Climate controls on savanna C3 and C4 expansion in Southern Africa during the last 36 kyr BP

Savannahs contain a mixture of C3 and C4 vegetation, accounting for more than a quarter of global primary production and are the second most important biome on the continents. However, our understanding on how savannahs will respond to rising CO2 concentration and temperatures or the IPCC estimated decrease in rainfall is not yet clear in spite of potential far reaching socio-economic consequences. In this study, we used the δD and δ13C of sedimentary long-chain n-alkanes (n-C27,29,31,33 ) in concert with reconstructions for sea surface temperatures and fluvial discharge from a marine sediment core (GIK16160-3, 18°14.47’S, 37°52.27’W, 1334m water depth), collected near the Zambezi river mouth to examine savannah responses under different hydrological and climate conditions in Southern Africa during the last 36 kyr BP. Our data show large variability in both δD and δ13C records of the four n-alkanes, with isotopic differences between individual n-alkanes being far more pronounced during the Glacial than during the Deglacial and Holocene. These large differences may be explained by proportionally higher contributions of C4 grasses over C3 trees to the n-C33,31, which seems to be opposite for n-C29. A strong anticorrelation between δD and δ13C from 36 to 16 kyr BP for n-C31 (R2=0.55) and n-C33 (R2=0.70) suggests that δD of these n-alkanes is strongly influenced by changes in vegetation types as well as physiological effects, rather than being directly related to evaporation/ precipitation balance. In contrast, no apparent relationship (R2=0.32) exists between δD and δ13C of n-C29, suggesting that n-C29 is the most promising hydrological proxy due to less variable vegetation type contributions to n-C29 throughout the core. The C4 plant contribution, which was estimated by taking into account the four n-alkanes δ13C signals and their abundance, implies dominance of C4 grass between 36 and 20 kyr BP, and more evenly distributed C3 and C4 vegetation from 20 kyr BP to Present. We further assume strong seasonal effects on δD of individual n-alkanes for the latter period. Changes in δD of n-C33,31 coincide with Latest Pleistocene to Holocene austral summer insolation, which is in agreement with modern observations that savannah C4 plants grow mainly during summer. Likewise, δD n-C29,27 records closely correspond with changes in austral spring insolation, also in line with modern observations that trees grow during spring and autumn. During glacial times less strong seasonal effects on the isotopic signature of individual n-alkanes are assumed due to predominance of C4 grasslands. The δD record of n-C29suggests wet conditions between 30 and 21 kyr BP and for the Holocene. Dry conditions prevailed during 35 to 31 kyr BP, part of the LGM and the Deglacial, with the driest climate at ~16kyr. According to our estimation, C4 relative to C3 plant abundance has remained rather stable throughout the entire Deglacial and Holocene. While rising temperature is generally assumed to stimulate C4 vegetation expansion, gradual increases in CO2 and humidity may have instead fostered C3 vegetation, counter-balancing expansion of C4 plants.