Mapping climate adaptation corridors for biodiversity—A regional-scale case study in Central America

duplicated polygons and remove protected areas with unreported area, “proposed” status, or ‘UNESCO-MAB Biosphere Reserve’ designation. We included terrestrial protected areas and land portions of marine or “partial” protected areas. We also added the ‘Belize Maya’ protected area, a recently created, large protected area (958 km2) not found in the WDPA at the time of access. We used the Shuttle Radar Topography Mission digital elevation data v4 [43] to identify protected areas below 500 m (mean), a threshold to define lowland forested habitats [44, 45]. We defined a single location as the start node within each lowland protected area using the ‘st_point_on_surface’ function in the ‘sf’ R package [46]. Because we operated at the regional scale, we only considered start nodes � 5 km2. Start nodes per country ranged from 14 (El Sal- vador) to 110 (Mexico) (S2 Table in S1 Appendix). In total, our analysis included 475 start nodes ranging from 5–7249 km2 (median = 51 km2) (S2 Table in S1 Appendix, Figs 1 and 2A). We defined end nodes (target habitats) as patches (� 5 km2) of upper-montane (� 1500 m) (mean) contiguous forest [44], with medium or high Forest Landscape Integrity Index values that represent relatively intact, potential highland climate change refugia [47]. This resulted in 529 end nodes ranging from 5–5938 km2 (median = 10 km2). End nodes were mostly located in Guatemala (183), Colombia (151), and Mexico (105), whereas Belize was the only country without end nodes (S2 Table in S1 Appendix). Similar to start nodes, we

used the ‘st_point_on_surface’ function [46] to define end nodes within each target habitat. Unlike start nodes, however, we included both protected and unprotected end nodes because any contiguous, high-elevation forest may represent potential climate change refu- gia or candidate areas for future conservation. To prevent least cost paths needing to travel further to end nodes within larger target habitats, we seeded additional end nodes using hexagonal sampling in the ‘st_sample’ function of the ‘sf’ R package [46]. This resulted in a suite of 779 end points across the 529 main end node polygons with 1–41 end points per polygon (median = 1 point). Conductance surface and least cost path modeling We developed a structural conductance surface (i.e., inverse of resistance) based on the Euro- pean Space Agency’s 10 m WorldCover 2020 v100 dataset [48]. We resampled the land cover data to 100 m resolution to optimize computational constraints. We assigned habitat-specific conductance values (S3 Table in S1 Appendix) based upon previous research conducted for three keystone, medium-large terrestrial mammals of conservation concern present across the study area: White-lipped peccary (crucial seed disperser and browser) [29]; Baird’s tapir (cru- cial seed disperser and browser) [49]; and jaguar (apex predator) [30]. As the WorldCover ‘for- est’ category encompasses a broad array of forest qualities and disturbances, we modified its conductance value by its standing biomass (derived from [50]), to penalize low quality (low

biomass) habitats. To create the modifier, we thresholded all forest biomass values > 100 Mt C to 100 and divided biomass values by 100 such that 0 represented locations with zero biomass and 1 represented locations with � 100 Mt C, then multiplied the conductance score by this modifier (S1 Fig in S1 Appendix). Like many previous regional- to continental-scale climate connectivity studies, we did not adjust conductance values for topography as our focus was long-term range shifts rather than short-term movements (e.g., [51–53]). Moreover, steeper corridors have greater capacity to help species overcome rapid climate change velocities, topography can influence individual species’ movements positively or negatively [54], and our study lacks a single focal species on which to base terrain-derived conductance corrections. Similarly, we did not adjust conductance values for climate dissimilarity as previous “climate corridor” studies have done (e.g., [55]). Whereas “climate corridors” typically are designed to minimize exposure to climatically unsuitable habitat along routes for long-term range shifts, PLOS ONE Mapping climate adaptation corridors for biodiversity PLOS ONE | https://doi.org/10.1371/journal.pone.0304756 May 31, 2024 5 / 19