Evaluating Error Propagation in Coupled Land–Atmosphere Models

Abstract

his study examines how land-use errors from the Land Transformation Model (LTM) propagate through to climate as simulated by the Regional Atmospheric Model System (RAMS). The authors conducted five simulations of regional climate over East Africa: one using observed land cover/land use (LULC) and four utilizing LTM-derived LULC. The study examined how quantifiable errors generated by the LTM impact typical land?climate variables: precipitation, land surface temperature, air temperature, soil moisture, and latent heat flux. Error propagation was not evident when domain averages for the land?climate variables of the yearlong simulation were examined. However, the authors found that spatial errors from the LTM propagate through in complex ways, temporally affecting the seasonal distributions of rainfall, surface temperature, soil moisture, and latent heat flux. In particular, rainy seasons exhibited greater precipitation in LTM-RAMS simulations than in the reference simulation and less precipitation occurred during the dry season. Complex interactions of precipitation and soil moisture were also evident. Overall, results indicate that small errors from a land change model could grow as a ?coupling drift? if both are used to forecast into the future; these couplings could create larger combined errors of land?climate interactions because of time-scale differences into the future. Thus, although land-use change projection is necessary for a more accurate climate projection, existing errors from a land change model will likely amplify through the climate simulation. This finding affects interpretation of large-scale versus land-use/land-cover feedbacks on climate projections.