Habitat patches providing south–north connectivity are under-protected in a fragmented landscape

1. Introduction

Species can be hampered in their ability to shift ranges as an adaptation to climate change [1] where there are synergistic negative impacts of anthropogenic land use [2,3]. We need to safeguard species’ ability to respond to climate change by incorporating regional and national connectivity into conservation planning [4]. Many studies look at how easily individuals can traverse landscapes [5,6], but modelling landscape connectivity across one or few generations is unlikely to predict long-term, large-scale responses to climate change. Studies need to assess multi-generational connectivity, i.e. whether there is enough habitat in the right places to facilitate long-distance range shifts. Landscape-scale decision-making is crucial to deliver climate-resilient landscapes [7], and losing habitat patches from critical regions between current and projected ranges will hamper species’ range expansion—potentially causing extinction [8].

Recently, we have seen a global shift towards promoting functionally connected networks, typified by Aichi biodiversity target 11 [9]. National examples of this include the UK Government's plan to develop a nationwide Nature Recovery Network to protect, restore and connect the country's wildlife sites [10]. Such initiatives cause stakeholders to reconsider where to prioritize conservation of priority habitats. For both pragmatic and strategic reasons, conservation may have historically favoured larger patches over small ones, thus avoiding fragmented regions [11,12]. However, there remains active debate on the value of several small patches for species richness, versus one contiguous patch of the same size [13,14]. Simulations of species persistence and expansion, using simple metapopulation models, highlight the strengths and weaknesses of different habitat creation strategies. In general, aggregation strategies are good for facilitating metapopulation persistence but not for range expansions, because large gaps are left between habitat aggregations in the direction of range advance [15,16].

Safeguarding habitats in protected areas (PAs) is a widespread, cost-effective tool for biodiversity conservation [17]. Many studies have demonstrated the representation of species’ projected future ranges in existing PA networks [18]. Others have shown existing PAs may facilitate species' range expansions by supporting high abundances of, and preferential colonization by, range expanding species [19]. However, colonization does not necessarily lead to successful range expansions, and an important subset of species are failing to shift their range. If protection was lost in patches critical to reaching the projected range, even more species could be vulnerable, and up to now, the protection of such critical patches has generally not been prioritized. Following intensive research, software can now incorporate connectivity in relation to climate change into the decision-making process [7,20–22]. Work is ongoing to put connectivity science into practice, and incorporate connectivity in a nuanced, ‘climate-wise’ context [7,23]. To that end, tools to identify and protect habitat patches that are crucial for range expansion need to be developed and disseminated.

Successful inclusion of connectivity in conservation decision-making also depends on legal and ecological context. In England, a 2006 Act of Parliament [24] provides for the conservation of listed priority habitats and species [25]. Specifically, legally recognized priority habitats, from lowland meadows to blanket bog, are platforms to protect ca 1000 priority species. Priority habitats are ecologically distinct from one another, providing for unique subsets of priority species including threatened and specialist plants, fungi, birds, beetles, butterflies, moths and several other taxa [26]. Some species depend on multiple priority habitats, but protection, restoration and conservation decisions are likely to consider each habitat individually. Beyond priority habitats, conservation practice in England now emphasizes building a ‘coherent and resilient ecological network’, and ensuring that wildlife sites are ‘joined up’ [27]. During contemporary climatic warming, South–North range shifts have been widely documented in England among many species [26]. In recent decades, those species undertaking range shifts have disproportionately colonized PAs, highlighting PAs' key role in protecting habitats—even in species’ potential future ranges [28].

Here, we use connectivity analysis to inform decision-making within the constraints of a specific policy context. We assess the capacity of England's PAs to secure long-distance connectivity in 16 national conservation priority habitat networks. We define habitat networks as assemblages of patches of a given priority habitat type, because priority habitats receive distinct legal recognition and underpin planning decisions in England, and are highly ecologically distinct, providing for unique subsets of priority species. We use Condatis [29,30], a landscape-scale decision-support software, to identify habitat patches (i.e. contiguous clumps of habitat) critical to long-distance connectivity and range expansion under climate change. Condatis uses circuit theory to efficiently calculate how quickly a species could reach a specified target location from a specified source. It has mathematical similarities to, but key conceptual differences from, the circuit theory models used by other landscape ecologists [31]: one link in the Condatis network represents a population sending colonists to an empty patch to found a new population (not a disperser stepping between one cell and its neighbour). Crucially, if a patch in Condatis has high ‘flow’, it is located on one of the likeliest routes for range expansion between the source and the target [30].

To better understand and conserve priority habitats under climate change, we ask: (i) to what extent are high-flow habitat patches represented in England's current PA network? (ii) How is the protection and/or high-flow status of habitat patches related to their area? (iii) How much extra network connectivity could be conserved through targeted conservation of high-flow habitat patches? We use generalized linear models to explore relationships between the flow, size and protection status of patches across priority habitat networks. We rank unprotected habitat patches based on their contribution to long-distance connectivity, and strategically add them to the PA network to demonstrate how targeted conservation could efficiently increase connectivity for a given increase in PA coverage.

2. Methods

(c) Condatis settings

We ran Condatis for each of the 16 priority habitat networks, and three exemplar mean dispersal distances. We did not attempt to make exact species-specific predictions; instead, we focused on habitat networks as platforms for conservation actions, using traits and processes relevant for multiple species. For the mean dispersal distance trait (2/α; equation (2.1)), 2, 4 and 8 km options were run, aiming to represent a broad range of plants, fungi, vertebrates and invertebrate species specialized to each priority habitat network. While many relevant species likely have dispersal abilities of less than 2 km, Condatis calculations encountered rounding errors if the average dispersal was several orders of magnitude lower than the largest gap in the network. Reproductive rate (R; equation (2.1)) was fixed at 100 throughout, equating to the production of one emigrant per hectare. This was not based on specific data but is plausible for a medium-bodied vertebrate, or an invertebrate with a low population density. Varying R would not have affected the relative performance of networks and patches, which were the focus of this study, because R modulates all flow and conductance values in proportion.

We identified sources and targets for Condatis on the premise that species are adjusting their ranges to higher latitudes [1]. Thus, a 10 km raster file was produced with sources along the south coast of England, and targets along the northern border with Scotland (figure 1).

• Download figure
• Open in new tab
• Download PowerPoint

3. Results

The networks of priority habitat in England range in extent from greater than 0.7 M ha (deciduous woodland) to 297 ha (calaminarian grassland), cover 13.1% of England's land (1.7 million ha total; table 1) and are highly fragmented (median patch size 0.75 ha; electronic supplementary material, table S2). Six habitat networks (salt marsh, maritime cliff and slope, coastal sand dunes, coastal vegetated shingle, saline lagoons and reedbeds) were spatially distributed such that they could not be analysed as electrical circuits at the scales and resolutions used in the study (see electronic supplementary material, appendix C). Of the remaining habitat networks, those covering a larger area facilitated significantly faster speeds of range expansion (regression of log conductance on log-transformed area; β = 3.655, 95% CI [1.371, 5.940], R² = 0.371, F_1,14 = 9.493, p = 0.0073; figure 2a). Habitat networks also varied widely in the extent to which they are currently protected, ranging from 0.3% (traditional orchard) to 94% (mudflats), with a mean of 53.5% (figure 2b). Although the majority of habitat area was protected in most of the habitat networks (figure 2b), most patches were unprotected (electronic supplementary material, table S2). This was possible because within each habitat network protected patches were, on average, larger than unprotected patches (overall protected mean area 20.98 ha (n = 32 253); overall unprotected mean area 3.58 ha (n = 287 737); electronic supplementary material, table S2), and tended to be less fragmented (protected GISfrag = 258.84, unprotected GISfrag = 84.84; electronic supplementary material, table S2). Proportionally, protection of flow was generally lower than the protection of area; in 12 of the 16 habitat networks, flow protection was, on average, 13.6% lower than area protection. The proportion of flow protected matched or exceeded the proportion of area protected in the remaining four habitat networks (blanket bog (+5.28%), traditional orchard (+0.01%), lowland raised bog (+ 2.19%) and upland hay meadow (+4.60%); figure 2b).

• Download figure
• Open in new tab
• Download PowerPoint

Larger patches of a given habitat network generally had higher flow. The Kendall rank-order correlations showed weak-to-moderate positive correlations between patch size and patch flow in most habitat networks (electronic supplementary material, table S2; overall τ = 0.309). However, small patches can contribute disproportionally to connectivity; there is wide variation in patch flow values among patches with low area (figure 3a; electronic supplementary material, figure S2). Of the top 10% of patches for flow in each habitat network, an average of 13.8% were patches with an area of less than or equal to 1 ha (electronic supplementary material, table S3).

• Download figure
• Open in new tab
• Download PowerPoint

Given a tendency for larger patches to have higher flow, and to be more often protected, we might expect flow to be well protected. Two results help to show why this is not the case. Firstly, among patches that have below-average area, protection level clearly declines with the increasing flow (figure 3b; electronic supplementary material, figure S3). Secondly, GLMs that include flow as a predictor of protection indicates that flow has generally negative effects, and those that include both area and flow as predictors of protection status show even more negative effects of flow (figure 4; electronic supplementary material, table S4). Effects of patch area in these GLMs tend to be positive and to become more positive when flow is included as a predictor. Just three habitat networks are exceptions, where the model shows positive effects on protection attributed to both the predictor's area and flow (figure 4; electronic supplementary material, table S4).

• Download figure
• Open in new tab
• Download PowerPoint

In our scenarios in which additional high-flow habitat patches were protected, increases in the proportion of flow protected were almost always greater than increases in the proportion of habitat area protected (figure 5). In a few cases, disproportionate improvements to overall flow protection were not possible due to insufficient unprotected high-flow patches: specifically, coastal floodplain grazing march (when adding 1, 10 and 25%), blanket bog (adding 10%) and lowland raised bog (adding 10%). However, most connectivity conservation gains were highly disproportionate to the areas of habitat selected for protection. Across all habitat networks, increasing the coverage of PAs by 1, 10 and 25% resulted in respective median increases of 8.0, 40.9 and 57.8% flow protection (figure 5b).

• Download figure
• Open in new tab
• Download PowerPoint

4. Discussion

Here, we highlight that the connectivity of the fragmented networks studied is vulnerable because patches critical for species range expansions are under-protected. Crucially, we found that, for the majority of habitat networks, protection is biased away from high-flow patches (figure 4); the median rate of protection of patches in each habitat network is 44.5% but drops to 37.5% when considering only patches above the 90th percentile for flow. This absence of designation increases the likelihood of degradation or destruction of habitat patches, which is expected to severely impact network connectivity.

Previous research has established that for the majority of countries, PA connectivity is lacking [39]. However, species will not directly respond to PA connectivity per se; it is the connectivity of the entire habitat network, whether or not protected, which affects the reproduction and dispersal of species, and is critical for range expansion under climate change [19]. Unlike previous work, we investigate patch connectivity and patch protection independently. Thus, we contribute ecological realism by focusing not on PA connectivity, but the connectivity of the habitat networks that PAs conserve. We identify the important routes a wide variety of species may take, using simplified dispersal assumptions, as they shift ranges from South to North in reaction to climate change, regardless of protection status (electronic supplementary material, figure S4). In this way, we identify a critical oversight in the design of England's PA network.

Patches that happen to be strategically located to act as South–North stepping stones (figure 6) may be small and may lack other attributes that were important for past PA designation. We found that existing PAs tended to be biased towards low-flow patches for most habitat networks, despite also being biased towards large patches. This is surprising because large patches typically have higher flow (figure 3a). The preferential protection of large patches over small is not a new finding [11]. However, that those same patches typically contribute more to connectivity, and yet connectivity is still under-represented, indicates a disconnect between past protection decisions and those needed to facilitate range shifts.

• Download figure
• Open in new tab
• Download PowerPoint

The patterns we observe in protection are probably not unique to England, given similar biases and lack of PA connectivity have also been evidenced in other regions [12,40]. We propose that comparable network vulnerabilities elsewhere probably result from similar habitat protection principles—and practical considerations—to those known in our study region [41]. For example, reserve selection might actively favour aggregation because, under a stable climate, species persistence is expected to be higher in aggregated networks than fragmented ones [42]. However, passive processes could also be at play. In the UK, many PAs arose from ‘Rothschild's Reserves’ [27], the selection criteria of which included ‘areas of land … which retain primitive conditions and contain rare and local species liable to extinction’ [43]. This led to reserves being clustered in areas of low economic and agricultural development, especially in the North and the uplands [44], a phenomenon not limited to the UK [41]. Furthermore, while climate change was not an issue of the time, it is unlikely any form of connectivity was a factor in historical designation decisions, given the growth of PAs in the UK has often occurred without consideration of their wider context [45].

We do not envisage contribution to connectivity—represented here as flow—to be the sole criterion to prioritize protection. Patches that contribute little to connectivity are often crucial to sustain metapopulations [46]. However, we argue that flow should form part of a nuanced prioritization process, accounting for land-use changes, habitat quality, climate suitability and landscape connectivity [44]. Nevertheless, considering the extent to which high-flow areas have been overlooked, it would not be unreasonable to ring-fence some future PA resources to specifically promote connectivity. Note particularly that flow distribution across patches is highly skewed (on average 31.2% of patches contained 75% of the flow), so future selection of high-flow patches by chance, or by a moderately correlated proxy such as area, is unlikely. By contrast, targeted patch selection on the basis of flow could be very efficient. For example, between 2014 and 2019, terrestrial PAs in the UK increased by 11 200 ha [47]; our analyses show that the addition of 714.25, 438.50 and 3544.50 ha to lowland dry acid grassland, purple moor grass and rush pasture, and calcareous grassland PAs (representing 5% increases in PA) would yield 15.6, 33.1 and 33.5% gains in flow protection, respectively. Such increases in connectivity protection are an urgent requirement if we are to help build more resilient networks for nature in the face of climate change [48].

As connectivity ascends the conservation agenda, we demonstrate the potential for efficient conservation of climate-resilient landscapes. We show that the inclusion of a connectivity measure into the planning process can facilitate the identification of patches important to climate change connectivity, resilience and adaptation. In most habitat networks studied here, substantial gains in connectivity protection can be made for relatively small increases in PA coverage (figure 5b). Only for a small number of habitat networks were proportional connectivity gains less than the proportional increase in area. In these instances, either a strong correlation existed between area and flow or existing protection coverage was high, such that the majority of high-flow patches were already protected (electronic supplementary material, table S3). Saura et al. [39] identified that the targeted designation of PAs to enhance connectivity was many countries’ most pressing priority for meeting PA conservation goals. The flow metric described here provides potential for proactive safeguarding of connected habitats and stepping stones, allowing conservation planners to target their designation and conservation activities to achieve substantial increases in connectivity protection. This could help to meet targets outlined in legislation such as the UK's 25-year environment plan [10] and the EU's biodiversity strategy, or international commitments, including the upcoming Post-2020 global biodiversity framework [49].

Our study uses cutting-edge methods to quantify protection of long-distance, multi-generational habitat connectivity. Our approach has limitations, but also clear avenues for progression. For example, we analyse 16 priority habitat networks individually, based on the UK Biodiversity Action Plan. Some species are of course reliant on multiple habitat types, and to differing extents, so future work could analyse composite networks of associated habitats used by different subsets of generalist species. However, a more comprehensive assessment might not show any additional crucial patches that had been missed in analyses of individual habitats. Furthermore, if actions increased connectivity for individual priority habitat networks, the connectedness of the composite networks they form part of would also improve. The assumption of a homogeneous matrix is another limitation and may lead Condatis to overestimate the importance of some regions for those species that are hindered by landscape barriers. However, this assumption reduces the computational burden of evaluating connectivity, which is itself a major limitation [50], while still maintaining the principles of isolation by resistance [51]. Another limitation is that while our choice of sources and targets follow the general trend of species moving away from the equator, it does not consider that climate refugia may be found at higher altitudes or different aspects. Furthermore, although the negative exponential kernel at the core of our analysis has been tried-and-tested for modelling animal movements, it may be a poor function for plant dispersal; for example, some studies suggest that log hyperbolic secant or exponential power probability density functions would be more appropriate [52]. Finally, due to our focus on habitats, the scale at which most conservation actions happen, we made use of theoretical species. It would be beneficial to validate our findings empirically with data from species that have already shifted ranges. However, data are not always available, while conservation guidance is needed immediately; many range shifts are ongoing, or have yet to start [53].

Our study quantifies how South–North connectivity is currently conserved within PAs across fragmented habitat networks, using England as an exemplar for application to other countries or regions. Although PAs tend to contain larger patches, which usually contribute more flow, they under-represent connectivity in the majority of habitats studied. The scientific community has been emphasizing the importance of incorporating connectivity into the planning process for at least 30 years [54], but the connectedness of habitats remains vulnerable to degradation and loss. We have shown that patches important to long-distance connectivity can be easily identified, allowing the proportion protected to be greatly increased with minimal additional resources. The decision-making tools demonstrated here help enable the change in conservation planning needed to protect the permeability of landscapes, allowing species to track changing climate and preventing extinction.

Data accessibility

All processed data and R code (including demonstration of Condatis and proportional flow assignment) are available via the Environmental Information Data Centre at https://doi.org/10.5285/a715112e-08ae-4d6e-943a-77933fd5ddd1 [55].

Authors' contributions

T.J.P.T.: conceptualization, data curation, formal analysis, methodology, writing—original draft, writing—review and editing; J.A.: conceptualization, data curation, formal analysis, methodology, writing—original draft, writing—review and editing; S.D.T.: conceptualization, writing—original draft, writing—review and editing; H.Q.P.C.: conceptualization, writing—original draft, writing—review and editing; J.A.H.: conceptualization, data curation, formal analysis, methodology, writing—original draft, writing—review and editing. All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Competing interests

We declare we have no competing interests.

Funding

T.J.P.T. was funded by an NERC studentship through the ACCE (Adapting to the Challenges of a Changing Environment) Doctoral Training Partnership (grant no. NE/L002450/1), and by Natural England. J.A. and J.A.H. were supported by a Memorandum of Agreement between Natural England and the University of Liverpool. Condatis development was funded by NERC grants (grant nos. NE/L002787/1 and NE/R009597/1).