Combined statistical and dynamical assessment of simulated vegetation–rainfall interactions in North Africa during the mid‐Holocene1

A negative feedback of vegetation cover on subsequent annual precipitation is simulated for the mid‐Holocene over North Africa using a fully coupled general circulation model with dynamic vegetation, FOAM‐LPJ (Fast Ocean Atmosphere Model‐Lund Potsdam Jena Model). By computing a vegetation feedback parameter based on lagged auto‐covariances, the simulated impact of North African vegetation on precipitation is statistically quantified. The feedback is also dynamically assessed through initial value ensemble experiments, in which North African grass cover is initially reduced and the climatic response analyzed. The statistical and dynamical assessments of the negative vegetation feedback agree in sign and relative magnitude for FOAM‐LPJ. The negative feedback on annual precipitation largely results from a competition between bare soil evaporation and plant transpiration, with increases in the former outweighing reductions in the latter given reduced grass cover. This negative feedback weakens and eventually reverses sign over time during a transient simulation from the mid‐Holocene to present. A similar, but weaker, negative feedback is identified in Community Climate System Model Version 2 (CCSM2) over North Africa for the mid‐Holocene.