Characteristics of Woods Used Recently and Historically by Lesser Spotted Woodpeckers Dendrocopos Minor in England

Ibis (2010), 152, 543–555

Characteristics of woods used recently and historically by Lesser Spotted Woodpeckers Dendrocopos minor in England

ELISABETH C. CHARMAN,1* KEN W. SMITH,1 DEREK J. GRUAR,1 STEPHEN DODD1 & PHILIP V. GRICE2 1Royal Society for the Protection of Birds, The Lodge, Sandy, Bedfordshire SG19 2DL, UK 2Natural England, Northminster House, Peterborough, Cambridgeshire PE1 1UA, UK

Lesser Spotted Woodpecker Dendrocopos minor numbers have declined greatly in England since the early 1980s for reasons that are not yet fully understood. It has been suggested that the species’ decline may be linked to the increase in Great Spotted Woodpeckers Dendrocopos major, changes in woodland habitat quality (such as deadwood abundance) and landscape-scale changes in tree abundance. We tested some of these hypotheses by comparing the characteristics of woods in southern England where the species is still relatively numerous with those of woods used in the 1980s before the major decline. In each time period, habitat, predator and landscape information from woods known to be occupied by Lesser Spotted Woodpeckers was compared with those found to be unoccupied during surveys. Before the main period of decline, Lesser Spotted Woodpeckers used oak-dominated, mature, open woods with a large amount of standing deadwood. Habitat use assessed from recent data was very similar, the species being present in mature, open, oak-dominated woodlands. There was a strong relationship between wood use probability and the extent of woodland within a 3-km radius, suggesting selection for more heavily wooded landscapes. In recent surveys, there was no difference in deadwood abundance or potential predator densities between occupied and unoccupied woods. Habitat management should focus on creating and maintaining networks of connected woodlands in areas of mature, open woods. Finer-scale habitat selection by Lesser Spot- ted Woodpecker within woodlands should be assessed to aid development of beneﬁcial management actions. Keywords: conservation, habitat, landscape, population declines, woodland birds, woodland structure.

National monitoring schemes have identiﬁed sub- further four woodland species to the Red List. stantial declines in the populations of several Species in decline include both specialist residents woodland birds in Britain during the period 1994– and long-distance migrants. For some species, there 2007 (Risely et al. 2009). The Repeat Woodland is evidence that a primary driver of decline is Bird Survey (RWBS; Amar et al. 2006, Hewson change in woodland structure because of decreased et al. 2007) investigated changes in bird numbers woodland management and increased deer brows- and habitat in British broadleaved and mixed ing (Amar et al. 2006). However, the causes of woodlands over a 20-year period and detected decline for resident woodland specialists remain steep declines in line with national monitoring largely unknown. schemes. In response to the declines, the latest UK Lesser Spotted Woodpeckers Dendrocopos minor bird priority listing (Eaton et al. 2009) elevated a have undergone a population decline in England since around 1980, having previously increased up *Corresponding author. to the late 1970s, possibly due to the increased Email: [email protected] availability of dying and dead trees following the

ª 2010 The Authors Journal compilation ª 2010 British Ornithologists’ Union 544 E. C. Charman et al.

epidemic spread of Dutch elm disease (Osborne than nest locations in England. Furthermore, the 1982). UK monitoring schemes show a decline of birds in Britain are the subspecies comminutus 73% in BTO Common Birds Census indices from (endemic to England and Wales), which may show 1974 to 1999, but the species is now too rare to differing habitat requirements to its European be monitored adequately by the BTO ⁄ counterparts. JNCC ⁄ RSPB Breeding Bird Survey. Because of Reasons for the decline in the UK and elsewhere these declines, the species is included in the UK in Europe are unclear, largely because despite its Biodiversity Action Plan priority species list as well widespread range the species is extremely difficult as on the UK Red List. Numbers may have fallen to study due to its low population density and more widely across Europe (Mikusinski & Angel- secretive behaviour. Fuller et al. (2005) proposed stam 1998, PECBM 2006) in recent years. three potential hypotheses for the Lesser Spotted Various aspects of the Lesser Spotted Wood- Woodpecker decline: (i) interactions, particularly pecker’s breeding ecology have been studied in Eur- in relation to nest-sites, with Great Spotted Wood- ope, such as reproductive success (Rossmanith pecker, which have increased greatly during the et al. 2007), parental care and social mating (Wikt- period of decline; (ii) changes in wood-scale habi- ander et al. 2000), nestling diet (Rossmanith et al. tat quality, especially the availability of deadwood 2007), foraging routines (Olsson et al. 2000, 2001) and deadwood invertebrates; and (iii) landscape- and home-range size (Wiktander et al. 2001). scale changes in tree abundance. To understand Habitat preferences have also been widely studied. factors instrumental in the decline, it is necessary Olsson et al. (1992) found that the species pre- to establish their requirements in areas where their ferred stands with many snags (standing dead trees) populations are still relatively abundant. In addi- and older stands in Sweden, and it was hypothe- tion, if a good population exists in these areas, sites sized that the loss and degradation of this preferred are less likely to be unoccupied due to low immi- habitat through logging could, in part, explain the gration probability and more likely to be unoccu- decline there. A Belgian study (Delahaye et al. pied due to other variables relating to habitat 2002) suggested that the species required the pres- requirements. ence of high tree species diversity, a high density of We present the results of analyses of two data- dead trees and stands with a basal area of more than sets describing characteristics of woods where 15 m2 ⁄ ha. A study comparing occupied and unoc- observations were made of Lesser Spotted Wood- cupied woods in central Europe found that forests peckers in England. These included an intensive were most likely to be occupied if more softwoods field project in 2007 that explicitly aimed to sur- but fewer snags (diameter at breast height vey areas for Lesser Spotted Woodpeckers and (dbh) > 15 cm) per area were available, where the measure habitat use across a wide geographical distance to lakes and rivers was small and the eleva- area, and data from surveys in the 1980s in south- tion was low (Miranda & Pasinelli 2001). ern England, which were used to describe occupied In Britain, an analysis of nest record cards up to woods in the period before the major decline 1989 described nesting preferences at the nest tree (Amar et al. 2006). The paper aims to identify the scale and basic breeding ecology in the period factors associated with wood use across England in before the decline in population (Glue & Boswell areas where Lesser Spotted Woodpeckers are still 1994). There is evidence that the species has relatively numerous and use data from the 1980s declined more in woods where there was a current to describe factors associated with wood use in the high density of Grey Squirrels Sciurus carolinensis period before or early in the decline. (Amar et al. 2006). Smith (2007) describes the use of dead wood for foraging and nesting by wood- METHODS peckers in four woods in Hertfordshire alongside changes in the deadwood resource over a period of Intensive surveys, 2007 more than 20 years, suggesting that stand and deadwood dynamics are likely to provide habitats Seventy-two woods were surveyed in 2007 for Les- more favourable to Great Spotted Woodpecker ser Spotted Woodpeckers in three regions of Eng- Dendrocopos major. With these few exceptions, land: South Yorkshire (SY; 17 woods in the area there have been no quantitative studies of Lesser around Sheffield), Worcestershire and Shropshire Spotted Woodpecker habitat use at a scale wider (WO; 27 woods around the Wyre Forest) and the

ª 2010 The Authors Journal compilation ª 2010 British Ornithologists’ Union Woods used by Lesser Spotted Woodpeckers 545

Wiltshire ⁄ Hampshire border (HA; 28 woods duration was scaled to the size of the block as fol- including parts of the New Forest). The study areas lows: < 30 ha, 3 h; 30–50 ha, 4 h; 50–70 ha, 5 h; were selected as areas where the species is still and > 70 ha, 6 h. known to occur, based on accounts from county Our objective was to assess whether each wood bird reports, the 1988–91 breeding atlas (Gibbons was used by Lesser Spotted Woodpeckers in March et al. 1993), the RSPB ⁄ FC ⁄ NE ⁄ BTO Bird Conser- and April and therefore could possibly be a breed- vation Targeting Project (http://www.rspb.org.uk/ ing site. Although Lesser Spotted Woodpeckers are targeting) and local knowledge through county known to have large territories at some times of bird recorders. the year, in March and April they begin to focus Areas of woodland 20 ha and larger (up to a on a smaller area for breeding (Wiktander et al. maximum of 85 ha) were selected for survey 2001). Once a bird had been observed in a wood within an approximate 10-km square centred on (by sight, calling or drumming), the site was classi- each study region. In Worcestershire, the area was fied as ‘occupied’. Woods were re-surveyed until a increased to an approximately 15-km square to bird was located or until the survey period ended, increase the sample of non-occupied woods. The whichever was sooner. Once sites had been con- National Inventory of Woodland and Trees firmed as ‘occupied’, they were not revisited for (NIWT) (Forestry Commission 2003) was used to formal surveys during this period of fieldwork. classify these woods, and only those that were However, subsequent records of birds in the same classed as predominantly broadleaved were chosen woods were achieved in the vast majority of cases for survey. It was not possible to survey all wood- during other phases of the project. Analysis of lands in each study area due to resource limitations pilot study data suggested that the number of visits and refused access permission. required to confirm occupancy varied between the Lesser Spotted Woodpeckers are notoriously three woods included in analyses but one visit gave secretive and it was necessary to devise a survey on average a 49% probability, two visits 73%, three method that optimized the chances of successfully visits 91% and four visits 96%. This was indepen- detecting birds when they were present at a site. dent of the size of the wood. As a compromise Data from pilot studies in 2005 and 2006 showed between covering a large number of sites and that there is a significant relationship between sur- obtaining reliable detection, at least three visits vey date and the probability of hearing a bird were made to each wood before it was declared (back-calculated once the site had been confirmed ‘unoccupied’ (range 3–8, mean 4.3). On repeat vis- as occupied) such that there is a burst of activity its, the route taken through the wood was reversed between the beginning of March and mid-April. where possible. Similar seasonality has been reported by Hontsch All occupied and unoccupied woods were (2004) and Miranda and Pasinelli (2001). After surveyed for habitat from June 2007. Habitat mid-April, the chances of detecting a Lesser Spot- monitoring largely followed the same protocols as ted Woodpecker by call appears to fall below 50%. the RWBS (Amar et al. 2006). Study areas were Therefore, in this study surveys were undertaken divided into 1-ha squares. Squares within the during daylight hours between 1 March and 20 surveyed area of the wood that contained more April 2007, avoiding days with heavy rain and than 25% woodland were numbered and random strong wind. number tables were used to generate a set of 10 Woods were surveyed by selecting a transect squares of 1 ha in each wood for habitat assess- route that covered as much of the wood as possi- ment. The centre of each selected square formed ble. Analyses of data from three woodlands with the centre of a 25-m-radius area in which habitat regular visits in the pilot study showed that a 6-h recording took place. Some measurements were period spent surveying a wood gave a 95% proba- recorded from the centre of the 25-m plot, and bility of Lesser Spotted Woodpeckers being heard others were taken from four 5-m-radius subplots or seen where they were known to be present. The centred 12.5 m north, east, south and west of the three woods in the pilot study were 40, 50 and centre of the edge of the larger plot (Fig. 1 and 80 ha in size. In the two smaller woodlands, there Table 1). was an indication that fewer hours were necessary. Great Spotted Woodpeckers were surveyed In the current study, the selected woodland blocks simultaneously with Lesser Spotted Woodpeckers. were between 20 and 85 ha, and therefore survey The area covered and the duration of the survey

ª 2010 The Authors Journal compilation ª 2010 British Ornithologists’ Union 546 E. C. Charman et al.

peckers present. Due to the small number of regions with Lesser Spotted Woodpeckers in the RWBS, analyses presented here were restricted to 1 southern England where the species remained widespread in the 1980s. The sites were surveyed in 1985 using a point 25 m count method. Point count locations were chosen 12.5 m at random, with no points closer than 50 m from 2 the edge of a wood or within 100 m of each other. 4 Points were marked onto a map, located in the 5 m Central tree field and marked with flagging tape to allow them to be relocated. Each point count lasted 5 min and was carried out during two visits to each site. First visits were in April or the first week in May, and 3 second visits were in the last 3 weeks in May or the first half of June. Woods were classified as occupied if there was a visual or audible contact with a Lesser Spotted Woodpecker at any point during the point counts on either visit. Habitat Figure 1. Habitat measurements were taken at various scales data were collected between the middle of May within a 25-m-radius area plot. Some measurements were recorded from the centre of the 25-m plot, and others were and the middle of July at the same points as bird taken from four 5-m-radius subplots centred 12.5 m north, point counts. Table 1 details the variables mea- east, south and west of the centre of the edge of the larger plot sured in the habitat surveys. (see Table 1 for details of habitat measurements and scales). Although RWBS surveys were carried out later in the season than the 2007 study, and so may have missed the peak time of Lesser Spotted Wood- were used to produce an index of Great Spotted pecker activity, the species’ densities in 1985 Woodpecker abundance (mean birds ⁄ ha ⁄ h) at the (occupied woods: mean number of detec- wood level. tions = 3.1, range = 1–8) may mean that detection Grey Squirrels were surveyed using methods probability was higher compared with woods with described in the RWBS (Amar et al. 2006). At a single pair (most woods in the 2007 intensive each site, a continuous 1000-m transect was estab- survey). lished through the wood. At sites where this was not possible, shorter individual transects totalling Statistical analysis 1000 m were established which were more than 100 m apart. Transects were divided in 100-m Habitat variables taken at the point level were sections and walked slowly, recording all visible summed and divided by the number of points in a active dreys and their perpendicular distance from wood to produce a mean for each variable. The the transect line (using a laser range finder). A means were then used as potential explanatory drey was classed as active if it was > 30 cm in variables in analyses to compare woods with obser- diameter and little light was visible through it. vations of Lesser Spotted Woodpeckers and those Squirrel drey density estimates were then com- with no observations during surveys. puted using the software DISTANCE (Buckland et al. We used logistic regression with a logit-link 1993). function and binomial errors to model the probability of an observation of Lesser Spotted Wood- pecker in a wood (hereafter referred to as Pre-decline bird and habitat surveys ‘probability of wood use’) in relation to its Full details of bird and habitat survey methods for attributes. Significance was assessed using deletion the RWBS are given in Hewson et al. (2007) and likelihood-ratio tests and examining change in Amar et al. (2006). For a sample of sites, habitat residual deviance using Chi-squared tests. data were collected in the 1980s when there were A three-stage filtering process was used with the still reasonable numbers of Lesser Spotted Wood- aim of reducing the number of covariates entering

ª 2010 The Authors Journal compilation ª 2010 British Ornithologists’ Union Woods used by Lesser Spotted Woodpeckers 547

Table 1. Descriptions of possible correlates of wood use by Lesser Spotted Woodpeckers surveyed in the 2007 intensive survey. Variables shown in bold were also surveyed in the 1980s survey. Variables shown in italics were tested for quadratic relationships in the analyses.

Category Scale (Fig. 1) Measurement Description

Wood and Area Area of wood surveyed (ha) landscape Percentage of woodland within 1, Percentages calculated from the National Inventory of 3 and 10 km Woodlands and Trees (NIWT) using a grid reference Percentage of broadleaved at the centre of the study wood as reference woodland within 1, 3 and 10 km Understorey Subplot Horizontal visibility A 2.4-m pole marked with alternate black and orange sections was placed in the centre of the plot and viewed from each of the subplots. The number of orange bands at least 50% visible through the vegetation from each subplot was recorded (Wilson et al. 2005) Plot ⁄ subplot Vegetation cover 0.5–2 m, Percentage of cover in height band estimated visually 2–4 m and 4–10 m as if viewed from above Tree size Plot ⁄ subplot Canopy cover Percentage of cover estimated visually using a standardized method. The number of 2-cm squares in a 4 · 4 wire grid that were at least 50% covered by canopy foliage when viewed directly from below were counted and converted to %. Canopy was defined as foliage at least 10 m high Plot Basal area Scored according to a standardized reloscope when viewed from the centre of the plot (Hamilton 1975) Plot Maximum diameter at breast dbh of the largest tree in the plot (cm) height (dbh) Plot Maximum height Height of the tallest tree in the plot (m) Field layer Plot ⁄ subplot Bracken, bramble, herbaceous, Percentage of cover in field layer (below 0.5 m) grass, moss, bare ground, leaf estimated visually litter Deadwood Plot Dead trees Count of all dead trees Plot ⁄ subplot Dead limbs Count of the number of dead limbs attached to trees > 20 cm in diameter Subplot Ground deadwood Count of pieces of ground deadwood > 10 cm in diameter and 1 m in length Other habitat Plot Dominant tree Three-level factor – oak, beech, birch. The most numerous tree species in the plot Plot Water features Two-level factor. Presence ⁄ absence of wet feature (stream, bog, flush, pond) Plot Slope Estimation of ground slope. Measured in degrees Plot Altitude Recorded from maps (m) Predators and Great Spotted Woodpecker See text for details competitors abundance Grey Squirrel density See text for details

the final models. First, we examined the signifi- variables and landscape variables (stage 2) and ran cance of each variable alone (stage 1). At this multivariate backwards stepwise models for each stage, any variable not significant at the 10% level group to identify terms to be entered into a final was discarded from all further analyses. We also model. Terms explaining significant variance were tested for non-linear relationships in those vari- entered together and removed in a stepwise pro- ables where we thought the response might be cess if not significant at the 10% level. Finally, all non-linear (Table 1) by including both the term variables remaining after stage 2, and those from and its square together. Secondly, we grouped the stage 1 which did not fall into the groupings and variables describing understorey structure, field enter stage 2, were combined and entered into a layer, tree structure, deadwood, large-scale final model (stage 3). We ran the full model,

ª 2010 The Authors Journal compilation ª 2010 British Ornithologists’ Union 548 E. C. Charman et al.

removing terms using backwards stepwise deletion pied, WO – 20 ⁄ 27, HA – 15 ⁄ 28). At the univariate until only variables significant at the 5% level stage, 10 variables were significant (at 10% level) remained. These variables then formed our chosen in predicting wood use by Lesser Spotted Wood- model. Given the number of potential explanatory peckers (Table 2). Figure 2 shows effects signifi- variables, a degree of co-linearity between variables cant (5% level) at stage 2 and those that remained is expected. To partially account for this, we to stage 3. At stage 2, Lesser Spotted Woodpeckers selected a backwards approach, as this is more were more likely to occur in woods with a gener- appropriate in situations where explanatory vari- ally open structure (Fig. 2a), dominated by oak, ables may be mutually correlated, as the variable with low herbaceous cover in the field layer that explains the greater amount of deviance is (Fig. 2b), and woods in a highly wooded landscape usually retained in the final model (Crawley 2002). (Fig. 2c). The final model contained locality (a Stepwise modelling approaches have been criti- fixed effect in all models), horizontal visibility and cized (e.g. Whittingham et al. 2006, Mundry & the proportion of land within 3 km that is Nunn 2009) although there is also suggestion that wooded. final models vary little between modelling proce- The relationship with horizontal visibility sug- dures (Murtaugh 2009). Given the large number gests an effect of the ‘openness’ of the wood, Les- of explanatory variables and the exploratory nature ser Spotted Woodpeckers having a > 50% of the study, we considered the method appropri- probability of occupancy from values of around ate. A benefit of the modelling approach we have eight on the horizontal visibility scale. A mature, chosen is that significant univariate relationships tall woodland with a closed canopy is likely to have are apparent, which may be useful in the discus- a poor shrub understorey layer through shading. sion stage, especially where significant variables are This process could also lead to the negative rela- considered proxies of other effects. Furthermore, tionship with herbaceous cover, which is also where multivariate models are run within groups shaded in closed-canopy woodland. The relation- of similar variables, the most important variable ships with wood ‘openness’ and herbaceous cover from that group (if any) is identified and taken for- are therefore likely to be proxy measures of the ward to the final model (thereby reducing poten- species’ use of mature, closed-canopy woodland. tial co-linearity in the data). Creating a cut-off at There was a > 50% probability of wood occu- the 10% level rather than at the 5% level at stages pancy where the proportion of woodland in the 1 and 2 also means that marginally non-significant surrounding landscape (at 3-km radius) was variables are included in the final modelling stage. > 30%, equating to around 800 ha of woodland in In the 2007 Lesser Spotted Woodpecker survey the vicinity. There was no significant relationship dataset, locality (SY, WO or HA) was retained in between wood occupancy and the amount of the models at all times as a fixed effect. This was woodland within a 10-km radius. As our survey to control to some degree the possible effects of woods were within a restricted area, there was a spatial autocorrelation, accounting for geographical large degree of overlap in woodland area at that variation in occupancy. Therefore, the effect of scale between woods. This is likely to have resulted any covariate will be identified over and above any in a small amount of variation in values and so any gross differences accounted for by locality. We effect of wooded area at that scale may have been accept that when these effects are fixed, the final lost. models will be conservative. However, given the number of variables tested and the exploratory Pre-decline bird and habitat surveys nature of the analyses we considered this appropriate. Analyses were carried out using R version Sixteen of the 34 woods were occupied during the 2.5.1. 1985 survey. Five of the measured habitat variables were associated univariately with woods used by Lesser Spotted Woodpeckers in the 1980s: domi- RESULTS nant species, maximum height, cover at 4–10 m, water features and the number of dead limbs Intensive studies, 2007 > 20 cm (Table 3). Lesser Spotted Woodpeckers Lesser Spotted Woodpeckers were recorded in 40 were significantly more likely to be present in of the 72 woods surveyed (SY – 5 ⁄ 17 sites occu- woods with tall trees dominated by oak, with more

ª 2010 The Authors Journal compilation ª 2010 British Ornithologists’ Union Woods used by Lesser Spotted Woodpeckers 549

Table 2. Results from analyses using the 2007 Lesser Spotted Woodpecker survey dataset. For each variable at each stage, the signed P-value is shown. Non-linear relationships are shown using U. Blank cells show that those variables were not part of that stage of analysis.

Stage 1 Stage 2 Stage 3 Category Variable Univariate Within group Final model

Locality (ﬁxed in all models) 0.01259 0.01259 0.000278 Size and landscape Surveyed area ns Woodland 1 km + 0.001867 ns Broadleaved woodland 1 km + 0.006913 ns Woodland 3 km + 0.0005075 + 0.0005075 + 0.007904 Broadleaved woodland 3 km + 0.003021 ns Woodland 10 km ns Broadleaved woodland 10 km ns Understorey Horizontal visibility + 0.0001593 + 0.0001593 + 0.002386 Vegetation cover 0.5–2m ns Vegetation cover 2–4m ns Vegetation cover 4–10 m ns Tree size Canopy cover ns Basal area ns Maximum dbh ns Maximum height U 0.072086 U 0.072086 Field layer Bracken ns Bramble ns Herbaceous )0.022473 )0.022473 Grass ns Moss ns Bare ground ns Leaf litter + 0.03053 ns Deadwood Dead trees ns Dead limbs ns Ground deadwood ns Other habitat Dominant tree Oak 0.08858 Oak 0.08858 Water features ns Mean slope ns Mean altitude ns Predators and competitors Great Spotted Woodpecker abundance ns Grey Squirrel density ns

water features, a greater number of dead limbs DISCUSSION < 20 cm on live trees and high cover at 4–10 m. Figure 3 shows effects significant (0.05 level) at We describe broad-scale habitat characteristics of stage 2 and those that remained to stage 3. The woods containing Lesser Spotted Woodpeckers in final model retained dead limbs and oak as the England. Woods with an open structure at least up most significant variables. to 2.4 m and set within a heavily wooded land- Table 4 compares the results from the 1980s scape are more likely to be used by the species dur- study and the intensive study. Cross-referencing of ing the March and April pre-breeding stage. Woods important variables between the two surveys at were more likely to be used if they were oak-domi- each stage of the analyses showed very similar key nated and had a higher abundance of dead limbs. variables, suggesting that habitat associations have Cross-referencing suggested that habitat associa- remained consistent throughout the period of tions have remained largely consistent throughout decline. Although some of the specific variables are the period of decline with the use of mature, open different, they generally describe mature open woodland with a preference for oak and a possible woodland with a preference for oak and a possible relationship with wet features. Furthermore, with relationship with wet features. the recent dataset we have been able to establish

ª 2010 The Authors Journal compilation ª 2010 British Ornithologists’ Union 550 E. C. Charman et al.

(a) (c) 1 1

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2 Lesser Spotted Woodpecker Lesser Spotted Probability of wood occupancy Probability of wood Woodpecker Lesser Spotted by

Probability of wood occupancy by occupancy by Probability of wood 0 0 45678910 0 0.2 0.4 0.6 Horizontal visibility Proportion of area wooded within 3 km (NIWT)

(b) 1

0.8

0.6

0.4

0.2 Lesser Spotted Woodpecker Lesser Spotted

Probability of wood occupancy by occupancy by Probability of wood 0 01020304050 % Herb cover

Figure 2. Relationships between variables significant at stage 2 of the analyses and the probability of Lesser Spotted Woodpeckers being detected during surveys. Solid line shows the modelled relationship and circles show the mean figure per site. (a) Mean horizontal visibility at the wood scale; (b) percentage of herbaceous cover in the field layer; (c) the proportion of the landscape within 3-km radius classified as woodland.

Table 3. Results from analyses using the 1980s Lesser Spotted Woodpecker survey dataset. For each variable at each stage, the signed P-value is shown. Blank cells show that those variables were not part of that stage of analysis.

Stage 1 Stage 2 Stage 3 Category Variable Univariate Within group Final model

Size and landscape Surveyed area ns Understorey Vegetation cover 0.5–2 m ns Vegetation cover 2–4m ns Vegetation cover 4–10 m + 0.03494 + 0.03494 ns Tree size Canopy cover ns Basal area ns Maximum height + 0.0001032 + 0.0001032 ns Field layer Bracken ns Bramble ns Herbaceous ns Deadwood Dead trees ns Dead limbs + 0.00007803 + 0.00007803 + 0.0003117 Other habitat Dominant tree Oak 0.0003812 Oak 0.0003812 Oak 0.0046474 Water features + 0.02086 + 0.02086 ns

the importance of the extent of woodland in the 10 m, low vegetation cover (below 2 m), canopy wider landscape. height and deadwood. If Lesser Spotted Wood- The RWBS (Amar et al. 2006) documented peckers are selecting open woods, this change may changes in woodland structure between the early have created sub-optimal conditions and may be 1980s and the early 2000s, during which time related to the observed change in population. there was an increase in sub-canopy cover at 2– However, wood openness could be a product of

ª 2010 The Authors Journal compilation ª 2010 British Ornithologists’ Union Woods used by Lesser Spotted Woodpeckers 551

1 1

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2 Probability of wood occupancy Probability of wood by Lesser Spotted Woodpecker Lesser Spotted by Probability of wood occupancy Probability of wood by Lesser Spotted Woodpecker Lesser Spotted by 0 0 0102030405060 0 0.1 0.2 0.3 0.4 0.5 0.6 % Cover 4 – 10 m Proportion of points with a wet feature

(b) 1980s data (d) 1980s data 1 1 0.8 0.8 0.6 0.6

0.4 0.4

0.2 0.2 Probability of wood occupancy Probability of wood by Lesser Spotted Woodpecker Lesser Spotted by Probability of wood occupancy Probability of wood by Lesser Spotted Woodpecker Lesser Spotted by 0 0 12 14 16 18 20 22 24 26 28 30 32 0 0.5 1 1.5 2 2.5 3 Maximum height (m) Number of dead limbs < 20 cm in diameter

Figure 3. Relationships between variables signiﬁcant at stage 2 of the analyses and the probability of Lesser Spotted Woodpeckers being detected during surveys in the 1980s in woods in southeast England. Solid line shows the modelled relationship and circles show the mean ﬁgure per site. (a) Percentage of vegetation cover at 4–10 m; (b) maximum height; (c) water features in a wood; (d) the number of dead limbs > 20 cm on live trees.

Table 4. Comparison of signiﬁcant results from the 1980s survey and the 2007 survey.

Agreement between two Category 1980s RWBS data Intensive survey 2007 time periods

Field layer No correlations More likely to be found during surveys in No woods with low herbaceous cover Understorey More likely in woods with high cover at More likely in woods with low cover Yes 4–10 m below 2.4 m Tree size More likely in woods with tall trees More likely in woods with tall trees Yes Deadwood More likely in woods with more dead No correlations with deadwood No. Needs further testing limbs at ﬁner scale Landscape Landscape not tested More likely to occur within a wooded n ⁄ a landscape Other habitat More likely in oak woods More likely in oak woods Yes More likely in woods with a greater No association with wet features No. Needs further proportion of wet features research

mature woodland with high canopy cover which The results do not suggest that understorey should has the effect of over-shading and therefore creat- be cleared in woodlands to assist Lesser Spotted ing an open horizontal structure. Lesser Spotted Woodpeckers, as that would be to the detriment of Woodpeckers spend most of their time foraging in other species and the effect will not alter the layer the canopy (our obs.) and it seems plausible that the species forages in. this variable acts as a proxy for the species’ selec- Lesser Spotted Woodpeckers favoured oak-dom- tion of mature stands with high canopy cover. Fur- inated woodlands. Oak is an important tree species ther work is required at a ﬁner scale to establish for foraging and nesting. From autumn to early the relationship between canopy requirements for spring, Lesser Spotted Woodpeckers feed almost Lesser Spotted Woodpeckers and wood openness. exclusively on wood-living insect larvae extracted

ª 2010 The Authors Journal compilation ª 2010 British Ornithologists’ Union 552 E. C. Charman et al.

from small-diameter dead branches on living trees off between energy demand and predation risk (Cramp 1985, Olsson 1998). From bud-burst and (Wiktander et al. 2001). Olsson and Wiktander through the breeding season a high proportion of (unpubl. data) report that in winter in Sweden, their diet comprises surface-living insects such as Lesser Spotted Woodpeckers spend more time for- aphids and caterpillars and bark-dwelling inverte- aging in spruce, possibly because energy demands brates, which are gleaned from the bark and leaves are lower at this time (Wiktander 1998). This (Wiktander et al. 1994, Olsson 1998). Aphids and allows the species to select habitats of greater especially caterpillars also make up the vast major- safety (e.g. no defoliation in spruce forests) rather ity of the food given to nestlings (Wiktander et al. than energetic proﬁtability (Olsson 1998); there- 1994). Breeding has been reported to coincide fore they select a large territory encompassing this with oak bud-burst (Wiktander et al. 1992, 2001). habitat type and minimize their predation risk as a Furthermore, laying date has been found to be result. strongly inﬂuenced by the availability of food in Woodland cover in Britain has undergone con- deadwood more than a month before laying (Ols- siderable changes over time. Results from the latest son et al. 1999). An alternative explanation of a Countryside Survey (Carey et al. 2008) show that preference for oak stands is that areas dominated whilst woodland cover is increasing currently in by oak contain other favoured species for nesting England (5.8% increase between 1998 and 2007 and foraging. In southern Sweden, Lime Tilia cor- and 10.6% between 1990 and 2007), this is largely data is a preferred species and forest stands are due to new plantings, which are unlikely to be selected on this basis. However, these stands are suitable for Lesser Spotted Woodpecker at the cur- also dominated by oak (Olsson 1998). Finer-scale rent time. There is some evidence from the same work on habitat selection will establish any prefer- survey that woodland quality is decreasing, as there ences such as these in Britain. has been a long-term decrease in plant species rich- The relationship found with the amount of ness in broadleaved woodland of 19% between woodland at the landscape level is supported by 1990 and 2007 in the areas targeted by the Coun- several studies from continental Europe. Wiktan- tryside Survey for their botanical interest. Further- der et al. (1992) found that the frequency of more, connectivity of woodlands may have occurrence of Lesser Spotted Woodpeckers in large decreased over time, with the woodland resource (200-ha) census areas increased with the total area becoming increasingly fragmented. The total of deciduous woodland. Where there was < 17 ha, length of woody linear features deceased by 1.4% the frequency of occurrence was 24%, rising to in England between 1998 and 2007, following an 62% where there was 17–38 ha, and those areas increase between 1990 and 1998, and a decrease with more than 38 ha had a frequency of Lesser between 1984 and 1990. The length of ‘managed’ Spotted Woodpecker occurrence of 80%. Opdam hedgerows decreased by 6.1% between 1998 and et al. (1985) suggested that isolation of woodlots 2007, with a large proportion of ‘managed’ hedges in The Netherlands may be a factor determining turning into lines of trees or relict hedges due to the distribution of Lesser Spotted Woodpeckers in lack of management. For a species with a require- the landscape. They found a weak correlation ment of a large amount of wooded area across a between the presence of the species and the area large area, this trend may be of concern. Lesser of forest within a 3-km radius. Spotted Woodpecker populations may be surviving This requirement for a wooded landscape is also in heavily wooded parts of the country where it is supported by the species’ large home-range. possible for them to retain their large territories Hontsch (2004) found that Lesser Spotted Wood- and cover large areas, making use of hedgerows pecker home-range averaged 211 ha in winter, and other types of woodland. 131 ha in the pre-breeding season and 27 ha in the Hontsch (2004) recommended that conserva- breeding season. Wiktander et al. (2001) found tion strategies for Lesser Spotted Woodpecker similarly large territories in the winter: 742 ha, should focus on creating and maintaining networks reducing to 355 ha in early spring, 103 ha in late of suitable habitats rather than large patches of spring and 43 ha during nesting. The reasons for uniform habitat, as the species probably tolerates a this seasonal variation in home-range size are not certain amount of fragmentation in its home-range. fully understood, but it has been proposed that it Miranda and Pasinelli (2001) concluded that the may relate to territorial behaviour and be a trade- distribution of Lesser Spotted Woodpecker was

ª 2010 The Authors Journal compilation ª 2010 British Ornithologists’ Union Woods used by Lesser Spotted Woodpeckers 553

affected by variables associated with forest struc- occurrence in a wood is related to Great Spotted ture and the relative position within the landscape Woodpecker abundance. There was also no sup- and, like Hontsch (2004), habitat connectivity was porting evidence that Grey Squirrel density had a suggested as a conservation measure. negative effect on Lesser Spotted Woodpeckers. It The importance of deadwood was only sug- is therefore suggested that the correlation found in gested in the 1980s analyses. Deadwood of the sort the RWBS is likely to be a product of a relation- measured in this study has increased over time ship with another variable. There was also little (Amar et al. 2006), and therefore lack of correla- suggestion that deadwood of the type measured tions in the recent dataset do not necessarily mean and at the scale reported here is related to Lesser a lack of importance. All the woods in the inten- Spotted Woodpecker occurrence within a wood, sive survey had higher levels of deadwood than the although as already discussed, we suspect that 1980s surveyed woods. The deadwood variables there may be associations at a finer scale. We measured in this and the 1980s study were dead know that deadwood has generally increased over trees, ground deadwood over 10 cm diameter and time but the resources used by Lesser Spotted 1 m in length, and dead limbs > 20 cm in diameter Woodpeckers may have declined. Furthermore, we on live trees. Lesser Spotted Woodpeckers forage have no information about deadwood inverte- on small-diameter dead branches in live trees brates in the study woods and how they have (Smith 2007) and nest in dead trees or areas of changed. The final hypothesis was that there may deadwood on live trees (Smith 2007, Kosinski & have been landscape-scale changes which have Kempa 2007), so the deadwood we measured may impacted the Lesser Spotted Woodpecker. Cur- not have been the correct sort to establish an asso- rently, due to what we know regarding connectiv- ciation. It is also possible that Lesser Spotted ity of woodlands and the suitability of new Woodpeckers may be selecting areas of deadwood plantings, this may be the most plausible hypothe- at a scale finer than the wood level. sis for influences on Lesser Spotted Woodpecker There is a suggestion that Lesser Spotted Wood- populations. However, this study has not been peckers prefer damp woodlands (Cramp 1985). exhaustive and other potential limiting factors Indeed, Wiktander et al. (1992) suggested that a remain untested. possible cause of the Lesser Spotted Woodpecker decline in Sweden could be due to the large-scale Management recommendations felling of riparian woodland. Miranda and Pasinelli (2001) also found a positive association between If our findings are representative of the British Lesser Spotted Woodpecker occurrence and dis- range of the Lesser Spotted Woodpecker, habitat tance to water features. In this study, the species management for the species needs to focus on was found to have a relationship with wet features maintaining a network of connected woodlands in the 1980s data. This relationship was not appar- within an area of around 3-km radius of mature ent in the 2007 dataset, although there was a posi- woodlands. The relationship with landscape should tive trend between soil wetness and Lesser Spotted be used to target remaining well-wooded areas Woodpecker occurrence in a small sample of within the species current range for finer-scale hab- woods. Soil moisture has been found to be associ- itat manipulations, and there should be measures ated with wood occupancy in the Willow Tit to encourage habitat connectivity in areas where Poecile montanus, another declining woodland Lesser Spotted Woodpeckers remain. Further work species (Lewis et al. 2007, 2009a,b). We suggest is required to establish the role of wet features, that this factor needs careful consideration in the deadwood and finer-scale habitat selection, such as decline of the Lesser Spotted Woodpecker and that required for breeding territories and specific further research is required. foraging requirements, as well as other factors Fuller et al. (2005) proposed three hypotheses potentially limiting populations. to explain the decline: interactions with Great Spotted Woodpecker, changes in wood-scale habi- This study was funded by the Royal Society for the Protection tat (especially deadwood and deadwood inverte- of Birds and Natural England (through the ‘Action for Birds in brates), and landscape-scale changes in tree England’ partnership). We are grateful to all landowners who gave us access to their land in Hampshire, Wiltshire, Worcester- abundance. There was no evidence from the inten- shire, Shropshire and South Yorkshire. Anne Heath and Paul sive studies that Lesser Spotted Woodpecker

ª 2010 The Authors Journal compilation ª 2010 British Ornithologists’ Union 554 E. C. Charman et al.

Toynton helped collect field data for which we are grateful. The Lewis, A.J.G., Amar, A., Cordi-Piec, D. & Thewlis, R. 2007. RWBS was funded by a consortium comprising Defra, English Factors influencing Willow Tit Poecile montanus site occu- Nature, Forestry Commission England, Forestry Commission pancy: a comparison of abandoned and occupied woods. Scotland, Forestry Commission Wales, The Woodland Trust, Ibis 149: 205–213. BTO and RSPB. Ola Olsson, Tom Charman and an anonymous Lewis, A.J.G., Amar, A., Charman, E.C. & Stewart, F.R.P. referee provided useful comments on the manuscript. 2009a. The decline of the Willow Tit in Britain. Br. Birds 102: 386–393. Lewis, A.J.G., Amar, A., Daniells, L., Charman, E.C., Grice, REFERENCES P. & Smith, K. 2009b. Factors influencing patch occupancy and within patch habitat use in an apparently stable popula- Amar, A., Hewson, C.M., Thewlis, R.M., Smith, K.W., Fuller, tion of British Willow Tits (Poecile montanus kleinschmidti). R.J., Lindsell, J.A., Conway, G., Butler, S. & MacDonald, Bird Study 56: 326–337. M. 2006. What’s Happening to our Woodland Birds? Long- Mikusinski, G. & Angelstam, P. 1998. Economic geography, Term Changes in the Populations of Woodland Birds. forest distribution and woodpecker diversity in central Eur- RSPB Research Report no. 19 and BTO Research Report ope. Conserv. Biol. 12: 200–208. no. 169. Miranda, B. & Pasinelli, G. 2001. Habitat requirements of the Buckland, S.T., Anderson, D.R., Burnham, K.P. & Laake, Lesser Spotted Woodpecker (Dendrocopos minor) in for- J.L. 1993. Distance Sampling: Estimating Abundance of ests of north-eastern Switzerland. J. Ornithol. 142: 295– Biological Populations. Chapman & Hall, London. 305. Carey, P.D., Wallis, S., Chamberlain, P.M., Cooper, A., Mundry, R. & Nunn, C.L. 2009. Stepwise model fitting and Emmett, B.A., Maskell, L.C., McCann, T., Murphy, J., statistical inference; turning noise into signal pollution. Am. Norton, L.R., Reynolds, B., Scott, W.A., Simpson, I.C., Nat. 173: 119–123. Smart, S.M. & Ullyett, J.M. 2008. Countryside Survey: UK Murtaugh, P.A. 2009. Performance of several variable-selec- Results from 2007. NERC Centre for Ecology & Hydrology. ⁄ tion methods applied to real ecological data. Ecol. Lett. 12: 105pp. (CEH Project Number: C03259). 1061–1068. Cramp, S. (ed.) 1985. The Birds of the Western Palearctic, Olsson, O. 1998. Through the Eyes of a Woodpecker: Under- Vol. 4. Oxford: Oxford University Press. standing Habitat Selection, Territory Quality and Reproduc- Crawley, M.J. 2002. Statistical Computing: An Introduction to tive Decisions from Individual Behaviour. PhD Thesis, Lund Data Analysis Using S-Plus. Chichester: John Wiley & University. Sons. Olsson, O., Nilsson, I.N., Nilsson, S.G., Pettersson, B., Delahaye, L., Derouaux, A. & Delvingt, W. 2002. Habitat Stagen, A. & Wiktander, U. 1992. Habitat preferences of modelling: a tool for forest ecology management. A case the Lesser Spotted Woodpecker Dendrocopos minor. Ornis study of Lesser Spotted Woodpeckers Dendrocopos minor Fenn. 69: 119–125. in Belgian Ardenne. Aves 39: 3–4. Olsson, O., Wiktander, U., Holmgren, N. & Nilsson, S.G. Eaton, M.A., Brown, A.F., Noble, D.G., Musgrove, A.J., 1999. Gaining ecological information about Bayesian forag- Hearn, R., Aebischer, N.J., Gibbons, D.W., Evans, A. & ers through their behaviour. II. A field test with woodpeck- Gregory, R.D. 2009. Birds of Conservation Concern 3: the ers. Oikos 87: 264–276. population status of birds in the United Kingdom, Channel Olsson, O., Wiktander, U. & Nilsson, S.G. 2000. Daily forag- Islands and the Isle of Man. Br. Birds 102: 296–341. ing routines and feeding effort of a small bird feeding on a Forestry Commission . 2003. National Inventory of Woodland predictable resource. Proc. R. Soc. Lond. B. Biol. Sci. 267: and Trees. Edinburgh: Forestry Commission. 1457–1461. Fuller, R.J., Noble, D.G., Smith, K.W. & Vanhinsbergh, D. Olsson, O., Wiktander, U., Malaqvist, A. & Nilsson, S.G. 2005. Recent declines in populations of woodland birds in 2001. Variability of patch type preference in relation to 98 Britain: a review of possible causes. Br. Birds : 116–143. resource availability and breeding success in a bird. Oeco- Gibbons, D., Reid, J.B. & Chapman, R.A. 1993. The New logia 127: 435–443. Atlas of Breeding Birds in Britain and Ireland: 1988–1991. Opdam, P., Rijsdijk, G. & Hustings, F. 1985. Bird communi- London: Poyser. ties in small woods in an agricultural landscape: effects of Glue, E. & Boswell, T. 1994. Comparative nesting ecology of area and isolation. Biol. Conserv. 34: 333–352. 87 the three British breeding woodpeckers. Br. Birds : 253– Osborne, P. 1982. Some effects of Dutch elm disease on 269. nesting farmland birds. Bird Study 29: 2–16. Hamilton, G.J. 1975. Forestry Commission Booklet No. 39 – PECBM. 2006. The State of Europe’s Common Birds, 2005. Forest Mensuration Handbook. London: HMSO. Prague: CSO ⁄ RSPB. Hewson, C.M., Amar, A., Lindsell, J.A., Thewlis, R.M., But- Risely, K., Noble, D.G. & Baillie, S.R. 2009. The Breeding ler, S., Smith, K. & Fuller, R.J. 2007. Recent changes in Bird Survey 2008. BTO Research Report 537. Thetford: 149 bird populations in British broadleaved woodlands. Ibis : British Trust for Ornithology. 14–28. Rossmanith, E., Hontch, K., Blaum, N. & Jeltsch, F. 2007. Hontsch, K. 2004, Der Kleinspecht (Picoides minor): Auto- Reproductive success and nestling diet in the Lesser Spot- kologie einer bestandsbedrohten Vogelart im hessischen ted Woodpecker (Picoides minor): the early bird gets the Vordertaunus. PhD Thesis, University of Frankfurt. caterpillar. J. Ornithol. 148: 323–332. Kosinski, Z. & Kempa, M. 2007. Density, distribution and Smith, K.W. 2007. The utilization of dead wood resources by nest-sites of woodpeckers picidae, in a managed forest of woodpeckers in Britain. Ibis 149: 183–192. Western Poland. Pol. J. Ecol. 55: 519–533.

ª 2010 The Authors Journal compilation ª 2010 British Ornithologists’ Union Woods used by Lesser Spotted Woodpeckers 555

Whittingham, M.J., Stephens, P.A., Bradbury, R.B. & Wiktander, U. 1998. Reproduction and Survival in the Lesser Freckleton, R.P. 2006. Why do we still use stepwise mod- Spotted Woodpecker. Effects of the life history, mating sys- elling in ecology and behaviour? J. Anim. Ecol. 75: 1182– tem and age. PhD Thesis, Lund University. 1189. Wiktander, U., Olsson, O. & Nilsson, S.G. 2001. Seasonal Wiktander, U., Nilsson, I.N., Nilsson, S.G., Olsson, O., Pet- variation in home-range size, and habitat area requirement tersson, B. & Stagen, A. 1992. Occurrence of the Lesser of the Lesser Spotted Woodpecker (Dendrocopos minor) in Spotted Woodpecker Dendrocopos minor in relation to area southern Sweden. Biol. Conserv. 100: 387–395. of deciduous forest. Ornis Fenn. 69: 113–118. Wilson, A.M., Fuller, R.J., Day, C. & Smith, G. 2005. Nightin- Wiktander, U., Nilsson, S.G., Olsson, O. & Stagen, A. 1994. gales Luscinia megarhynchos in scrub habitats in the south- Breeding success of a Lesser Spotted Woodpecker Dend- ern fens of East Anglia, England: associations with soil type rocopos minor population. Ibis 136: 318–322. and vegetation structure. Ibis 147: 498–511. Wiktander, U., Olsson, O. & Nilsson, S.G. 2000. Parental care and social mating system in the Lesser Spotted Received 17 February 2009; Woodpecker Dendrocopos minor. J. Avian Biol. 31: 447– revision accepted 8 February 2010. 456.

ª 2010 The Authors Journal compilation ª 2010 British Ornithologists’ Union