by combining the
214 MONARCH 2 Report – Chapter 8 ______bracken class for LCM2000 and Phase 1 survey, as there is poor agreement between the two. The C. bigelowii distribution was based on field survey data provided by Alex Turner (CCW). These different methods of constructing the 1-km distributions mean that there is a lack of consistency in deriving the observed distribution inputs for the dispersal model and their accuracy often depends on the degree of association between the species and the habitat classes and the extent to which the species occurs in each suitable habitat grid square. As was noted for upland oak woodland, this could lead to an underestimation of the distribution where the species occurs in other habitat classes and an overestimation where it does not fill its entire suitable habitat. Until a 1-km database becomes available in this area an accurate picture of the distributions for several of the species cannot be gained.
The dispersal model shows that C. vulgaris has the potential to cover all of the National Park, apart from the coastal area to the north and south of Barmouth and V. myrtillus shows a similar response with the model outputs suggesting that it is absent from the western edge of the National Park based on suitability of land cover and that this does not become suitable under the future scenarios. The two recruitment species, U. gallii and P. aquilinum, and in particular the former, have potential to disperse under the future scenarios. In the case of U. gallii, all of Snowdonia, apart from the higher parts, represents suitable space under the future scenarios (Figure 8.27). However, there could be issues of competition with other species. The future potential distribution of P. aquilinum is restricted on the sandy soils of Morfa Harlech, the wetter ground in Migneint and the higher slopes of the mountains (Figure 8.28). These two species have the potential to move into new areas and thus their invasive capacity within the habitat needs to be considered. The montane species, C. bigelowii, is primarily found in the north of the park, but it does occur near Cader Idris, Rhinog Fawr, Aran Fawddwy and Moel Llyfnant. Dispersal modelling suggests that it has the potential to disperse from all these locations under the climate change scenarios to cover quite a large proportion of the higher ground in these areas (Figure 8.29). This result is surprising as currently C. bigelowii is confined to montane heath, but the downscaled SPECIES model suggests a wider suitability surface based on land cover classes. Dispersal does not seem to be a limiting factor for these species fulfilling their suitable space, with even U. gallii, with its patchy distribution, having the same potential.
None of the species modelled for upland/montane heath are seen to lose suitable space in Snowdonia in the future and thus they do not reflect the identified bioclimate sensitivity (Chapter 2). It is possible this is a function of the downscaled SPECIES model being dominated by the land cover or that more sensitive species should have been chosen to represent this habitat. The modelling does, however, confirm the potential recruitment of U. gallii and P. aquilinum and the implications of this need further exploration, especially through their effect on the habitats.
8.6.2 Implications for the species composition of upland heath
The dispersal model outputs indicate that the two dominant upland heath species modelled will continue to encounter suitable space within areas that support upland heath at present (Table 8.5). The following section will briefly discuss possible direct impacts of climate change on these two species.
In addition, the dispersal outputs indicate that two new species, Pteridium aquilinum and Ulex gallii will also find suitable climate space in these areas (Table 8.5). P. aquilinum is already a major conservation problem in grassland and heaths at lower altitudes within the Snowdonia National Park. Therefore it is important to determine the possible impacts of P. aquilinum colonisation on community composition within these areas. Similarly, U. gallii colonisation may also have impacts on community composition. The severity and possible outcomes, of these impacts will be assessed using the conceptual Arriver model (see Chapter 4). Unfortunately climate change is not the only pressure on upland heaths within the Snowdonia National Park, and other stresses may interact with climate change and alter the outcomes predicted. Of particular importance are nitrogen deposition, grazing management, and burning.
MONARCH 2 Report – Chapter 8 215 ______
Figure 8.24: Model outputs for Quercus petraea: (a) climate suitability surface from the European- trained network, (b) observed 10km distribution, (c) climate and land cover suitability surface from the downscaled model, and (d) presence/absence surface from the downscaled model.
(a) (b)
(c) (d)
216 MONARCH 2 Report – Chapter 8 ______
Figure 8.25: Presence/absence suitability surfaces for Ulex gallii from the European-trained SPECIES model, under the UKCIP02 scenarios.
Baseline
2020s Low scenario 2020s High scenario
2050s Low scenario 2050s High scenario
MONARCH 2 Report – Chapter 8 217 ______
Figure 8.26: Model outputs for Ulex gallii: (a) climate suitability surface from the European-trained network, (b) observed 10km distribution, (c) presence/absence surface for baseline climate from the downscaled model, and (d) presence/absence surface for the 2050s High scenario from the downscaled model.
(a) (b)
(c) (d)
218 MONARCH 2 Report – Chapter 8 ______
Figure 8.27: Ulex gallii dispersal model outputs.
Observed 1km distribution
2020s Low scenario 2020s High scenario
2050s Low scenario 2050s High scenario
MONARCH 2 Report – Chapter 8 219 ______
Figure 8.28: Pteridium aquilinum dispersal model outputs.
Observed 1km distribution
2020s Low scenario 2020s High scenario
2050s Low scenario 2050s High scenario
220 MONARCH 2 Report – Chapter 8 ______
Figure 8.29: Carex bigelowii dispersal model outputs.
Observed 1km distribution
2020s Low scenario 2020s High scenario
2050s Low scenario 2050s High scenario
MONARCH 2 Report – Chapter 8 221 ______
Table 8.5: Summary of the dispersal model results. Species Category Predicted space change Calluna vulgaris heather Dominant No change Vaccinium myrtillus bilberry Dominant No change Ulex gallii western gorse Recruitment Increase Pteridium aquilinum bracken Recruitment Increase
8.6.2.1 Climate change and the current community
The main impacts of elevated carbon dioxide levels, warming and drought recorded in experiments for C. vulgaris, and to a lesser extent for V. myrtillus, are displayed in Box 8.B. Although the most impressive responses are reported for C. vulgaris plants grown under elevated carbon dioxide concentrations, it should be noted that greater biomass was not accompanied by an increase in nutrient uptake, and therefore these plants generally had lower nutrient concentrations in the leaves (Woodin et al., 1992). Upland heath plants are usually nutrient limited, and therefore this increased growth may not be sustainable (although warming may increase nutrient availability in the future).
The most pronounced effect of climate change on the case study area by the 2050’s is expected to be an increase in temperature. This warming will accelerate rates of nutrient cycling and thus have large impacts on upland heathland plant communities, through changes in nutrient availability, as mentioned above. Also temperature seems to be one of the most important environmental factors affecting insect herbivores (Bale et al., 2002).
All experiments to date, have suggested that C. vulgaris growth is unresponsive to warming (Box 8.B) but this may be a reflection of the short time period over which experiments were run, although earlier shoot growth has been recorded (Gordon et al., 1999a). Earlier bud break has also been reported in the year following on from a drought (Gordon et al., 1999b) (Box 10.2). Although the resultant extended growing season may enhance plant growth, it may also be highly detrimental in years, in which late spring frosts occur, as much new growth may be damaged and as too early dehardening may be induced by mild temperatures (Ogren, 1996). Accelerated dehardening leading to a greater likelihood of frost damage has also been demonstrated in V. myrtillus (Taulavouri, 1997). Furthermore it has been hypothesised that warmer temperatures may interact with increased tissue water concentrations, (due to the slightly higher levels of winter precipitation predicted for Snowdonia) and may exacerbate the amounts of frost damage suffered by plants such as V. myrtillus (Ogren, 1996). Such frost damage will influence the ability of plants like C. vulgaris and V. myrtillus to compete with other potentially less susceptible species within the same community.
The effects of drought will be dependent on the soil drainage characteristics of the heath in question and it is possible that at least in some upland heath sites drought may be a relatively unimportant factor, unless it occurs over prolonged periods. On a C. vulgaris dominated heath at Clocaenog in North Wales, it was only after 60 days of experimentally imposed drought that the soil water content was significantly reduced. Furthermore, after the cessation of the drought treatments it took only four days for soil water levels to return to normal (Jensen et al., 2003).
8.6.2.2 Pteridium aquilinum
Pteridium aquilinum is a herbaceous rhizamatous perennial fern and represents a functional group that is not currently present within upland heath communities. On this basis alone, the Arriver conceptual model predicts (Figure 8.30) that this species is likely to have a large impact on species composition of the community. In addition, the interaction web produced for this species suggests that P. aquilinum supports interactions to many species that will not currently be present within the community, as they are not associated with the plant species growing here (Figure 8.31). This again, suggests that P. aquilinum has the potential to have a large impact on the ecosystem.
222 MONARCH 2 Report – Chapter 8 ______
Box 8.B: The reported impacts of elevated carbon dioxide concentrations, warming and drought on the two modelled dominant plant species.
Vaccinium myrtillus Calluna vulgaris
Carbon dioxide • No increase in photosynthetic rates (in • Increased photosynthetic rates. boreal forest in Norway). • Increased shoot biomass and extension • Increased water use efficiency. of shoots (and sometimes shoot production). • Unchanged shoot:root ratio • Earlier and greater amounts of Temperature flowering. • Accelerated dehardening. • No impact on density, shoot vigour, current year’s growth, total stem dry weight, water use efficiency or photosynthetic rates. • Increase in foliage production and shoot length.
• Increase in number of flowers.
• Earlier shoot growth and flowering. Drought • Physiological stress.
• More resistant to water deficits than C. • Increased water use efficiency (linked to
vulgaris but sustains greater damage decreased shoot growth). through drought. • Increased shoot density (over longer term). • Earlier bud-break.
1, Beerling, (1999); 2, Whitehead et al., (1997); 3, Woodin et al., (1992); 4, Ogren, (1996); 5, Werkman and Callaghan, (2002); 6, Gordon et al., (1999a); 7, Llorens et al., (2002); 8, Bannister, (1971); 9, Gordon et al., (1999b).
However P. aquilinum may also be able to achieve dominance (as opposed to sub-dominance) in the future and whether this occurs will be dependent on biotic interactions, such as competition with the existing community, on the upland heath sites. These interactions will be modified by climate change, other abiotic factors, as well as by site management. Luckily attempts to study the effects of climate change on the competitive interaction, between a dominant upland heath species, C. vulgaris and P. aquilinum have been made, and these can help us to predict future changes to upland heath in the study area (Box 8.C).
Box 8.C: Will P. aquilinum out-compete C. vulgaris under climate change?
• a A warmer climate will allow P. aquilinum to continue to spread at the expense of C. vulgaris, with drier summers reducing the rate of encroachment (Werkman and Callaghan, 2002).
• r A study looking at water use of P. aquilinum and C. vulgaris concluded that under droughted conditions C. vulgaris outcompeted P. aquilinum (Gordon et al., 1999b)
• r Whitehead et al., (1997) found that elevated carbon dioxide produced a larger response in C. vulgaris than in P. aquilinum, again suggesting that C. vulgaris may become more competitive relative to P. aquilinum in the future.
• ? The timing and nature of extreme events is critical in determining the competitive balance between C.
vulgaris and P. aquilinum (Gordon et al., 1999a)
MONARCH 2 Report – Chapter 8 223 ______
Werkman and Callaghan (2002) found that the positive effects of warming on the thickness and length of 1-year and 2-year old C. vulgaris shoots, the biomass of 2-year old shoots, shoot vigour, the proportion of flowering 1-year old shoots and the abundance of flowers on them, were reversed when C. vulgaris was in competition with P. aquilinum. This effect was attributed to increased shading by P. aquilinum, which also performed better under increased temperatures at the Upper Teesdale field- site (Werkman and Callaghan, 2002). On the basis of these experiments the authors concluded that a warmer climate would allow P. aquilinum to continue to spread at the expense of C. vulgaris, but with drier summers reducing the rate of encroachment.
Figure 8.30: The predicted pathway of Pteridium aquilinum through the Arriver model.
224 MONARCH 2 Report – Chapter 8 ______
Figure 8.31: An interaction web showing the main interactions for P. aquilinum. Red lines show predator-prey interactions, blue lines competitive interactions, brown lines host-parasite interactions, and black lines all other interactions. Other interactions
Pathogens Cryptomycina Camarographium Litter provides Cover and Chalara pteridina Ascochyta pteridis Shelter for Cover and Crocicreas pteridis stephensii goodfood breeding areas (Fungi) (Fungi) Vulpes vulpes breeding area for cyathoideum (Discomycetes) (Fungi) source of for (Mammalia) Saxicola rubetra (Discomycetes) Invertebrates for Phylloscopus spp. Phomatospora Dasyscyphus Leptopeltis Didymella (Aves) Sorex spp. and (Aves) Melittosporium endopteris pteridialis litigiosa prominula Shelter for Erinaceus pteridinum (Ascomycetes) (Discomycetes) (Ascomycetes) (Ascomycetes) Bracken litter as Oryctolagus europaeus hibernation sites (Discomycetes) cuniculus Diaporthopsis Dasyscyphus Leptopeltis Didymella for reptiles burrows Coniothyrium pantherina pteridis pteridis lophospora pteridis (Ascomycetes) (Discomycetes) (Ascomycetes) (Ascomycetes) Shelter and (Fungi) bedding Hyaloscypha Monographos Rhopographus Psilachnum material for flaveola fuckelii filicinus pteridigenum Cover for deer Meles meles setts Mollisia pteridina (Discomycetes) (Ascomycetes) (Ascomycetes) (Discomycetes) (Dama dama, (Discomycetes) Capreolus Scirrhia capreolus and aspidiorum Cover for raptors Cervus elaphus) Cover for (Ascomycetes) Caprimulgus Microscypha Micropodia Mycosphaerella Pezizella europaeus grisella pteridina pteridis chrysostigma (Aves) (Discomycetes) (Discomycetes) (Ascomycetes) (Discomycetes) Pteridium Sheep ticks Nesting areas for Nesting areas for Cover for Aneugmenus (bracken litter is a Acanthis Turdus torquatus Stromboceros Anthus trivialis Bourletiella fürstenbergensis major habitat flavirostis (Aves) delicatulus aquilinum (Aves) viridescens (Hymenoptera) (Hymenoptera) for them) (Aves) (Collembola)
Dasineura filicina Olethreutes Phytoliriomyza (Diptera) lacunana hilarella competitors Chirosia (Lepidoptera) (Diptera) albitarsis Vesicular Chirosia (Diptera) arbuscular grossicauda Chirosia Macrosiphum histricina ptericolens mycorrhizae Calluna vulgaris (Diptera) Phlogophera (Diptera) (Hemiptera) meticulosa Paltodora (Lepidoptera) Mycorrhizae cytisella Phytoliriomyza Lacanobia (Lepidoptera) Oryctolagus pteridii oleracea cuniculus (Diptera) (Lepidoptera) Tenthredo sp. (Mammalia) Petrophora (Hymenoptera) Ditropis pteridis chlorosata Dasineura (Hemiptera) (Lepidoptera) Euplexia pteridicola (Diptera) Key: lucipara Monalocoris Philaenus Polyphagous Monophagous (Lepidoptera) filicis spumarius (feeds on other Chirosia on bracken (probably) (Hemiptera) (Hemiptera) heathland plants) Aneugmenus albifrons temporalis (Diptera) Host specificity (Hymenoptera) Feeds on Not present Aneugmenus padi Eriophyes pteridis and/or distribution other ferns in test area Strongylogaster (Hymenoptera) (Acari) unknown Ceramica pisi lineata (Lepidoptera) (Hymenoptera) ‘predators’
In contrast, a study looking at water use of P. aquilinum and C. vulgaris concluded that drought would have a far larger impact on their competitive balance than increasing temperatures. C. vulgaris roots could invade P. aquilinum dominated areas depleting water from its rooting zone, but P. aquilinum roots were unable to compete for water in C. vulgaris dominated areas, suggesting that under these conditions C. vulgaris outcompeted P. aquilinum (Gordon et al., 1999b).
Whitehead et al., (1997) found that elevated carbon dioxide produced a larger response in C. vulgaris than in P. aquilinum, again suggesting that C. vulgaris may become more competitive relative to P. aquilinum in the future.
Furthermore, a fourth study reached the conclusion that interactions between environmental change variables and extreme events such as droughts and frosts are crucially important in determining the competitive balance between C. vulgaris and P. aquilinum. The timing of extreme events, in relation to the growing season was also found to exert a major influence. For example, although P. aquilinum tends to be more disadvantaged by drought than C. vulgaris, a drought before the croziers have emerged fully is much more detrimental to P. aquilinum than one later in the growing season (Gordon et al., 1999a).
Currently, P. aquilinum is abundant up to an altitude of 300m but may reach altitudes of 600m. Its altitudinal range is limited by its susceptibility to spring frosts and strong winds, rather than by temperature. Spring frosts kill or severely damage fronds, whereas autumn frosts may kill fronds
MONARCH 2 Report – Chapter 8 225 ______which emerge later in the year and shorten the growing season (Watt, 1976). If fronds emerge earlier in the future, due to increases in temperature, this would make them more susceptible to frost. Replacement of fronds damaged or killed by frosts can only take place through the initiation of new or actively dormant buds on the rhizome and it is energetically costly (Gordon et al., 1999a). Increased temperatures may lengthen the other end of the growing season, and this also may make P. aquilinum more susceptible to frost, as it decreases the autumn hardening of P. aquilinum (Gordon et al., 1999a).
Wind speed is not predicted to change in the case study area and therefore exposure may still limit the distribution and vigour of P. aquilinum. Strong winds can cause damage to fronds, which tend to be short and sparsely distributed (Watt, 1976), and replacement costs in these under-nourished plants are disproportionately high, making them less competitive in such environments.
The effect of extreme events and their timing on competition between P. aquilinum and C. vulgaris makes is difficult, if not impossible, to predict the outcome. It is likely that additional factors, such as nitrogen deposition, the growth phase of heather, the amount of exposure, grazing pressure, soil moisture content and many other biotic variables may also play an important role. However P. aquilinum may invade and become a dominant species at the expense of other upland heath vegetation (C. vulgaris and V. myrtillus) in at least some of the upland heath communities.
Therefore on the basis of the above discussion, the Arriver conceptual model predicts that P. aquilinum has the potential to have a large impact on the Snowdonia landscape, through its colonisation and dominance of large areas of heathland, as well as more subtle changes in other areas on community composition, which may result in the creation of a new community (Figure 8.30).
Communities in which P. aquilinum is already a dominant species generally have a very impoverished ground flora, through the effects of the dense shading by fronds, the dense litter layer and the allelopathic chemicals that the fern produces. The problem of shading may be further exacerbated by the predicted lengthening of the growing season due to elevated temperature (Werkman et al., 1996), potentially making it harder for understory plants to survive, unless they are also able to capitalise on the increased temperature predicted for Snowdonia and commence growth sufficiently early in the year. C. vulgaris is unable to tolerate the shading by P. aquilinum fronds and will therefore be lost from communities in which P. aquilinum is a dominant component. Other species such as V. myrtillus, Erica spp., and U. gallii, may be able to persist under the P. aquilinum canopy, but with much reduced plant size and abundance.
Although P. aquilinum has a fairly sizeable herbivore community associated with it, it is very different from that observed for either the C. vulgaris or V. myrtillus that it is replacing. Furthermore, there is evidence that this herbivore community is less productive than the one associated with C. vulgaris (Pakeman and Marrs, 1993). Invasion by P. aquilinum will also change the vegetation structure, creating a two canopy levels, the top level being made up of P. aquilinum fronds and the lower one consisting of the ground flora, and this will affect invertebrate composition and abundance.
In addition, through the exclusion of other plant species by the dense bracken canopy, there is a reduced diversity of food plants for other invertebrate herbivores to utilise in comparison with upland heath today, and this will have knock on effects on the invertebrate predator populations as well as on insectivorous vertebrate species. For example, there is a reduction in abundance of an insectivorous bird species, wheatear (Oenanthe oenanthe), in areas with dense P. aquilinum stands. Such effects will also feed further up the food chain and affect top predators.
P. aquilinum also has strong influences on the microclimate, especially in terms of humidity and temperature, which may again have strong effects on the co-occurring plant and animal communities. In addition P. aquilinum may play an important role in providing animals with cover and nesting sites, within the mountain environment, which is generally lacking in tall vegetation. This role of P. aquilinum is probably more pronounced when it occurs in patches as a mosaic with other shorter heathland vegetation types. It can provide cover for warblers (Oenanthe spp.), tree pipit (Anthus
226 MONARCH 2 Report – Chapter 8 ______trivialis), nightjar (Caprimulgus europaeus) (a Snowdonia Biodiversity Action Plan species), and raptors, as well as nesting sites for willow warbler (Phylloscopus trochilus), whinchat (Saxicola rubetra) and ring ouzel (Turdus torquatus). However, golden plover (Pluvialis apricaria), curlew (Numenius arquata), red grouse (Lagopus lagopus) and short-eared owl (Asio flammeus) avoid P. aquilinum as nesting sites. P. aquilinum also provides cover for many mammals, including foxes, rabbits, badgers, fallow deer and roe deer (Pakeman and Marrs, 1993).
8.6.2.3 Ulex gallii
In contrast to P. aquilinum, relatively little information is available on U. gallii, and this makes the task of assessing it’s potential impact on the habitat much more difficult and the predicted outcomes of the Arriver model more uncertain.
U. gallii is presently found growing in Britain in a range of moisture conditions, on a range of soil types and is frost tolerant, although severe frosts may cause shoot damage (Stokes et al., 2003). It therefore, seems that with climate change conditions within upland heath areas will become increasingly suitable for U. gallii. However this species may be dispersal limited, with seed dispersal occurring by explosive dehiscence of the pods, with the result that the vast majority of seeds fall within close proximity of the adult plant (Stokes et al., 2003).
Information displayed in the interaction web (Figure 8.32) indicates that the number of species associated with U. gallii, is relatively small. Therefore, this species appears to support only a very limited number of interactions within the lowland heath community in which it is currently present within the Snowdonia National Park, and this will also hold true if it colonises the upland heath communities. On this basis, it cannot be thought of as a dominant species.
However, this species can be considered to be of a functional group, not represented within the present-day upland heath community, as it is a leguminous fast growing woody shrub. As the interaction web suggests (Figure 8.32), U. gallii is a nitrogen fixer, and on this basis the Arriver model predicts that it will have a significant impact on upland heath community composition if it colonises (Figure 8.33).
However, whether U. gallii actually achieves sub-dominance within this community will probably depend largely on whether or not there is a high abundance of P. aquilinum within the community, as well as on other environmental and management factors.
8.6.3 Implications for the species composition of montane heath
The discussion in this section is focussed on the montane areas, especially the Carex bigelowii – Racomitrium lanuginosum moss heath and the Nardus stricta- Carex bigelowii grass heath, which are of particular conservation concern. The dispersal model predicts that space will continue to exist within the montane areas, for the species that are currently dominant, and that space may become suitable for Ulex gallii (Table 8.6), in some areas and this may impact on community composition along with changes in the abundance of species already present within the montane communities.
Table 8.6: A summary of the predictions of the dispersal model. Species Category Predicted space change Carex bigelowii stiff sedge Dominant Increase Vaccinium myrtillus bilberry Dominant No change Calluna vulgaris heather Sub-dominant Slight increase Ulex gallii western gorse Recruitment Increase
MONARCH 2 Report – Chapter 8 227 ______
Figure 8.32: The interaction web for U. gallii, showing the main interactions. Pink boxes indicate species that also interact with other Ulex species; blue boxes represent species which interact more widely; grey boxes indicate that the species is not currently present on upland heath in the test study area and white boxes indicate species for which this information is not known. Red lines show predator-prey interactions, blue lines competitive interactions, brown lines host-parasite interactions, and black lines all other interactions.
Flower visitors Andrena ovatula Sphecodes geoffrellus Apis mellifera (Hymenoptera) (Hymenoptera) (Hymenoptera) Parasitic plants Bombus terrestris Other interactions (Hymenoptera) Sphaerophoria scripta Cuscuta epithymum Nitrogen fixing bacteria (Diptera) in root nodules - Bombus lucorum Rhizobium species (Hymenoptera) Episyrphus sp. competitors (Diptera) Bombus humilis (Hymenoptera) Ulex minor Bombus lapidarius (Hymenoptera) Ants (Myrmica ruginodis, Bombus ruderarius M. scabrinodis and (Hymenoptera) Tetramorium cespitum) Eristalis sp. carry seeds in the lab. (Diptera) Syritta pipiens (Diptera) Ulex gallii ‘predators’ Agonopterix ulicetella (Lepidoptera)
deer (Mammalia) Cladosporium Chaetomium sp. Hypoxylon deustum tenuissimum (Ascomycete) (Hyphomycete) (Hyphomycete) cattle Gelasinospora (Mammalia) Coniothyrium fuckelii Daldinia vernicosa reticulispora (Deuteromycete) (Hyphomycete) (Ascomycete)
Coniothyrium ponies Pleospora herbarum Fusarium lateritium Coniothyrium olivaceum (Mammalia) (Ascomycete) (Hyphomycete) sphaerospermum (Deuteromycete) (Hyphomycete) Sporormiella australis Phomopsis ligulata Hypoxylon fragiforme Rabbits (Ascomycete) (Deuteromycete) (Hyphomycete) (Mammalia) Coniothyrium sp. Phomopsis sp. Ramularia deusta Apion ulicis (Deuteromycete) (Deuteromycete) (Hyphomycete) (Coleoptera) Fungal pathogen Geniculosporium sp. Alternaria alternata Rhizictonia sp. Apion scutellare Scythris gradipennis Callophrys rubi (Hypkomycete) (Hyphomycete) (Hyphomycete) (Coleoptera) (Lepidoptera) (Lepidoptera) Sporormiella minima (Ascomycete) Endophytes
Unfortunately most of the research into the impacts of climate change on montane species has been carried out in the Arctic, and is therefore of little relevance when considering communities at or near the southern limit of their distribution within Britain and Ireland. Within the case study area, the montane areas on mountain summit plateaux and ridges constitute one of the harshest and most extreme terrestrial environments present in Britain and Ireland (Bardgett and Leemans, 1996). Organisms existing in these areas have to withstand exposure to wind, extremes of temperature, periods of mist, periods of high insolation, frosts and snow (Evans, 1932). Although increases in temperature are expected within the montane parts of the area under climate change, Box 8.D indicates that many of the factors listed above are unlikely to be ameliorated, and this will have an important impact on future community composition and structure.
228 MONARCH 2 Report – Chapter 8 ______
Figure 8.33: The predicted pathway for Ulex gallii through the Arriver model.
Box 8.D: A summary of the predicted changes in climate for the Snowdonia area (Hulme et al., 2002).
Temperature Increase
Summer rainfall Decrease
Winter rainfall Increase Wind speed No change
Amount of precipitation falling as snow Decreased by 30-40% (low scenario) and 40-50% (high scenario) by the 2050s Length of snow lie Decrease
8.6.3.1 Climate change and the current community
Increases in temperature are known to lead to increases in nitrogen mineralization rates and plant productivity (Rustad et al., 2001). Furthermore warming may lengthen the growing season for many montane species including V. myrtillus, which is currently limited in the Nardus-Carex grass heath by the short growing season (Rodwell et al., 1992). Higher temperatures will allow earlier growth of shoots and leaves of this deciduous shrub, although this may also be accompanied by earlier de- hardening (Taulavuori et al., 1997), which may make it highly susceptible to late frosts that are likely to become more common in the montane environment.
MONARCH 2 Report – Chapter 8 229 ______
As Box 8.D indicates, climate change will affect the proportion of precipitation that will fall as snow, and the length of snowlie (warmer temperatures will lead to more rapid melting of any snow that does fall). Snow cover leads to a redistribution of water availability and protects plants from extreme winter temperatures, fluctuations in temperature as well as winter drought, but can curtail the growing season. Currently, it is thought that V. myrtillus can successfully out-compete mat grass (Nardus stricta) under conditions of reduced snowlie, especially if drainage is also improved (Rodwell et al., 1992), suggesting that a switch in dominance may occur in some areas under climate change. This may have substantial effects on community composition, as the two species belong to different functional groups. However, although the stems of V. myrtillus are fairly resistant to desiccation, warming related accelerated bud-burst together with decreased snow cover will make this species more vulnerable to desiccation.
In contrast to V. myrtillus, a study by Bannister (1971) on water relations indicated that C. vulgaris is better able to recover from water deficits and may therefore be more successful under future conditions. Currently, C. vulgaris is commonly found in montane areas that closely mirror possible future conditions (areas with little snow cover and moderate exposure). However, in order for C. vulgaris to increase in abundance and thereby affect community composition, it needs to compete successfully with other species within the community.
Competition between C. vulgaris and grasses, especially N. stricta (a dominant in the montane community) has been investigated using pot experiments, and results indicate that C. vulgaris is the superior competitor, especially where organic matter is concentrated at the top of the soil horizon (Genney et al., 2002). However, the fact that C. vulgaris does concentrate its roots in the surface horizons of the soil make it more susceptible to drought. Summer drought can cause both increased frost damage to plants in the following year and a decrease in canopy height (Gordon et al., 1999). Furthermore, although C. vulgaris is present in montane communities, frosts and strong winds, followed by warm sunshine in spring can lead to severe dieback of plants (Gimingham, 1960). This makes it questionable whether C. vulgaris would in reality be able to survive at high frequency, and compete successfully with grasses, such as N. stricta, especially as at these high altitudes C. vulgaris plants tend to be stunted and therefore less effective at shading out the grasses.
8.6.3.2 U. gallii colonisation
It is also likely that the harsh conditions, which are expected to continue to be a feature of the high mountain tops within the case study area, together with the poor seed dispersal mechanism, of U. gallii, will prevent it from becoming a prominent feature of the montane community. As has been discussed in the upland heath section, U. gallii is a nitrogen fixer and supports novel interactions and furthermore its colonisation would add a new functional group to the Carex bigelowii- Racomitrium lanuginosum moss heath and the Nardus stricta- Carex bigelowii grass heath communities. Its colonisation would thus be predicted to impact significantly on the functioning of montane heath.
8.6.4 Conclusions for montane/upland heath
• The European SPECIES model performed well, although the modelled results for species with lower prevalence, C. bigelowii and U. gallii, have lower agreement statistics. • The downscaled SPECIES model led to improvement in the match between observed national species’ distributions and simulated model outputs, although the level of agreement between the two is lower as land cover can lead to the over-restriction of suitable area. • The lack of observed 1-km distribution data led to a mix of methods being used to derive such distributions and the effect of this on dispersal model outputs is unknown. • All the species showed a high potential to disperse into the simulated suitable areas and in the case of the recruitment species the issue of invasion needs considering. • No dominant species in upland and montane heaths will loose suitable climate space.
230 MONARCH 2 Report – Chapter 8 ______
• The Arriver model suggests that colonisation by P. aquilinum and/or U. gallii will have far- reaching effects on community composition and structure. • The limited information available on U. gallii restricts the ability to assess the potential impacts upon community composition and species interaction using the Arriver model. • The order of colonisation will influence the exact nature of any changes to community composition. If P. aquilinum is first to colonise upland heath areas, then it is possible that strong competition from this species will prevent colonisation by U. gallii. • Management factors and nitrogen deposition may exacerbate the problems experienced with P. aquilinum.
8.7 Discussion and conclusions
8.7.1 Bioclimatic classification
This case study area was demonstrated to be sensitive to climate change based on the bioclimatic classification using the UKCIP98 dataset. The recalibration of this Baseline98 classification to the UKCIP02 data has shown that there is some difference in the prediction of the baseline climate conditions in this area. As with the Central Highlands (Chapter 7), in the Baseline02 classification some (eight squares out of 115) squares lie beyond the range of the Baseline98 classification. Given the improved resolution and data used within the UKCIP02 dataset, these misclassified squares imply that the Baseline98 classification is not necessarily an accurate reflection at this higher spatial resolution of the bioclimate for some parts of Snowdonia.
The most climatically sensitive parts of the study area are in the north and east of Snowdonia. They show the greatest squared Mahalanobis distance from the Baseline98 classification. These areas should be key locations for monitoring environmental change whether climatic variables or impacts on species, habitats or ecosystems. These sites, Carnedd range and Arenig Fawr, had the greatest concentration of conservation sites and were the most climatically sensitive. Hence, the ECN site on Snowdon can be regarded to be in a prime location for monitoring environmental change, being within both a climatically sensitive area and a conservation site of high importance.
8.7.2 Upland oak woodland
These results support those of MONARCH 1 (Berry et al., 2001; Berry et al., 2003) suggesting that the dominant trees are not vulnerable to climate change within the time frame of the study (up to 2050s), but that some of the herb species, such as H. non-scripta, may start to decline in abundance in Snowdonia, especially if climate change is combined with other environmental pressures. It is the interaction between climate change and other pressures (combination punches on ecosystems, after Masters and Midgley, 2004), such as land management, air pollution, nitrogen deposition that could be key in driving species and habitat responses. Adaptive woodland management may also be an important factor in either speeding up or reducing the predicted impact of climate change.
Although dramatic changes in dispersal and climate space are not predicted for the tree species, what is certain is that climate change will affect the growth and phenology of the dominant tree species, Q. petraea, and this will have consequences for the remainder of the upland oak woodland community. Such subtle changes in a species “performance” within a community can easily go unnoticed without detailed (and expensive) monitoring, however, the consequences can be extreme (for a grassland example see Grime et al., 2000). Over time, subtle changes that accumulate gradually can cause an imbalance within the community. After a critical time, when sufficient accumulation has occurred, the community can undergo a collapse, often leading to a change in ecosystem state (Scheffer et al., 2001; Scheffer and Carpenter, 2003). Such a change can occur quickly and is often irreversible. Additionally, the outcome of a change in states is difficult if not impossible to predict. These subtle, often unobserved, changes may be mediated through changes to food availability and quality for herbivorous insects, which may also affect higher trophic levels or through changes to the ground
MONARCH 2 Report – Chapter 8 231 ______flora, for example H. non-scripta, as a consequence of alterations in light intensity, humidity and litter production.
8.7.3 Montane/upland heath
Of the two species likely to recruit into this habitat, the expansion in suitable climate space for P. aquilinum, within areas of current upland heath communities will be of concern for conservation. Climate change, nitrogen deposition, grazing and burning may all favour P. aquilinum over the characteristic heathland species, leading to a large alteration in the habitat reflected through a change in the structure, diversity and composition of the entire community. However, to drive these changes P. aquilinum needs to recruit and become at least a sub-dominant within the community and this is dependent upon biotic interactions such as competition. The strength and direction of such interactions will also be modified by climate change (c.f. the subtle changes described for the oak woodland previously) and by management or other abiotic factors. A key factor here will be how climate and other factors affect the performance, and hence competitive ability, of C. vulgaris, potentially the main competitor for P. aquilinum. In areas in which P. aquilinum is limited by frosts or exposure, upland heath communities may survive, especially if grazing pressure is reduced. However, the effects of climate change together with grazing and continued nitrogen deposition will encourage the continued conversion of montane heaths to acid grasslands.
In contrast to P. aquilinum, the potential for and impact of colonisation of heathland communities by U. gallii is less clear, as knowledge on this species is lacking in the literature. However, it is likely that the impact will be less dramatic, especially as nitrogen deposition is already having an impact on this nutrient limited community over and above any impact that the nitrogen fixation capability of U. gallii, may induce. Continuing extreme weather conditions seem to preclude the colonisation of U. gallii in montane habitat
8.8 References
Backshall, J. (2001). Moorland. In: The upland management handbook. English Nature.
Bale, J.S., Masters, G.J., Hodkinson, I.D., Awmack, C., Bezemer, T.M., Brown, V.K., Butterfield, J., Buse, A., Coulson, J.C., Farrar, J., Good J.E.G., Harrington, R., Hartley, S., Jones, T.F., Lindroth, R.L., Press, M.C., Symrnioudis, I. Watt, A.D. and Whittaker, J.B. (2002). Herbivory in global climate change research: direct effect of rising temperature on insect herbivores. Global Change Biology, 8, 1- 16.
Bannister, P. (1971). The water relations of heath plants from open and shaded habitats. Journal of Ecology, 59, 51-64.
Beerling, D.J. (1999). Long-term responses of boreal vegetation to global change: an experimental and modelling investigation. Global Change Biology, 5, 55-74.
Berry, P.M., Dawson, T.P., Harrison, P.A, Pearson, R.G. and Butt, N. (2003). The sensitivity and vulnerability of terrestrial habitats and species in Britain and Ireland to climate change. Journal for Nature Conservation, 11, 15-23.
Berry, P.M., Vanhinsbergh, D., Viles, H.A., Harrison, P.A., Pearson, R.G., Fuller, R., Butt, N. and Miller, F. (2001). Impacts on terrestrial environments. In Harrison, P.A., Berry, P.M. and Dawson, T.P. (eds.) Climate Change and Nature Conservation in the UK and Ireland: Modelling natural resource responses to climate change (the MONARCH project). UKCIP Technical Report, Oxford, pp43-150.
232 MONARCH 2 Report – Chapter 8 ______
Berthold, P. (1990). Patterns of avian migration in light of current global “greenhouse” effects: a central European perspective. Acta XX Congressus Internationalis Ornithologici, 780-786.
Blackman, G.E. and Rutter, A.J. (1947). Physiological and ecological studies in the analysis of plant environment. II The interaction between light intensity and mineral nutrient supply on the growth and development of bluebell, (Scilla nonscripta). Annals of Botany, London N.S., 11, 126-158.
Both, C. and Visser, M.E. (2001). Adjustment to climate change is constrained by arrival date in a long- distance migrant bird. Nature, 411, 296-298.
Braid, K.W. and Conway, E. (1943). Rate of growth of bracken. Nature, 152, 751-752.
Broadmeadow, M.S.J. and Jackson, S.B. (2000). Growth responses of Quercus petraea, Fraxinus excelsior and Pinus sylvestris to elevated carbon dioxide, ozone and water supply. New Phytologist, 146, 437-451.
Bullock, J.M., Edwards, R.J., Carey, P.D. and Rose, R.J. (2000). Geographical separation of two Ulex species at three spatial scales: Does competition limit species' ranges? Ecography, 23, 257-271.
Buse, A., Dury, S.J., Woodburn, R.J.W., Perrins, C.M. and Good, J.E.G. (1999). Effects of elevated temperature on multi-species interactions: the case of Pendunculate Oak, Winter Moth and Tits. Functional Ecology, 13, 74-82.
Buse, A. and Good, J.E.G. (1996). Synchronization of larval emergence in winter moth (Operophtera brumata L.) and budburst in pendunculate oak (Quercus robur L.) under simulated climate change. Ecological Entomology, 21, 335-343.
Carroll, J.A., Capron, S.J.M., Cawley, L., Read, D.J. and Lee, J.A. (1999). The affect of increased deposition of atmospheric nitrogen on C. vulgaris in Upland Britain. New Phytologist, 141, 423-431.
Conway, E. (1957). Spore production in bracken (Pteridium aquilinum (L.) Kuhn). Journal of Ecology, 45, 273-284
Conway, E. (1953). Spore and sporeling survival in bracken (Pteridium aquilinum (L.) Kuhn). Journal of Ecology, 41, 289-294
Coppack, T. and Both, C. (2002). Predicting life-cycle adaptation of migratory birds to global climate change. Ardea, 90, 369-378.
Corbet, S.A. (1998). Fruit and seed production in relation to pollination and resources in bluebell, Hyacinthoides non-scripta. Oecologia, 114, 349-360.
Crick, H.Q.P. (In press). The impacts of climate change on birds. Ibis.
Crick, H.Q.P. & Sparks, T.H. (In press). Changes in the phenology of breeding and migration in relation to global climate change. Proceedings of the 23rd International Ornithological Congress, Beijing.
Drent, R., Both, C. Green, M., Madsen, J. and Peirsma, T. (2003). Pay-offs and penalties of competing migratory schedules. Oikos, 103, 274-292.
Dury, S.J., Good, J.E.G., Perrins, C.M., Buse A. and Kaye, T. (1998). The effects of increasing CO2 and temperature on oak leaf palatability and the implications for herbivorous insects. Global Change Biology, 4, 55-61.
MONARCH 2 Report – Chapter 8 233 ______
Edington, J.M. and Edington, M.A. (1972). Spatial patterns and habitat partition in the breeding birds of an upland wood. Journal of Animal Ecology, 41, 331-357.
Edwards, M.E. and Birks, H.J.B. (1986). Vegetation and ecology of four western oakwoods Blechno- Quercetum-Petraea in North Wales. Phytocoenologia, 14, 237-262.
Gallet, S., Lemauviel, S. and Roze, F. (2004). Responses of three heathland shrubs to single or repeated experimental trampling. Environmental Management, 33, 821-829.
Gimingham, C.H. (1960). Calluna Salisb. Biological flora of the British Isles. Journal of Ecology, 48, 455-483.
Gordon, C., Woodin, S.J., Mullins, C.E. and Alexander, I.J. (1999a). Effects of environmental change, including drought, on water use by competing Calluna vulgaris (heather) and Pteridium aquilinum (bracken). Functional Ecology, 13(SUPPL. 1), 96-106.
Gordon, C., Woodin, S.J., Alexander, I.J. and Mullins, C.E. (1999b). Effects of increased temperature, drought and nitrogen supply on two upland perennials of contrasting functional type: Calluna vulgaris and Pteridium aquilinum. New Phytologist, 142, 243-258. Gonzalez, S.G., Penas, M.A. and Alonso, H.E. (1996). Phenology of Hyacinthoides non-scripta (L.) Chouard, Melittis melissophyllum L. and Symphytum tuberosum L. in two deciduous forests in the Cantabrian mountains, Northwest Spain. Vegetatio, 122, 69-82.
Grabham, P.W. and Packham, J.R. (1983). A comparative study of the bluebell Hyacinthoides non- scripta (L.) Chouard in two different woodland situations in the West Midlands, England. Biological Conservation, 26, 105-126.
Grime, J.P., Brown, V.K., Thompson, K., Masters, G.J., Hillier, S.H., Clarke, I.P., Askew, A.P., Corker, D. and Kielty, J.P. (2000). The response of two contrasting limestone grasslands to simulated climate change. Science, 289, 762-765.
Hossell, J.E., Briggs, B. and Hepburn, I.R. (2000). Climate Change and Nature Conservation: A Review of the Impact of Climate Change on UK Species and Habitat Conservation Policy. HMSO DETR MAFF, London 73pp + Appendices.
Järvinen, A. (1996). Correlation between egg size and clutch size in the Pied Flycatcher Ficedula hypoleuca in cold and warm summers. Ibis, 138, 620-623.
Järvinen, A. (1994). Global warming and egg size of birds. Ecography, 17, 108-110.
Järvinen, A. (1989). Patterns and causes of long-term variation in reproductive traits of the Pied Flycatcher Ficedula hypoleuca in Finnish Lapland. Ornis Fennica, 66, 24-31.
Jarvis, P.G. (1964). The adaptability to light intensity of seedlings of Quercus petraea (Matt.) Liebl. Journal of Ecology, 52, 545-571.
Jensen, K.D., Beier, C., Michelsen, A. and Emmett, B.A. (2003). Effects of experimental drought on microbial processes at contrasting water conditions. Applied Soil Ecology, 24, 165-176.
Jones, E.W. (1959). Quercus L. Biological flora of the British Isles. Journal of Ecology, 47, 169-222.
Jones, M.L.M., Oxley, E.R.B. and Ashenden, T.W. (2002). The influence of nitrogen deposition, competition and desiccation on growth and regeneration of Racomitrium lanuginosum (Hedw.) Brid. Environmental Pollution, 120, 371-378.
234 MONARCH 2 Report – Chapter 8 ______
Knight, G.H. (1964). Some factors affecting the distribution of Endymion nonscriptus (L.) Garcke in Warwickshire woods. Journal of Ecology, 52, 405-421.
Lehtila, K. and Syrjanen, K. (1995). Compensatory responses of two Melampyrum species after damage. Functional Ecology, 9, 511-517.
Llorens, L., Penuelas, J. and Emmett, B. (2002). Developmental instability and gas exchange responses of a heathland shrub to experimental drought and warming. International Journal of Plant Sciences, 163, 959-967.
Lundberg, A. and Alatalo, R.V. (1992) The pied flycatcher. T & AD Poyser, London. 267pp.
Marrs, R.H., Pakeman, R.J., Le-Duc, M.G. and Paterson, S. (1997). Bracken invasion in Scotland. Botanical Journal of Scotland, 49, 347-356.
Mead, C. (2000) The state of the nation’s birds. Whittet Books.285pp.
Nygaard, P.H. and Odegaard, T. (1999). Sixty years of vegetation dynamics in a south boreal coniferous forest in southern Norway. Journal of Vegetation Science, 10, 5-16. Nygaard, P.H. and Abrahamsen, G. (1991). Effects of long-term artificial acidification on the ground and soil in a 100-year old stand of Scots Pine, Pinus sylvestris. Plant and Soil, 131, 151-160.
Ogren, E. (1996). Premature dehardening in Vaccinium myrtillus during a mild winter: a cause for winter dieback? Functional Ecology, 10, 724-732.
Pakeman R.J., Le-Duc, M.G. and Marrs, R.H. (2000). Bracken distribution in Great Britain: Strategies for its control and the sustainable management of marginal land. Annals of Botany, 85(Suppl. B), 37- 46.
Pakeman, R.J. and Marrs, R.H. (1996). Modelling the effects of climate change on the growth of bracken (Pteridium aquilinum) in Britain. Journal of Applied Ecology, 33, 561-575.
Pakeman, R.J. and Marrs, R.H. (1992). The conservation value of bracken Pteridium aquilinum (L.) Kuhn – dominated communities in the UK, and an assessment of the ecological impact of bracken expansion or its removal. Biological Conservation, 62, 101-114.
Paterson, J. (2000). Modelling potential distribution changes of European tree species in response to climate change and the consequences for W12 Fagus sylvatica-Mercuralis perennis woodland in south east England. M.Sc. dissertation, Oxford Forestry Institute, University of Oxford.
Pearce, S.K. and van der Wal, R. (2002). Effects of nitrogen deposition on growth and survival of montane Racomitrium lanuginosum heath. Biological Conservation, 104, 83-89.
Preston, C.D., Pearman, D.A. and Dines, T.D. (2002). New Atlas of the British Flora. Oxford University Press, Oxford, 910pp.
Proctor, M.C.F. (1965). The distinguishing characteristics and geographical distributions of Ulex minor and Ulex gallii. Watsonia, 6, 177-187.
Przybylo, R., Sheldon, B.C. and Merila, J. (2000). Climatic effects on breeding and morphology: evidence for phenotypic plasticity. Journal of Animal Ecology, 69, 395-403.
Ratcliffe, D.A. (1968). An ecological account of Atlantic bryophytes in the British Isles. New Phytologist, 67, 365-439.
MONARCH 2 Report – Chapter 8 235 ______
Rhind, P. and Jones, B. (2003). The vegetation history of Snowdonia since the late glacial period. Field Studies, 10, 539-552.
Rhind, P. and Evans, D. (Eds.) (2001). The plant life of Snowdonia. Gormer Press, Llandysul.173 pp
Rieley, J.O., Richards, P.W. and Bebbington, A.D.L. (1979) The ecological role of bryophytes in a North Wales woodland. Journal of Ecology, 67, 497-527.
Rodwell, J.S., Pigott, C.D., Ratcliffe, D.A., Malloch, A.J.C., Birks, H.J.B., Proctor, M.C.F., Shimwell, D.W., Huntley, J.P., Radford, E., Wigginton, M.J and Wilkins, P. (1992), British Plant Communities. Volume 3. Grasslands and montane communities. Cambridge University Press, Cambridge, 540pp.
Sanz, J.J. (2003). Large-scale effect of climate change on breeding parameters of pied flycatchers in Western Europe. Ecography, 26, 45-50.
Saxe, H., Cannell, M.G.R., Johnsen, O., Ryan M.G. and Vourlitis, G. (2001). Tree and forest functioning in response to global warming. New Phytologist, 149, 369-400.
Scheffer, M. and Carpenter, S.R. (2003). Catastrophic regime shifts in ecosystems: linking theory to observations. Trends in Ecology and Evolution, 18, 648-656.
Scheffer, M., Carpenter, S. Foley, J.A., Folkes, L. and Walker, B. (2001). Catastrophic shifts in ecosystems. Nature, 413, 591-596.
Shaw, M.W. (1968). Factors affecting the natural regeneration of sessile oak (Quercus petraea) in North Wales: I. A preliminary study of acorn production, viability and losses. Journal of Ecology, 56, 565-583.
Slater, F.M. (1999). First-egg date fluctuations for the Pied Flycatcher Ficedula hypoleuca in the woodlands of mid-Wales in the twentieth century. Ibis, 141, 497-499.
Snow, C.S.R. and Marrs, R.H. (1997). Restoration of Calluna heathland on a bracken Pteridium - infested site in north west England. Biological Conservation, 81, 35-42.
Snowdonia National Park Authority (1999). Biodiversity in Snowdonia. Biodiversity Action Plan.
Sparks, T.H., Brown, D, Slater, F.M. and Wright, J. (in press) Advances in Pied Flycatcher (Ficedula hypoleuca) nesting phenology synchronise across sites in Wales. Ibis.
Spidso, T.K. and Stuen, O.H. (1988). Food selection by Capercaillie chicks in Southern Norway. Canadian Journal of Zoology, 66, 279-283.
Stokes, K.E., Bullock, J.M. and Watkinson, A. R. (2003). Ulex gallii Planch. and Ulex minor Roth. Biological flora of the British Isles. Journal of Ecology, 91, 1106-1124.
Sydes, C. and Grime, J.P. (1981). Effects of tree litter of herbaceous vegetation in deciduous woodland: II. An experimental investigation. Journal of Ecology, 69, 249-262.
Taulavuori, K., Laine, K., Taulavuori, E., Pakonen, T. and Saari, E. (1997).Accelerated dehardening in bilberry (Vaccinium myrtillus L.) induced by a small elevation in air temperature. Environmental Pollution, 98, 91-95.
Thomas, F.M. and Ahlers, U. (1999). Effects of excess nitrogen on frost hardiness and freezing injury of above-ground tissue in young oaks (Quercus petraea and Q. robur). New Phytologist, 144, 73-83.
236 MONARCH 2 Report – Chapter 8 ______
Thompson, P.A. and Cox, S.A. (1978). Germination of the bluebell Hyacinthoides non-scripta in relation to its distribution and habitat. Annals of Botany, 42, 51-62.
Turrill, W.B. (1952). Endymnion hispanicus. Curtis’s Botanical Magazine, N.S., t. 176.
Visser, M.E., van Noordwijk, A.J., Tinbergen, J.M. and Lessells, C.M. (1998). Warmer springs lead to mistimed reproduction in great tits (Parus major). Proccedings of the Royal Society, London, B, 265, 1867-1870.
Warren, M.S. (1987). The ecology and conservation of the heath fritillary butterfly, Mellicta athalia. I. Host selection and phenology. Journal of Applied Ecology, 24, 467-482.
Watt, A.S. (1976) The ecological status of bracken. Botanical Journal of the Linnean Society, 73, 217-239.
Welch, D. (1998). Response of Bilberry Vaccinium myrtillus L. stands in the Derbyshire Peak District to sheep grazing and implications for moorland conservation. Biological Conservation, 83, 155-164.
Werkman, B.R. and Callaghan, T.V. (2002). Responses of P. aquilinum and heather to increased temperature and nitrogen addition, alone and in competition. Basic and Applied Ecology, 3, 267-276.
Werkman, B.R., Callaghan, T.V. and Welker, J.M. (1996). Responses of bracken to increased temperature and nitrogen availability. Global Change Biology, 2, 59-66.
Whitehead, S.J., Capron, S.J.M. and Press, M.C. (1997). Effects of elevated CO2, nitrogen and phosphorus on the growth and photosynthesis of two upland perennials: C. vulgaris and Pteridium aquilinum. New Phytologist, 135, 201-211.
Woodin, S., Graham, B., Killick, A., Skiba, U. and Cresser, M. (1992). Nutrient limitation of the long term response of heather (Calluna vulgaris (L.). Hull) to CO2 enrichment. New Phytologist, 122, 635- 642.
Winkel, W. and Hudde, H. (1997). Long-term trends in reproductive traits of tits (Parus major, P.caeruleus) and Pied Flycatchers Ficedula hypoleuca. Journal of Avian Biology, 28, 187-190.
MONARCH 2 Report – Chapter 9 237 ______
9 Impacts for the Cuilcagh/Pettigo Peatlands case study area
P.M. BERRY, G.J. MASTERS, J.E. HOSSELL, N. BUTT, S. FREEMAN, P.A. HARRISON AND N. WARD
Summary
The Cuilcagh/Pettigo case study area was selected as a cross-border site in an area identified by the bioclimatic modelling as potentially sensitive to climate change and containing extensive areas of nature conservation interest. It was decided to choose only one habitat, blanket bog. A bog moss (Sphagnum cuspidatum), hare’s-tail cotton grass (Eriophorum vaginatum), deergrass (Trichophorum cespitosum), white beak-sedge (Rhynchospora alba) and crowberry (Empetrum nigrum) were selected as dominant/sub dominant species; lesser twayblade (Listera cordata) and golden plover (Pluvialis apricaria) as rare and flagship species respectively; and downy birch (Betula pubescens) and bracken (Pteridium aquilinum) as recruitment species.
The research predicted that:
1. The Cuilcagh/Pettigo peatlands would experience little change by the 2050s, although some of the area would become warmer. This is surprising given the apparent sensitivity of the area under the UKCIP98 scenarios. It may be that the reduced number of variables used in this dataset has masked the extent of the changes.
2. This would lead to little change in land cover within the case study area.
3. Under the climate change scenarios, Ireland would become less suitable for all the modelled species, apart from R. alba and P. aquilinum, with predicted loss of suitable climate space particularly occurring in south-west Ireland.
4. The addition of Corine land cover data in the downscaled model did not provide a good match between the simulated suitability space and observed distributions of species. This could be because peat bog is present in most squares at the 10-km resolution but its observed distribution is very patchy at a smaller scale.
5. The downscaled model performed poorly because the addition of the Corine land cover data led to a poor statistical agreement between modelled suitability space and actual distribution. As a result the dispersal model was not run.
6. One potential implication of reduced rainfall for the community composition is thought to be a shift towards a community similar to that of drier areas by the 2050s, perhaps with a reduction in the area of bog and an increase in the area of wet heath type vegetation, as occurs on areas of drained bog.
9.1 Introduction
This area and habitat were chosen because of their potential sensitivity to climate change as revealed through the application of the UKCIP98 climate change scenarios. Schouten (1984) included, as one of several reasons for the necessity to conserve the range of variation in bog types in Ireland, the fact that ombrotrophic bogs represent ideal objects for the study of relationships between climate and ecosystem, since these relationships are very direct ones, such as the adaptation of Sphagnum spp. to varying wetness, without the interference of any secondary factors like soil influences. Thus unlike the other case study areas it was decided to focus on one habitat type, blanket bog, across an
238 MONARCH 2 Report – Chapter 9 ______altitudinal gradient of about 500 m. Within the case study area this habitat type is represented within the Cuilcagh Mountain and Cuilcagh-Anierin upland SACs, the Dunragh Loughs/Pettigo Plateau SAC/SPA/Ramsar site and other designated areas. The study area boundary was defined to include them.
9.2 Bioclimatic classification
9.2.1 Baseline classification
The pattern of classes between the classification using the UKCIP98 data (Baseline98) and that using the UKCIP02 data (Baseline02) shows a considerable difference (Figure 9.1; Table 9.1). Whilst two of the three Baseline98 classes are still represented there is a large-scale shift into classes 12 and 9, with class 5 being lost. Figure 9.1 shows the distribution of the classes under both baseline datasets.
In the Baseline98 classification, the area is dominated by classes 1 and 4, with some squares in class 5. Classes 1 and 5 are both cool climate classes with short growing season lengths (<250 days) with quite high winter rainfall and high winter wind speeds. Class 4 shows warmer conditions, with lower wind speeds and a growing season length >300 days. Under the Baseline02 classification, the study area is dominated by class 12, which has temperatures intermediate to classes 1 and 4, but has lower rainfall totals particularly in autumn and winter. However, in the Baseline98 classification class 12 is still associated with raised bog areas in both Ireland and over the Pennines in England (See Table 2.5). It is also clear from Table 9.2 that the climate of the Baseline02 data set is both drier and slightly warmer than that described by the Baseline98. For the squares assigned to Class 4 in the case study area the Baseline98 data for the average autumn and winter rainfalls are 409mm and 385mm respectively. Under the Baseline02 data the figures are 394 mm and 371mm respectively. Since both baseline assessments used data from the same period (1961-90) the changes to the classification for the Baseline02 data cannot be considered to be a climatic shift. They represent a change in data resolution not a change in climate. It was recognised in Chapter 2 that the reduction in the number of variables used in the Baseline02 might affect the ability of the classification to match the Baseline98 classification. However, it is more likely that the change in class is simply a reflection of the improved resolution of the Baseline02 dataset.
Table 9.1: Percentage of grid squares within each of the classes of the bioclimatic classification for the Baseline98 and Baseline02 datasets and for the 5km 2050s Low and High emissions scenarios. Class 1 4 5 7 9 11 12 13 Baseline98 52.38 33.33 14.29 Baseline02 19.0 4.8 3.6 72.6 2050s Low 9.5 71.4 1.2 6.0 11.9 2050s High 71.4 1.2 27.4
Table 9.2: Comparison of mean values of key variables averaged over the case study area from each of the data sets. The variables selected represent those related to the first three components of the PCA in the derivation of the bioclimatic classification. Spring precipitation Growing Degree January total (mm) Days >5°C wind speed (m/s) Baseline98 201.64 1293.90 5.57 Baseline02 179.18 1324.61 5.57 2050s Low 164.25 1627.41 5.60 2050s High 155.47 1823.08 5.62
MONARCH 2 Report – Chapter 9 239 ______
Figure 9.1: Distribution of bioclimate classes. UKCIP Baseline02 classes (in the centre of a 5 km grid cell) are compared with Baseline98 classes (at the intersection of grid cells).
9.2.2 Climate change scenarios – 2050s Low and High
Figure 9.2 shows the projections for bioclimatic classes under the 2050s Low and 2050s High scenarios respectively. Unlike in the other case study areas, all squares within both future scenarios have climates recognised in the UKCIP98 baselines, possibly reflecting the lower level of climate change projected for this area as compared to Britain. Class 1 declines under 2050s Low and disappears under 2050s High. The main change is an extension of the warmer class 4 across most of the study area, with the cooler parts of the upland areas in the south-east of the study area being assigned to the cooler class 13.
9.3 Land cover changes
The land cover models for Ireland were based on Corine (Chapter 3.4) and showed great variation in their ability to predict the baseline land cover. Across the case study area most of the ubiquitous land cover classes, such as peat bog, would be little affected by climate change, with the climate envelope for peat bog being maintained across the area. In contrast, Figure 9.3 shows how natural grassland is projected to decrease in extent under the future climate change scenarios. However, the model showed only moderate predictive power in the baseline assessment (Chapter 3), suggesting that the results should be viewed with some caution.
240 MONARCH 2 Report – Chapter 9 ______
Figure 9.2: The projections for bioclimatic classes under the (a) 2050s Low scenario, and (b) 2050s High scenario.
(a) (b)
9.4 Species and dispersal modelling
The validation statistics for the European trained SPECIES models show a very good discrimination ability for all species modelled, based on the AUC (Table 3.3). The Kappa statistic shows that there is an excellent agreement for Betula pubescens, Eriophorum vaginatum, Pteridium aquilinum and Pluvialis apricaria, while the others show very good agreement, as their distributions are slightly more patchy. Sphagnum cuspidatum has the lowest kappa (0.747), as the model fails to predict most of the German distribution and over-predicts for Ireland and parts of Britain (Figure 9.4). The whole of Ireland is simulated as suitable climate space for all the species at the 5-km resolution, except Listera cordata and S. cuspidatum which are predicted to have less suitable climate space in the south-west (Figure 9.5). A number of the species are quite sensitive to the climate change scenarios, with climate space being progressively lost from southern Ireland, as indicated for L. cordata and Trichophorum cespitosum (Figures 9.6 and 9.7). The pattern of loss is slightly different for each species, indicating their individual response to climate change, as was found in MONARCH 1 (Berry et al, 2001).
MONARCH 2 Report – Chapter 9 241 ______
Figure 9.3: Changes in natural grassland under future climate change scenarios.
242 MONARCH 2 Report – Chapter 9 ______
Figure 9.4: The European observed distribution (a) and simulated climate space (b) for Sphagnum cuspidatum.
(a) (b)
Figure 9.5: Suitable climate space under UKCIP02 baseline climate for (a) Listera cordata and (b) Sphagnum cuspidatum
(a) (b)
MONARCH 2 Report – Chapter 9 243 ______
Figure 9.6: Progressive loss of suitable climate space for Listera cordata under the UKCIP02 scenarios.
2020s Low scenario 2020s High scenario
2050s Low scenario 2050s High scenario
244 MONARCH 2 Report – Chapter 9 ______
Figure 9.7: Progressive loss of suitable climate space for Trichophorum cespitosum under the UKCIP02 scenarios.
2020s Low scenario 2020s High scenario
2050s Low scenario 2050s High scenario
The validation statistics for the downscaled SPECIES modelling based on climate and land cover are much lower than most of those for the climate-driven SPECIES model (Table 3.5), with six of the eight species having an AUC lower than 0.8 and the two bog dominants S. cuspidatum and E. vaginatum lower than 0.6. P. apricaria still has a high Best Available Match although this is calculated differently (Chapter 3). The only species in the other case study areas that have an AUC below 0.8 are common cow-wheat (Melampyrum pratense) (0.761), hairy wood ant (Formica lugubris) (0.747) and sessile oak (Quercus petraea) (0.740). In the case of M. pratense (Chapter 8.4), the close association of its distribution with woodland was not picked up well by the downscaled model, so when the dispersal model was run a relatively large number of sites where it does occur were simulated as unsuitable and thus dispersal from these locations was unsuccessful. The poor match between the observed distribution and suitable space at the 10-km resolution for F. lugubris led to a gross over-simulation of its 1-km suitable space, making the dispersal results potentially over-
MONARCH 2 Report – Chapter 9 245 ______optimistic (Chapter 7.4.1). This experience from the other case study areas, suggested that where the AUC was less than 0.8 for the downscaled SPECIES model the suitability surface was not sufficiently robust to run the dispersal model. The poor agreement between the observed and downscaled simulated suitability surfaces for all the Irish case study species meant that they would not provide a good test for the dispersal model, especially when combined with the different methods of data collection, so the dispersal modelling was not undertaken. The reasons for the poor agreement need much further investigation, but possible avenues include the effect of the different and greater number of land cover classes within Corine and a possible greater role of climate in affecting species’ distributions in Ireland compared to Britain.
A visual comparison between the observed distribution and downscaled simulated suitability surfaces shows that, not surprisingly, B. pubescens appears to have the best match, as it is a widespread species. Empetrum nigrum and L. cordata have a very low prevalence, but the whole (or nearly the whole for the latter) of Ireland is simulated as suitable (Figure 9.5). The downscaled model thus over predicts suitable space. The same is true, but to a lesser extent for R. alba and S. cuspidatum. It is apparent, therefore, that the presence/absence of land cover types at a 10-km resolution does not provide a good correlate with distribution for these species in Ireland. This is likely to be due to the fact that at this resolution nearly all 10-km cells incorporate at least one small patch of suitable land cover (i.e. a ‘presence’), leading to blanket coverage throughout the case study area (Pearson et al., 2004). Analysis of the land cover variables used (44 classes from the Corine dataset) supports this (Figure 9.8). In order to better identify correlations between land cover type and distribution of these species it would be necessary to adopt a finer resolution of analysis at which patterns in the distribution of suitable land cover are apparent in the dataset. These results support the proposal that the search for environmental correlates with species’ distributions must be addressed at an appropriate spatial scale, whereby distribution patterns for environmental variables and species distributions are similar at the resolution of analysis.
Figure 9.8: ‘Peat bog’ land cover class in Ireland at 10-km resolution (left panel) and 250 m resolution (right panel). Aggregating the data to 10-km is shown to remove much of the patterning that is apparent at the finer resolution.
246 MONARCH 2 Report – Chapter 9 ______
9.5 Implications for the composition of species communities
The study area contains extensive areas of both lowland (e.g. Dumragh Loughs/Pettigoe Plateau) and upland blanket bogs (e.g. Cuilcagh-Anierin Upland/Cuilcagh Mountain), and the original basis of the species’ selection was to give an indication of changes in these two ecosystem types. However, the resolution of the model outputs (10km2) is too coarse to attempt this at present, as it is not possible to distinguish between areas of the two habitat types. Furthermore, the outputs from the models are based on presence or absence of a species from each of the 21 squares that form the case study area, and they do not give any indication of species’ abundance. Therefore, the following discussion is limited to a general, broad-brush assessment of how climate change may affect the blanket bog community.
Table 9.3: A summary of the predictions from the downscaled models for the case study area. Species Selection Downscaled model Comments category (climate change and land Latin name Common name cover) Betula pubescens downy birch Recruitment No change
Empetrum nigrum crowberry Dominant Loss of half of suitable Model over estimates squares by 2050s from current distribution centre and SW of case study area
Eriophorum hare’s-tail cotton Dominant No change (gain of one Fairly widely vaginatum grass square (same one) under distributed in case all scenarios) study area (15 squares today)
Listera cordata lesser twayblade Rare Loss of space under all Model over estimates scenarios. No suitable current distribution. climate space under 2050s Fairly widely high. distributed today.
Pluvialis apricaria golden plover Flagship No change
Pteridium aquilinum bracken Recruitment No change (widely Model for current distributed in Northern conditions under- Ireland) estimates, compared to the actual distribution.
Rhynchospora alba white beak-sedge Dominant No change (widely Not a good match distributed) between current distribution and simulated distribution.
Sphagnum Dominant Increase in space Modelling of climate cuspidatum (central part of case study alone indicates loss of area) space, but with inclusion of land cover this is not predicted. Hence, this may not be a good prediction overall. Trichophorum deergrass Dominant No change Very widespread cespitosum
MONARCH 2 Report – Chapter 9 247 ______
The results from the downscaled model are summarised in Table 9.3, and indicate that there is no large-scale loss of suitable space in the study area, for any of the dominant species modelled. Although, the rare species, Listera cordata is predicted to lose all suitable space from within the case study area under the 2050s High scenario. Both of the recruitment species (B. pubescens and P. aquilinum) continue to experience suitable space, in the vicinity of at least some of the blanket bogs. It is tempting to conclude that within this case study area the dominants fare well and the rare species (L. cordata) is under threat of being lost from the habitat. However caution is needed with this interpretation, because of the poor modelling results and the use of presence-absence data, which may mask significant changes in potential relative abundance. Also, there was only one rare species in this study, making it difficult to generalise.
9.5.1 Climate change and the case study area
Annual mean temperatures in the case study area are predicted to increase by a maximum of 2ºC under the 2050s High scenario (Hulme et al., 2002), but the largest climate change effect will probably be on rainfall. The overall reduction in annual precipitation (0 to –10% under the 2050s High scenario) and the changes in rainfall patterns, resulting in up to 30% drier summers under the 2050s High scenario, may have substantial impacts on the bog habitat for which water relations are crucial.
In the case study area, there is a concern that the risk of peat slides may increase due to climate change. Drying out of peat leads to the filling up of pores with air and causes a switch from anaerobic to aerobic decomposition, which occurs at a rate approximately 50 times faster (Holden et al., 2004), and warmer climatic conditions may also accelerate decomposition. This faster decomposition of peat, together with collapses of the upper peat and of macropores (which can be important drainage channels) causes peat subsidence. This subsidence causes cracking in the peat surface and prolonged drying of peat can switch it from being hydrophilic to hydrophobic in nature, and reduces its ability to absorb rainfall. As a consequence, a greater proportion of the rain falling on desiccated peat may flow over the surface causing erosion of the peat surface (Evans et al., 1999), while ingress into the cracks may result in the development of gullies. Sheet erosion of dried peat by wind can also occur (Wein, 1968). While climate change may exacerbate erosion problems on blanket bog, it should not be forgotten that it is most commonly human impacts, such as overgrazing, that initiate the problem and subsequently render areas vulnerable to such effects under climate change.
9.5.2 Current blanket bog species
Blanket bogs provide a gradient of soil moisture conditions, ranging from wet pools to drier hummocks, and many species occupy distinct regions within this gradient. Even very small changes and fluctuations in water table depth can cause marked changes in vegetation composition in these highly stress-sensitive ecosystems (Holden et al., 2004). For example, Sphagnum moss species are a critical component of the blanket bog ecosystem as they are responsible for the continued production of peat and are important for retaining moisture. However, with peat desiccation (seasonal, annual and long term) Sphagnum may have reduced growth rates, or may even die.
Sphagnum cuspidatum, the species modelled in this study, is characteristic of the wetter areas of blanket bogs, commonly found in the pool and wet hollow areas of patterned bog, as well as in wet lawns. This species gives way progressively to Sphagnum papillosum, Sphagnum tenellum, Sphagnum magellanicum and Sphagnum capillifolium as conditions become drier (Rodwell et al., 2002). S. cuspidatum is known to be adversely affected by the drainage of blanket bogs (Rodwell et al., 2002), and it is probable that climate change may have a similar influence, resulting in a loss of habitat for this species. The climate space modelling predicted such a reduction in suitable space for S. cuspidatum but the inclusion of the land cover data resulted in a projected increase in suitable space.
248 MONARCH 2 Report – Chapter 9 ______
This contradictory result suggests that inclusion of land cover data may have masked totally the climatic effect.
A second species that has been modelled in this study, Rhynchospora alba, is also associated with a similar environment within blanket bogs and it too is known to be adversely affected by bog drainage schemes. This species is intolerant of competition and grows in shallow standing water, as well as on areas of bare wet peat (Preston et al., 2002). Like Sphagnum cuspidatum, it is probable that R. alba, will lose space due to the loss of bog pools (Holden et al., 2004), although this species may be able to colonise some eroded, bare peat surfaces, perhaps in gully bottoms. It is worth noting, though, that all of Ireland is climatically suitable for R. alba under future climate scenarios. However, more local effects, such as land cover and interactions with other species may well add up to a loss of environmental space overall.
Predicting the impact of climate change on Eriophorum vaginatum is more complicated. It has been described as highly dependent on waterlogging and is associated with wet areas of stagnant water (Wein, 1968). However it is tolerant of summer drying of peat, as it is deep rooting (over 50cm). It has been observed to persist for long periods of time after former bog communities have been drained. This species usually regenerates from seeds and as seedlings have shallow rooting systems they will be highly susceptible to drought (Wein, 1968) and this may influence its distribution over the longer term.
Both Trichophorum cespitosum and Empetrum nigrum represent species inhabiting the drier areas of blanket bogs, although E. nigrum has a much more restricted distribution than T. cespitosum. In the case study area, it is uncommon in the lowlands but common in the uplands. Indeed experiments have shown that prolonged waterlogging of the roots of E. nigrum leads to a loss of turgor in the plants and eventual death (Bell and Tallis, 1974), although no significant trend with drainage was observed in a study in the Northern Pennines (Stewart and Lance, 1991). It is likely that these species may expand their distribution on blanket bogs as they dry out. Both these species are found in wet heaths and thus may form important components of the vegetation in the future. However, Table 9.3 suggests that E. nigrum loses half of its suitable space while T. cespitosum maintains its space.
Although the models predict a marked reduction in suitable space for Listera cordata, and its loss from areas of blanket bog, this is unlikely to impact greatly on the bog community. This species is currently rare in the blanket bogs of the case study area (although it may be under-recorded).
This case study area contains a Ramsar site and SPA sites and so is an internationally important location for birds. P. apricaria numbers are low, with a nationally important population estimated at only 12 pairs currently breeding within the Pettigoe Plateau (representing 4% of the Irish population) (Ramsar sites database). Furthermore, although P. apricaria breed on blanket bogs, much of the adult feeding takes place on nearby agricultural land (Pearce-Higgins and Yalden, 2003). Hence, any changes in the distribution or population levels of breeding Pluvialis apricaria is unlikely to have a marked impact on the composition and structure of blanket bogs.
If seasonal drying out of the peat occurs, there may be a shift in the composition of the community towards that of drier areas, perhaps with a reduction in the area of bog and an increase in the area of wet heath vegetation. A shift in the dominance of species would have far reaching impacts on community composition, as would the invasion of B. pubescens or P. aquilinum.
9.5.3 Recruitment species
The model predicts no change in suitable space for B. pubescens, although colonisation may occur in the bogs from nearby seed sources, especially if areas of bog become drier. This species has the ability to quickly colonise bare areas (Atkinson, 1992), such as those that may occur through peat erosion. This tree would represent a new functional type within the bog ecosystem and thus it may be expected to have a large impact on ecosystem function if it were to spread. In addition, it supports
MONARCH 2 Report – Chapter 9 249 ______many species of invertebrates and fungi that do not yet occur in the blanket bog. However, experimental evidence from Clara bog in Ireland, a midland raised bog, suggests that successful colonisation of this species was precluded by phosphorus limitation on a desiccated section of the bog. Irish bogs in general, have very low levels of phosphorus concentrations in the rhizosphere (Tomassen et al., 2004) and this may be an additional limit to colonisation.
Pteridium aquilinum was the second recruitment species modelled, and again the models do not predict any change in suitable space. However, P. aquilinum already grows very close to the edge of the bog at Cuilcagh Mountain and encroachment may occur as areas of the bog become less waterlogged. But, this assumes that the poor mineral content of peat is not too limiting for colonisation by bracken. Although bracken does grow on shallow peat its roots may penetrate through to mineral layer. Like B. pubescens, P. aquilinum would represent an entirely new functional group, associated with a new suite of species, should it prove capable of invading the bog community.
9.6 Discussion and conclusions
It is surprising that there is so little change in the bioclimate classification under the UKCIP02 climate change scenarios, given the apparent sensitivity of the area under the UKCIP98 scenarios. It may be that the reduced number of variables used in this dataset has masked the extent of potential changes.
The species modelling concluded that:
• The European-trained SPECIES model performed well to very well for all species and simulated the whole of Ireland as being suitable climate space for all, but two, of the plant species. • The downscaled species model outputs had particularly low agreement statistics, when compared with those for the other case study areas. The reasons for this need further investigation. • This meant that a decision was made not to run the dispersal model and this meant that no “arrivers” or “leavers” could be predicted so the species community conceptual models could not be applied.
The community composition interpretation concluded that:
• Prolonged seasonal drying of the peat habitat may favour those species that prefer slightly drier conditions in the long term and may drive the community towards a wet heath type of community composition.
9.7 References
Atkinson, M.D. (1992). Betula pendula Roth (B. verrucosa Ehrh.) and B. pubescens Ehrh. Biological Flora of the British Isles. Journal of Ecology, 80, 837-870.
Bell, J.N.B. and Tallis, J.H. (1974). The response of E. nigrum L. to different mire water regimes with special reference to Wybunbury Moss, Cheshire and Featherbed Moss, Derbyshire. Journal of Ecology, 62, 75-95.
Berry, P.M., Vanhinsbergh, D., Viles, H.A., Harrison, P.A., Pearson, R.G., Fuller, R., Butt, N. and Miller, F. (2001). Impacts on terrestrial environments. In: Harrison, P.A., Berry, P.M. and Dawson, T.P. (Eds.) Climate Change and Nature Conservation in the Britain and Ireland: Modelling natural resource responses to climate change (the MONARCH project). UKCIP Technical Report, Oxford.
Dykes, A.P. and Kirk, K.J. (2001). Initiation of a multiple peat slide on Cuilcagh Mountain, Northern Ireland. Earth Surface Processes and Landforms, 26, 395-408.
250 MONARCH 2 Report – Chapter 9 ______
Evans, M.G., Burt, T.P., Holden, J. and Adamson, J.K. (1999). Runoff generation and water table fluctuations in blanket peat: evidence from UK data spanning the dry summer of 1995. Journal of Hydrology, 221, 141-160.
Holden, J., Chapman, P.J. and Labadz, J.C. (2004). Artificial drainage of peatlands: hydrological and hydrochemical process and wetland restoration. Progress in Physical Geography, 28, 95-123.
Hulme, M., Jenkins, G.J., Lu, X., Turnpenny, J. R., Mitchell, T.D., Jones, R.G., Lowe, J., Murphy, J.M., Hassell, D., Boorman P., McDonald, R. & Hill, S. (2002). Climate Change Scenarios for the United Kingdom: The UKCIP02 Scientific Report. Tyndall Centre for Climate Change Research, School of Environmental Sciences, University of East Anglia, Norwich, UK, 120 pp.
Pearce-Higgins J.W. and Yalden, D.W. (2003). Variation in the use of pasture by breeding European Golden Plovers, Pluvialis apricaria in relation to prey availability. Ibis, 145, 365-381.
Pearson, R.G., Dawson, T.P. and Lui, C. (2004). Modelling species distributions in Britain: a hierarchical integration of climate and land-cover data. Ecography: in press.
Preston, C.D., Pearman, D.A. and Dines, T.D. (2002). New Atlas of the British Flora. Oxford University Press, Oxford, 910pp.
RAMSAR Sites Database. A directory of wetlands of international importance. United Kingdom 7UK105. Pettigoe Plateau. www.wetlands.org/RDB/Ramsar_Dir/UnitedKingdom/UK105D02.htm.
Rodwell, J.S. (Ed.), Pigott, C.D., Ratcliffe, D.A., Malloch, A.J.C., Birks, H.J.B., Proctor, M.C.F., Shimwell, D.W., Huntley, J.P., Radford, E., Wigginton,M.J. and Wilkins, P. (1991). British plant communities. Volume 2. Mires and heaths. Cambridge University Press, pp628.
Schouten, M.G.C. (1984). Some Aspects of the Ecogeographical Gradient in the Irish Ombrotrophic Bogs. Proceeding of the 7th International Peat Congress, Dublin. Vol 1, pp 414-432.
Stewart, A.J.A. and Lance, A.N. (1991). Effects of moor-draining on the hydrology and drainage of Northern Pennine blanket bog. Journal of Applied Ecology, 28, 1105-1117.
Tomassen, H.B.M., Smolders, A.J.P., Limpens, J., Lamers, L.P.M. and Roelofs, J.G.M. (2004). Expansion of invasive species on ombrotrophic bogs: desiccation or high N deposition? Journal of Applied Ecology, 41, 139-150.
Warburton, J., Holden, J. and Mills, A.J. (2004). Hydrological controls of surficial mass movements in peat. Earth-Science Reviews, 67, 139-156.
Wein, R.W. (1973). Eriophorum vaginatum L. Biological Flora of the British Isles. Journal of Ecology, 61, 601-615.
MONARCH 2 Report – Chapter 10 251 ______
10 Conclusions and Future Research
P.M. BERRY, G.J. MASTERS, P.A. HARRISON, J.E. HOSSELL, H.Q.P. CRICK S.N. FREEMAN, N.E. ELLIS, H.G. ORR, C.A. WALMSLEY AND O. WATTS
10.1 Introduction
In this chapter, a synthesis of the outputs of MONARCH 2 is provided by: • summarising the methodological developments; • drawing conclusions for each of the case studies and identifying generic conclusions; • cataloguing the model limitations and uncertainties; • assessing the impact of downscaling on the models and the availability of suitable data inputs; • identifying key areas for future research while distinguishing between those issues that will be addressed during MONARCH 3 and those that require consideration outside the project.
Finally, key conclusions derived from the MONARCH 2 outputs are provided.
10.2 Methodological Developments
MONARCH 1 used the SPECIES model to simulate the impacts of climate change on the potential climate space for selected species in Britain and Ireland (Hossell et al., 2001). Other researchers (e.g. Midgley et al., 2002; Midgley et al., 2003; Peterson, 2003; Thomas et al., 2004;) have developed approaches similar to MONARCH using bioclimatic classification of the landscape, the application of climate change scenarios, generation of a climate envelope and some form of dispersal. However, MONARCH used a neural network to generate the climate envelope, which compares very favourably with more traditional modelling approaches (Pearson et al., 2004). Future research needs were identified during MONARCH 1, including better climatic data, improved understanding of the relationship between climatic parameters and species’ distributions and a consideration of climate change in a broader context (Hossell et al., 2001). Some of these issues were addressed through three main methodological developments in MONARCH 2: • The addition of land cover and land cover change as a surrogate for habitat availability, since for species’ distributions at a finer scale this may be a more important control than climate on species’ presence and absence (Pearson et al., 2004). • Land cover change scenarios were applied for selected species together with a dispersal model to assess the ability of species to move into new climate space thus indicating their potential vulnerability to climate change. • The results of the species modelling were applied to conceptual models of community composition to consider the potential impacts of local extinction or species dispersal on local communities.
The revised MONARCH 2 methodology was tested and evaluated in four case study areas. These areas were selected using the MONARCH 1 bioclimate classification to identify areas sensitive to climate change. A key element of the testing was to examine the feasibility of using the method at a regional rather than national scale to determine impacts upon habitats of nature conservation interest at a landscape level. In this way it would be possible to assess impacts on suites of statutory sites, e.g. montane habitats in the Welsh and Scottish uplands. The revised method produced a suitability surface based on land cover and climate, and while this did not always prove successful when simulating future scenarios, it does indicate the importance of land cover for current distributions at finer resolutions. The dispersal model is population based, as opposed to the MIGRATE model (Collingham et al., 1996) which is individual based, facilitating a larger scale approach and potentially enabling the linkage with the species interaction models.
252 MONARCH 2 Report – Chapter 10 ______
10.3 Key impacts on species in the case study areas
10.3.1 Conclusions from the case study areas The four case study areas (Hampshire, Central Highlands, Snowdonia and Cuilcagh/Pettigo) represent lowland and upland systems, a north-south gradient and to a lesser extent a west-east gradient. All the selected areas were identified as regions where the bioclimate in the 2050s (High emissions scenario) was predicted to be beyond the baseline classification for Britain and Ireland. Key conclusions are presented for each of the case study areas followed by a summary of the generic conclusions.
10.3.1.1 Hampshire case study Hampshire may experience climatic conditions well beyond those currently seen in Britain and Ireland by the 2050s, more akin to continental Europe. The land cover change model showed that there could be a large reduction in grassland extent. This may be due to the climate conditions exceeding those on which the model was trained and may not be a reflection of the fact that the land cover cannot exist in these areas. The sensitivity of the area to changes in climate under the 2050s High emissions scenario, as shown in the bioclimate work, is not so strongly reflected in the plant species modelling results. This is partly due to the choice of long-lived tree species within the Beech Hangers habitat and to the robustness of many of the dominant heath species. Dog’s mercury (Mercurialis perennis), yellow-necked mouse (Apodemus flavicollis) and the wet heathland plants may be able to expand their distribution due to unoccupied suitable environmental space, while the birds, hawfinch (Coccothraustes coccothraustes) and curlew (Numenius arquata), show potentially significant losses of suitable space under the 2050s High emissions scenarios; as does the bog bush- cricket (Metrioptera brachyptera) in the New Forest. On the estuaries there are no predicted changes in the number of oystercatcher (Haematopus ostralegus) and redshank (Tringa totanus) as coastal defences are maintained into the future so that the area of intertidal mudflats remain similar.
10.3.1.2 Central Highlands case study Central Highlands may experience a loss of montane environments with an increase in extent of areas classified as upland. The land cover model indicated a decline in the extent of montane habitats and a concurrent increase in extent of neutral grassland under future climate change scenarios. The fact that it also predicted a complete loss of dwarf shrub is contrary to the SPECIES and dispersal modelling, which indicated little or no change in the suitable space for ling (Calluna vulgaris), bilberry (Vaccinium myrtillus) and cowberry (Vaccinium vitis-idaea). This may be because of the only moderate match between dwarf shrub land cover data and modelled space or because these species are already extensive in the area. The SPECIES and dispersal modelling indicated suitable space remains extensive across the study area for species characteristic of montane heaths, both stiff sedge (Carex bigelowii) and ptarmigan (Lagopus mutus), even under the 2050s High emissions scenario. This may be due to the coarse scale of modelling (climatic data at 5km2) lacking sensitivity to indicate changes in an area of varied altitude. When land cover change was incorporated for stiff sedge then loss of montane habitat led to a decrease in suitable area.
Scots pine (P. sylvestris) is predicted to remain a dominant species in Caledonian pinewoods. The SPECIES and dispersal modelling indicated very little change in extent for P. sylvestris or silver birch (Betula pendula) within the next 50 years owing to their longevity and limited capacity for dispersal resulting in an inherent time lag in responding to an altered distribution of suitable climate space. Hairy wood ant (Formica lugubris) is currently rare in the study area (and possibly under-recorded), yet the modelling indicated considerable potential for dispersal under future climatic scenarios.
10.3.1.13 Snowdonia case study The bioclimatic modelling showed Snowdonia to be sensitive to climate change. However, these sensitivities are not reflected, as a whole, by the results for the species modelled. None of the dominant species in upland and montane heaths are predicted to lose suitable space. This lack of response may be a result of the reduced climatic sensitivity of the downscaled SPECIES model due to
MONARCH 2 Report – Chapter 10 253 ______the inclusion of land cover. All the upland/montane species showed a high potential to disperse into the simulated suitable space and in the case of the recruitment species, western gorse (Ulex gallii) and bracken (Pteridium aquilinum), their potential invasion of montane and upland heath habitat at higher altitudes is of particular concern for nature conservation. The Arriver model suggests that colonisation by P. aquilinum and/or U. gallii will have far-reaching effects on community composition and structure. The order of colonisation and competition effects will influence the exact nature of any changes to community composition.
Within the upland oak woods, the dispersal model showed that while sessile oak (Quercus petraea) had some potential to disperse, this was limited by its long time to reproductive maturity. Other factors that constrained species’ dispersal were the availability of suitable new areas (bluebell, Hyacinthoides non-scripta and common cow-wheat, Melampyrum pratense) and altitude (M. pratense). In upland oak woodlands, the loss of pied flycatcher (Ficedula hypoleuca) from some woods due to restricted climate space is not predicted to significantly impact upon the species composition of the community, according to the Leaver model.
For a few species, most notably bracken, the results are contradictory. The land cover modelling predicts a decline in suitable space for bracken while the downscaled SPECIES model predicts no significant change in extent, and yet experimental evidence suggests that with reduced frosts bracken will be able to spread altitudinally. There are a number of potential reasons for the apparently contradictory results including: i) frost and wind exposure are not explicitly included in model inputs, ii) the land cover change model is at a different resolution (5km as opposed to 1km) and is working with only climate data as opposed to climate and land cover in the downscaled SPECIES model iii) the addition of land cover in the downscaled SPECIES model may suppress some of the climate response iv) the distribution for P. aquilinum was derived by combining the bracken classes for LCM2000 and the Phase 1 survey, but there is poor agreement between them, so once again the two models are not directly comparable as they utilise different data. It is likely that a combination of these factors is responsible for the differences observed.
10.3.1.4 Cuilcagh/Pettigo Peatlands case study The modelling predicts that the Cuilcagh/Pettigo peatlands will experience little change by the 2050s, although some of the area would become warmer. This is surprising given the apparent sensitivity of the area under the UKCIP98 scenarios. It may be that the reduced number of variables used in this bioclimate dataset has masked the extent of the changes. This would lead to little change in land cover. Under the climate change scenarios, Ireland as a whole would become less suitable for all the species, apart from white beak-sedge (Rhynchospora alba) and P. aquilinum, with loss of suitable climate space particularly occurring in the south-west. The addition of CORINE land cover in the downscaled model did not provide a good match between the simulated suitable space and observed distributions. This could be because peat bog is present in most squares at the 10-km resolution. The poor performance of the downscaled model meant that the dispersal model was not run. It is possible that if warming leads to seasonal drying out of peat, which the SPECIES modelling does not pick up, there may be a shift in the composition of the community towards that of drier areas, perhaps with a reduction in the area of bog and an increase in the area of wet heath vegetation.
10.3.2 Generic conclusions from the case study areas Overall, there was no or little change in the area suitable for many dominant species in the case study areas over the time frame of the project’s scenarios (up to the 2050s). This may reflect the life history of the dominant plant species involved; some have longer life spans than the time frame of the scenarios, the majority are long-lived perennial species, often in already harsh conditions (temperature and rainfall), and so are very slow to respond (e.g. Grime et al., 2000). However, the limited response also reflects the fact that most of the selected species are not at their ‘edge of range’ at the case study sites. It is also important to appreciate that these bioclimate, land cover and dispersal models predict presence-absence but most species are likely to show significant and ecologically important abundance changes ahead of any change in distribution. It is also important to recognise that climate space changes are based on mean changes in bioclimate while it is likely that changes to the scale and
254 MONARCH 2 Report – Chapter 10 ______frequency of extremes of drought, flooding and wind speed will be significant in changing community composition.
Some species did show the potential for dispersal, e.g. B. pendula, F. lugubris, and A. flavicollis. Their effect on the communities they recruit into will depend largely on whether other species already fill the functional niche they could occupy and their relative competitive abilities. The dispersal model indicated significant spread among some species to higher altitudes at the montane sites, e.g. U. gallii and P. aquilinum in Snowdonia, but whether this is realistic is uncertain. Existing distributions suggest that exposure rather than climate per se is an important factor controlling distributions at altitude and that this is not adequately covered by the bioclimatic data used for modelling. Montane species, such as Carex bigelowii, that are predicted to spread to lower altitudes according to the downscaled SPECIES model, are currently altitudinally limited so with rising temperatures their spread to lower altitudes seems highly unlikely. Use of climatic data at 5 km2 in topographically and altitudinally varied environments is not sensitive enough to reveal local shifts in species uphill. It would appear that the downscaling has failed to adequately take account of altitude and importantly exposure due to the lack of wind variables within the bioclimate modelling.
The land cover modelling showed a high potential for loss in, for example, the montane and dwarf shrub land cover classes, but gain in neutral grassland. These changes are not reflected so clearly in the species modelling, partly as a result of few sensitive species confined to these habitats being chosen for analysis (and partly because the downscaled SPECIES model which can incorporate land cover change was only run for a few species). The conceptual models of community composition and the ecological literature indicate potential direct and indirect effects from climate in the selected habitats. Unfortunately, in some cases species responses are contradictory, for example, the distribution of bracken is shown to decline in Snowdonia according to the land cover modelling, remain fairly unchanged based on the SPECIES model outputs while experimental evidence and the conceptual model of community composition suggest that bracken may spread attitudinally. While other species show more consistent responses, it would appear that the modelling at the regional scale has failed to adequately integrate some of the key factors influencing distribution at the landscape scale; this is particularly so in the mountainous case study areas. The uncertainties and limitations that have led to these problems are explored more fully in the following section.
Some similar species and habitats were chosen for different case study areas and a comparison of their responses indicates the regional nature of climate change impacts. Calluna vulgaris was chosen to represent lowland heath (Hampshire), as well as upland heath (Central Highlands and Snowdonia). In Hampshire, it has a restricted distribution (Chapter 6.5.1) and its dispersal is limited by available suitable space (Figure 6.32). In the other two case study areas its suitable area occupies almost all of the study area and thus there is little change in the future. In all three, it is not sensitive to climate change and could continue as a dominant. The same is true of V. myrtillus, which was chosen to represent upland heath in the Central Highlands and Snowdonia. In both cases, it is observed throughout the study area and is simulated to occur in all suitable areas in the future. Q. petraea was selected as a component of Caledonian pine woodland (Central Highlands) and upland oak woodland (Snowdonia). In both cases it showed a limited potential for expansion. Carex bigelowii was chosen for the montane habitat in the Central Highlands and Snowdonia. In the former it already occupies most of the land over 600m and thus there is limited opportunity for dispersal, whereas it has a much more restricted distribution above 700m in Snowdonia, where it is at its southern range margin in Britain. The modelling suggests that it has the potential to disperse to cover quite a large proportion of the higher ground (Figure 8.29). This is uncertain given that species, such as C. bigelowii, are at their range margin and thought to be at risk from climate change.
The only common habitat within the case study areas was montane/upland heath, which was chosen for both Central Highlands and Snowdonia. The models predict little shift in the suitable space for the modelled species, although there is some opportunity for expansion to higher altitudes. The conclusions for both areas were similar in that changes in the relative abundance of the various heathland species may occur, with increasing temperatures allowing them to become dominant at
MONARCH 2 Report – Chapter 10 255 ______higher altitudes, although exposure may become a limiting factor and changes in soil moisture levels may result in an increase in drier communities. Other environmental and management factors will play a role in affecting the outcome of climate change. Overall, where comparisons could be made, there were similar species’ and habitat responses once differences in the availability of suitable space were taken into account.
In the case study areas, at least one recruitment species was chosen to test the Arriver conceptual model (Chapter 4). In most cases, species already present in the ecosystem were chosen (C. vulgaris – Hampshire and Snowdonia; Q. petraea – Central Highlands; Pteridium aquilinum and U. gallii – Snowdonia; Betula pubescens and P. aquilinum – Cuilcagh/Pettigo). The dispersal model indicated they had potential to spread, where there was unoccupied suitable space and the Arriver model suggested that several of them could have a significant impact on community composition. Although willow tit (Parus montanus) was selected as a recruitment species for Caledonian pine woodland, it did not reach the case study area under the UKCIP02 scenarios.
10.4 Uncertainties and limitations of the models and data inadequacies
10.4.1 Model uncertainties
Uncertainty in the modelling work has not been explicitly addressed in MONARCH 2. Uncertainty can come from a number of sources, such as the quality of data about a species, scenarios, ecological models and their parameterisation.
10.4.1.1 Uncertainty in distribution maps There is an accuracy/uncertainty issue associated with the European distribution maps used for training the SPECIES model, as they were derived in different ways. Given the sensitivity of the model to presences and absences, any inaccuracy could affect model training; this needs further exploration. Also the size of the sample training set and species’ prevalence for model-building data can have effects on model performance (Liu et al., submitted).
10.4.1.2 Scenario uncertainty The climate scenario uncertainty stems from all UKCIP02 scenarios being based on a single regional climate model (HadRM3), but these are compared to other GCM scenarios in Chapter 2.9 and in the UKCIP02 scientific report (Hulme et al., 2002). The magnitude and pattern of change in many parameters is similar to the UKCIP02 scenarios, but there are some notable differences, especially in summer precipitation. Incorporating the results from these other GCMs illustrates a wider range of scenarios than is captured by the UKCIP02 scenarios and thus model outputs must be seen as not encompassing the full range of possible outcomes.
10.4.1.3 Algorithmic and neural network uncertainties There are a number of possible sources of algorithmic uncertainty coming from the ecological models. Firstly, the decision threshold for converting the SPECIES suitability surface to a presence/absence map can be derived and determined in different ways (Chapter 3.3.1). In MONARCH 2 the Receiver Operating Characteristic (ROC) curve procedure was used to identify three probability thresholds for the likelihood of suitable climate space under any scenario, and thus indicate the uncertainty in this modelling component. Secondly, the apparent insensitivity of the downscaled model to climate change (very few species showed a change in suitable space) could be a function of the climate suitability surface being a single input variable compared with the 22 land cover types input for Great Britain and 38 for Ireland. It is also likely to reflect the importance of land cover in species’ distribution at this scale.
Although using a neural network to generate a climate envelope compares favourably with more traditional modelling approaches (Pearson et al., 2004), neural networks are generally perceived to be a ‘black-box’ technique, meaning that the workings of the model are difficult to disentangle, so that
256 MONARCH 2 Report – Chapter 10 ______the relative importance of the different land cover classes in determining species’ distribution is not readily identifiable from the network structure. The main limitation is that it is not easy to work out the internal workings of the network, so using more traditional statistics is generally preferred if this is the aim. An analysis of the network structure to identify key driving variables and comparing the results to those from other statistical techniques, such as General Linear Models (GLM) is a possible area for future research.
10.4.2 Model limitations
10.4.2.1 Bioclimatic classification The bioclimatic classification did not consider any climates beyond Britain and Ireland. In some parts of the country, such as the Hampshire case study area, it is expected that conditions in the future might be unlike anything that is currently experienced. Similarly, the technique did not allow any new classes to be created.
The selection of the case study areas using the bioclimatic sensitivity identified by the classification work was made not solely on the basis of climate, but also considered the interests of the Steering Group, the conservation interest of the case study areas and the availability of data, thus introducing a known, selective bias. The bioclimatic work started from the assumption that the baseline data provided by the UKCIP98 and UKCIP02 scenarios would be reasonably similar and that the shift to the new datasets would not greatly affect the baseline classification of the case study areas. The work undertaken shows that this is not the case and that differences in the spatial scale and method of interpolation of baseline datasets can add to the uncertainty inherent in climate change impacts modelling.
10.4.2.2 SPECIES modelling The SPECIES model assumes that the current distribution used for model training is in equilibrium with the climate and this relationship occurs in the future, and that the climate is the dominant factor affecting species’ distributions. The validity and limitations of this bioclimate envelope approach have been fully discussed in Pearson and Dawson (2003) and Pearson (2004). They demonstrated that for certain species and particularly at the continental-scale, the bioclimate envelope modelling approach can provide useful results. At smaller resolutions, other factors, such as habitat availability (land cover) and competitive interactions become important. It was this hierarchy of factors affecting species’ distributions that led to the methodological developments in MONARCH 2.
The failure of the inclusion of land cover to always improve the relationship between the observed and simulated species’ distribution, may show that at finer resolutions climate, or variables closely associated with climate, can still be important. Or, as in the Irish case study area, the critical land cover(s) may be too widespread for the model to discriminate the relationship. However, the lack of relationship is also due to important variables not being included in the model explicitly, such as soil pH and competitive interactions. The failure to include species’ interactions, which may be particularly significant at southern range margins, is seen as serious problem by Davis et al. (1998 a, b) and Hampe (2004). This limitation is particularly pertinent to those ‘limit of range’ species examined in Snowdonia and Central Highlands, e.g Carex bigelowii, and may in part explain the anomalous migration of montane species down the altitudinal gradient under climate change. A further caveat is that the model also includes no element of evolutionary change, as discussed in Pearson and Dawson (2003).
10.4.2.3 Land cover modelling The land cover change model utilised data from the Land Cover Map 2000 and CORINE, which was available at a 1-km resolution. However, climate data was only available at a 5-km resolution. Analysis can only be done at the level of the lowest resolution dataset, otherwise spurious accuracy is introduced into the analysis. Hence, the land cover data were aggregated to this level. Only if a land cover was present in more than 5% of the area was it marked as being present in a 5-km square. This
MONARCH 2 Report – Chapter 10 257 ______was to remove errors in interpretation of the satellite imagery. The base data classified in this manner was then used to create the 5 x 5 km surfaces on which the models were based. The land cover work assumes, therefore, that using this method the LCM2000 map provides a reasonable prediction of the presence-absence of all land classes modelled. Unfortunately, the comprehensive Phase 1 vegetation survey data available in Wales showed that even using this 5% cut off threshold serious errors in the classification of habitats within LCM2000 introduced a major limitation to the modelling. For example, calcareous grassland was mis-identified by LCM2000 throughout Snowdonia providing a false baseline.
The modelling assumed that all land cover types are affected to a similar degree by climate change; and the influence of other drivers interacting with climate change to influence land use/cover was not addressed. These simplifications are unlikely to hold true as climate change drives developments in countryside management, from new farming opportunities that aim to exploit climate change to projects that build resilience to climate change. Similarly the assumption was made that any changes projected are assumed to occur instantaneously and without interaction/influence from neighbouring land cover types and that land cover types excluded from the analysis are not significantly affected by climate change. These issues are beyond the scope of the modelling but should be considered in interpretation of the model outputs.
Accepting that many of the case study areas are predicted to have bioclimatic conditions outside of the current British and Irish range by the 2050s, it is a significant weakness that the land cover models were based solely on data for Britain and Ireland. There is a need to extend the range of climatic conditions over which the land cover model was developed using European climate and land cover data in order to capture the full range of class types that may exist under future climatic conditions. Further, some land cover types could not be modelled using the presence-absence approach, as their coverage is too great for a climatic restriction on their distribution to be detected within the bounds of Britain and Ireland, and thus changes in landscape within an area could not be modelled.
For certain land cover classes, such as those dominated by agricultural production or forestry production, it is clear that the Common Agricultural Policy (CAP), Rural Development Plan measures and other land use and agricultural policies will have at least as great as influence on the future location of such land cover types as climate. This is not a weakness of the modelling work as it never sought to address this issue, but it is an important caveat.
The land cover modelling took no account of the effect of climate on soil itself. The implications for land cover types occurring on peat soils, for example, also need to examine how peat loss or formation will be affected under climate change. The predicted changes in distribution of bog do not adequately take account of the hydrological and topographical requirements for changes in the distribution of this habitat.
10.4.2.4 Composition of species communities In order to address some of the caveats outlined above models of community composition were developed to consider the competitive interactions and impacts of ‘arrivers’ and ‘leavers’. However, the conceptual models and the interactive species matrices can only explore scenarios based on available autecological data. As a result, it was difficult to quantify impacts due to data limitations.
Extreme events such as storms and droughts may have large impacts on species success and hence community composition, but only influence the models through averaged climate data. It is likely that the interaction between climate change and other pressures, such as land management and pollution in combination with extreme events could be key in driving responses and community composition.
258 MONARCH 2 Report – Chapter 10 ______
10.4.3 Downscaling limitations
10.4.3.1 Reduction in climatic sensitivity by the inclusion of land cover The addition of land cover in the downscaled SPECIES model improved model results in a number of cases, by providing a better match between the observed and simulated current species’ distribution. In the case of species like Calluna vulgaris, purple moor grass (Molinia caerula) and Vaccinium myrtillus that are simulated to have suitable climate space throughout Britain, the addition of land cover restricts the simulated suitable space, recognising that land management is an important factor affecting their distribution. Other land-use and land management factors are driving habitat existence and playing a significant part in distribution at a local/regional scale. Similarly, the observed distribution of Melampyrum pratense is more limited than suggested by climate alone and the addition of land cover leads to a better, but slightly over-restricted simulation. Hyacinthoides non-scripta also occurs throughout Britain, except in the parts of the Highlands of Scotland and round the Wash. The downscaled model picks up the pattern slightly better than the climate-only simulation. The addition of land cover, therefore, seems to be important in constraining the simulated current suitable space of species that are climatically simulated to occur (almost) throughout Britain. This, however, is not the case for the three trees, beech (Fagus sylvatica, including its planted range), ash (Fraxinus excelsior), Betula pendula and also for Pteridium aquilinum, where there is little difference between the two simulations and for Mercurialis perennis, where the two simulations produce different patterns at the northern range margin that do not match the observed distribution particularly closely. For northern species, such as Carex bigelowii, Vaccinium vitis-idaea and Scots pine (Pinus sylvestris), the addition of land cover leads to a restriction in the simulated current suitable space in Snowdonia and northern England and for Ulex gallii it provides a more appropriate western simulation. Another group of species are those with a more fragmented distribution, like Formica lugubris and Metrioptera brachyptera, which have a very scattered distribution and while most of Britain is climatically suitable the addition of land cover restricts this, but still provides too broad an envelope (e.g. Figure 7.16). For Quercus petraea, which has a wider distribution, it leads to an over-restriction in southern and eastern England. Nevertheless, in the majority of cases the addition of land cover has led to a closer match between the observed distribution and simulated current suitable space, but this improvement is not always so good in the case of fragmented distributions.
The addition of land cover, however, led to the masking of the response of species to the climate change scenarios outputs from the SPECIES model, as discussed in Chapter 3. This resulted in no change in future suitable space for many species, although some showed loss in East Anglia and Central England (see Figures 7.16 and 8.25). An alternative means of including land cover at the finer resolution, needs to be devised, such as applying masks, so that the climate signal is not obscured.
The land cover change scenarios were applied to a few species, in order to test the methodology. This was not possible in the case of Hampshire, as the relevant land cover types were not modelled (Chapter 3). In the Central Highlands, Carex bigelowii was selected, as it is associated with montane habitats and loss of this habitat has the potential to restrict its future distribution. This is illustrated by the model under the 2050s High emissions scenario (Figure 7.24), where suitable space is lost in the south-east of the case study area. The application of land cover change scenarios in the downscaled SPECIES model did, therefore, for the species selected, lead to changes in their suitable space.
10.4.3.2 Parameterisation of the dispersal model The apparently different abilities of species to migrate and thereby track changes in their climate space was a function of the parameterisation of the dispersal model. Many species were credited with long distance dispersal on the basis of producing small light seeds (e.g. Betula pendula and hare’s-tail cotton grass, Eriophorum vaginatum) or their inherent mobility (e.g. birds). Their maximum and mean dispersal distance, however, varied. The largest maximum dispersal distance, for non-bird species, was for Apodemus flavicollis and this led to it potentially spreading through much of Hampshire and even onto the Isle of Wight (Figure 6.20). Other species with a maximum dispersal distance of 10 km or more (Metrioptera brachyptera, Formica lugubris, Quercus petraea, Vaccinium myrtillus and
MONARCH 2 Report – Chapter 10 259 ______
Vaccinium vitis-idaea) did not always have high potential dispersal, because other factors constrained the process. Q. petraea, for example, was constrained by the long time to reach reproductive maturity. The importance of long-distance and a high maximum dispersal can best be seen in the case of M. brachyptera, because many of the other species have little new space to colonise. The sensitivity of the model to the different parameters needs more exploration. Some form of validation could also be attempted against the palaeoecological record, but it should be borne in mind that in the past species were moving relatively unhindered across the barrier free landscape. Also at the case study area scale, suitable space can be an important constraint, as was the case with B. pendula.
10.4.3.3 Case study area location and species selection Beech hangers (Hampshire), native pine woodland (Central Highlands) and upland oak woodland (Snowdonia) showed very limited direct change, partly due to the longevity of the dominant tree species over the relatively short time-scales considered. The species modelled for each habitat were selected on the basis of their importance within the case study areas and habitats studied (e.g. dominant, flagship, rare) and the majority are not located at their ‘edge of range’. Given the limited number of species that could be modelled, the species are considered representative of the habitats chosen. It would not have been appropriate to select ‘edge of range’ species a priori to demonstrate regional changes in climate space. Most species show greater sensitivity at their range margins, although, at the national scale, eastern central England and East Anglia consistently are the areas likely to show loss of suitable climate space. No case study areas were located near here, although it was identified as an area of high sensitivity to climate change.
10.4.4 Data limitations Differences in input data or their derivation is one possible source of the differences between the different resolution of models and also between case study areas.
10.4.4.1 Climate Data The 10-km grid for the UKCIP98 scenarios provided an appropriate resolution to derive the Baseline98 classification for MONARCH1; the more detailed 5-km dataset (UKCIP02) would have been computationally unmanageable. However, UKCIP02 was used for the more detailed case studies. Unfortunately, the two UKCIP baseline datasets have been developed using different techniques with different source climate data; this makes it impossible to determine why the datasets produce such different climate classifications over the case study areas. The provision of baseline climate data for Ireland has largely been resolved through the British Irish Council (BIC), again at a 5-km spatial resolution. However, not all the climate variables required for modelling were available from the BIC, e.g. wind speed or cloud cover data, so these variables were simply downscaled from the 10-km Irish climatology data developed during MONARCH 1. Monthly values of solar radiation and potential evapotranspiration also had to be derived. It was surprising that there was so little change in the bioclimate classification under the climate change scenarios in Ireland, given the apparent sensitivity of the area under the UKCIP98 scenarios. It may be that the reduced number of variables used in the 5-km Irish dataset has masked the extent of the changes. The need to compute these climatic variables, poses an issue of compatibility with other UKCIP studies. These variables are used widely in climate change impact studies and their provision as part of the baseline datasets would be a valuable resource and ensure standard methods are utilised for deriving such variables throughout UK studies.
The UKCIP02 climate change scenarios were provided on a 5-km grid for Britain. However, compatible scenarios had to be created for Ireland by downscaling the HadRM3 change fields provided by UKCIP on a 50-km grid to the Irish 5-km baseline climatology. A simple downscaling technique was applied whereby higher resolution climate change scenarios were produced by directly applying the 50-km climate change fields to the higher resolution baseline climatology 5-km grid. This method adds no new meteorological information and assumes that the spatial pattern of current (i.e. 1961-90) climate remains the same into the future. However, this method is consistent with that used to create the UKCIP02 5-km scenarios for Britain.
260 MONARCH 2 Report – Chapter 10 ______
10.4.4.2 Land cover data Land cover data were available (LCM2000 for England, Scotland and Wales and CORINE for Ireland) for input into the downscaled SPECIES model. The use of two different land cover data sets means that different land cover classes were used for the land cover change work and as inputs to the downscaled SPECIES model. This affects the transferability of results between Ireland and the other countries. The usefulness of LCM2000 is limited by poor discrimination and mis-matching of classes, particularly at satellite image tile boundaries for certain cover classes (e.g. set-aside grass, arable cereals, arable non-rotational and inland bare ground). Data from tiles could have been merged but in this study they were omitted. Data to verify LCM2000 are limited but there is very poor agreement between LCM2000 and the comprehensive Phase 1 vegetation survey data for Snowdonia (Chapter 8). The poor discrimination of LCM2000 classes is insignificant at the Britain and Ireland scale while at the regional case study scale these errors become highly significant resulting in substantial errors in the land cover baseline, e.g. calcareous grassland.
10.4.4.3 Soils data At the 10-km resolution, there are issues of compatibility, availability of derived data (e.g. available water-holding capacity), and matching of different datasets (see Chapter 1). The availability of soils data at a resolution finer than 10-km is an ongoing problem in Scotland and Ireland. There are reservations about extrapolating the 10-km resolution MLURI soils data to finer resolutions. There is still no digitised soil database for Ireland and thus no advance has been made on this since MONARCH 1 (Harrison et al., 2001).
10.4.4.4 Species distribution data Some taxa are not well mapped at European scales. Fine resolution data at 1-km scale were available for the Hampshire case study but had to be derived by association between species and land cover types, expert opinion and field knowledge for other case study areas (see Chapters 7-9). The impacts of different methods of deriving the 1-km resolution species’ distributions have not been explored. Caution, therefore, needs to be exercised in interpreting and comparing the results of the dispersal models, as these derivation methods for populating the landscape for the dispersal model could lead to differences between species’ results and between case study areas. It would be beneficial to focus biodiversity recording efforts, at least for priority species, to a 1-km resolution, to match many other datasets. It would enhance research at the mesoscale into factors affecting species’ distributions.
Data on the habitats within the case study areas were mainly required to indicate which species were present within a habitat. Good data was available for the Snowdonia National Park and appropriate data was obtained for the Hampshire and the Central Highlands case study areas. No data was readily available for the part of the Irish case study area in the Republic. Species lists for statutory sites were very patchy, with groups other than higher plants and birds, often not included. This lack of information for important community components made it extremely difficult to determine whether interactions between species were likely to occur. Furthermore, species lists were often presented for an entire site with no indication on within which habitats each species occurred.
10.5 The way forward
10.5.1 MONARCH 3 The downscaling of the MONARCH 1 SPECIES model to a landscape scale uncovered some key limitations, principally due to data issues at the 1-km square resolution. The third phase of MONARCH will therefore focus on maximizing its practicality and usefulness as a tool for conservation planning through its application at the 5-km square resolution for Britain and Ireland. MONARCH 3 aims to improve the understanding of the potential impacts of climate change on biodiversity through four project objectives:
• Objective 1: Automate the SPECIES modelling process One serious constraint during the MONARCH research programme to date has been the limited number of species selected for modelling. Modelling has been a time-consuming and, therefore, costly
MONARCH 2 Report – Chapter 10 261 ______process. Hence, the application of the climate envelope ‘SPECIES’ model will be automated, enabling a much larger number of species to be studied. It will broaden the potential application of the outputs in relation to the development of nature conservation policy and practice.
• Objective 2: Apply MONARCH methodology to a wide range of taxa in Britain and Ireland The targets set within the UK Habitat Action Plans (HAPs) and Species Action Plans (SAPs) are being reviewed during 2005/6. MONARCH outputs could inform the review process. In order to provide generic guidance on the impacts of climate change and inform future reviews, the MONARCH methodology will be applied to a selection of some 120 UK and Irish SAP species. The model will provide both unconstrained simulated changes in suitable climate space and masked simulations based on the suitability of land cover classes.
• Objective 3: Validate ‘SPECIES’ model through hindcasting Validation of models against real data provides the only objective test of model performance. In order to validate the SPECIES model, twelve species, including birds, plants and invertebrates, will be examined under both current and past climate to identify where observed distributions are outside their predicted climate space. Distributional discrepancies will be related to the historical distribution data. • Objective 4: To assess and quantify the uncertainty involved in the modelling process Uncertainty is inherent in any form of modelling and has been recognised by the IPCC as a key developmental area in its assessment process. In 10.4, we have sought to identify the different sources of uncertainty in the MONARCH modelling methodology. Further assessment of uncertainties will focus on (a) greenhouse gas emission scenarios, (b) climate model based uncertainty, and (c) natural variability. The impact of these uncertainties upon the SPECIES model will be assessed for six species. This should help identify ways in which the model can be improved and, by providing greater transparency, should enable conservation policy-makers and practitioners to make more informed decisions based on the MONARCH outputs.
10.5.2 Other research challenges Research challenges that need to be addressed to enable the application of MONARCH at the local/regional level as tested in this second project phase are outlined below. The main areas requiring further consideration are: (1) lack of suitable data, (2) incorporating land cover and climate space at the landscape scale, (3) including the effects of random dispersal events, and (4) addressing community effects and other drivers for change.
10.5.2.1 Lack of suitable data Weaknesses of the data sets used (see Box 10.1) should be disseminated to organisations involved in data gathering with a view to improving the collection and presentation of datasets to further their use in modelling studies. The paucity of European range data for a number of taxa is a major issue while at the opposite end of the scale there is very limited 1-km scale distribution data for many areas. Information on species interactions at the local or site level are usually not adequate to assess the impact of species immigration and emigration on the composition and function of the existing community.
10.5.2.2 Incorporating land cover and climate space at the landscape scale An alternative means of including land cover at the finer resolution, needs to be devised, such as applying masks, so that the climate change response is not obscured. It is a significant weakness that the land cover models were based solely on data for Britain and Ireland. There is a need to extend the range of climatic conditions over which the land cover model was developed using European land cover data in order to capture the full range of class types that may exist under future climatic conditions.
262 MONARCH 2 Report – Chapter 10 ______
Box 10.1 A list of data weaknesses preventing the full exploitation of models predicting the impact of climate change on the distributional ranges of species
• Many species have distributions described at a European range but for taxa, like mosses and scarce species distribution data is often poor or absent at the continental scale. This is essential to obtain statistically robust species-climate associations. • Within the UK, most datasets describing distributions of species are at a 10km2 level. A 1km2 resolution is necessary for downscaling. • There is scant data on the dispersal capabilities of most species. • There is little information at large on the direct physiological responses of species to climate change. • Climate scenario data used in MONARCH 2 was at a 5km2 level, which constrained the usability of other data, particularly the land cover data.
10.5.2.3 Including stochasticity in dispersal modelling The inherent stochasticity of long distance dispersal events, which although rare have a large impact on long-term dispersal, and the limited availability of data on dispersal meant that many of the dispersal parameters had to be derived from knowledge of other species or based on expert opinion. The sensitivity of dispersal modelling to the input parameter values has only undergone limited testing and future work could include the use of various parameterisations for each species, so that different ‘scenarios’ of potential migration ability can be derived. For birds, at least, reliable data on abundance (as well as simple presence), and also on demographic variables (survival and recruitment) are available. An extension of the approach would be to estimate numbers of birds per square kilometre, and incorporate these, along with available information on demography and dispersal (c.f. Baillie et al., 2000) into the modelling framework.
10.5.2.4 Addressing community effects and other drivers for change There remains potential to link the dispersal modelling with the descriptive community composition work using a meta-community framework approach. To achieve this, a species interaction model would need to be linked with the climate envelope and dispersal models, so that the consequences for a community on a species arriving and/or leaving can be assessed.
More generally, the interactions between other key drivers, like air pollution and land management, and climate change are important but poorly understood. Moreover, the importance of species competition needs exploration in order that its impact on species’ distributions and community composition under climate change scenarios can be determined. The processes of immigration and emigration are crucial determinants of a species’ persistence in a community. It is also apparent that community composition can be greatly affected by the number and type of species present, the strength, type and direction of the direct and indirect interactions between species, species extinction and species immigration and emigration. Ecological research has not yet resulted in holistic understanding of habitats. The synthesis of studies in population biology, community ecology, ecosystem ecology or landscape ecology towards this end would improve understanding of likely community responses to climate change.
10.6 Conclusions
MONARCH 2 has sought to advance the science and understanding of the potential impacts of climate change on biodiversity by downscaling the resolution of the modelled climate and land cover suitability surfaces to the 1-kilometre square scale. It has also extended the scope of modelling to
MONARCH 2 Report – Chapter 10 263 ______consider the role of land cover, the potential for dispersal and impacts on community structure. The original objective was to predict likely changes to species distributions at a (1-km) landscape-scale, because it is more appropriate to land managers seeking to consider climate change. However, MONARCH 2 has discovered that reducing the scale of application of climate space and associated modelling to a local or regional level requires data, in particular species distribution data, that is frequently unavailable or of an inadequate quality to enable the models to be run effectively.
The great complexity of natural systems suggests that there are fundamental limits to the prediction of potential future species’ distributions using climate envelope modelling. The importance of the model predictions undertaken by MONARCH should not be underestimated, but in downscaling the SPECIES model the inadequacies of data sources, such as LCM2000, have been exacerbated while the sensitivity of the model to climatic changes has been reduced due to the inclusion of the land cover data. In addition, the limited range of variables examined by the model and its inability to adequately consider factors such as land management, competitive interactions and exposure have become more significant due to their greater importance at the landscape scale. Hence, the case study predictions should be interpreted with due caution and should be viewed as first approximations indicating the potential magnitude and broad pattern of future impacts, rather than as accurate simulations of future species’ distributions. The larger the geographic scale of output, the more reliable the generalisations, whereas interpretations at a regional scale (less than 1000 km2) should be viewed with caution because of the complexities within the natural systems; and in particular climatic factors become secondary to habitat or land management requirements.
In essence, MONARCH 2 tested the capacity for downscaling the climatic envelope modelling approach and concluded that, owing chiefly to data limitations at the finer scale, it was more appropriate to provide generic indications of likely changes within exemplar species for a full range of taxa at the Britain and Ireland scale, rather than further develop the downscaled approach. MONARCH 3 will therefore refine the modelling at the Britain and Ireland scale as in the original MONARCH model, but with added refinements and a faster procedure that will improve its usefulness in helping nature conservation to adapt to the impacts of climate change.
10.7 References
Allen, T.F.H. and Hoekstra, T.W. (1992). Toward a unified ecology. University of Columbia Press. Conservation, 11, 15-23.
Berry, P.M., Vanhinsbergh, D., Viles, H.A., Harrison, P.A., Pearson, R.G., Fuller, R., Butt, N. and Miller, F. (2001). Impacts on terrestrial environments. In Harrison, P.A., Berry, P.M. and Dawson, T.P. (eds.) Climate Change and Nature Conservation in the UK and Ireland: Modelling natural resource responses to climate change (the MONARCH project). UKCIP Technical Report, Oxford, pp43-150.
Brose, U., Ostling, A., Harrison, K., Martinez, N.D. (2004). Unified spatial scaling of species and their trophic interactions. Nature, 428(6979), 167-171.
Brown, J.H. (1995). Macroecology. Chicago University Press, 269pp.
Cottenie, K. and De Meester, L. (2004). Metacommunity structure: Synergy of biotic interactions as selective agents and dispersal as fuel. Ecology, 85, 114-119.
Collingham Y.C., Hill M.O. and Huntley B. (1996). The migration of sessile organisms: a simulation model with measurable parameters. Journal of Vegetation Science, 7, 831-846.
Cottenie, K. and De Meester, L. (2004). Metacommunity structure: Synergy of biotic interactions as selective agents and dispersal as fuel. Ecology, 85, 114-119.
264 MONARCH 2 Report – Chapter 10 ______
Davis, A.J., Jenkinson, L.S., Lawton, J.L., Shorrocks, B. and Wood, S. (1998a). Making mistakes when predicting shifts in species range in response to global warming. Nature, 391, 783–786.
Davis, A.J., Lawton, J.L., Shorrocks, B. and Jenkinson, L.S. (1998b). Individualistic species responses invalidate simple physiological models of community dynamics under global environmental change. Journal of Animal Ecology, 67, 600–610.
Emmerson, M., Bezemer, T.M., Hunter, M.D., Jones, T.H., Masters, G.J., Van Dam, N.M. (2004) Trophic interactions in a changing world. Altering of interaction strength. Basic and Applied Ecology (In Press).
Grime, J.P., Brown, V.K., Thompson, K., Masters, G.J., Hillier, S.H., Clarke, I.P., Askew, A.P., Corker, D. & Kielty, J.P. (2000). The response of two contrasting limestone grasslands to simulated climate change. Science, 289, 762-765.
Hampe, A. (2004). Bioclimate envelope models: what they detect and what they hide. Global Ecology and Biogeography, 13, 469-476.
Harrison, P.A., Berry, P.M. and Dawson, T.P. (Eds.) (2001). Climate Change and Nature Conservation in the Britain and Ireland: Modelling natural resource responses to climate change (the MONARCH project). UKCIP Technical Report, Oxford.
Hossell, J.E., MacGillivray, A., Hepburn, I.R., Ellis, N. and Harley, M. (2001). Implications for policy needs and future research. In: Harrison, P.A., Berry, P.M. and Dawson, T.P. (Eds.) (2001) Climate Change and Nature Conservation in the Britain and Ireland: Modelling natural resource responses to climate change (the MONARCH project). UKCIP Technical Report, Oxford, pp257-271.
Hulme, M., Jenkins, G.J., Lu, X., Turnpenny, J. R., Mitchell, T.D., Jones, R.G., Lowe, J., Murphy, J.M., Hassell, D., Boorman P., McDonald, R. & Hill, S. (2002). Climate Change Scenarios for the United Kingdom: The UKCIP02 Scientific Report. Tyndall Centre for Climate Change Research, School of Environmental Sciences, University of East Anglia, Norwich, UK, 100 pp.
Kneitel, J.M. and Chase, J.M. (2004). Trade-offs in community ecology: linking spatial scales and species coexistence. Ecology Letters, 7, 69-80.
Loreau, M., Mouquet, N., Holt, R.D. (2003). Meta-ecosystems: a theoretical framework for a spatial ecosystem ecology. Ecology Letters, 6, 673-679.
Liu, C., Dawson T. P., Berry P.M. Pearson, R.G. (2004). The effect of sample size and prevalence on species distribution modelling. Ecological Modelling, submitted.
Masters, G.J. and Midgley, G.F. (2004) Analysis of invasive species and climate change. Report to the World Bank and the Global Invasive Species Programme, 35pp.
Midgley, G.F., Hannah, L., Millar, D., Thuiller, W. and Booth, A. (2003). Developing regional and species-level assessments of climate change impacts on biodiversity in the Cape Floristic Region. Biological Conservation, 110 (1-2), 87-97.
Midgley, G.F., Millar, D., Rutherford, M.C. and Powrie, L.W. (2002). Assessing the vulnerability of species richness to anthropogenic climate change in a biodiversity hotspot. Global Ecology and Biogeography, 11(6), 445-451.
O'Neill, R.V., DeAngelis, D., Waide, J.B. and Allen, T.F.H. (1986). A Hierarchical Concept of Ecosystems. Princeton University Press, Princeton.
MONARCH 2 Report – Chapter 10 265 ______
Pearson, R.G., Dawson, T.P. and Lui, C. (2004). Modelling species distributions in Britain: a hierarchical integration of climate and land-cover data. Ecography, 27, 285-298.
Pearson, R. G. and Dawson, T. P. (2003). Predicting the impacts of climate change on the distribution of species: are bioclimate envelope models useful? Global Ecology and Biogeography, 10, 361-371.
Peterson, P.A. (2003). Predicting the geography of species' invasions via ecological niche modeling. Quarterly Review of Biology, 78(4), 419-433.
Sparks, T.H., Jeffree, E.P. and Jeffree, C.E (2000). An examination of the relationship between flowering times and temperature at the national scale using long-term phenological records from the UK. International Journal of Biometeorology, 44, 82-87.
Sparks, T.H. and Carey, P.D. (1995). The responses of species to climate over two centuries: An analysis of the Marsham phenological record, 1736-1947. Journal of Ecology, 83, 321-329.
Thomas, C.D., Cameron, A., Green, R.E., Bakkenes, M., Beaumont, L.J., Collingham, Y.C., Erasmus, B.F.N., Ferreira de Siqueira, M., Grainger, A., Hannah, L., Hughes, L., Huntley, B., van Jaarsveld, A.S., Midgley, G.F., Miles, L., Ortega-Huerta, M.A., Peterson, A.T., Phillips, O.L. and Williams, S.E. (2004). Extinction risk from climate change. Nature, 427, 145-148.
Willows, R. and Connell, R. (2003). Climate Adaptation: Risk, uncertainty and decision-making. UK Climate Impacts Programme, Technical Report, 154pp.