Reconstructing savanna tree cover from pollen, phytoliths and stable carbon isotopes

Aim: To calibrate a model of the relationship between bio-proxies (pollen, phytoliths and δ¹³C of soil organic matter) and woody cover, measured as the leaf area index (LAI). This relationship, applied in palaeosequences, enables reconstruction of past savanna tree cover. Location: The samples are from tropical Africa. Modern soil samples are from the Central African Republic and past samples are from sediments of lakes in Senegal and Congo. Methods: We analysed the pollen and phytolith content and stable carbon isotope values of 17 soil samples taken from three short transects in the Central African Republic; LAI was measured on the same transects. The indices used were the APINAP ratio of arboreal (AP) to non-arboreal (NAP) pollen, the D/P ratio of ligneous dicotyledons (D) to Poaceae (P) phytoliths, and the δ¹³C of soil organic matter, i.e. the ¹³C/¹²C ratio. Results: A multi-proxy model was calibrated. The best model included only a combination of pollen and phytolith as proxies, excluding organic matter δ¹³C because of its long mean residence time in the soil. The model was then applied to two palaeosequences in Africa, and a time series of relative LAI changes was obtained, providing new information about vegetation changes. Conclusion: This model can be applied in palaeosequences to reconstruct relative time series of LAI in African savannas and can help interpret vegetation changes quantitatively. This approach is complementary to the description of pollen and phytolith assemblages.