Increasing livestock densities and more severe drought events challenge sustainable management in South Africa’s semi-arid savannas. Effective mitigation strategies require accurate assessments of livestock foraging behaviour. By utilising high-throughput sequencing technology, this study evaluated the use of a dual-locus metabarcoding approach (trnL and rbcL) together with study-area-specific reference libraries, to analyse cattle diets in two bioregions of the eastern semi-arid South African savanna. Both markers demonstrated the ability to identify various plant families, but trnL exhibited a higher diversity in terms of family and genus identification at both sampling sites. Forbs, although comprising a diverse component of savanna plant communities, have relatively small above-ground biomass, but can still serve as crucial forage items, especially during dry periods. Our study underscores the significant role of forbs in cattle diets, demonstrating a shift in cattle foraging preferences from grass-based diets to higher inclusions of forbs and woody taxa during the drier season. Although grasses, such as Setaria, were still prevalent, forbs, belonging to the genera Malvastrum, Asparagus, Pollichia and Ipomoea were also important food items for cattle as well as woody taxa belonging to Fabaceae, Combretaceae, Ebenaceae, and Malvaceae with a selection of food items from trees and shrubs from genera Albizia, Combretum, Euclea and Vachellia. Furthermore, our study highlights the value of a dual-locus metabarcoding approach for understanding herbivorous diets. Using trnL and rbcL markers, with study-area-specific reference libraries, improves taxonomic resolution for accurately reconstructing cattle diets in semi-arid savannas. This study may improve biodiversity estimates and inform sustainable rangeland management strategies in semi-arid African savanna ecosystems.
