Modeling the potential contribution of land cover changes to the late twentieth century Sahel drought using a regional climate model: impact of lateral boundary conditions

This paper investigates the potential impact of idealized-but-realistic land cover degradation on the late twentieth century Sahel drought using a regional climate model (RCM) driven with lateral boundary conditions (LBCs) from three different sources, including one re-analysis data and two global climate models (GCMs). The impact of land cover degradation is quantified based on a large number of control-and-experiment pairs of simulations, where the experiment features a degraded land cover relative to the control. Two different approaches of experimental design are tested: in the 1st approach, the RCM land cover degradation experiment shares the same LBCs as the corresponding RCM control, which can be derived from either reanalysis data or a GCM; with the 2nd approach, the LBCs for the RCM control are derived from a GCM control, and the LBCs for the RCM land cover degradation experiment are derived from a corresponding GCM land cover degradation experiment. When the 1st approach is used, results from the RCM driven with the three different sources of LBCs are generally consistent with each other, indicating robustness of the model response against LBCs; when the 2nd approach is used, the RCM results show strong sensitivity to the source of LBCs and the response in the RCM is dominated by the response of the driving GCMs. The spatiotemporal pattern of the precipitation response to land cover degradation as simulated by RCM using the 1st approach closely resembles that of the observed historical changes, while results from the GCMs and the RCM using the 2nd approach bear less similarity to observations. Compared with the 1st approach, the 2nd approach has the advantage of capturing the impact on large scale circulation, but has the disadvantage of being influenced by the GCMs' internal variability and any potential erroneous response of the driving GCMs to land degradation. The 2nd approach therefore requires a large ensemble to reduce the uncertainties derived from the driving GCMs. All RCM experiments based on the 1st approach produce a predominantly dry signal in West Africa throughout the year, with a dipole pattern found in the peak monsoon season that features a slight increase of precipitation over the Guinea Coast and strong decrease in the north; a similar spatiotemporal distribution is found for temperature changes, with warming (cooling) coinciding with precipitation decrease (increase). The model precipitation changes in West Africa are dominated by evapotranspiration changes in the north and by atmospheric moisture convergence changes in the south; in temperature changes, surface warming due to the decrease of evaporative cooling dominates over the albedo-induced radiative cooling.