Elsevier, Quaternary Science Reviews, 13-14(26), p. 1818-1837
DOI: 10.1016/j.quascirev.2007.03.002
Full text: Download
We present the first synchronously coupled transient simulation of the evolution of the northern Africa climate-ecosystem for the last 6500 years in a global general circulation ocean–atmosphere–terrestrial ecosystem model. The model simulated the major abrupt vegetation collapse in the southern Sahara at about 5 ka, consistent with the proxy records. Local precipitation, however, shows a much more gradual decline with time, implying a lack of strong positive vegetation feedback on annual rainfall during the collapse. The vegetation change in northern Africa is driven by local precipitation decline and strong precipitation variability. In contrast, the change of precipitation is dominated by internal climate variability and a gradual monsoonal climate response to orbital forcing. In addition, some minor vegetation changes are also simulated in different regions across northern AfricaThe model also simulated a gradual annual mean surface cooling in the subtropical North Atlantic towards the latest Holocene, as well as a reduced seasonal cycle of SST. The SST response is caused largely by the insolation forcing, while the annual mean cooling is also reinforced by the increased coastal upwelling near the east boundary. The increased upwelling results from a southward retreat of the North Africa monsoon system, and, in turn, an increased northeasterly trade wind. The simulated changes of SST and upwelling are also largely consistent with marine proxy records, albeit with a weaker magnitude in the model.The mismatch between the collapse of vegetation and gradual transition of rainfall suggests that the vegetation collapse is not caused by a strong positive vegetation feedback. Instead, it is suggested that the Mid-Holocene collapse of North African vegetation is caused mainly by a nonlinear response of the vegetation to a precipitation threshold in the presence of strong climate variability. The implication to the modeling and observations is also discussed.