Introduction

Climate change is causing broad poleward and uphill shifts in species distributions (Chen et al. 2011; Lenoir et al. 2020; Parmesan et al. 2022). However, ranges appear not to be shifting as fast as climate change, especially over latitudinal gradients on land (Chen et al. 2011; Lenoir et al. 2020). Given evidence that range expansions are delayed by a lack of suitable habitat (Platts et al. 2019; Hodgson et al. 2022), initiatives to adapt conservation to climate change often prioritise strategies to facilitate poleward and uphill movements (Heller and Zavaleta 2009; Littlefield et al. 2019; LEDee et al. 2021). However, despite evidence of localised extirpations (Franco et al. 2006) and upward range retractions in mountainous areas (Wilson et al. 2005) the understanding of large-scale latitudinal range retractions remains sparse compared with poleward range expansions, even though range retractions are of greater conservation urgency and concern. Moreover, a focus on broad latitudinal range shifts (Taheri et al. 2021), and on shepherding species to higher latitudes, risks overlooking the interacting drivers of species distributions and the importance of in situ strategies to enhance the persistence of existing populations (Greenwood et al. 2016). An understanding of where and how species ranges contract is needed to identify and protect rear-edge populations against the impacts of climate change (Hampe and Petit 2005).

We contend that trailing edge shifts in response to climate change may rarely correspond to marked latitude or elevation gradients, because variation in topography, geology and vegetation influence local climate and rates of climatic change (Vanderwal et al. 2013; Maclean et al. 2017; Suggitt et al. 2018; Oldfather et al. 2020). This climatic variation also occurs over landscapes containing spatial variation in habitat quality and degree of fragmentation, which together result in patchy patterns of climate-driven population decline and metapopulation contraction that could be masked in coarse-scale distribution assessments (Thomas et al. 2008; Rumpf et al. 2019). Furthermore, spatial variability in climate and habitat availability interact with temporal climate variability (Ummenhofer and Meehl 2017) to influence range dynamics via regionally synchronised population and metapopulation declines (Kahilainen et al. 2018). Local declines in response to climate change in turn represent direct responses to thermal or hydrological extremes, or indirect effects of climate through changing abundance or phenology of either resources (Marcelino et al. 2020; Salgado et al. 2020) or antagonistic biotic interactions (Hargreaves et al. 2014; Vilà-Cabrera and Jump 2019). We expect the combined effects of spatio-temporal variation in climate and habitat to cause patchy demographic and evolutionary responses to climate change in processes such as local extinction, dispersal, gene flow and colonisation that are critical to regional persistence or expansion. Understanding these processes, and how they cause patchy retractions across species ranges, is needed to inform in situ strategies to prevent the loss of populations whose local adaptations are potentially important for the future adaptive capacity of species to climate change (Melero et al. 2022; Parmesan and Singer 2022).

Here, we consider how patchy retractions can form throughout terrestrial species’ geographic ranges, and the implications for biodiversity assessment and conservation. We first describe how climate is changing spatio-temporally, with geology, topography and vegetation interacting with macroclimatic changes to result in fine-resolution spatial heterogeneity in rates of warming. We next outline species’ demographic and adaptive responses to climate change, and how these interact with habitat quality to result in patchy range retractions. We then present evidence of cases where climate change and environmental variability over space and time have led to the hollowing out of species ranges. In the second part of the review, we show how understanding these processes can inform approaches to terrestrial species conservation: identifying species or populations at risk from population collapse; pinpointing areas in the landscape more resistant to climate change; and measures to increase the resilience of species and communities across wide geographic ranges by protecting key landscapes or parts of species ranges that will gain the most time to protect species in situ before they colonise new regions.

Climate and environmental variability over the geographic range

Environmental heterogeneity, as a result of terrain (e.g. elevation, slope, aspect), geology (e.g. soil type) and biotic features (e.g. land use and vegetation type) influences the climatic conditions experienced by organisms (Dobrowski 2011; Potter et al. 2013). This heterogeneity has two important effects on ecological responses to climate change. On the one hand, it provides localised cool microclimates in which species might persist in the face of climate warming (Suggitt et al. 2018) or move to if necessary (on the assumption that species are not already restricted to the coolest microclimates, and thus able to find cooler conditions locally as the climate warms). Secondly, heterogeneity results in the decoupling of localised microclimatic changes from macroclimatic changes, which can amplify or buffer exposure to climate change (Ashcroft et al. 2008, 2009; Suggitt et al. 2011). Where climate warming goes hand in hand with changes in weather conditions, terrain and vegetation influence rates of localised warming profoundly both in terms of average microclimates experienced and microclimatic dynamics (Maclean et al. 2017; De Frenne et al. 2019). In temperate forests, greater canopy cover buffers the understorey against temperature extremes, and microclimate dynamics over time resulting from changes in tree cover (rather than broader macroclimate changes) determine the rates at which understorey plant communities have changed (Zellweger et al. 2020). Hydrological processes related to local relief, geology and weather conditions can also generate hydrological refugia which are vital for the persistence of water-limited plants (McLaughlin et al. 2017), that may therefore provide wooded or wetland habitats and cool, moist microclimates and resources for a wider range of plants and animals.

Spatial variation in climate and rates of climate change mean that equatorward margins are not necessarily the first places in species ranges to exceed climatic limits (Oldfather et al. 2020). Hotspots of decline and range retractions can occur in habitats where rates of climate change are amplified (Crossley et al. 2021). In contrast, species might persist in geographical locations where rates of warming are slower: areas receiving less solar radiation such as below the canopy or on poleward facing slopes under increasingly sunny conditions (Suggitt et al. 2011; Zellweger et al. 2020), or places exposed historically to greater solar radiation if conditions become cloudier (Maclean et al. 2017). Therefore refugia are not necessarily static through time: the places experiencing slower rates of warming historically may not do so indefinitely, owing to changes in prevailing weather conditions (Maclean 2020).

Spatially variable rates of change, whether spatially static in the long-term or not, occur over landscapes differing in habitat quality and the extent of habitat fragmentation. Species’ demographic and evolutionary responses (fitness, survival, reproduction and population persistence) to climate change interact with landscape habitat fragmentation and quality (Polechová 2018; Sherpa et al. 2022; Johansson et al. 2022). For example, exposure to climate change can exacerbate contractions from low quality or fragmented habitat networks (Fritts et al. 2018; Johansson et al. 2020; Billman et al. 2021). Throughout their ranges, species may be affected by synergistic (or antagonistic) effects of habitat fragmentation and climate change (Angert et al. 2018; Latimer and Zuckerberg 2021) and threshold requirements of habitat availability for survival may be greater where exposure to climate change is greater (Oliver et al. 2015). Similarly, the exposure to climate change needed to cause regional extirpation could be reduced in regions where habitat is more fragmented (Northrup et al. 2019).

The abundant-centre hypothesis, that population density is highest at the core of species’ geographic ranges responding to gradients of environmental (including climatic) variation (Brown 1984), has only equivocal support in many species (Sagarin and Gaines 2002). Textures of abundance are known to occur throughout species ranges (Lawton 1993), and ecological marginality (because of the presence of competitors or natural enemies, or the absence of resources) can cause abrupt changes in occupancy over continuous environmental gradients across the range (Hargreaves et al. 2014). Ecological marginality can therefore impose climatic limitations on species occupancy near the range core (Parmesan and Singer 2022), in spite of apparently favourable prevailing climates. In fact, populations of some declining species appear more resistant to extinction at the limits than at the core of their ranges both because geographic marginality does not correspond to ecological marginality (Vilà-Cabrera and Jump 2019) or because of the spread of additional human-driven extinction forces from within the range (Channell and Lomolino 2000).

Species adaptive responses

Climate, in particular temperature, directly affects an organism’s fitness through physiological function (e.g. movement, reproduction and development), and therefore rates of birth, death and population growth. The effects of temperature on fitness can be described by thermal performance curves, whereby fitness is relatively constant over a range of temperatures, but diminishes rapidly as temperatures approach upper or lower critical limits (Angilletta et al. 2002) (Fig. 1a). This non-linearity in fitness means that extreme temperatures can cause sudden and abrupt population fluctuations over short time-scales when organisms are exposed to conditions near or outside critical limits (Vasseur et al. 2014). Conservation initiatives based on an assumption of gradual attrition of biodiversity may overlook these stochastic population responses to climate variation (Kahilainen et al. 2018; Marcelino et al. 2020; Trisos et al. 2020). However, the extent to which such responses are abrupt or occur simultaneously across wide areas depends largely on spatiotemporal variability in the critical limits and how intrinsic and extrinsic factors affect them.

Adaptive responses to climate change depend on the ability of species to persist in place or shift in space (Thurman et al. 2020) shaped by intrinsic factors such as local adaptations and dispersal capacity, and extrinsic factors such as microclimatic variation and habitat quality. The availability of buffering microclimates can facilitate thermoregulation as organisms track suitable climatic conditions within their thermal limits by shifting to suitable microclimates (Ashton et al. 2009; Kearney et al. 2009) (Fig. 1b). The grassland bird Tetrax tetrax makes greater use of cool microclimatic refugia during hotter years and in hotter landscapes, but can also undertake short-distance migrations to cooler landscapes before returning to landscapes lacking cool microrefugia when conditions improve (Ramos et al. 2023). Microclimates are highly dynamic in many terrestrial habitats (especially those modified by humans), and a key requirement for persistence is the availability of favourable microclimates within the movement range of the species concerned. Short-distance movements by montane forest birds track climatic variation within breeding seasons (Frey et al. 2016), whilst migratory birds can alter migration strategy, distance, location of breeding sites, or the timing of arrival or departure to avoid climate extremes (Acker et al. 2021). In North American landbirds, the range cores of permanent residents have more fragmented distributions than neotropical migrants, suggesting that the greater mobility of migrants can help overcome patchy habitat distributions (Linder et al. 2000). However, unlike permanent residents, neotropical migrants have shown poleward contractions in their low-latitude range margins: permanent residents may have greater opportunities for local adaptation or small-scale redistribution in response to changing climatic conditions and resource phenology, whereas breeding site selection by migrants may need to respond more directly to broad geographic gradients in climatic conditions and the phenology of prey species (Rushing et al. 2020).

Behavioural thermoregulation and other strategies such as dormancy and aestivation enable persistence in some landscapes even when regional climatic conditions deteriorate, but can put constraints on time for foraging and breeding, with implications for fitness, population growth and recovery (e.g. see Tuberville et al. 2015 for reptiles and amphibians; Nicolai and Ansart 2017 for gastropod molluscs). In large and high-quality habitat patches there are likely to be more readily available resources (e.g. food, breeding habitat), and time constraints imposed by behavioural thermoregulation are less likely to lead to reduced fitness or lower reproductive success. In smaller, fragmented habitat patches not only are resources and microclimatic heterogeneity lower but organisms are more likely to be exposed to extreme conditions through edge effects (Tuff et al. 2016; Latimer and Zuckerberg 2017): the effects of human-modification and habitat fragmentation on microclimate dynamics are important because they determine the frequency and intensity of exposure to extremes of temperature and hydrology (Latimer and Zuckerberg 2017). Across species ranges, populations are therefore more likely to contract from fragmented, low-quality habitats, irrespective of geographic location.

Fig. 1
figure 1

Species response to an extreme climatic event (in this example, a heatwave). The top two panels (a) show individual responses under both mean temperature conditions (top left) and extreme conditions (top right), illustrated through the thermal performance curve (CTmin = minimum critical temperature thresholds, CTmax = maximum critical temperature thresholds; Opt = optimum thermal conditions). The bottom two panels (b) show the population response under mean temperatures (bottom left) and extreme conditions (bottom right) as populations track suitable climatic conditions

Adaptive capacity can vary over space even for the same species due to local adaptations (Anstett et al. 2021), exacerbating patchy occupancy patterns. Populations previously exposed to a wider range of climatic conditions are more likely to harbour a broader range of local climatic adaptations (Bay et al. 2018; Graae et al. 2018; Hargreaves and Eckert 2019; Bontrager et al. 2021; Melero et al. 2022) potentially enabling them to withstand climatic variability. This includes populations persisting in locations with a history of climate variability, such as towards the rear climatic edge (Hampe and Petit 2005) or those in environmentally heterogeneous landscapes where the flow of individuals between climatically dissimilar locations is promoted (Graae et al. 2018). Trophic adaptations of animals across their geographic ranges depend on the local abundance, phenology and habitat of potential resource species under prevailing climatic conditions, thus influencing distributional responses to climate change (Braschler and Hill 2007). In the butterfly Euphydryas editha, geographic variation in larval host plant adaptation causes patchy climate-driven occupancy patterns near the core of the range (Parmesan and Singer 2022).

Ecological traps can occur either when local climatic adaptations come at the expense of dispersive traits (Hargreaves and Eckert 2019), reducing the flow of adaptations across a species’ range (Graae et al. 2018; Hargreaves and Eckert 2019; Sherpa et al. 2022), or in response to local biotic or abiotic factors resulting in maladaptive life-history strategies (Turner and Maclean 2022; Parmesan and Singer 2022). Ecological traps exacerbate the long-term threat of climate change to isolated populations adapted to a narrow range of climatic conditions (Hargreaves and Eckert 2019; Billman et al. 2021; Parmesan and Singer 2022), and need to be considered when implementing conservation recovery programmes. For example, nest site provisioning for the lesser kestrel Falco naumanii increases juvenile mortality in hot breeding seasons if birds are attracted to nest sites that are vulnerable to overheating due to their compass orientation, location or construction (Catry et al. 2011).

Adaptive capacity is further reduced where habitat fragmentation impedes gene flow across landscapes, leaving isolated populations more prone to genetic drift and inbreeding depression (Polechová 2018; Sherpa et al. 2022). However, gene flow from some parts of the range could be maladaptive depending on the prevailing climatic conditions in emigrant and immigrant populations (Van Dyck et al. 2015; Angert et al. 2020); thus it could be counterproductive to encourage range shifts from some populations. As local adaptations vary across species ranges, populations in the geographic range centre may be less well adapted to climate change than those at the range periphery, thus range centre populations could be exposed earlier than range margin populations to extremes (Hampe and Petit 2005) causing contractions in the interior rather than at latitudinal edges.

Metapopulation responses to spatiotemporal climate variability

Metapopulation dynamics provide a useful framework for species range responses to heterogeneity in climate warming, based on an understanding of the factors driving colonisation (when conditions improve) and local extinction (when conditions deteriorate). Range dynamics depend on changes in local climate that tip the balance from equilibria to non-equilibria between local extinction and colonisation. Improving climatic conditions in a landscape trigger net colonisation and range expansion (Bennie et al. 2013), whilst deteriorating climates cause net local extinction and range contraction (Kahilainen et al. 2018; van Bergen et al. 2020). Populations typically contract to large, high-quality habitat patches (Thomas et al. 2008; Fourcade et al. 2021; Johansson et al. 2022) or those containing favourable microclimates (Oliver et al. 2010; Suggitt et al. 2018) (Fig. 2). This results in extirpation from small, isolated and low-quality habitats (Piessens et al. 2009; Johansson et al. 2022) or those with microclimates where climatic changes are amplified (van Bergen et al. 2020; Crossley et al. 2021). This illustrates two points: first, decline hotspots and range contractions can occur in any parts of species ranges that are closer to tolerance limits in climate or habitat (Oldfather et al. 2020) or exposed to faster rates of climatic change (Suggitt et al. 2011; Crossley et al. 2021). Second, whether larger scale retractions or expansions occur depends largely on the overall prevalence of suitable conditions. On one hand, a deteriorating climate can cause an increase in the area or quality of habitat needed to sustain viable regional populations. On the other, an increase in the area or quality of habitat will buffer species against the adverse effects of climate.

Metapopulation dynamics also help understand responses to temporal climatic variability, which may interact with spatial dynamics. Stochastic extremes in climate and events such as droughts and heatwaves are expected to occur at an increased intensity and frequency (IPCC 2012; Perkins-Kirkpatrick and Lewis 2020). Extreme events suddenly expose organisms to conditions exceeding critical limits or impacting organisms indirectly e.g. through effects on habitat condition or food resources. When populations across a network of sites are exposed to extremes simultaneously, population declines, and synchronised local extinctions occur (Fig. 2). These synchronous declines can cause the extirpation of entire metapopulations (Kahilainen et al. 2018; Johansson et al. 2022; Parmesan and Singer 2022). Extinction risk is highest in highly fragmented habitat networks comprising small or low-quality habitat patches (Piessens et al. 2009; Johansson et al. 2020) whereas persistence is promoted in microrefugia or large areas of high-quality habitat. Moreover, recovery from intermittent climate-driven declines is dependent on the extent of functional and direct habitat connectivity, with populations able to recolonise habitat patches more quickly after extreme events in networks with higher connectivity, or where more local populations are able to withstand exposure to a climatic extreme (Piessens et al. 2009; Oliver et al. 2013; Latimer and Zuckerberg 2017; Johansson et al. 2020, 2022). Conversely, recovery is unlikely in isolated and small metapopulations (Piessens et al. 2009; Oliver et al. 2013) especially where there is no source of immigration to facilitate recovery. Temporal variability can therefore exacerbate declines from parts of species ranges, resulting in the clustered disappearance of populations from some regions or landscapes, and survival only in core landscapes with large areas of high-quality habitat or habitats which have not been exposed to extremes.

Fig. 2
figure 2

Metapopulation response to an extreme climatic event following individual and population response (Fig. 1). Showing patch occupancy in a habitat network comprising microclimatic variation through habitat type before an extreme event (left column) and following an extreme event (right column). The top panels (A) show a resilient network with a large habitat area, microclimatic variation, and aggregated habitat patches, bottom panels for a fragmented network (B)

If the frequency and intensity of extreme events increases, threshold habitat requirements for landscape-scale persistence may also increase, because recovery from climate-driven extirpations is not instantaneous (Oliver et al. 2013, 2015) and faster in large and well-connected networks (Oliver et al. 2013; Johansson et al. 2022). Hence species may become limited to networks where populations can recover in time for the next perturbation (Oliver et al. 2015) or where there is sufficient heterogeneity for persistence in microrefugia (Suggitt et al. 2018). Understanding how environmental heterogeneity helps moderate the interacting effects of spatiotemporal climate variability and habitat fragmentation on metapopulations is thus vital to identify locations at risk from range contraction in response to extreme events.

Evidence for patchy range retractions

If climate change resulted in deteriorating conditions along smooth latitude or elevation gradients, we would expect uniform upward or poleward retractions rather than localised extirpations. However, there is mounting evidence for patchy range retractions (Table 1), particularly for species where long-term metapopulation dynamics have been well studied (e.g. some butterfly species and the American Pika Ochotona princeps). These retractions can be overlooked in assessments of species range shifts (Thomas and Abery 1995; Pöyry et al. 2018) as species might persist in diminishing numbers of patches under an extinction lag, with contractions remaining undetected by coarse-scale assessments, since all populations within a grid square must go extinct before it is considered unoccupied (Thomas et al. 2006).

An example of a species experiencing patchy range retractions is the American Pika. The drivers of occupancy for American Pika including extirpation of metapopulations and retractions from lower elevations are the focus of ongoing debate (Stewart et al. 2017; White and Smith 2018; Smith 2020; Billman et al. 2021). Local extinctions often occur from small, isolated and low-quality habitat patches (Stewart et al. 2017; White and Smith 2018; Smith 2020; Billman et al. 2021), whilst exposure to extreme temperatures causes synchronous short-term fluctuations in abundance, particularly at lower elevations (Billman et al. 2021), though populations persist where there are thermoregulation opportunities (Rodhouse et al. 2018). In parts of the range reduced functional connectivity following abundance decline and warming of low-elevation dispersal corridors impedes rescue from core metapopulations, affecting metapopulation recovery and gene flow (Smith and Nagy 2015; Stewart et al. 2017; Billman et al. 2021). The American Pika represents an example of how extremes in climate occurring in topographically heterogenous landscapes interact with habitat availability affecting occupancy, dispersal and range shifts in the centre of a species’ geographic range, causing retractions over multiple elevation gradients (Billman et al. 2021).

Table 1 Patchy range retractions: distribution responses to climate variability in landscapes comprising spatial variation in habitat fragmentation and local microclimate

Conserving terrestrial species undergoing patchy range retractions

Initiatives to adapt conservation to climate change often include site protection and strategies to increase connectivity and facilitate movements poleward and uphill (Heller and Zavaleta 2009; Littlefield et al. 2019; LEDee et al. 2021). Though potentially effective for facilitating expansion at the leading edge (Thomas et al. 2012) they are not specifically tailored to conserve trailing edge populations and may be ineffective at doing so (Araújo et al. 2011). Based on evidence that species experience patchy range retractions, we outline a framework: first, to set conservation priorities; then strategies for management at trailing edges to dampen population variability associated with climate change and promote metapopulation persistence. The strategies aim to slow the hollowing out of species ranges, and prevent loss of potentially important local adaptations for future adaptive capacity to climate change (Hargreaves and Eckert 2019; Bontrager et al. 2021; Melero et al. 2022).

Targeting conservation management and protection

Identifying at-risk species and populations

Climate Change Vulnerability Assessments (CCVAs) use species traits and current or predicted distributions to assess risk of extinction from climate change, but the incorporation of effects of spatio-temporal climate variability into CCVAs is limited (Wheatley et al. 2017; Foden et al. 2019; Rocha-Ortega et al. 2020). For herpetofauna using ephemeral wetlands, for example, climate change vulnerability assessment may underestimate vulnerability to changing frequency and severity of climatic extremes, especially for species that lack adaptations, such as aestivation, underground retreats, or the ability to shift diel activity patterns, that enable them to withstand these threats (Tuberville et al. 2015).

In Table 2, we suggest improvements for three vulnerability assessment methods in the context of patchy range retractions: (i) distribution assessments; (ii) trait-based assessments; (iii) Species Distribution Modelling (SDM). We also propose methods to detect at-risk species and populations for protection and targeted conservation interventions (Table 2). Though limited data on fine-resolution changes to species distributions are available for most species in most parts of the world, making use of this available information based on an understanding of how environmental variation influences climate change and hence species distributions could help to inform future assessments of threat.

Table 2 Climate change vulnerability assessment (CCVA) approaches: limitations and suggestions for improvements to identify species and populations vulnerable to patchy range retractions

Identifying microrefugial landscapes for protection and management

Microrefugial landscapes where populations are likely to be more resilient to climate change have high microclimatic heterogeneity (so species can withstand extreme events by local survival or short-distance redistribution) and are landscapes where microclimatic or habitat variability promote asynchronous population dynamics. A range of modelling approaches can be utilised to help identify such landscapes (Table 3), including metapopulation models accounting for effects of habitat fragmentation on extinction-colonisation dynamics in a changing climate. Outputs from metapopulation models can estimate habitat thresholds at which metapopulations shift from equilibria to expected dynamic extinction (with time lags), helping to identify and manage landscapes close to extinction thresholds. Specific modelling approaches include Stochastic Patch Occupancy Modelling (SPOMs) (Bennie et al. 2013; Johansson et al. 2020, 2022) or other population simulation models (White and Smith 2018) (Table 3). Incorporation of habitat quality or local climate suitability into SPOMs can improve predictions of range expansions and contractions in the context of spatial climatic and habitat variability (Bennie et al. 2013; Johansson et al. 2020) identifying core habitat networks where persistence and metapopulation recovery remain viable as habitat or climatic suitability deteriorate. Modelling population dynamics can also identify core patches that species contract to and that promote regional persistence in years of metapopulation contraction. Improved understanding of the location of core networks and minimum habitat thresholds for persistence under different climate scenarios can help facilitate the protection of key networks, core sites and plan conservation to maintain stable dynamics and increase speed of recovery following extremes.

Environmentally heterogeneous or microrefugial landscapes including stable or unique climatic conditions can encourage regional persistence as macroclimatic conditions decline (Ashcroft et al. 2012). Identifying microrefugial landscapes (Table 3) can be based on; (a) areas supporting cooler microclimates (e.g. valley bottoms where cold air pooling occurs, or sub-canopy temperatures in forests) (De Frenne et al. 2019; Scherer et al. 2021), where species can persist as macroclimate warms; and (b) areas where microclimatic conditions are decoupled from macroclimate resulting in slower rates of change and more stable climates (Lenoir et al. 2017), though these areas are not static and can change dependent on prevailing weather (Maclean 2020). The spatial resolution of microclimate modelling needed to identify microrefugia and assess landscape resilience may range from 100 m for microclimate modelling based on topographic and habitat heterogeneity (Suggitt et al. 2018; Maclean and Early 2023) to < 30 m (Potter et al. 2013) for within-site climate variation, or < 1 m if modelling sub canopy temperatures (Lenoir et al. 2017). The availability of microclimate models, high resolution remote sensing data and climate forcing data means that modelling spatio-temporal climate variation is becoming more accessible (Maclean et al. 2019; Kearney et al. 2020; Maclean 2020). Fine-scale microclimate data has applications in SPOMs (Bennie et al. 2013) or microclimate SDMs, and for modelling future climate and weather patterns to show how microrefugia might change (Maclean 2020) with the most resilient landscapes providing some continuity of microrefugia over space and time (Table 3).

Modelling microclimate, metapopulation dynamics or species distributions can help identify microrefugial landscapes, or strategies that prolong persistence. These landscapes are likely to include heterogeneous microclimates, unique climates, or large and aggregated habitat networks for management and protection. They can be identified or monitored based on features such as stable species trends or high probability of persistence under warming; or areas where the microclimate is decoupled from macroclimate (Table 3).

Table 3 Features for identification of microrefugial landscapes in the context of climate change

Conservation management in patchy regions of decline

Practical conservation of species experiencing patchy range retractions aims to promote stable metapopulation dynamics and prolong range-wide persistence, through enhancing the resilience of landscapes to climate change and variability. We propose five strategies for targeting depending on the conservation needs of the species, extent of habitat and the thermal and hydrological conditions in the landscape (Table 4). These complementary strategies aim to slow climate-driven population declines, promote stable metapopulation dynamics and population recovery.

Table 4 Summary of conservation management approaches for species experiencing patchy range retractions

Maximise microclimatic heterogeneity at a range of spatial scales

Many species can buffer the effects of climate change by shifting to warmer or cooler microclimates to regulate temperatures (Kearney et al. 2009). These small shifts to thermoregulate include between short and tall vegetation (Kindvall 1995; Fritts et al. 2018), scrub and bare ground (Ashton et al. 2009; Rytteri et al. 2021), shaded and unshaded leaves in a tree canopy (Pincebourde and Suppo 2016), or patches of shrubs and trees in grasslands (Ramos et al. 2023). Thermoregulation is constrained by the availability of shade and cooler microclimates and therefore management to maximise microclimatic variation can help reduce time constraints and risks associated with thermoregulation, and thus help promote persistence (Suggitt et al. 2018). Vegetation management within human-used habitats has important effects on microclimate dynamics and, as a result, can be used to modify the responses of ecological communities to climate change (Zellweger et al. 2020).

Maximising microclimatic heterogeneity can include in-situ conservation management within a habitat patch or targeted management at the landscape level. Specific management depends on a species’ requirements, local habitat or climatic conditions and understanding how organisms might shift in response to climatic extremes. Management can be informed by field-based observations across thermal or humidity gradients within a species range or by modelling species’ fine-scale climatic associations to identify management strategies which maximise thermoregulation opportunities (e.g. modelling how changing vegetation structure might help to prolong availability of suitable conditions). In situ interventions include promoting patchy scrub or secondary woodland growth to increase canopy shading (Ashton et al. 2009; González-del-Pliego et al. 2020), creating microtopographic variation (e.g. banks and ditches) to create warmer as well as cooler spots (Greenwood et al. 2016), or manipulating vegetation height through changing grazing intensity (Kindvall 1995; Rytteri et al. 2021). At the landscape-level managing habitat heterogeneity across areas of topographic diversity can help maximise the range of climatic conditions e.g. along elevational gradients, valley bottoms or different aspects (Oliver et al. 2010; Lenoir et al. 2017).

This approach requires a balance between provision of thermoregulation opportunities and maintaining sufficient high-quality habitat for foraging, breeding and ongoing persistence (Kearney et al. 2009; Graae et al. 2018). Though for many species the priority is to increase the availability of suitable high-quality habitats, for those species persisting near the upper edges of their critical thermal limits there will be limited scope to increase habitat area, especially in locations of rapid climate change. In such circumstances, maximising microclimatic heterogeneity to increase thermoregulatory opportunities is particularly important to reduce the risk of population decline and synchronous extinctions across whole networks owing to sudden exposure to temperatures exceeding critical limits (Oliver et al. 2010).

Minimising temporal variability in microclimate

As climate changes some regions might experience more variable and unpredictable precipitation patterns, including more frequent and intense periods of drought (IPCC 2012), and promoting hydrological stability can reduce the impacts of precipitation extremes on species. Precipitation extremes can affect species fitness, habitat condition and adaptive responses (Weiss et al. 1988; McLaughlin et al. 2002a; Fritts et al. 2018; Johansson et al. 2020; Anstett et al. 2021). Contractions can occur from small, drought prone sites (Oliver et al. 2013; van Bergen et al. 2020) or from cooler sites in years of heavy precipitation, such as in the case of Euphydryas editha bayensis (Weiss et al. 1988). Management that increases stability in water availability can help promote survival, metapopulation stability and the rescue effect following precipitation extremes (Zylstra et al. 2019; van Bergen et al. 2020; Johansson et al. 2022).

Spatial targeting of management depends on the habitat and water requirements of target species, expected changes to precipitation in the landscape and how this is affected by environmental heterogeneity. Modelling approaches that determine run-off as a function of soil infiltration capacity (land use, topography, soil type, hydrological conductivity) and evapotranspiration (vegetation cover) can predict future hydrological conditions (Douglas-Mankin et al. 2010), facilitating management targeting. Where high variability in precipitation is expected then interventions should aim to promote stability in hydrological conditions over time. In areas expected to become drier and more drought-prone then species may benefit from interventions that maintain soil moisture or that result in wetter areas where suitable habitat conditions can be maintained (Carroll et al. 2011). Where there is less certainty in predictions of future hydrological conditions then maximizing spatial heterogeneity in water availability can increase landscape resilience to variable rainfall. In human modified landscapes, many wetlands (in both open and forested areas) have been ditched, drained or canalized, and restoration of hydrological variability in such areas could restore cooler or moister places in the landscape for localized biota (Raney et al. 2014; Horsák et al. 2018; Beranek et al. 2022).

Specific practical management interventions could include the following (see also Greenwood et al. 2016). In areas expected to become drier, tussocky vegetation structure (e.g. through livestock grazing), and promoting more natural features in landscapes (scrub, woodland, taller vegetation) and along hydrological systems (e.g. trees and debris along and in rivers) can attenuate quick run-off and reduce soil erosion. Alternatively, ditch blocking and manipulation of catchment hydrology (reducing drainage, raising water levels or pond creation) can help prolong suitable conditions in drought events, particularly for wetland species (Franks et al. 2018; Zylstra et al. 2019; Mathwin et al. 2021). To promote soil water stability, mulching (where compacted) or adding organic matter to soil affects the rate and direction of water flow, improving hydraulic connectivity. Maximising heterogeneity in water availability can be achieved through mechanical interventions to create scrapes, ditches or pools or through grazing by heavy livestock. Such approaches can be complemented by habitat restoration in areas of hydrological refugia (Nimmo et al. 2016; van Bergen et al. 2020) thus promoting survival of species adversely affected by drier or more variable hydrological conditions.

Protect, restore and manage intact landscapes

Promoting aggregated networks of high-quality habitat is especially important in slowing climate-driven declines as it helps increase metapopulation capacity and speed up metapopulation recovery. As the climate changes, species contract to core habitat networks where populations persist and recover more quickly from climate extremes (Thomas et al. 2008; Oliver et al. 2015; Fourcade et al. 2021; Johansson et al. 2022). These core networks often contain bigger, high-quality habitat patches with larger and less vulnerable populations due to greater resource availability and microclimatic heterogeneity, and reduced edge effects (Tuff et al. 2016; Latimer and Zuckerberg 2017; Fourcade et al. 2021). Habitat patches within core networks are often aggregated, encouraging dispersal between populations, and promoting gene flow and the rescue effect even at low levels of occupancy and abundance following stochastic extreme events (Johansson et al. 2020, 2022). Protecting and managing intact landscapes is not only key to promote regional persistence but also to help facilitate the recovery of dependent smaller, fragmented subnetworks (Wilson et al. 2002; Johansson et al. 2022) thus slowing contractions and rates of decline at multiple locations.

Outputs from mechanistic niche-based models of population dynamics or SPOMs which incorporate local variation in habitat and climate can assist in setting goals in conservation strategies and the spatial targeting of management. Models can first be used to identify core sites and networks with higher probability of occupancy under deteriorating conditions for protection. Secondly, models can help identify minimum threshold habitat requirements for persistence under climate change and how managing habitat quality can increase habitat availability and population resilience (Bulman et al. 2007). Lastly, models can show how metapopulation dynamics and habitat thresholds are affected by changing local climatic conditions (or as habitat quality deteriorates under climate change (Johansson et al. 2020) to identify how much high-quality habitat is required to ensure metapopulation persistence and recovery under varying scenarios. This can include increasing frequency or intensity of climatic extreme events and where only the largest and most aggregated habitat networks are likely to survive (Oliver et al. 2015).

Core habitat networks comprising large, aggregated areas of native and semi-natural habitats should be prioritised for protection from future habitat loss, enabling habitat quality to be maintained or enhanced. In wider, peripheral habitat networks management can aim to restore habitat to increase the area and aggregation of patches. If resources or space for management are limited then enhancing patch aggregation or habitat quality could be effective (Oliver et al. 2015; Johansson et al. 2017). Improving habitat quality can boost population densities and functional connectivity (Hodgson et al. 2011) and therefore metapopulation resilience and recovery. Managing habitat quality is important in core and peripheral networks, but particularly where land use conflicts limit scope for habitat restoration (Piessens et al. 2009; Johansson et al. 2020, 2022; van Bergen et al. 2020).

Management of habitat variability and microclimate dynamics over space and time

Managing habitat variability over time and space to vary organisms’ exposure to the effects of climate change can promote asynchronous metapopulation dynamics, spreading extinction risk from climate extremes across populations. Synchronous population declines, when all organisms across a network are simultaneously exposed to conditions that exceed critical limits, can cause the extirpation of entire metapopulations in fragmented habitats (McLaughlin et al. 2002a; Kahilainen et al. 2018; Johansson et al. 2022) and push large and well connected sites below viability thresholds (van Bergen et al. 2020). Varying how organisms experience climate between sites and over different seasons or years through habitat management to modulate indirect and direct ecological impacts can help reduce the synchrony in population declines (Oliver et al. 2010).

During climate extremes, declines in species abundance can vary between patches depending on habitat condition and how this affects exposure to climate change. Habitat condition directly influences exposure (e.g. through microclimate as a function of vegetation height) or indirectly due to impacts on habitat quality, which can impact breeding habitat and food abundance (Piessens et al. 2009; Fritts et al. 2018). Populations in lower quality habitat patches might have a higher chance of survival under some climatic conditions (Rytteri et al. 2021), for example, sub-optimum habitat with taller vegetation might support low population densities in average years but more resilient populations in years of drought as taller vegetation shelters food plants or nests (Fritts et al. 2018; Johansson et al. 2020; van Bergen et al. 2020). Field-based observations and mechanistic niche-based models (e.g. incorporating patch habitat or thermal quality) can help understand the limits of a species niche breadth, how species densities are influenced by habitat and climate, and how manipulating habitat quality in some patches might affect metapopulation dynamics under a changing climate. The understanding gained from observations and models can help determine specific management interventions and how to target these over space and time.

Managing habitat variability requires a co-ordinated approach to ensure sufficient availability of high-quality habitat is maintained to support metapopulation dynamics (Graae et al. 2018). Delivery over space involves maintaining habitats within networks across the range of a species habitat breadth (e.g. shorter and taller vegetation alongside patches containing optimum habitat), including management on known patches, but also currently unoccupied patches, or those supporting small or transient populations. Managing variability over time also requires reactivity in response to climatic conditions, for example, grazing intensity could be reduced in years of drought, then restored as habitats and populations recover from extremes (Fritts et al. 2018; Johansson et al. 2020). This approach might be suitable when there is limited scope to manage variable habitats over space (e.g. due to limited high quality habitat). Managing variability is not an appropriate strategy for all species (Table 4) but could be particularly suitable where topographic heterogeneity is lacking and there are limited options to maximise heterogeneity or restore habitat on microrefugia.

Management across a range of microrefugia

Targeted habitat management across a range of microrefugia in environmentally heterogenous landscapes can help provide continuity in climatic conditions over space and time, to slow rates of decline as regional macroclimate deteriorates (Suggitt et al. 2018) or even reverse expected distribution shifts (Tellería 2020). This targeted management can help facilitate persistence at trailing edges, firstly, as species can adapt gradually to deteriorating climates thus facilitating climate adaptation (Graae et al. 2018; Hargreaves and Eckert 2019). Secondly, regional metapopulation persistence is promoted in microrefugia (Oliver et al. 2010; Massimino et al. 2020), as rates of climatic change are slower or more stable than regional macroclimate, thus preserving important local adaptations and gaining time to develop conservation strategies.

An understanding of what constitutes microrefugia and how they are expected to shift in space and time is required to target management interventions that prolong suitable conditions or allow species to track suitable conditions. Modelling species’ fine-scale climatic associations (e.g. using microclimate SDMs), can show the distribution of thermally suitable habitats in a landscape, how long microrefugia might provide suitable conditions (González-del-Pliego et al. 2020) and how locations of microrefugia might shift in space and time. For species occupying the coolest locations, microrefugia may only be available for a short time period, and the species may benefit from management interventions which prolong cool microrefugia in the landscape where rates of climatic change are slower (e.g. increasing woodland cover in valley bottoms, as long as target species do not require open habitats). For other species, locations of microrefugia might change depending on changes in exposure to prevailing weather (e.g. as a result of changing patterns of wind and cloud cover) (Maclean 2020) and management could aim to promote continuity in climatic conditions so species do not have to travel far to track suitable climate (Suggitt et al. 2018).

Management interventions will depend on the requirements of target species and whether the goal of management is to prolong availability of cooler conditions or promote continuity of climatically stable habitats. For the former, interventions could include increasing canopy shading (Suggitt et al. 2011; De Frenne et al. 2019) including in areas of topographic microrefugia (Lenoir et al. 2017), targeted habitat management along elevation gradients (Tellería 2020) or in cool valley bottoms (Scherer et al. 2021). Management to promote continuity in microrefugia can involve restoring habitats (e.g. by grazing, cutting, planting food plants, pond creation) across areas of topographic heterogeneity within the dispersal capacity of the species, including poleward facing slopes (Suggitt et al. 2011, 2018), towards coastlines (Ashcroft et al. 2008; Van Dyck et al. 2015) or up, down or orthogonally along elevation gradients (Oldfather et al. 2020), or habitat restoration in areas of hydrological refugia including habitats with lower rates of evapotranspiration, or a higher or more reliable water holding capacity (Nimmo et al. 2016; Zylstra et al. 2019; Johansson et al. 2020; van Bergen et al. 2020). Though these approaches may be difficult for species with highly specialised and inflexible habitat requirements (Salgado et al. 2020), they could help promote persistence in landscapes for longer as regional macroclimates deteriorate.

Conclusion

In this review, we explain and evidence how species adaptive and demographic responses to climate change vary across environmentally heterogenous landscapes depending on the extent of habitat fragmentation and fine-scale spatial variation in climate and rates of climate change. As a result, patchy range retractions occur throughout species ranges, with contractions to microrefugia or high-quality, large, aggregated habitat networks. Given the extent of evidence for effects of climate change and habitat fragmentation on species, it is therefore surprising these patchy retractions are not more widely considered in conservation management, policies for climate change adaptation and resilience, or research on climate-driven range shifts.

Patchy range retractions could have been overlooked in conservation as until recently many studies of climate-driven range shifts have focused on large-scale, latitudinal range shifts, and especially expansions (Taheri et al. 2021). Improving our understanding requires; (i) more fine-scale distribution data across a range of taxa; (ii) improved understanding of other drivers of abundance and distribution trends in the context of climate change (e.g. nitrogen deposition, biotic interactions); (iii) field-based and genetic research on adaptive and thermal limits which shape species responses to climate change (Bay et al. 2018; Parmesan and Singer 2022). Increasing the evidence base and understanding of adaptive management strategies, species responses and drivers of range retractions is a priority to ensure trailing edges occurring across species ranges are represented in conservation and policy.

Slowing declines is important to provide time to amass the evidence and understanding of species’ adaptive and thermal limits, to thus develop long-term strategies to adapt conservation to climate change. A conservation framework that acknowledges the role of local climatic variation in creating patchy declines across landscapes and regions can help inform adaptive management strategies in the context of other interacting threats. Contrary to initiatives which rush to shift species northwards, this approach helps reduce the risks associated with regionally maladaptive strategies or ecological traps (Van Dyck et al. 2015; Parmesan and Singer 2022), and prolongs persistence of local adaptations which might be important for long-term conservation throughout species ranges. Through protecting species and increasing resilience of landscapes, our framework aims to slow declines and avoid a trajectory of hollowed out range centres, as a complement to any focus on promoting range shifts and leading-edge expansions.