Agricultural expansion in African savannas: effects on diversity and composition of trees and mammals

Abstract Land use change (LUC) is the leading cause of biodiversity loss worldwide. However, global understanding LUC's impact on mainly based comparisons land endpoints (habitat vs non-habitat) in forest ecosystems. Hence, it may not generalise to savannas, which are ecologically distinct from forests, as they inherently patchy, and disturbance adapted. Endpoint also cannot inform management intermediate mosaic landscapes. We aim address these gaps by investigating species- community-level responses mammals trees along a gradient small scale agricultural expansion miombo woodlands northern Mozambique. Thus, case study represents most common pathway LUC world's largest savanna. Tree abundance, mammal occupancy, tree- mammal-species richness showed non-linear relationship with (characterised Division Index, LDI). These occurrence diversity metrics increased at LDI (0.3 0.7), started decreasing beyond > 0.7, underwent high levels decline extreme (LDI 0.9). Despite similarities species responses, two taxonomic groups contrasting ?-diversity patterns response increasing LDI: dissimilarity among tree communities (heterogenisation) similarity (homogenisation). Our analysis landscape-scale intensification allows novel impacts different conversion, can help guide restoration policy. Biodiversity this landscape was lower than would be inferred existing syntheses biodiversity-land relations for Africa or tropics, probably because such take fully converted endpoint. As, currently, African savanna landscapes habitats agriculture, current estimates, albeit trend towards further conversion. 0.9 < 15% habitat cover) appears more sensitive meta-analyses. To mitigate worst effects biodiversity, our results suggest that should retain 25% cover avoid 0.75—after both begin decline. findings indicate easier restore natural diversity, became spatially homogeneous.