Hierarchical eco-zone delineation in data-scarce Afrotropical River landscapes using global datasets

This repository contains data, statistics and maps on the hierarchical ecological zonation of Uganda using machine learning algorithms and global environmental datasets. The framework employs a systematic five-step approach integrating GIS analysis, statistical processing, and unsupervised hierarchical clustering to delineate ecological boundaries based on climate, topography, and vegetation patterns. Methodology: The eco-zonation process utilised 16 environmental parameters derived from global datasets, including WorldClim bioclimatic variables (1970-2000), SRTM elevation data, and surface water extent maps. Zonal statistics were calculated for micro-watersheds based on HydroSHEDS level-10 basins intersected with 1 km elevation contours. Hierarchical clustering analysis using Euclidean distance and agglomerative clustering identified optimal cluster numbers through silhouette analysis. Random forest models determined key environmental drivers at each hierarchical level. Validation:The framework was validated against 34 Potential Natural Vegetation (PNV) types using supervised random forest classification. Cross-validation achieved accuracies exceeding 59.87% across all hierarchical levels, with the highest performance (88.68%) at the ecoregion level. Indicator species analysis identified statistically significant vegetation associations within each ecological zone. Results:Uganda's 241,147 km² terrestrial area was stratified into five ecoregions: Central-Northern Plains (CNP), Southern Savanna-Woodland Landscapes (SSWL), Karamoja Dry Steppe (KDS), Lake Victoria-Kyoga Wetland-Woodlands (VKWW), and Montane-Alpine Ranges (MAR). These were further subdivided into 8, 12, and 20 bioregions at three successive hierarchical levels. Monthly precipitation patterns and wetland distribution emerged as primary drivers across all spatial scales. Repository Contents: This repository contains the datasets, scripts, and workflows for the hierarchical delineation and validation of eco-zones in Uganda’s Afrotropical river landscapes. The project integrates global environmental datasets with open-source machine learning algorithms to produce spatially consistent, ecologically meaningful eco-zonation maps. Workflow > Abiotic Delineation: Hierarchical clustering of climate, topography, and hydrology variables to define eco-zones at multiple scales. > Biotic Validation: Random Forest classification and Indicator Species Analysis (ISA) using Potential Natural Vegetation (PNV) to assess ecological coherence. Included Materials Raw and processed datasets for both abiotic and biotic steps R scripts and R Markdown workflows for reproducible analyses Output files, including maps, tables, and validation metrics This framework provides a reproducible, open-data-driven approach for eco-zone mapping in data-scarce tropical landscapes.