Improving long-term rangeland aboveground biomass estimation in the middle Awash basin, Ethiopia, using a locally calibrated Carnegie–Ames–Stanford Approach

Reliable long-term monitoring of rangeland productivity in tropical drylands is constrained by the scarcity of high-resolution, temporally continuous biomass datasets. This study developed a locally calibrated, uncertainty-aware Carnegie–Ames–Stanford Approach (CASA) framework to reconstruct monthly aboveground biomass production at 30 m resolution for the middle Awash rangelands, Ethiopia, from 2001 to 2025. The framework integrates spatiotemporal fusion of Landsat–MODIS vegetation indices using the Gap Filling–Savitzky–Golay (GF–SG) method, reconstruction of Landsat land surface temperature using Harmonic Analysis of Time Series implemented in Google Earth Engine (HANTS–GEE), and a Temperature–Vegetation Dryness Index (TVDI)-based water stress formulation. The fraction of absorbed photosynthetically active radiation (fAPAR) and maximum light-use efficiency (ε max ) were jointly optimized through spatially blocked cross-validation and parameter equifinality analysis. The optimized EVI-based parameterization (fAPAR = 0.944 × EVI − 0.160; ε max = 1.881 g C MJ -1 ) achieved high validation accuracy (R 2 = 0.768; RMSE = 0.606 t ha -1 ) and strong agreement with the independent FAO WaPOR biomass product (R 2 = 0.723). Seasonal aboveground biomass, reconstructed by accumulating monthly AGB production during the growing season, averaged 1.68 t ha -1 . Trend-free pre-whitened Mann–Kendall analysis revealed widespread long-term productivity declines, with the strongest reductions occurring during August, indicating degradation of peak-growing-season productivity. Although model evaluation was constrained by a single peak-growing-season field campaign, the resulting high-resolution monthly AGB production dataset and uncertainty-aware calibration framework provide a robust foundation for drought monitoring, forage assessment, adaptive rangeland management, and future biomass reconstruction in data-scarce pastoral ecosystems.