South Manchurian mixed forests (SMMFs): distribution maps (2001 & 2021) and habitat suitability models

This repository contains data and code accompanying the study "Satellite-based mapping of South Manchurian mixed forests highlights critical conservation needs", which assesses the status and threats to one of the most biodiverse temperate forests globally. Background and Significance The South Manchurian mixed forests (SMMF), spanning northeastern China, southern Primorsky Krai (Russia), and the Korean Peninsula, are experiencing significant degradation. They are being replaced by secondary growth, agricultural land, and anthropogenic vegetation. Precise quantification of the remaining SMMF patches is critical for their conservation. Materials and Methods Mapping: Integration of original geobotanical data (relevés, literature, regional maps) to train classifiers using Landsat time-series (GLAD ARD, 30m). Ensemble decision tree methods were employed. Potential Distribution Modelling: Application of habitat suitability modelling (ecological niche modelling) using community-level presence/absence data, topographic (elevation, slope, aspect), and bioclimatic predictors (a proxy for precipitation, warmth and continentality indices). Key Outputs: - Actual Distribution Maps: Digital forest cover and South Manchurian mixed forest (SMMF) maps at 30m resolution for 2001 and 2021. - Validation Data: Data for assessing the mapping accuracy. - Stable Forest Paches: A composite 2021 "actual" SMMF map identifying stable forest patches that persisted for 20 years. - Potential Distribution: Models of the potential (pre-anthropogenic) extent of SMMF, generated using habitat suitability modelling. Data Structure: File / Dataset Format Description 1. Source Training Data forest-non-forest_2001_2021.zip Shapefile Polygons for creating binary "forest/non-forest" masks (target and background). smmf-other_forest_2001.zip Shapefile Polygons for classifying SMMF within the 2001 forest mask. smmf-other_forest_2021.zip Shapefile Polygons for classifying SMMF within the 2021 forest mask. 2. Resulting Raster Maps forest2001mask.tif, forest2021mask.tif GeoTIFF Binary forest cover masks (1 = forest, 0 = non-forest). smmf_2001_binary.tif, smmf_2021_binary.tif GeoTIFF Binary SMMF maps (1 = SMMF, 0 = other forest). smmf_combined.tif GeoTIFF Actual SMMF distribution for 2021. Composite map integrating patches stable over 20 years. 3. Vector Data & Statistics actual_SMMF_map_area.zip Shapefile Area distribution of actual SMMF (2021) per 10×10 km grid cell. actual_SMMF_map_patches.zip Shapefile Spatial boundaries and attributes (area) of individual identified SMMF patches. grid10x10_potential.zip Shapefile Zonal statistics for the actual and all potential distribution models in a 10×10 km grid. 4. Validation Data Sampling.csv CSV Stratified sampling points for map accuracy assessment. sampling_sum.tif GeoTIFF Raster aggregation layer for error quantification. 5. Potential Distribution Models proj_Current_SMMF.tb[...].tif GeoTIFF Ensemble of potential distribution models (based on random sampling from target/background polygons). proj_Current_SMMF.sample[...].tif GeoTIFF Ensemble of potential distribution models (based on stratified sampling presence/absence data). Results_by_algo_SMMF_tb.csv CSV Model evaluation metrics (target-background sampling ensemble). Results_by_algo_SMMF_sample.csv CSV Model evaluation metrics (stratified sampling ensemble). 6. Analysis Scripts (R) 01_biomod_potential_distribution.R R Script for modelling the potential distribution of SMMF. 02_potential_area_calc.R R Script for comparing actual and potential SMMF distribution.