AUTHOR=Muthoka James M. , Rowhani Pedram , Salakpi Edward E. , Balzter Heiko , Antonarakis Alexander S. 

TITLE=Classification of grassland community types and palatable pastures in semi-arid savannah grasslands of Kenya using multispectral Sentinel-2 imagery

JOURNAL=Frontiers in Sustainable Food Systems

VOLUME=Volume 9 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/sustainable-food-systems/articles/10.3389/fsufs.2025.1543491

DOI=10.3389/fsufs.2025.1543491

ISSN=2571-581X

ABSTRACT=Semi-arid grassland ecosystems are crucial for biodiversity, carbon sequestration, and animal fodder; however, they are increasingly threatened by overgrazing degradation and climate variability. Understanding their spatial distribution and palatability is essential for sustainable land management and maintenance of pastoralist livelihoods. This study aimed to map grassland communities and assess their palatability in semi-arid Kenya using Multiple Endmember Spectral Mixture Analysis (MESMA) and Sentinel-2 satellite imagery, integrating species abundance with forage quality metrics. Sentinel-2 imagery was processed using MESMA to classify the fractional cover of four key grass species (Cynodon, Setaria, Themeda, and Kunthii) along with non-grass land cover types (bare ground, forests, shrubs, and water). An iterative endmember selection method optimized the classification, achieving a root mean square error (RMSE) of 23.5% and a 6% improvement in the overall accuracy compared to the unoptimized models. Palatability was assessed based on literature-derived chemical analyses and pastoralists’ perceptions of the forage quality. In the study area, medium and low-palatable species (Setaria and Kunthii) predominated lowland and midland areas, whereas highly palatable Cynodon was found in small, scattered areas across varied elevations. Mixed-grass communities were found in the central areas. The optimized MESMA model effectively identified overgrazed areas and areas vulnerable to degradation by observing grass palatability with grazing pressure from wildlife and livestock. The MESMA model utilized Sentinel-2 imagery and successfully characterized grassland communities’ spatial distribution and palatability in the study area. These findings provide actionable insights for sustainable grazing management and land protection, assisting pastoralists in identifying optimal grazing areas and enabling land managers to implement targeted restoration measures.