A multi-scaled analysis of avian response to habitat amount and fragmentation in the Canadian dry mixed-grass prairie

Previous research has suggested that ducks and songbirds may benefit from prairie landscapes that consist primarily of contiguous grasslands. However, the relative importance of landscape-level vs. local characteristics on mechanisms underlying observed patterns is unclear. We measured effects of grassland amount and fragmentation on upland and wetland songbird and duck density and nest success, and on some nest predators, across 16 landscapes in southern Alberta, Canada. We compared these landscape-level effects with local-scale responses, including distance to various edges and vegetation characteristics. We also evaluated several statistical approaches to comparing effects of habitat characteristics at multiple spatial scales. Few species were influenced by grassland amount or fragmentation. In contrast, distance to edge and local vegetation characteristics had significant effects on densities and nest success of many species. Previous studies that reported effects of landscape characteristics may have detected patterns driven by local mechanisms. As a corollary, results were very sensitive to statistical model structure; landscape level effects were much less apparent when local characteristics were included in the models.     

Landscape patterns as habitat predictors: building and testing models for cavity-nesting birds in the Uinta Mountains of Utah,USA

The ability to predict species occurrences quickly is often crucial for managers and conservation biologists with limited time and funds. We used measured associations with landscape patterns to build accurate predictive habitat models that were quickly and easily applied (i.e., required no additional data collection in the field to make predictions). We used classification trees (a nonparametric alternative to discriminant function analysis, logistic regression, and other generalized linear models) to model nesting habitat of red-naped sapsuckers (Sphyrapicus nuchalis), northern flickers (Colaptes auratus),tree swallows (Tachycineta bicolor), and mountain chickadees (Parus gambeli) in the Uinta Mountains of northeastern Utah, USA. We then tested the predictive capability of the models with independent data collected in the field the following year. The models built for the northern flicker, red-naped sapsucker, and tree swallow were relatively accurate (84%, 80%, and 75% nests correctly classified,respectively)compared to the models for the mountain chickadee (50% nests correctly classified). All four models were more selective than a null model that predicted habitat based solely on a gross association with aspen forests. We conclude that associations with landscape patterns can be used to build relatively accurate, easy to use, predictive models for some species. Our results stress, however, that both selecting the proper scale at which to assess landscape associations and empirically testing the models derived from those associations are crucial for building useful predictive models. This revised version was published online in July 2006 with corrections to the Cover Date.     

Residuals cannot distinguish between ecological effects of habitat amount and fragmentation: implications for the debate

Habitat amount and fragmentation usually covary in natural and simulated landscapes. A common way of distinguishing between their effects is to take the residuals of the fragmentation index or indices regressed on habitat amount, as the index of habitat fragmentation. We used data on prairie songbird relative abundances from southern Alberta, Canada to compare this approach with the reverse: taking the residuals of habitat amount regressed on habitat fragmentation as the index of habitat amount. We used generalized additive models (GAMs) to derive residuals, and modeled relative abundances using linear mixed-effects models. The modeling approach used strongly influenced the statistical results. Using residuals as an index of fragmentation resulted in an apparently stronger effect of habitat amount relative to habitat fragmentation. In contrast, habitat fragmentation appeared more influential than habitat amount when residuals were used as an index of habitat amount. Regression of residuals may eliminate statistical collinearity, but cannot distinguish between the ecological effects of habitat amount and fragmentation. Habitat fragmentation may therefore have a larger effect on species than previously studies have shown, but experimental manipulations of underlying mechanisms are ultimately required to address this debate.     

Connectivity change in habitat networks

Habitat management is essential for safeguarding important flora and fauna. Further, habitat connectivity is a crucial component for maintaining biodiversity given that it is known to have implications for species persistence. However, damage to habitat due to natural and human induced hazards can alter spatial relationships between habitats, potentially impacting biodiversity. Therefore, the susceptibility of spatial relationships to patch loss and associated connectivity degradation is obviously an important factor in maintaining existing or planned habitat networks. Identifying patches vital to connectivity is critical both for effectively prioritizing protection (e.g., enhancing habitat connectivity) and establishing disaster mitigation measures (e.g., stemming the spread of habitat loss). This paper presents a methodology for characterizing connectivity associated with habitat networks. Methods for evaluating habitat network connectivity change are formalized. Examples are presented to facilitate analysis of connectivity in the management of biodiversity.

Comparison and development of new graph-based landscape connectivity indices: towards the priorization of habitat patches and corridors for conservation

The loss of connectivity of natural areas is a major threat for wildlife dispersal and survival and for the conservation of biodiversity in general. Thus, there is an increasing interest in considering connectivity in landscape planning and habitat conservation. In this context, graph structures have been shown to be a powerful and effective way of both representing the landscape pattern as a network and performing complex analysis regarding landscape connectivity. Many indices have been used for connectivity analyses so far but comparatively very little efforts have been made to understand their behaviour and sensitivity to spatial changes, which seriously undermines their adequate interpretation and usefulness. We systematically compare a set of ten graph-based connectivity indices, evaluating their reaction to different types of change that can occur in the landscape (habitat patches loss, corridors loss, etc.) and their effectiveness for identifying which landscape elements are more critical for habitat conservation. Many of the available indices were found to present serious limitations that make them inadequate as a basis for conservation planning. We present a new index (IIC) that achieves all the properties of an ideal index according to our analysis. We suggest that the connectivity problem should be considered within the wider concept of habitat availability, which considers a habitat patch itself as a space where connectivity exists, integrating habitat amount and connectivity between habitat patches in a single measure.     

The Implications of Metalandscape Connectivity for Population Viabilityin Migratory Songbirds

Landscape connectivity is considered important for species persistence, but linkages among landscape populations (metalandscape connectivity) may be necessary to ensure the long-term viability of some migratory songbirds at a broader regional scale. Because of regional source-sink dynamics, these species can maintain steady populations within extensively fragmented landscapes (landscape sinks) owing to high levels of immigration from source landscapes. We undertook a modeling study to identify the conditions under which immigration, an index of metalandscape connectivity, could rescue declining populations of songbirds in heavily disturbed landscapes. In general, low to moderate levels of immigration (m = 0–20%) were sufficient to rescue species with low edge-sensitivity in landscapes where<70% habitat had been destroyed. At the other extreme, moderate to high levels of immigration (m = 11–40%) were usually required to rescue highly edge-sensitive species in these same landscapes. Very high levels of immigration (m>40%) were required to rescue highly edge-sensitive species in extensively fragmented landscapes that had lost >50% habitat, or when any landscape lost ≥50% habitat gradually over a period of 100 or more years (r = 0.5% habitat lost/year). Paradoxically higher levels of immigration were thus necessary to offset population declines when habitat was lost gradually than when it was lost quickly, where population response lagged behind landscape change. This implies that the importance of metalandscape connectivity for population viability may not be fully appreciated in landscapes undergoing rapid rates of change. Natural immigration rates for migratory songbirds match the very high levels (>40%) we found necessary to sustain populations in heavily disturbed landscapes, which underscores the importance of metalandscape connectivity for the continued persistence of many migratory songbirds in the face of widespread habitat loss and fragmentation.     

Spatial models for inferring topographic controls on historical low-severity fire in the eastern Cascade Range of Washington,USA

Fire regimes are complex systems that represent an aggregate of spatial and temporal events whose statistical properties are scale dependent. Despite the breadth of research regarding the spatial controls on fire regime variability, few datasets are available with sufficient resolution to test spatially explicit hypotheses. We used a spatially distributed network of georeferenced fire-scarred trees to investigate the spatial structure of fire occurrence at multiple scales. Mantel’s tests and geostatistical analysis of fire-occurrence time series led to inferences about the mechanisms that generated spatial patterns of historical fire synchrony (multiple trees recording fire in a single year) in eastern Washington, USA. The spatial autocorrelation structure of historical fire regimes varied within and among sites, with clearer patterns in the complex rugged terrain of the Cascade Range than in more open and rolling terrain further north and east. Results illustrate that the statistical spatial characteristics of fire regimes change with landform characteristics within a forest type, suggesting that simple relationships between fire frequency, fire synchrony, and forest type do not exist. Quantifying the spatial structures in fire occurrence associated with topographic variation showed that fire regime variability depends on both landscape structure and the scale of measurement. Spatially explicit fire-scar data open new possibilities for analysis and interpretation, potentially informing the design and application of fire management on landscapes, including hazardous fuel treatments and the use of fire for ecosystem restoration.

Accessible habitat: an improved measure of the effects of habitat loss and roads on wildlife populations

Habitat loss is known to be the main cause of the current global decline in biodiversity, and roads are thought to affect the persistence of many species by restricting movement between habitat patches. However, measuring the effects of roads and habitat loss separately means that the configuration of habitat relative to roads is not considered. We present a new measure of the combined effects of roads and habitat amount: accessible habitat. We define accessible habitat as the amount of habitat that can be reached from a focal habitat patch without crossing a road, and make available a GIS tool to calculate accessible habitat. We hypothesize that accessible habitat will be the best predictor of the effects of habitat loss and roads for any species for which roads are a major barrier to movement. We conducted a case study of the utility of the accessible habitat concept using a data set of anuran species richness from 27 ponds near a motorway. We defined habitat as forest in this example. We found that accessible habitat was not only a better predictor of species richness than total habitat in the landscape or distance to the motorway, but also that by failing to consider accessible habitat we would have incorrectly concluded that there was no effect of habitat amount on species richness.     

Quantifying patch distribution at multiple spatial scales: Applications to wildlife-habitat models

Multiscale analyses are widely employed for wildlife-habitat studies. In most cases, however, each scale is considered discrete and little emphasis is placed on incorporating or measuring the responses of wildlife to resources across multiple scales. We modeled the responses of three Arctic wildlife species to vegetative resources distributed at two spatial scales: patches and collections of patches aggregated across a regional area. We defined a patch as a single or homogeneous collection of pixels representing 1 of 10 unique vegetation types. We employed a spatial pattern technique, three-term local quadrat variance, to quantify the distribution of patches at a larger regional scale. We used the distance at which the variance for each of 10 vegetation types peaked to define a moving window for calculating the density of patches. When measures of vegetation patch and density were applied to resource selection functions, the most parsimonious models for wolves and grizzly bears included covariates recorded at both scales. Seasonal resource selection by caribou was best described using a model consisting of only regional scale covariates. Our results suggest that for some species and environments simple patch-scale models may not capture the full range of spatial variation in resources to which wildlife may respond. For mobile animals that range across heterogeneous areas we recommend selection models that integrate resources occurring at a number of spatial scales. Patch density is a simple technique for representing such higher-order spatial patterns.     

Quantifying patch distribution at multiple spatial scales: Applications to wildlife-habitat models

Multiscale analyses are widely employed for wildlife-habitat studies. In most cases, however, each scale is considered discrete and little emphasis is placed on incorporating or measuring the responses of wildlife to resources across multiple scales. We modeled the responses of three Arctic wildlife species to vegetative resources distributed at two spatial scales: patches and collections of patches aggregated across a regional area. We defined a patch as a single or homogeneous collection of pixels representing 1 of 10 unique vegetation types. We employed a spatial pattern technique, three-term local quadrat variance, to quantify the distribution of patches at a larger regional scale. We used the distance at which the variance for each of 10 vegetation types peaked to define a moving window for calculating the density of patches. When measures of vegetation patch and density were applied to resource selection functions, the most parsimonious models for wolves and grizzly bears included covariates recorded at both scales. Seasonal resource selection by caribou was best described using a model consisting of only regional scale covariates. Our results suggest that for some species and environments simple patch-scale models may not capture the full range of spatial variation in resources to which wildlife may respond. For mobile animals that range across heterogeneous areas we recommend selection models that integrate resources occurring at a number of spatial scales. Patch density is a simple technique for representing such higher-order spatial patterns.     

Responses of Chilean forest birds to anthropogenic habitat fragmentation across spatial scales

Although it is recognized that anthropogenic forest fragmentation affects habitat use by organisms across multiple spatial scales, there is uncertainty about these effects. We used a hierarchical sampling design spanning three spatial scales of habitat variability (landscape > patch > within-patch) and generalized mixed-effect models to assess the scale-dependent responses of bird species to fragmentation in temperate forests of southern Chile. The abundances of nine of 20 bird species were affected by interactions across spatial scales. These interactions resulted in a limited effect of within-patch habitat structure on the abundance of birds in landscapes with low forest cover, suggesting that suitable local habitats, such as sites with dense understory cover or large trees, are underutilized or remain unused in highly fragmented landscapes. Habitat specialists and cavity-nesters, such as tree-trunk foragers and tapaculos, were most likely to exhibit interactions across spatial scales. Because providing additional sites with dense understory vegetation or large habitat trees does not compensate the negative effect of the loss of forest area on bird species, conservation strategies should ensure the retention of native forest patches in the mixed-use landscapes.     

The influence of research scale on bald eagle habitat selection along the lower Hudson River,New York (USA)

As the concepts of landscape ecology have been incorporated into otherdisciplines, the influence of spatial patterns on animal abundance anddistribution has attracted considerable attention. However, there remains asignificant gap in the application of landscape ecology theories and techniquesto wildlife research. By combining landscape ecology techniques withtraditionalwildlife habitat analysis methods, we defined an organism-centeredperspectivefor breeding bald eagles (Haliaeetus leucocephalus) alongthe Hudson River, New York, USA. We intensively monitored four pairs ofbreedingeagles during the 1999 and 2000 breeding seasons, and collected detailedinformation on perch and forage locations. Our analysis focused on threecritical habitat elements: available perch trees, access to foraging areas, andfreedom from human disturbance. We hypothesized that eagle habitat selectionrelative to each of these elementswould vary with the spatial scale of analysis, and that these scalingrelationships would vary among habitat elements. We investigated two elementsofspatial scale: grain and local extent. Grain was defined as the minimum mappingunit; local extent was defined by the size of an analysis window placed aroundeach focal point. For each habitat element, we quantified habitat use over arange of spatial scales. Eagles displayed scale-dependent patterns of habitatuse in relation to all habitat features, including multi-scale andthreshold-like patterns. This information supports the existence ofscale-dependant relationships in wildlife habitat use and allowed for a moreaccurate and biologically relevant evaluation of Hudson River breeding eagle habitat.This revised version was published online in May 2005 with corrections to the Cover Date.     

Gap-crossing decisions by forest birds: an empirical basis for parameterizing spatially-explicit,individual-based models

Spatially-explicit, individual-based models are increasingly used to evaluate the effects of habitat loss and fragmentation on habitat use and population persistence. Yet, they are criticized on the basis that they rely on little empirical data, especially regarding decision rules of moving individuals. Here we report the results of an experiment measuring the gap-crossing decisions of forest birds attracted to a recording of chickadee(Poecile atricapillus) mobbing calls, and provided with options to travel to the speaker by either crossing an open area (short cut) or taking a longer route under forest cover (detour). We performed the experiment in winter and late summer near Québec City, Québec, Canada. We recorded 1078 travel paths from 6 resident and 12 migratory species in 249 experimental sites. In both seasons, birds preferred to travel under forest cover rather than cross open areas, even when the forested detour conveyed a substantially longer route than the short cut in the open. Only when the detour under forest cover. This was considerably longer than the short-cut in the open, in both relative and absolute terms, were birds more likely to take short cuts, indicating that gap-crossing decisions are scale dependent. However, birds rarely ventured >25 m from forest edges despite having the opportunity to do so. Except for Hairy Woodpeckers (Picoides villosus) which ventured further into the open, all species showed similar gap-crossing decisions. Residents remained marginally closer to forest edges in late summer as compared to in winter. Conspecific group size had no influence on gap-crossing decisions. This experiment supports the hypothesis that forest bird movements are constrained in fragmented landscapes, and provides opportunities to calibrate spatially-explicit, individual-based models addressing the influence of landscape composition and configuration on dispersal. This revised version was published online in July 2006 with corrections to the Cover Date.     

Effects of structural and functional habitat gaps on breeding woodland birds: working harder for less

The effects of habitat gaps on breeding success and parental daily energy expenditure (DEE) were investigated in great tits (Parus major) and blue tits (Cyanistes caeruleus) in urban parkland (Cardiff, UK) compared with birds in deciduous woodland (eastern England, UK). Tree canopy height, the percentage of gap in the canopy and the percentage of oak (in the wood only) within a 30 m radius of nest boxes were obtained from airborne remote-sensed data. Breeding success was monitored and parental DEE (great tits: both habitats; blue tits: park only) was measured using doubly labelled water in birds feeding young. In the park, mean (±SD) tree height (7.5 ± 4.7 m) was less than in the wood (10.6 ± 4.5 m), but the incidence of gaps (32.7 ± 22.6%) was greater (9.2 ± 14.7%). Great tits and blue tits both reared fewer young in the park and chick body mass was also reduced in park-reared great tits. Park great tits had a higher DEE (86.3 ± 12.3 kJ day⁻¹) than those in the wood (78.0 ± 11.7 kJ day⁻¹) and, because of smaller brood sizes, worked about 64% harder for each chick reared. Tits in the park with more than about 35% gap around their boxes had higher DEEs than the average for the habitat. In the wood, great tits with less oak around their boxes worked harder than average. Thus structural gaps, and functional gaps generated by variation in the quality of foraging habitat, increased the costs of rearing young. Electronic supplementary material The online version of this article (doi:) contains supplementary material, which is available to authorized users.     

Conservation of northern bobwhite on private lands in Georgia, USA under uncertainty about landscape-level habitat effects

Large-scale habitat enhancement programs for birds are becoming more widespread, however, most lack monitoring to resolve uncertainties and enhance program impact over time. Georgia’s Bobwhite Quail Initiative (BQI) is a competitive, proposal-based system that provides incentives to landowners to establish habitat for northern bobwhites (Colinus virginianus). Using data from monitoring conducted in the program’s first years (1999–2001), we developed alternative hierarchical models to predict bobwhite abundance in response to program habitat modifications on local and regional scales. Effects of habitat and habitat management on bobwhite population response varied among geographical scales, but high measurement variability rendered the specific nature of these scaled effects equivocal. Under some models, BQI had positive impact at both local farm scales (1, 9 km²), particularly when practice acres were clustered, whereas other credible models indicated that bird response did not depend on spatial arrangement of practices. Thus, uncertainty about landscape-level effects of management presents a challenge to program managers who must decide which proposals to accept. We demonstrate that optimal selection decisions can be made despite this uncertainty and that uncertainty can be reduced over time, with consequent improvement in management efficacy. However, such an adaptive approach to BQI program implementation would require the reestablishment of monitoring of bobwhite abundance, an effort for which funding was discontinued in 2002. For landscape-level conservation programs generally, our approach demonstrates the value in assessing multiple scales of impact of habitat modification programs, and it reveals the utility of addressing management uncertainty through multiple decision models and system monitoring.     

Area utilization of gulls in a coastal farmland landscape: habitat mosaic supports niche segregation of opportunistic species

The intensively farmed coastal lowland landscape of Germany, adjacent to the North Sea, provides important foraging opportunities for Black-headed, Common, Herring and Lesser Black-backed gull (Larus ridibundus, L. canus, L. argentatus and L. fuscus). We expected that spatial and temporal utilization of the landscape mosaic as well as behavioural traits and utilization of food resources would differ between these closely related species, facilitating niche segregation. We recorded habitat types and their utilization by the four species over a whole year. Furthermore, we related species abundance to several abiotic parameters. Black-headed and Common gulls were the most numerous species in the study area throughout the year. In general, the former species preferred bare fields with recently prepared soils and was often associated with tractors in the fields, whereas the latter species was most often found on pastures. Black-headed gulls seem to have a higher ability to exploit ephemeral, food sources associated with human activities whereas common gulls prefer habitats with low human activity and with naturally distributed prey. The most prominent abiotic parameter influencing gull abundance was presence of tractors. Black-headed gulls have most likely benefited from recent changes in agricultural practice, particularly the increase in cropped land, while Common gulls may have suffered from a decline in pastures. At present, utilization of the farmland habitat mosaic leads to niche segregation and supports coexistence, as two of the four gull species mainly forage in the marine environment, while there is significant habitat partitioning between the other two temporally, spatially and behaviourally.     

Are hedgerows effective corridors between fragments of woodland habitat? An evidence-based approach

Anthropogenic modification of the countryside has resulted in much of the landscape consisting of fragments of once continuous habitat. Increasing habitat connectivity at the landscape-scale has a vital role to play in the conservation of species restricted to such remnant patches, especially as species may attempt to track zones of habitat that satisfy their niche requirements as the climate changes. Conservation policies and management strategies frequently advocate corridor creation as one approach to restore connectivity and to facilitate species movements through the landscape. Here we examine the utility of hedgerows as corridors between woodland habitat patches using rigorous systematic review methodology. Systematic searching yielded 26 studies which satisfied the review inclusion criteria. The empirical evidence currently available is insufficient to evaluate the effectiveness of hedgerow corridors as a conservation tool to promote the population viability of woodland fauna. However, the studies did provide anecdotal evidence of positive local population effects and indicated that some species use hedgerows as movement conduits. More replicated and controlled field investigations or long-term monitoring are required in order to allow practitioners and policy makers to make better informed decisions about hedgerow corridor creation and preservation. The benefits of such corridors in regard to increasing habitat connectivity remain equivocal, and the role of corridors in mitigating the effects of climate change at the landscape-scale is even less well understood.     

Interpreting habitat distribution models of an elusive species,the marbled murrelets: a response to Burger and Page

Burger and Page (this volume) evaluated our models of habitat preferences and breeding success of a threatened seabird, the marbled murrelet (Brachyramphus marmoratus), based on the largest available set of confirmed nest-sites found in coastal old-growth forest of the Pacific North-West. We believe our study documented novel and unexpected patterns of landscape-level distribution of marbled murrelets in both heavily logged and relatively intact old-growth landscapes and provided insights into how these patterns influence their reproduction, and, eventually, management. Considering the importance of the issue and to ensure appropriate and responsible use of the information we welcome discussion, detailed scrutiny and evaluation of our original results. Burger and Page claim to have identified flaws with model interpretation, data quality, statistical approaches, presentation and interpretation of our results that would invalidate our conclusions. We respond that most of their critique is irrelevant and/or misdirected with respect to our study and the interpretation of GIS data models, and that valid aspects of their claims do not critically affect our conclusions.     

Estimating the consequences of habitat fragmentation on extinction risk in dynamic landscapes

Analyzing the population dynamic consequences of spatio-temporal changes in landscape structure is a formidable challenge for spatial ecology. One key population dynamic process in fragmented landscapes is the influence of isolation on colonization rate and thereby on the occurrence of species in habitat fragments, but it is not obvious how isolation should be measured in landscapes that are affected by on-going habitat loss and fragmentation. We suggest the following procedure for the measurement of spatio-temporal isolation. First, a historical record of habitat loss and fragmentation in the landscape is prepared based on snapshots of the extent of the suitable habitat for the focal species. Second, a metapopulation model is used to simulate the occurrence of the species in this landscape, assuming the empirically observed landscape change. The model-predicted pattern of habitat occupancy at a particular point in time (usually the present time) is then compared with empirical observations on the occurrence of the species. We describe a metapopulation model that has been constructed for this purpose, and we apply it to a changing landscape of boreal forests in eastern Finland. We give an example on the occurrence of four threatened polyporous fungi in 18 small fragments of old-growth forest. In none of the species does the current isolation of the fragments nor the time since their isolation explain the occurrence of the species in the study fragments, but in three species the model-predicted occupancy probability had a significant effect on the observed abundance of the species. The model-predicted occupancy probabilities were also calculated by ignoring past landscape changes, that is, by assuming that the landscape had remained in the present configuration for a long time. These probabilities had a significant effect on the abundance of only one of the four species, suggesting that the occurrence of the species tracks landscapes changes with a noticable time lag.This revised version was published online in May 2005 with corrections to the Cover Date.     

Effects of habitat loss on populations of white-footed mice: testing matrix model predictions with landscape-scale perturbation experiments

Habitat loss is the leading cause of decline in wildlife diversity and abundance throughout the world, but its effects on wildlife are not always predictable. Matrix population modeling is an increasingly common tool used to predict the effects of habitat loss. In spite of the growing number of studies using this approach, and its wide use in conservation practice, the predictions generated by matrix population models are rarely explicitly tested in the field. We compared the ability of a suite of spatially explicit demographic matrix models to predict the response of white-footed mice to loss of high quality habitat at mosaic sites in northeast Connecticut, USA. We tested short-term model predictions with landscape scale habitat perturbation experiments, including clear-cut logging or prescribed burning of high quality habitat at two study sites. Comparison of each model’s predictions with the observed responses at both sites qualitatively supported predictions that perturbation of high quality habitat would have negative effects extending into the surrounding landscape. The best-supported model assumed that evicted residents of the perturbed habitat would successfully resettle in nearby intact habitats, and allowed for gradual population recovery in the perturbed habitat. Similarly, long-term simulations (20 years) revealed how loss of a single habitat could trigger population declines throughout a mosaic site. This study shows that careful consideration of model assumptions such as those pertaining to resettlement behavior is crucial if predictions are to be reliable, and highlights the role of experiments for comparing alternative model predictions.