Data and code for "Mapping Leopard Conservation Priorities in Southeast Asia"

Authors # Author Affil. ORCID 1 Atzeni, Luciano 1 0000-0002-4573-7431 2 Kamler, Jan 1 0000-0003-4148-2088 3 Ash, Eric 1 0000-0002-8099-269X 4 Kyaw, Pyae Phyoe 2 0000-0002-2763-5706 5 Rasphone, Akchousan 3 — 6 Pin, Chanratana 4, 5 — 7 Tan, Cedric Kai Wei 6 0000-0001-6505-2467 8 Rostro-García, Susana 7 0000-0002-1926-4861 9 Jantz, Patrick 8 0000-0001-5103-2270 10 Gonzalez, Ivan 8 — 11 Burnham, Dawn 1 0000-0001-6577-2640 12 Cushman, Samuel A. 1, 9 — 13 Macdonald, David W. * 1 0000-0003-0607-9373 * Corresponding author — [e-mail] Affiliations # Affiliation 1 Wildlife Conservation Research Unit, Department of Biology, University of Oxford, Life & Mind Building, South Parks Road, Oxford, UK 2 Wildlife Conservation Society Myanmar Program, Yangon, Myanmar 3 WWF-Laos, Vientiane, Lao PDR 4 Ministry of Environment, Cambodia 5 Conservation Ecology Program, King Mongkut's University of Technology Thonburi, Thailand 6 School of Environmental and Geographical Sciences, University of Nottingham, Malaysia 7 Panthera, 8 West 40th St, New York, NY 10018, USA 8 School of Informatics, Computing, and Cyber Systems, Northern Arizona University, Flagstaff, AZ, USA 9 Department of Biology, University of Southern Denmark, Odense, Denmark Description This deposit contains the data and R code to reproduce the landscape connectivity analysis and its sensitivity assessment for the Critically Endangered Indochinese leopard (Panthera pardus delacouri) in Southeast Asia, accompanying the manuscript "Mapping Leopard Conservation Priorities in Southeast Asia" (submitted to Ecology and Evolution). Connectivity is modelled with cumulative resistant kernels (CRK) grown from random source points across a land-cover resistance surface, on a ~1 km grid using block-average (mean) aggregation. The sensitivity axis is the functional form of the resistance transform: three surfaces — p07 (concave), p10 (linear) and p15 (convex) — each run with ten random source-point replicates. A finer ~500 m grid is also produced and used to confirm that the ~1 km grid is adequate. From these kernels the analysis delineates priority patches at two thresholds of the kernel-value distribution (Connectivity Kernels, P = 0.2; Core Areas, P = 0.8), computes per-patch priority metrics (Kernel Extent, Kernel Sum, Protected Areas, Intensity and a composite Mean), ranks the 14 extant forest complexes and 11 proposed reintroduction areas, and tests the robustness of the ranking to the resistance transform (Kendall's W, replicate coefficient of variation, and per-complex rank stability). The archive is organised as a five-stage, numbered pipeline; every script uses relative paths, and each folder carries its own README. A top-level README gives the full run order and where to start. 01_resistance_variants/ builds the three resistance surfaces by power-transforming a categorical land-cover resistance raster and runs the cumulative resistant kernels (cola) for every surface × scenario × study area × replicate, at ~1 km and ~500 m; it also holds the random source-point sets. 02_analyses_batch/ holds the resulting cumulative resistant-kernel rasters (data only) as two parallel grid trees (~1 km and ~500 m), one GeoTIFF per surface × scenario × area × replicate with per-run diagnostics, and is the input to stages 3–5. 03_correlation/ correlates the connectivity surfaces across resistance transforms and across the two grids, per study area and replicate, with the per-area summary figures and tables. 04_COV/ computes the per-cell coefficient of variation across the ten replicates (global and per area), the cross-transform correlation of the replicate-mean surfaces, and the CoV maps. 05_analyses_landscapes/ carries out the priority analysis: landscape surfaces, priority kernels and metrics, ranking tables, robustness analyses and figures, pairwise metric differences, and the publication maps. To reproduce the analysis from scratch, start at 01_resistance_variants/ (this stage needs the cola package and its Python interpreter and is compute-heavy); to reproduce only the analyses and figures, start from the cumulative resistant-kernel trees in 02_analyses_batch/ and run stages 3–5. The code requires R (≥ 4.2) with terra and sf throughout, cola for stage 1, and ggplot2, viridisLite, scales, tidyterra, ggspatial, ggrepel, patchwork, maptiles, flextable and officer for the figures, maps and formatted tables. Keywords Indochinese leopard; Panthera pardus delacouri; habitat connectivity; conservation translocation; reintroduction; CoLa DSS