Data for: Modeling Multifunctional Landscape Change: Diagnosing Spatial Trade-offs and Drivers in Arid and Semi-Arid Systems

This repository provides the spatial datasets, model outputs, and documentation supporting the study “Modeling Multifunctional Landscape Change: Diagnosing Spatial Trade-offs and Drivers in Arid and Semi-Arid Systems.” The data support a spatially explicit land‑use and land‑cover (LULC) change modeling analysis conducted for Narok County, Kenya, a multifunctional arid and semi‑arid landscape where conservation, agropastoral livelihoods, and climate stress co‑occur on the same land units. The modeling framework integrates biophysical, spatial, and institutional drivers of LULC change using the Land Change Modeler (LCM) within TerrSet. The repository includes processed and harmonized input data, transition potential surfaces, predicted LULC maps, vulnerability (soft prediction) maps, and spatial error diagnostics (hits, misses, false alarms, and wrong hits). All spatial datasets are provided in open, interoperable GIS formats (GeoTIFF, RST) compatible with QGIS, R, and other open‑source platforms. Extensive documentation (see metadata file) describes data provenance, preprocessing steps (including class aggregation and spatial resolution harmonization), model parameterization, calibration, and evaluation. While LCM is a proprietary modeling environment, all intermediate and final outputs are openly shared to enable independent inspection, evaluation, and reuse. Full replication of calibration requires access to TerrSet; however, transparency and reproducibility are supported through detailed workflow documentation and exported GIS products. Collectively, these materials enable reuse of the datasets for comparative modeling, scenario analysis, or policy‑relevant assessments of land‑use trade‑offs under climate and governance constraints in sub‑Saharan Africa. The project data are organized into structured directories to support reproducibility, clarity, and ease of use. The folder structure and contents are described below to guide users in navigating input datasets, reference files, and model outputs. 1. ComplementaryFiles This folder contains reference datasets used to ensure spatial consistency across all analyses and for figure production. Study Area Boundary (narok_county_utm.shp): A vector shapefile defining the geographic extent of the study area. This file was used for clipping all spatial datasets and for map visualization in figures. Extent Raster (extentRaster.tif): A template raster defining the coordinate reference system (CRS), spatial resolution, and grid alignment. All raster datasets were standardized to match this template to ensure pixel-level consistency across inputs. 2. Preprocessed This directory contains all input datasets used for modeling, including land-use maps and explanatory (driver) variables. All rasters have been clipped to the study area and resampled to a common spatial resolution of 90 m. LULC Baseline Maps: Historical land-use/land-cover rasters (e.g., lu2010_4cls_90m.tif, lu2018_4cls_90m.tif) reclassified into four primary categories: Forest, Rangeland, Cropland, and Urban. Driver Variables: Spatial predictor layers used in the modeling framework, including: Topographic variables (e.g., elevation, slope) Proximity metrics (e.g., Euclidean distance rasters such as dist2_cropland2000.tif, dist2_forest2000.tif) Environmental variables (e.g., annual average temperature, rainfall, vegetation indices) Socio-political variables (e.g., land tenure/policy layers) Landscape structure variables (e.g., forest fragmentation) All datasets were harmonized to ensure consistent projection, resolution, and spatial alignment for analysis. 3. LCM_Outputs This folder contains model outputs generated from the Land Change Modeler (LCM), representing results from both the calibration and prediction stages. Hard Prediction Maps: Categorical land-use maps representing the most likely land-cover state for the prediction year (e.g., 2018). Soft Prediction Maps: Continuous surfaces (e.g., _soft.rst) indicating the likelihood or susceptibility of land-use change, used to assess model uncertainty and confidence. Transition Potential Maps: Raster outputs representing the modeled probability of transition between specific land-use classes, generated by the multi-layer perceptron (MLP) modeling process.