Designing Effective Protected Area Networks for Multiple Species

Biological Conservation 258 (2021) 109125

Contents lists available at ScienceDirect

Biological Conservation

journal homepage: www.elsevier.com/locate/biocon

Policy analysis Designing effective protected area networks for multiple species

Lynda Donaldson a,*,1, Jonathan J. Bennie b, Robert J. Wilson c,d, Ilya M.D. Maclean a a Environment & Sustainability Institute, University of Exeter, Penryn Campus, Cornwall TR10 9FE, UK b Centre for Geography and Environmental Science, University of Exeter, Penryn Campus, Cornwall TR10 9FE, UK c College of Life and Environmental Sciences, University of Exeter, Exeter EX4 4PS, UK d Museo Nacional de Ciencias Naturales (MNCN-CSIC), Madrid 28006, Spain

ARTICLE INFO ABSTRACT

Keywords: Protected area networks seek to ensure the persistence of multiple species, but their area and extent are limited Metapopulation by available land and conservation resources. Prioritising sites based on their quality, quantity, size, or con Cyperus papyrus nectivity is often proposed; potentially using the occupancy and metapopulation dynamics of individual Habitat network threatened species as surrogates for network effectiveness. However, the extent to which the dynamics of species Conservation planning with overlapping habitat requirements differ, and the implications of this for the optimal network designs for Connectivity Habitat quality multiple species, are rarely tested. We parameterise metapopulation models for 5 papyrus-specialist birds occupying a network of papyrus swamp in Uganda, each of which possess subtly different ecological charac teristics and habitat preferences. We estimate how each responds to different strategies based on prioritising patch size, number, quality and connectivity. The optimal approach differed depending on the metapopulation structure and characteristics of each species. The rank order of strategies also varied with the overall wetland area available and the desired persistence threshold. For individual species, prioritising habitat quality achieved the highest levels of persistence and population size for an equivalent amount of land area conserved. However, connected patches showed greatest overlap across species, thus the most effective strategy to conserve multiple species in the same network prioritised habitat connectivity. This emphasises the importance of individual species’ characteristics using the same habitat networks in conservation planning, and demonstrates the utility of prioritising protected sites based on the spatial connectivity of habitat patches, when aiming to conserve multiple species with differing or uncertain habitat requirements.

1. Introduction populations (D’Aloia et al., 2019), methods need to be developed and tested that consider the effects of habitat configuration on the persis The theories of island biogeography and metapopulation dynamics tence of species across protected area networks. (MacArthur and Wilson, 1967; Hanski, 1999), indicate that the occur Some general guidelines can be applied to the effects of habitat rence and abundance of species, and hence the likely effectiveness of configuration on species persistence. Metapopulation theory predicts protected areas, are related to the size and isolation of habitat islands. that colonization probability is greater in more connected patches (i.e. The principles proposed by these theories have thus been pivotal for the those closer to occupied patches), while local extinction rates are higher design of landscape-scale conservation initiatives (Donaldson et al., in small, low quality patches (Hanski, 1991, 1994; Moilanen and Hanski, 2017), and have led to recommendations for networks of protected areas 1998). Thus, protected area networks are recommended to consist of as a means of conserving biodiversity (Butchart et al., 2012). However, large, high quality, well connected sites (Lawton et al., 2010). But, with criteria for the designation of protected areas focus primarily on existing limited resources for conservation (McCarthy et al., 2012) and a land taxonomic irreplaceability (Dudley, 2008), and rarely consider the ef scape increasingly dominated by humans (Foley et al., 2005), optimising fects of the spatial configurationof habitats on likely persistence. Given all of these criteria is often impractical. Trade-offs may need to be made that drivers of environmental change, including habitat loss and climate between site size, quality and connectivity (Donaldson et al., 2017). change, produce dynamic changes in the distributions of species Metapopulation models have been used in fragmented landscapes to

* Corresponding author. E-mail address: [email protected] (L. Donaldson). 1 Present address: Wildfowl & Wetlands Trust, Slimbridge, Gloucestershire, GL2 7BT, UK. https://doi.org/10.1016/j.biocon.2021.109125 Received 4 August 2020; Received in revised form 2 April 2021; Accepted 9 April 2021 Available online 18 May 2021 0006-3207/© 2021 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). L. Donaldson et al. Biological Conservation 258 (2021) 109125 identify the most beneficial management strategies for the long-term responds to prioritising sites based on their size, quality and connec survival of individual species (Hanski et al., 1996b; Gutierrez, 2005), tivity. By comparing the outcomes among species, we test the suitability including the impact of enhancing the size, number or quality of patches of single-species metapopulation approaches to multi-species conser for persistence (Hanski and Thomas, 1994). Despite increasing interest vation, and assess the components of optimal planning for multiple in modelling the dynamics of two or more species residing in similar species sharing a habitat network. habitats (Gutierrez´ et al., 2001), or local communities of interacting species (Tilman et al., 1994), empirical tests of metapopulation ap 2. Materials and methods proaches to landscape-scale conservation planning typically continue to focus on the effects of single species alone (Etienne, 2004) rather than 2.1. Study system the influenceon multiple species. Yet protected area networks strive to represent whole communities of species (CBD, 2011), and chosen stra Work was conducted in papyrus swamps fringing Lake Bunyonyi, tegies should not negatively impact co-occurring species (Gutierrez´ south-west Uganda (01o17’S; 29o55’E; Fig. 1), where five species of et al., 2001) but instead balance the needs for all (Lawson et al., 2012). papyrus-specialist passerines are found: white-winged swamp-warbler Importantly, species often possess slightly different habitat re (Bradypterus carpalis), greater swamp-warbler (Acrocephalus rufescens quirements, even if associated with the same broad habitat type race foxi), papyrus canary (Crithagra koliensis), papyrus yellow warbler (Howard et al., 2000) and often vary in ecological characteristics, such (Calamonastides gracilirostris) and Carruthers’s cisticola (Cisticola carru as dispersal ability (Thomas, 2000) and dynamic responses to habitat thersi). Currently only papyrus yellow warbler is listed as threatened on availability (Glorvigen et al., 2013), that can lead to heterogenous re the IUCN Red List (IUCN, 2021), although all are estimated to be in sponses to different environmental factors (Falaschi et al., 2021). decline owing to extensive habitat loss and degradation (Maclean et al., However, detailed information on the habitat requirements or popula 2014) and are among the most inadequately protected bird species in the tion dynamics of most populations of species is lacking, and it is ques region (Maclean et al., 2011b). The presence of these birds has led to the tionable whether the dynamics of individual species can be used as a designation of an Important Bird and Biodiversity Area (IBA) at the reliable proxy for others because of variable needs (Meurant et al., north of the lake (BirdLife International, 2020), while Lake Bunyonyi 2018), and what the direct implications of this are for conservation and associated wetlands are proposed to be designated a Ramsar site for planning. Although prioritising species with distributions that encom their ecosystem service value and importance for biodiversity (Byar pass those of rare species may allow for a multi-species approach uhanga et al., 2001). All five species studied are primarily restricted to (Gutierrez´ et al., 2001), this may not be sufficient in networks con papyrus, though differ in ecological characteristics, such as feeding taining multiple threatened species that do not directly overlap. preferences (Britton, 1971), behaviour (Britton, 1978) and vegetation- Here we explicitly test the implications for selecting priority sites in structure preference (Maclean et al., 2003a, 2006; Donaldson et al., habitat networks based on the metapopulation dynamics of individual 2016, 2019). Their specific habitat requirements also differ slightly: versus multiple species, using a suite of specialist passerines primarily papyrus yellow warbler and Carruthers’s cisticola inhabit a broader restricted to papyrus (Cyperus papyrus) swamps in East and Central Af range of wetland vegetation types (Maclean et al., 2006). Although rica. Papyrus is a naturally fragmented wetland habitat because of its greater swamp-warbler is widely distributed in wetland habitats across limitation to wetland areas, and sites have become increasingly isolated Africa, race foxi is restricted to papyrus (Vande weghe, 1981). Together, as a result of local and widespread drainage and encroachment for these species provide a useful exemplar to explore the characteristics subsistence and commercial agriculture (Hartter and Ryan, 2010; that could drive interspecific differences in dynamics, and the broader Maclean et al., 2011b). Growing demands for agricultural land, resulting consequences of such differences for multi-species conservation from rapid population growth and climate uncertainties, will likely strategies. exacerbate these threats (Kipkemboi and van Dam, 2016). Birds restricted to papyrus are declining due to the extensive loss and degra 2.2. Distribution survey dation of their habitat (Maclean et al., 2014), leading to the inclusion of some of these species on the International Union for Conservation of Presence-absence surveys were conducted between May and August Nature (IUCN) Red List (IUCN, 2021). Yet resources available for con over two consecutive years at 518 discrete patches of papyrus (Fig. 1; servation in sub-Saharan Africa are severely limited (Howard et al., and see Donaldson et al., 2019). False absences caused by imperfect 2000) and currently offer papyrus-specialist species little protection detectability in occupancy surveys could over-estimate the probability (Maclean et al., 2006), necessitating clear priority setting. Previous of population turnover (local extinction and colonization). Nevertheless, work highlights the importance of habitat networks for sustaining due to the highly vocal nature of the passerines studied (Maclean et al., populations of these birds, at least over the short-term (Donaldson et al., 2006), the probability of recording a species as absent when it is present 2019) and the need for a consideration of metapopulation dynamics during an average survey is low (Appendix S1 in Donaldson et al., 2019). (Maclean et al., 2011b). But although these species have broadly over Wetlands consisting of broader wetland vegetation were included in the lapping distributions, they possess subtly different habitat requirements study for papyrus yellow warbler and Carruthers’s cisticola only, (Donaldson et al., 2016) and life history traits (Vande weghe, 1981). shoreline fringing patches were surveyed for the presence of greater Understanding the implications of these differences for conservation swamp-warbler and papyrus canary (the remaining three species were strategies will not only provide the evidence-base to support this system never recorded in this patch-type), and wetland recently converted to in urgent need of protection, but more broadly provides a unique op agriculture was surveyed for Carruthers’s cisticola (Fig. 1; Donaldson portunity to explore the fundamental principles of reserve design from et al., 2016). GPS points (GARMIN GPSMAP 64, Lenexa, KS) and sketch the perspective of multiple species residing in overlapping, fragmented maps drawn to scale with topographical maps were used to mark the habitats threatened by habitat loss, and determine if there are common boundary of each patch. Five vegetation categories were created relating strategies across species that can guide conservation action (Falaschi to the physical characteristics of swamps (see Donaldson et al., 2019, et al., 2021). 2016), and the proportion within each swamp was recorded. We assess the effectiveness of fivestrategies for prioritising areas for conservation: protecting either (1) a few large patches, (2) several small 2.3. Metapopulation model patches, (3) high quality patches, (4) well-connected patches, and (5) the single largest patch. We parameterise metapopulation models using SPOMSIM software (Moilanen, 2004), which is an implementation of occupancy data collected from a network of swamps in Uganda, and the Incidence Function Model (Hanski, 1994), was used for parameter perform metapopulation simulations to estimate how each species estimation and model simulations. The key feature is that local dynamics

2 L. Donaldson et al. Biological Conservation 258 (2021) 109125

Fig. 1. Location of wetlands surveyed surrounding Lake Bunyonyi (a), Uganda (b), Africa (c). Displayed using ArcGIS v 10.2.2.

3 L. Donaldson et al. Biological Conservation 258 (2021) 109125 are omitted and only patch occupancy is modelled (Moilanen, 2004), 2.5. Model simulations with the assumption that discrete habitat patches exist in a matrix of unsuitable habitat, and the probability of patch occupancy is determined Metapopulation simulations were carried out in SPOMSIM to iden by extinction and colonization events that can be inferred from spatial tify differences in population persistence between five distinct ap occupancy patterns (Ozgul et al., 2006). The probability of colonization proaches associated with reserve design: preferentially conserving (Ci) of an empty patch i is given by: either bigger, a greater number, better (higher quality ~ population

2 density), more connected sites, or the single biggest patch. For com Si Ci = (1) parison between approaches, results are shown with respect to a given S 2 + y2 i total area of habitat conserved. Metapopulation dynamics were simu lated for each species, all individual “networks” run for 100 iterations where y is a parameter defininghow fast the probability of colonization over 100 years, starting with the 2015 occupancy and habitat data. To approaches unity with increasing connectivity (S ) based on a sigmoid i approximate the minimum viable metapopulation (Hanski et al., function to represent a potentially strong Allee effect in colonization 1996a), a threshold of 95% likelihood of persistence after 100 years was (Hanski, 1994; Ovaskainen and Hanski, 2001). Si is given by: ∑ ) used (Shaffer and Samson, 1985). The relative population size after 100 Si = pjexp αdij Aj (2) years was estimated using the average proportion of occupied area after the time period, since area within our models represented population = = Here pj is the occupancy of patch j (1 occupied, 0 unoccupied), dij size (see Parameter estimation). This approach enabled us to explore is the edge-to-edge distance between patches i and j (i ∕= j) and α defines persistence and population size separately. All scenarios were repeated a negative exponential distribution of dispersal distances (1/α = average until 100% persistence over 100 years was obtained: dispersal distance) (Hanski, 1994; Moilanen, 2004). Aj is the population carrying capacity of patch j. Protecting large patches: Simulations began with the two largest The extinction probability (Ei) of an occupied patch is defined as: patches available in the network for each species, sequentially μ introducing patches to the network by size (largest first).As the area Ei = x (1 Ci) (3) Ai of the two largest patches was often high compared to the minimum areas conserved in the other approaches, their size was artificially where Ai is the carrying capacity of patch i, μ is a parameter that defines reduced to enable comparison with other approaches at equivalent the probability of extinction of a patch, x defines the scaling of extinc levels of habitat. tion risk with patch area (Hanski, 1998) and 1 - Ci is a rescue effect Protecting a large number of small patches: Patches were ordered by lowering the extinction risk of well-connected patches (Ozgul et al., size (smallest first)and initial simulations were run with a network of 2006). the smallest patches making up 10% of the full network area for each species, sequentially adding patches from the smallest to the largest (and hence increasing the total area) to identify how much area from 2.4. Parameter estimation the smallest patches combined was required to reach the threshold for persistence. Two consecutive years of survey data were used for parameter esti Protecting high quality patches: Patches with the highest density for mation (Moilanen and Hanski, 1998; Moilanen et al., 1998; Moilanen, that species were sequentially added to the network (highest density 1999); ensuring that habitat area had not altered considerably between first).Since densities were capped at the observed upper limit, a set years (Wilson et al., 2009). Since water levels remain stable year-round of patches with the initial highest density had equal values. There at this site (Denny, 1972), changes in patch sizes between surveys are fore, this set of patches was reduced in size until the metapopulation minimal, and no substantial differences in weather, habitat extent or became extinct, retaining the initial capped density values quality were recorded over the two years of study (Donaldson et al., throughout. 2019). Protecting most connected patches: Connectivity values for all patches Estimates of patch carrying capacity were based on patch size and within the network were calculated based on a simplifiedversion of predicted densities of each species (Hoyle and James, 2005), determined the full connectivity formula (2): by relating point count density estimates to patch geometry and/or the ∑ ) Si = αdij (4) proportion of particular vegetation categories (see Donaldson et al., 2016). Densities were capped at the highest observed densities for each species since predictions were made outwith the data range (Conn et al., This accounts for the distance between patches and the dispersal 2015). capabilities of each species, but assumes that all patches within the α Parameters , y, x and e were estimated using the Markov Chain network are occupied and equal in size. This enabled us to establish Monte Carlo (MCMC) method (Moilanen, 1999, 2004), assuming pop the patches with the highest connectivity values when all patches ulations are in equilibrium (Moilanen, 1999) and no colonization from were present, and start our simulations with these patches. outside the network. Though it is difficult to guarantee that patch net Removing the lower connectivity patches from the initial starting works are at equilibrium (Hanski et al., 1995), there was a balance be conditions did not alter the rank order of connectivity of patches. tween the overall number of occupied swamps over the two years, Hence we could determine the importance of connectivity during our consistent with a possible steady-state (Hanski, 1999; Franken and Hik, simulation scenarios with various combinations of patches, avoiding 2004). Due to the steep topography surrounding the study region, and intractable circularity in the modelling process which would result if with the closest stretch of papyrus outside of this network being ~40 km there was dependence on occupancy and abundance in other away, colonizations from outside this isolated network are thus unlikely. patches. Patches were added into the network from most to least All patches below the minimum observed occupied patch area from the connected. To calculate total wetland area required for levels of two years of field observations were removed. 19,200 iterations were persistence prior to the two most connected patches, these remaining performed for each estimation (as specified by high effort level in patches were reduced in size, keeping density levels constant as SPOMSIM), repeated at least three times with the same starting pa above. rameters to check for convergence (Moilanen, 2004). The parameter set Protecting the single largest patch: Metapopulation dynamics in with the lowest AIC value was selected for use in model simulations SPOMSIM can only be modelled with a minimum of two patches. (Burnham and Anderson, 2002).

4 L. Donaldson et al. Biological Conservation 258 (2021) 109125

Thus, the probability of survival of the single biggest patch in the the highest density (Table 1). In 2015, the proportion of newly colonized network was calculated using Eq. (3). Population size was estimated swamps was the highest for papyrus canary (52%) and the lowest for using the initial population size (at time 0) multiplied by the survival white-winged swamp-warbler (7%) (Table 1). The proportion of locally rate after 100 years. As previously, this patch was reduced in size extinct swamps in the network since 2014 were broadly similar between with density held constant, to enable comparisons. species in 2015, ranging from 11% of total patches for Carruthers’s cisticola, to 29% for papyrus canary (Table 1). Metapopulation param To explore the most effective strategy for individual versus multiple eter estimates exhibited variation among species (Table 2). species persistence, the area of habitat required for each species to exceed 95% likelihood of persistence was noted for each strategy, and the minimum area of habitat required overall (across all species) was 3.2. Simulations used to estimate the amount of wetland required for a given strategy to ensure each papyrus-specialist persisted over the long-term. Considering 3.2.1. Metapopulation persistence that the optimal network of patches differed between species for each The relationships between persistence and wetland area for each strategy, the area of habitat required in total to protect the individual strategy differed among species (Fig. 2a-e). With more wetland across networks for all species was calculated. In situations where optimal the network, investing in the highest quality patches for each species networks consisted of papyrus patches (for the papyrus-only species) consistently achieved 95% probability of persistence with the least total and broader wetland patches (for the broad wetland species), the area area. Retaining a high-quality habitat network gave similar results to including the broader wetland was used. protecting large sites for papyrus canary (Fig. 2b) and papyrus yellow warbler (Fig. 2e), the species with lower population densities (Table 1), and was similar to conserving more connected sites for white-winged 2.6. Regional stochasticity swamp-warbler (Fig. 2d), which had the lowest observed turnover (Table 1). Investing in numerous smaller sites was the least favourable Estimating regional stochasticity in metapopulations (the extent to option for most species, requiring a large amount of habitat before 95% which colonization and local extinction probability fluctuate synchro persistence was reached (ranging from ~82 ha across 492 patches for nously among patches from year to year) would require data from greater swamp-warbler, to ~332 ha across 131 patches for papyrus multiple years (Moilanen, 1999). Thus, although we were able to infer yellow warbler). The two species with higher carrying capacities but a the effects of patch area and isolation on colonization and extinction low population turnover, Carruthers’s cisticola and white-winged from the spatial distribution of occupied and vacant patches in the two swamp-warbler (Table 1), were an exception to this; investing in a few consecutive years, we parameterised the models without estimating large sites required the most amount of habitat to ensure a 95% chance regional stochasticity. However, because synchrony among populations of persistence (>300 ha; Fig. 2c, 2d). However, investing in the single can increase metapopulation extinction risk, we conducted simulations largest site for white-winged swamp-warbler never reached the 95% including moderate-high levels (0.2) of regional stochasticity (Ozgul likelihood of persistence (Fig. 2d), and the single largest wetland in the et al., 2006; Poos and Jackson, 2012). All simulations were repeated overall network for Carruthers’s cisticola must remain intact for this without regional stochasticity to confirmthat our conclusions were not species to exceed the 95% threshold (Fig. 2c). Meanwhile, papyrus sensitive to the inclusion of stochasticity (Fig. A1). yellow warbler, the most dispersive species (Table 2), required a large amount of habitat before connectivity (~250 ha across 62 patches) or a 3. Results network of small sites exceeded the % persistence threshold (Fig. 2e).

3.1. Species-specific data 3.2.2. Enhancing population size The relationships between relative population size and wetland area The number of suitable habitat patches in the network ranged from differed depending on the strategy for conserving patches (Fig. 2g-j), 77 for Carruthers’s cisticola to 518 for greater swamp-warbler (Table 1). with the exception of greater swamp-warbler which exhibited a similar Papyrus canary and papyrus yellow warbler occupied the lowest pro relationship for all strategies (Fig. 2f). The greatest differences among portion of suitable patches (11.3% and 14.5% in 2014 respectively) and strategies were shown for the broader wetland species which inhabit persisted at the lowest densities, while greater swamp-warbler occupied larger patches within the network: Carruthers’s cisticola and papyrus the highest proportion of suitable sites (53.1% in 2015) and persisted at yellow warbler (Table 1). Protecting large and high-quality patches

Table 1 2014–2015 survey data for greater swamp-warbler (GSW), papyrus canary (PC), Carruthers’s cisticola (CC), white-winged swamp-warbler (WWW), papyrus yellow warbler (PYW).

Species Suitable patches Habitat (ha) Mean patch Mean patch distance (km) (SD) Max patch density Relative carrying Occupied 2014–15 size (ha) (SD) (per ha) capacity 2014 2015 Ext Col

GSW 518* 550.7 1.06 (8.21) 10.7 (7.3) 355 269 275 (53.1%) 63 69 420 (51.9%) (23%) (25%) PC 495* 550.6 1.11 (8.4) 11.5 (7.7) 188 56 84 16 44 0.5 (11.3%) (17.0%) (29%) (52%) †‡ CC 77 1829.8 23.76 (119.21) 11.4 (7.4) 1320 35 39 4 8 17.6 (45.5%) (50.6%) (11%) (21%) WWW 197 537.1 2.73 (13.17) 12.2 (8.0) 638 53 44 12 3 394 (26.9%) (22.3%) (23%) (7%) † PYW 138 1068.0 7.74 (31.54) 10.8 (6.8) 330 20 27 3 10 1 (14.5%) (19.6%) (15%) (37%)

Habitat (ha) is the total available suitable habitat across the network per species. Mean patch distance is the average distance between all occupied patches across the 2 years of survey. Max patch density is the maximum ∑predicted density for a single patch for each species. Relative carrying capacity represents the predicted maximum possible population size for each species, based on (patch area x patch density). The percentage of occupied patches in 2014 and 2015 (from the total number of patches), and newly extinct (“Ext”) and colonized (“Col”) patches in 2015 (from the total number of populations in 2014 and 2015 respectively), are shown in brackets. † ‡ *Includes shoreline fringing patches; Includes broader wetland vegetation; Includes agricultural wetland.

5 L. Donaldson et al. Biological Conservation 258 (2021) 109125

Table 2 Metapopulation parameter values (α, y, μ, x) for the model with the lowest AIC for greater swamp-warbler (GSW), papyrus canary (PC), Carruthers’s cisticola (CC), white-winged swamp-warbler (WWW), papyrus yellow warbler (PYW). α is the dispersal parameter (km), y relates to colonization probability, μ and x refer to extinction risk (see Material and methods for details). 95% confidence intervals shown in brackets.

Species α y μ x Dispersal ability‡

GSWa 0.204 226.017 0.012 0.864 Low PCa 0.190 185.753 0.012 0.935 Low CC 0.070 1998.430 0.061 0.734 Intermediate (0.000–0.151) (1164.079–5417.371) (0.037–0.072) (0.523–1.031) WWW 0.021 5512.051 0.059 0.488 Intermediate (0.003–0.051) (3399.745–8138.029) (0.059–0.059) (0.361–0.509) PYW 0.001 1446.647 0.041 1.340 High (0.000–0.021) (1192.023–1984.177) (0.041–0.067) (0.720–2.579)

a Parameters settled at local minima and prevented estimation of 95% confidence intervals; estimates were calculated at least 3 times (Table A1) and confirmed similar values were obtained. ‡Relative dispersal ability based on the values calculated for the dispersal parameter for each species. generated a higher population size than smaller and more connected depend also on the total area protected. Designing optimal protected sites for these species, at least until the maximum amount of wetland in networks based on one species does not consistently lead to an optimal the network was reached (Fig. 2h, 2j). In contrast, connectivity was a network design for all, even for specialist species occupying the same beneficial strategy, alongside large and high-quality networks, for general habitat type. For the conservation of individual species, focusing achieving high populations of papyrus canary and greater swamp- on habitat quality is generally the most effective approach in our study warbler (Table 2; Fig. 2g; 2f), the least dispersive of the species. The system, but should the aim be to conserve multiple species, focusing on optimal strategy for achieving high population size differed depending connectivity is the most efficient method, due to greater congruence in on the amount of habitat available for white-winged swamp-warbler the patches that are important for species. only (Fig. 2i). This species was predicted to reside at relatively high densities (Table 1), thus benefited more from high quality sites compared to the other options when a small amount of papyrus was 4.1. Protected networks for single species available. However, as habitat increased, preserving a few larger sites enhanced the population size more than for the other strategies. It has become more common for conservation policy to aim to manage viable metapopulations than to concentrate resources on small, 3.2.3. Achieving persistence of multiple species isolated populations (Hanski et al., 1996b). Guidance for creating more Overall, the amount of high-quality habitat required to achieve effective ecological networks recommends that sites should first be ≥95% likelihood of persistence under the optimal strategy varied be made better, followed by bigger, then new sites created and finallysites tween species: approximately 25 ha, 144 ha, 61 ha and 31 ha for greater joined more (Lawton et al., 2010). However, our results suggest that swamp-warbler, white-winged swamp-warbler, Carruthers’s cisticola variation in the ecology of species changes the rank order of which and papyrus yellow warbler respectively, and 48 ha of large patches of strategy is best. habitat for papyrus canary (Fig. 2a-e). In terms of probability of metapopulation persistence and overall To ensure persistence of the single most demanding species for each metapopulation size, preservation of large sites was shown to be the strategy (regardless of the individual identity of patches in the network), most effective strategy for avifauna with the highest and lowest conserving a network of high-quality patches required the least area dispersal abilities. However, protecting larger sites is only effective at (144 ha), followed by connectivity (255 ha) (Fig. 3), driven by the boosting population sizes for species with lower (regional) carrying required area for white-winged swamp-warbler and papyrus yellow capacities (Griffen and Drake, 2008) and under a scenario of scarce warbler respectively (Fig. 2d; 2e). No single sites were of a sufficientsize habitat. Under the same scenario, investing in high quality sites was, to ensure 95% probability of persistence of white-winged swamp-war conversely, the best option for obtaining high population sizes and bler over 100 years (Fig. 2d), and concentrating on the single biggest enhancing the persistence of species with high regional carrying ca patch required the most amount of habitat to be suitable for the other pacities. Though some species occupied smaller habitat patches than species (Fig. 3), due to the large amount of wetland required for a single others, these smaller patches alone could not sustain populations in the patch to be suitable for Carruthers’s cisticola (Fig. 2c). long-term, hence persistence was enhanced by the maintenance of large However, when the total area required to conserve the optimal sites. Nevertheless, for the least dispersive species, smaller sites could networks for all species is considered (accounting for the identity of act as useful stepping stones to reach large sites (Saura et al., 2014). In individual patches in the network), the most favourable strategy overall contrast, enhancing connectivity was the least effective strategy, (requiring the least amount of habitat) differs (see ‘All species’ in Fig. 3). though, consistent with previous work (e.g. Thomas, 2000), it assisted The most connected sites overlapped between species to a greater extent species with intermediate dispersal capabilities to utilise desirable sites than the other strategies, implying that investment in a series of the most within the network, at least when more habitat was available. Species connected patches required the least amount of wetland overall (~299 with comparatively higher dispersal capabilities also have the freedom ha) (Fig. 3). High quality patches, on the other hand, intersected the to choose which sites to utilise (Glorvigen et al., 2013), hence why these least between species and required the conservation of approximately species benefited from higher quality habitat at lower levels of area 528 ha of wetland overall (Fig. 3). Investing in the single largest site compared to the other options explored. remained the least favourable strategy, for it could only ensure the Papyrus-specialists also vary in their preference to occupy the edge persistence of four of the study species and would require the protection and interior parts of swamps (Britton, 1971; Donaldson et al., 2016). of at least 812 ha of additional wetland (Fig. 3). Smaller, more fragmented sites have a higher edge to area ratio than larger sites (Fahrig, 2003). As expected, interior species (white-winged ’ 4. Discussion swamp-warbler, Carruthers s cisticola and papyrus yellow warbler) had notably smaller population sizes when multiple smaller sites were pre We show that differences in the ecology of species lead to differences served, while those using the edge of swamps (greater swamp-warbler in the strategy that would be most effective for their conservation, which and papyrus canary [Britton, 1978; Donaldson et al., 2016]) displayed a steady increase in population regardless of smaller or bigger sites being

6 L. Donaldson et al. Biological Conservation 258 (2021) 109125 e c n e t s i s r e P Population size

Area (ha) protected

Fig. 2. Output from metapopulation simulations for the study species: protecting the single largest, largest (2+), smallest, most connected and highest quality patches in the current network. a-e display the mean proportion of replicates that persisted after 100 years, and f-j show the mean relative population size after 100 years, against the total area of suitable wetland habitat available across the network (a-e are plotted on the log+1 scale for clarity). Solid lines show simulation results from habitat configurationas it was in 2015, dashed lines represent results from scenarios modelled by artificiallyreducing area to allow for a comparison between strategies at equivalent levels of habitat (see Model simulations in Materials and methods for full explanation). All simulation results shown here include regional stochasticity (see Fig. A1 for equivalent simulations without regional stochasticity).

7 L. Donaldson et al. Biological Conservation 258 (2021) 109125

Fig. 3. Area required for each network considered to achieve ≥95% probability of persistence over 100 years for the study species: protecting the single most connected patches, 2+ largest patches, highest quality patches, sin gle large patches* and greater number of the smallest patches across the network. Dark grey: the smallest area required to predict 95% chance of persistence for the single most demanding species; light grey: the smallest area needed to predict 95% chance of persistence for all species combined (considering patch overlap between All species species). *as it stands, no single large papyrus swamps are Single most demanding species sufficient to ensure ≥95% probability of persistence for

Area (ha) white-winged swamp-warbler.

Patches conserved

maintained. Similarly, the threshold level of persistence for edge species 2013; Heinrichs et al., 2016), the elements that represent high quality differed marginally between strategies, while for papyrus yellow war are species-specific (Mortelliti et al., 2010). Thus, when the optimal bler, a species commonly located within the interior of swamps, pre quality networks for each species were merged, the amount of habitat serving smaller sites was the least favourable option. Previous work by required for multiple species overall was higher than preserving con Vande weghe (1981) in swamps in nearby Rwanda and Burundi showed nected or large sites. Investing in a single large site became the least that competitive exclusion of the species studied here is unlikely, thus favourable strategy as not only does this require a large amount of species interactions are not expected to have altered the results habitat to sustain populations, for those with low turnover, protecting a discussed. single site was not sufficient to maintain a high probability of persistence.

4.2. Protected networks for multiple species 4.3. Conservation implications Protected area networks are usually designed for multiple species (Margules and Pressey, 2000), though making decisions for multiple Our results confirm that conservation planning can ensure the species is challenging (Albert et al., 2017). To combat this, conservation persistence of all species in a landscape if sufficient habitat is available managers often select ‘umbrella’ species, using one species with similar and in such situations, species-specific management is not necessarily needs as a surrogate for others (Akcakaya et al., 2007), thereby required. Networks should instead be designed with consideration of assuming that the species community as a whole will benefit (Bennett more practical constraints, such as ownership and finances (Donaldson et al., 2015). Previous studies have demonstrated the capacity of pa et al., 2017). However, because the resources available for conservation rameters in metapopulation models to predict dynamic patterns of other worldwide are limited and the demand for land particularly in devel related species (e.g. Wahlberg et al., 1996), or the potential of a nested oping countries is high (Fisher and Christopher, 2007), it is not apt or community approach where a more common species is chosen to feasible to conserve all suitable space for wildlife. We have demon represent rare species with nested subset distribution (Patterson, 1987). strated that, where basic knowledge gaps for many species remain (as is However, our results imply that using a single species as a surrogate commonly the case in the Tropics), and the characteristics of individual could be problematic when they fail to capture the needs of the wider species residing in a network do not directly overlap (Meurant et al., suite of specialist species using this habitat (Meurant et al., 2018), with 2018), guidelines that prioritise connectivity in protected area networks slightly opposing distributions (Guti´errez et al., 2001). should be developed and applied, in order to maintain multi-species Considering the optimal networks for all species together, we show persistence and population size (Albert et al., 2017). that concentrating on the most connected or a few large sites for papyrus Papyrus swamps have suffered high rates of loss over the past few birds requires comparatively less area than preserving either high decades (van Dam et al., 2014) and the limited protection offered to quality wetlands, or investing in lots of small patches across the species occupying these wetlands tends to focus on large sites hosting network. Larger sites offer numerous benefits,such as the availability of high numbers of specific species (e.g. BirdLife International, 2020). greater habitat heterogeneity that can support a wider range of species However, focusing on large or species-rich sites in isolation would (Donaldson et al., 2017). However, securing or restoring sites of a suf require an infeasible amount of wetland to be protected for the persis ficient size for persistence is often not possible in modern landscapes tence of all species. Securing large sites alone in this region is impractical (Doerr et al., 2011). In this case, spreading resources across the network (DeFries et al., 2007) and restoration can be expensive and time- and enhancing connectivity essentially serves to create a functionally consuming (Possingham et al., 2015). Moreover, predicted extreme bigger site, offering the same benefits,providing that the configuration weather events across this region and others (Doherty et al., 2010; of sites reflects dispersal capabilities. Although habitat quality is Ponce-Reyes et al., 2017) undermine the likely effectiveness of conser increasingly incorporated into landscape-scale conservation initiatives vation approaches based on individual, extinction-prone populations in (Thomas et al., 2001) to provide high source populations for recoloni single sites (Schnell et al., 2013; D’Aloia et al., 2019). Instead, habitat zation and prevent the network from overall extinction (Glorvigen et al., specialists such as those of papyrus swamps will benefit from a

8 L. Donaldson et al. Biological Conservation 258 (2021) 109125 consideration of metapopulation dynamics, with investment spread Data accessibility across a connected network of appropriately sited, well-managed, large sites. Enhancing connectivity can have unintended effects such as Data are available at https://doi.org/10.6084/m9.figshare increasing movement of invasive species, disease, and predators, or the .14485242. propagation of environmental perturbations such as fire, drought or flooding across wider areas of relatively homogeneous habitat (Donaldson et al., 2017). The benefits of increasing connectivity there Declaration of competing interest fore need to be assessed carefully for communities of species that are vulnerable to a common risk such as predation by invasive species The authors declare that they have no known competing financial (Falaschi et al., 2021). In our study system, the naturally fragmented interests or personal relationships that could have appeared to influence nature of papyrus swamps and the surrounding land help to minimise the work reported in this paper. the spread of threats such as pollution and alien species (Kipkemboi and van Dam, 2016), so that maintaining or increasing habitat connectivity Acknowledgements largely have beneficial effects on the focal species through increased dispersal or colonization (Williams et al., 2005). This research was supported by a Natural Environment Research Further efforts are required to improve the connectivity of protected Council (NERC) CASE studentship, in partnership with the Royal Society areas globally, and Uganda has been identified as a priority country for the Protection of Birds (RSPB; grant number NE/L501669/1). where protected sites should be strategically designated in locations that Additional financial support for fieldwork was provided by The Ex improve connectivity (Saura et al., 2018). In practice, maintaining plorers Club, British Ornithologists’ Union, Royal Geographic Society, connectivity across a network of protected sites requires strategies to John Muir Trust, and Gilchrist Educational Trust. Permission to conduct maintain and manage habitats, which in the case of papyrus swamps, this research was granted by the Uganda National Council of Science and necessitates the reduction of ongoing habitat loss (Maclean et al., Technology (UNCST). We are grateful for the valuable field assistance 2011b) and permitting moderate disturbance (Donaldson et al., 2016). provided by Steven Katungi, Johnson Ruhakana, Hilary Mwakire, In general, protected areas tend to be biased towards habitats not Columban Kamunyu and Anna Woodhead, and the in-country support threatened by conversion (Joppa and Pfaff, 2009). Wetland networks from Nature Uganda. We also thank Daniel Padfieldand James Duffy for such as those surrounding Lake Bunyonyi, where extensive networks of technical advice, Stephen Willis and Annette Broderick for helpful dis habitat remain vulnerable to loss and degradation, warrant wider scale cussions, and Vincent Devictor and anonymous referees for useful protection than currently in place, to help secure populations of those comments on the manuscript. threatened species relying on these habitats. Implementing participatory approaches to conservation is vital to ensure that this can be achieved Appendix A. Supplementary data (Kipkemboi and van Dam, 2016). The highly decentralised governance structure established in Uganda, for example, provides a key mechanism Supplementary data to this article can be found online at https://doi. with which to implement this strategy (Maclean et al., 2011a), thereby org/10.1016/j.biocon.2021.109125. recognising that those living within proximity to swamps stand to gain most from their sustainable conservation (Maclean et al., 2003b). References

5. Conclusions Akcakaya, H.R., Mills, G., Doncaster, C.P., Service, K., 2007. The role of metapopulations in conservation. In: Macdonald, D. (Ed.), Key Topics in Conservation Biology. Blackwell Publishing, Oxford, pp. 64–84. This study demonstrates the importance of focusing on habitat con Albert, C.H., Rayfield, B., Dumitru, M., Gonzalez, A., 2017. Applying network theory to nectivity in landscape-scale conservation planning to meet the needs of prioritize multispecies habitat networks that are robust to climate and land-use species with overlapping habitat requirements. We show that differ change. Conserv. Biol. 31, 1383–1396. https://doi.org/10.1111/cobi.12943. ences in ecological characteristics can influencethe optimal strategy for Bennett, J.R., Maloney, R., Possingham, H.P., Bennett, J.R., 2015. Biodiversity gains from efficient use of private sponsorship for flagship species conservation. Proc. R. maintaining persistence in ecological networks. For single species con Soc. B Biol. Sci. 282, 1–7. https://doi.org/10.1098/rspb.2014.2693. servation programmes, managing sites to maintain high quality is an BirdLife International, 2020. Important bird areas factsheet: Nyamuriro swamp. effective approach. However, because habitat characteristics that yield Downloaded from. http://www.birdlife.org. (Accessed 28 May 2020). Britton, P.L., 1971. Two sympatric canaries, Serinus koliensis and S. citrinelloides, in the highest densities differ among species, multi-species conservation Western Kenya. Auk 88, 911–914. https://doi.org/10.2307/4083848. programmes may gain most from strategies that are more congruent Britton, P.L., 1978. Seasonality, density and diversity of birds of a papyrus swamp in between species, such as enhancing connectivity across the network. western Kenya. Ibis 120, 450–466. https://doi.org/10.1111/j.1474-919X.1978. tb06811.x. With this in mind, utilising surrogate species to predict the conservation Burnham, K.P., Anderson, D.R., 2002. Model Selection and Multimodel Inference: A outcomes for other species occupying the same landscape may not be Practical Information-theoretic Approach. Springer-Verlag. https://doi.org/ sufficient,particularly when faced with differences between the habitat 10.1002/1521-3773(20010316)40:6<9823::AID-ANIE9823>3.3.CO;2-C. Butchart, S.H.M., Scharlemann, J.P.W., Evans, M.I., Quader, S., Arico,` S., Arinaitwe, J., requirements of individual species, and the extreme habitat destruction Balman, M., Bennun, L.A., Bertzky, B., Besançon, C., Boucher, T.M., Brooks, T.M., facing biodiversity. But by investing in the management of a few large Burfield, I.J., Burgess, N.D., Chan, S., Clay, R.P., Crosby, M.J., Davidson, N.C., de sites and maintaining sufficient levels of connectivity across the land Silva, N., Devenish, C., Dutson, G.C.L., Fernandez,´ D.F.D., Fishpool, L.D.C., Fitzgerald, C., Foster, M., Heath, M.F., Hockings, M., Hoffmann, M., Knox, D., scape, the long-term persistence of multiple species can be sustained. Larsen, F.W., Lamoreux, J.F., Loucks, C., May, I., Millett, J., Molloy, D., Morling, P., Parr, M., Ricketts, T.H., Seddon, N., Skolnik, B., Stuart, S.N., Upgren, A., CRediT authorship contribution statement Woodley, S., 2012. Protecting important sites for biodiversity contributes to meeting global conservation targets. PLoS One 7 (3), e32529. https://doi.org/10.1371/ journal.pone.0032529. Lynda Donaldson: Conceptualization, Methodology, Formal anal Byaruhanga, A., Kasoma, P., Pomeroy, D., 2001. Important Bird Areas in Uganda. Nature, ysis, Investigation, Data curation, Writing – original draft, Writing – Uganda, Kampala. review & editing, Visualization, Funding acquisition. Jonathan J. CBD, 2011. Aichi Biodiversity Targets. http://www.cbd.int/sp/targets/ (Accessed 10 January 2015). Bennie: Conceptualization, Methodology, Formal analysis, Writing – Conn, P.B., Johnson, D.S., Boveng, P.L., 2015. On extrapolating past the range of review & editing, Supervision. Robert J. Wilson: Conceptualization, observed data when making statistical predictions in ecology. PLoS One 10, 1–16. – & https://doi.org/10.1371/journal.pone.0141416. Methodology, Writing review editing, Supervision. Ilya M.D. ’ – & D Aloia, C.C., Naujokaitis-Lewis, I., Blackford, C., Chu, C., Curtis, J.M.R., Darling, E., Maclean: Conceptualization, Methodology, Writing review editing, Guichard, F., Leroux, S.J., Martensen, A.C., Rayfield, B., Sunday, J.M., Xuereb, A., Supervision, Funding acquisition. Fortin, M.J., 2019. Coupled networks of permanent protected areas and dynamic

9 L. Donaldson et al. Biological Conservation 258 (2021) 109125

conservation areas for biodiversity conservation under climate change. Front. Ecol. tropics: Uganda’s national system of Forest Nature Reserves. Conserv. Biol. 14, Evol. 7, 1–8. https://doi.org/10.3389/fevo.2019.00027. 858–875. https://doi.org/10.1046/j.1523-1739.2000.99180.x. DeFries, R., Hansen, A.J., Turner, B.L., Reid, R., Liu, J., 2007. Land use change around Hoyle, M., James, M., 2005. Global warming, human population pressure, and viability protected areas: management to balance human needs and ecological function. Ecol. of the world’s smallest butterfly. Conserv. Biol. 19, 1113–1124. https://doi.org/ Appl. 17, 1031–1038. https://doi.org/10.1890/05-1111. 10.1111/j.1523-1739.2005.00166.x. Denny, P., 1972. Lakes of south-western Uganda I. Physical and chemical studies on Lake IUCN, 2021. The IUCN red list of threatened species. https://www.iucnredlist.org/ Bunyonyi. Freshw. Biol. 2, 143–158. https://doi.org/10.1111/j.1365-2427.1972. (Accessed 20 January 2021). tb00367.x. Joppa, L.N., Pfaff, A., 2009. High and far: biases in the location of protected areas. PLoS Doerr, V.A.J., Barrett, T., Doerr, E.D., 2011. Connectivity, dispersal behaviour and One 4, 1–6. https://doi.org/10.1371/journal.pone.0008273. conservation under climate change: a response to Hodgson et al. J. Appl. Ecol. 48, Kipkemboi, J., van Dam, A.A., 2016. Papyrus wetlands. In: Finlayson, C.M., et al. (Eds.), 143–147. https://doi.org/10.1111/j.1365-2664.2010.01899.x. The Wetland Book Vol. II: Distribution, Description and Conservation. Springer, Doherty, R.M., Sitch, S., Smith, B., Lewis, S.L., Thornton, P.K., 2010. Implications of Dordrecht, pp. 1–15. https://doi.org/10.1007/978-94-007-6173-5. future climate and atmospheric CO2 content for regional biogeochemistry, Lawson, C.R., Bennie, J.J., Thomas, C.D., Hodgson, J.A., Wilson, R.J., 2012. Local and biogeography and ecosystem services across East Africa. Glob. Chang. Biol. 16, landscape management of an expanding range margin under climate change. J. Appl. 617–640. https://doi.org/10.1111/j.1365-2486.2009.01997.x. Ecol. 49, 552–561. https://doi.org/10.1111/j.1365-2664.2011.02098.x. Donaldson, L., Woodhead, A.J., Wilson, R.J., Maclean, I.M.D., 2016. Subsistence use of Lawton, J., Brotherton, P.N.M., Brown, V.K., Elphick, C., Fitter, A.H., Forshaw, J., papyrus is compatible with wetland bird conservation. Biol. Conserv. 201, 414–422. Haddow, R.W., Hilborne, S., Leafe, R.N., Mace, G.M., Southgate, M.P., https://doi.org/10.1016/j.biocon.2016.07.036. Sutherland, W.J., Tew, T.E., Varley, J., Wynne, G.R., 2010. Making Space for Donaldson, L., Wilson, R.J., Maclean, I.M., 2017. Old concepts, new challenges: adapting Nature : a review of England’s wildlife sites and ecological network. In: Report to landscape-scale conservation to the twenty-first century. Biodivers. Conserv. 26, Defra. 527–552. https://doi.org/10.1007/s10531-016-1257-9. MacArthur, R.H., Wilson, E.O., 1967. The Theory of Island Biogeography. Princeton Donaldson, L., Bennie, J.J., Wilson, R.J., Maclean, I.M., 2019. Quantifying resistance and University Press. resilience to local extinction for conservation prioritization. Ecol. Appl. 29, e01989 Maclean, I.M.D., Hassall, M., Boar, R., Nasirwa, O., 2003a. Effects of habitat degradation https://doi.org/10.1002/eap.1989. on avian guilds in East African papyrus Cyperus papyrus swamps. Bird Conserv. Int. Dudley, N., 2008. Guidelines for applying protected area management categories, IUCN, 13, 283–297. https://doi.org/10.1017/S0959270903003216. Gland, Switzerland. https://doi.org/10.2305/IUCN.CH.2008.PAPS.2.en. Maclean, I.M.D., Tinch, R., Hassall, M., Boar, R., 2003b. Social and economic use of Etienne, R.S., 2004. On optimal choices in increase of patch area and reduction of wetland resources: a case study from Lake Bunyoni, Uganda, ECM 03-09. CSERGE interpatch distance for metapopulation persistence. Ecol. Model. 179, 77–90. Working Paper ECM 03-09, Norwich. https://doi.org/10.1016/j.ecolmodel.2004.05.003. Maclean, I.M.D., Hassall, M., Boar, R.R., Lake, I.R., 2006. Effects of disturbance and Fahrig, L., 2003. Effects of habitat fragmentation on biodiversity. Annu. Rev. Ecol. Evol. habitat loss on papyrus-dwelling passerines. Biol. Conserv. 131, 349–358. https:// Syst. 34, 487–515. https://doi.org/10.1146/132419. doi.org/10.1016/j.biocon.2005.12.003. Falaschi, M., Giachello, S., Parrino, E. Lo, Muraro, M., Manenti, R., Ficetola, G.F., 2021. Maclean, I.M.D., Boar, R.R., Lugo, C., 2011a. A review of the relative merits of Long-term drivers of persistence and colonization dynamics in spatially structured conserving, using, or draining papyrus swamps. Environ. Manag. 47, 218–229. amphibian populations. Conserv. Biol. https://doi.org/10.1111/cobi.13686. https://doi.org/10.1007/s00267-010-9592-1. Fisher, B., Christopher, T., 2007. Poverty and biodiversity: measuring the overlap of Maclean, I.M.D., Wilson, R.J., Hassall, M., 2011b. Predicting changes in the abundance of human poverty and the biodiversity hotspots. Ecol. Econ. 62, 93–101. https://doi. African wetland birds by incorporating abundance-occupancy relationships into org/10.1016/j.ecolecon.2006.05.020. habitat association models. Divers. Distrib. 17, 480–490. https://doi.org/10.1111/ Foley, J.A., Defries, R., Asner, G.P., Barford, C., Bonan, G., Carpenter, S.R., Chapin, F.S., j.1472-4642.2011.00756.x. Coe, M.T., Daily, G.C., Gibbs, H.K., Helkowski, J.H., Holloway, T., Howard, E.A., Maclean, I.M.D., Bird, J.P., Hassall, M., 2014. Papyrus swamp drainage and the Kucharik, C.J., Monfreda, C., Patz, J.A., Prentice, I.C., Ramankutty, N., Snyder, P.K., conservation status of their avifauna. Wetl. Ecol. Manag. 22, 115–127. https://doi. 2005. Global consequences of land use. Science. 309, 570–574. https://doi.org/ org/10.1007/s11273-013-9292-8. 10.1126/science.1111772. Margules, C.R., Pressey, R.L., 2000. Systematic conservation planning. Nature 405, Franken, R.J., Hik, D.S., 2004. Influence of habitat quality, patch size and connectivity 243–253. https://doi.org/10.1038/35012251. on colonization and extinction dynamics of collared pikas Ochotona collaris. McCarthy, D.P., Donald, P.F., Scharlemann, J.P.W., Buchanan, G.M., Balmford, A., J. Anim. Ecol. 73, 889–896. https://doi.org/10.1111/j.0021-8790.2004.00865.x. Green, J.M.H., Bennun, L.A., Burgess, N.D., Fishpool, L.D.C., Garnett, S.T., Glorvigen, P., Andreassen, H.P., Ims, R.A., 2013. Local and regional determinants of Leonard, D.L., Maloney, R.F., Morling, P., Schaefer, H.M., Symes, A., Wiedenfeld, D. colonisation-extinction dynamics of a riparian mainland-island root vole A., Butchart, S.H.M., 2012. Financial costs of meeting global biodiversity metapopulation. PLoS One 8, e56462. https://doi.org/10.1371/journal. conservation targets: current spending and unmet needs. Science 338 (80), 946–950. pone.0056462. https://doi.org/10.1126/science.1229803. Griffen, B.D., Drake, J.M., 2008. Effects of habitat quality and size on extinction in Meurant, M., Gonzalez, A., Doxa, A., Albert, C.H., 2018. Selecting surrogate species for experimental populations. Proc. R. Soc. B 275, 2251–2256. https://doi.org/ connectivity conservation. Biol. Conserv. 227, 326–334. https://doi.org/10.1016/j. 10.1098/rspb.2008.0518. biocon.2018.09.028. Gutierrez, D., 2005. Effectiveness of existing reserves in the long-term protection of a Moilanen, A., 1999. Patch cccupancy models of metapopulation dynamics: efficient regionally rare butterfly. Conserv. Biol. 19, 1586–1597. https://doi.org/10.1111/ parameter estimation using implicit statistical inference. Ecology 80, 1031–1043. j.1523-1739.2005.00210.x. https://doi.org/10.1890/0012-9658(1999)080[1031:POMOMD]2.0.CO;2. Gutiérrez, D., Leon-cort´ és, J.L., Menéndez, R., Wilson, R.J., Cowley, J.R., Thomas, C.D., Moilanen, A., 2004. SPOMSIM: software for stochastic patch occupancy models of 2001. Metapopulations of four lepidopteran herbivores on a single host plant, Lotus metapopulation dynamics. Ecol. Model. 179, 533–550. https://doi.org/10.1016/j. corniculatus. Ecology 82, 1371–1386. https://doi.org/10.1890/0012-9658(2001) ecolmodel.2004.04.019. 082[1371:MOFLHO]2.0.CO;2. Moilanen, A., Hanski, I., 1998. Metapopulation dynamics: effects of habitat quality and Hanski, I., 1991. Single-species metapopulation dynamics: concepts, models and landscape structure. Ecology 79, 2503–2515. https://doi.org/10.2307/176839. observations. Biol. J. Linn. Soc. 42, 17–38. https://doi.org/10.1111/j.1095- Moilanen, A., Smith, A.T., Hanski, I., 1998. Long-term dynamics in a metapopulation of 8312.1991.tb00549.x. the American pika. Am. Nat. 152, 530–542. https://doi.org/10.1086/286188. Hanski, I., 1994. A practical model of metapopulation dynamics. J. Anim. Ecol. 63, Mortelliti, A., Amori, G., Boitani, L., 2010. The role of habitat quality in fragmented 151–162. https://doi.org/10.2307/5591. landscapes: a conceptual overview and prospectus for future research. Oecologia Hanski, I., 1998. Connecting the parameters of local extinction and metapopulation 163, 535–547. https://doi.org/10.1007/s00442-010-1623-3. dynamics. Oikos 83, 390–396. https://doi.org/10.2307/3546854. Ovaskainen, O., Hanski, I., 2001. Spatially structured metapopulation models: global and Hanski, I., 1999. Metapopulation Ecology. Oxford Univeristy Press. local assessment of metapopulation capacity. Theor. Popul. Biol. 60, 281–302. Hanski, I., Thomas, C.D., 1994. Metapopulation dynamics and conservation: a spatially https://doi.org/10.1006/tpbi.2001.1548. explicit model applied to butterflies. Biol. Conserv. 68, 167–180. https://doi.org/ Ozgul, A., Armitage, K.B., Blumstein, D.T., Vanvuren, D.H., Oli, M.K., 2006. Effects of 10.1016/0006-3207(94)90348-4. patch quality and network structure on patch occupancy dynamics of a yellow- Hanski, I., Poyry, J., Pakkala, T., Kuussaari, M., 1995. Multiple equilibria in bellied marmot metapopulation. J. Anim. Ecol. 75, 191–202. https://doi.org/ metapopulation dynamics. Nature 377, 618–621. https://doi.org/10.1038/ 10.1111/j.1365-2656.2006.01038.x. 377618a0. Patterson, B.D., 1987. The principle of nested subsets and its implications for biological Hanski, I., Moilanen, A., Gyllenberg, M., 1996a. Minimum viable metapopulation size. conservation. Conserv. Biol. 1, 323–334. https://doi.org/10.1111/j.1523- Am. Nat. 147, 527–541. https://doi.org/10.1086/285864. 1739.1987.tb00052.x. Hanski, I., Moilanen, A., Pakkala, T., Kuussaari, M., 1996b. The quantitative incidence Ponce-Reyes, R., Plumptre, A.J., Segan, D., Ayebare, S., Fuller, R.A., Possingham, H.P., function model and peristence of an endangered butterfly metapopulation. Conserv. Watson, J.E.M., 2017. Forecasting ecosystem responses to climate change across Biol. 10, 578–590. https://doi.org/10.1046/j.1523-1739.1996.10020578.x. Africa’s Albertine Rift. Biol. Conserv. 209, 464–472. https://doi.org/10.1016/j. Hartter, J., Ryan, S.J., 2010. Top-down or bottom-up? Decentralization, natural resource biocon.2017.03.015. management, and usufruct rights in the forests and wetlands of western Uganda. Poos, M.S., Jackson, D.A., 2012. Impact of species-specific dispersal and regional Land Use Policy 27, 815–826. https://doi.org/10.1016/j.landusepol.2009.11.001. stochasticity on estimates of population viability in stream metapopulations. Landsc. Heinrichs, J.A., Bender, D.J., Schumaker, N.H., 2016. Habitat degradation and loss as Ecol. 27, 405–416. https://doi.org/10.1007/s10980-011-9683-2. key drivers of regional population extinction. Ecol. Model. 335, 64–73. https://doi. Possingham, H.P., Bode, M., Klein, C.J., 2015. Optimal conservation outcomes require org/10.1016/j.ecolmodel.2016.05.009. both restoration and protection. PLoS Biol. 13, e1002052 https://doi.org/10.1371/ Howard, P.C., Davenport, T.R.B., Kigenyi, F.W., Viskanic, P., Baltzer, M.C., Dickinson, C. journal.pbio.1002052. J., Lwanga, J., Matthews, R.A., Mupada, E., 2000. Protected area planning in the

10 L. Donaldson et al. Biological Conservation 258 (2021) 109125

¨ Saura, S., Bodin, O., Fortin, M.J., 2014. Stepping stones are crucial for species’ long- Tilman, D., May, R.M., Lehman, C.L., Nowak, M.A., 1994. Habitat destruction and the distance dispersal and range expansion through habitat networks. J. Appl. Ecol. 51, extinction debt. Nature 371, 65–66. https://doi.org/10.1038/371065a0. 171–182. https://doi.org/10.1111/1365-2664.12179. van Dam, A.A., Kipkemboi, J., Mazvimavi, D., Irvine, K., 2014. A synthesis of past, Saura, S., Bertzky, B., Bastin, L., Battistella, L., Mandrici, A., Dubois, G., 2018. Protected current and future research for protection and management of papyrus (Cyperus area connectivity: Shortfalls in global targets and country-level priorities. Biol. papyrus L.) wetlands in Africa. Wetl. Ecol. Manag. 22, 99–114. https://doi.org/ Conserv. 219, 53–67. https://doi.org/10.1016/j.biocon.2017.12.020. 10.1007/s11273-013-9335-1. Schnell, J.K., Harris, G.M., Pimm, S.L., Russell, G.J., 2013. Estimating extinction risk Vande weghe, J.-P., 1981. L’avifaune des papyraies au Rwanda et au Burundi. Le Gerfaut with metapopulation models of large-scale fragmentation. Conserv. Biol. 27, 71, 489–536. 520–530. https://doi.org/10.1111/cobi.12047. Wahlberg, N., Moilanen, A., Hanski, I., 1996. Predicting the occurrence of endangered Shaffer, M.L., Samson, F.B., 1985. Population size and extinction: a note on determining species in fragmented landscapes. Science 273 (80), 1536–1538. https://doi.org/ critical population sizes. Am. Nat. 125, 144–152. https://doi.org/10.1086/284332. 10.1126/science.273.5281.1536. Thomas, C.D., 2000. Dispersal and extinction in fragmented landscapes. Proc. R. Soc. Williams, J.C., ReVelle, C.S., Levin, S.A., 2005. Spatial attributes and reserve design Lond. B 267, 139–145. https://doi.org/10.1098/rspb.2000.0978. models: a review. Environ. Model. Assess. 10, 163–181. https://doi.org/10.1007/ Thomas, J.A., Bourn, N.A.D., Clarke, R.T., Stewart, K.E., Simcox, D.J., Pearman, G.S., s10666-005-9007-5. Curtis, R., Goodger, B., 2001. The quality and isolation of habitat patches both Wilson, R.J., Davies, Z.G., Thomas, C.D., 2009. Modelling the effect of habitat determine where butterflies persist in fragmented landscapes. Proc. R. Soc. Lond. B fragmentation on range expansion in a butterfly. Proc. R. Soc. B 276, 1421–1427. 268, 1791–1796. https://doi.org/10.1098/rspb.2001.1693. https://doi.org/10.1098/rspb.2008.0724.