AbstractAimOwing to its diverse bioclimatic zones, long human history and intense anthropogenic impacts, Africa provides a model system for studying how global terrestrial ecosystems might respond to accelerated socio‐environmental stress. Africa is particularly vulnerable to climate change and human impact, and insufficient baseline data hamper current environmental management efforts. Using palaeoecological data, we seek to identify the timing, pace and drivers of change in African biomes on a long‐term scale to inform current ecosystem management frameworks on the continent.LocationAfrica.Time period0–12 ka.Major taxa studiedAfrican biomes.MethodsSixty‐four pollen records across Africa and nearby sites were retrieved from multiple databases/sources and grouped into biomes. Turnover (quantified using the squared chord distance dissimilarity metric) and rarefaction analyses were conducted on pollen records in each biome group to reconstruct regional temporal vegetation turnover and richness. Reconstructed vegetation turnover and richness were compared with independent records of climate, fire and human activity to identify possible drivers of change.ResultsWe found that the most stable biomes were those with the greatest floristic richness. Southern Africa's mediterranean‐type (SAM) ecosystems were the most stable and northern Africa's mediterranean‐type (NAM) ecosystems were the most unstable (mainly owing to fire). Tropical savannas (TS) and SAM ecosystems expressed the most sensitivity to climatic shifts from ≥6 ka, whereas tropical forests (TF) were relatively stable before human activities intensified from c. 2 ka. Floristic richness also declined across the tropics from c. 2 ka.Main conclusionsOur analysis pinpoints NAM ecosystems as undergoing the fastest acceleration in turnover on the continent in response to fire, whereas TF and TS have been undergoing unprecedented changes in biodiversity in the last 2,000 years. We expect further changes in biodiversity where climate becomes warmer and drier and where human impacts are novel and rapid in comparison to long‐term baselines.