Abstract

Regional cooling resulting from increases in surface albedo has been identified in several studies as the main biogeophysical effect of past land use-induced land cover changes (LCC) on climate. However, the amplitude of this effect remains quite uncertain due to, among other factors, (a) uncertainties in the extent of historical LCC and, (b) differences in the way various models simulate surface albedo and more specifically its dependency on vegetation type and snow cover. We derived monthly albedo climatologies for croplands and four other land cover types from the Moderate Resolution Imaging Spectroradiometer (MODIS) satellite observations. We then reconstructed the changes in surface albedo between preindustrial times and present-day by combining these climatologies with the land cover maps of 1870 and 1992 used by seven land surface models (LSMs) in the context of the LUCID ("Land Use and Climate: identification of robust Impacts") intercomparison project. These reconstructions show surface albedo increases larger than 10% (absolute) in winter, and larger than 2% in summer between 1870 and 1992 over areas that experienced intense deforestation in the northern temperate regions. The historical surface albedo changes estimated with MODIS data were then compared to those simulated by the various climate models participating in LUCID. The inter-model mean albedo response to LCC shows a similar spatial and seasonal pattern to the one resulting from the MODIS-based reconstructions, that is, larger albedo increases in winter than in summer, driven by the presence of snow. However, individual models show significant differences between the simulated albedo changes and the corresponding reconstructions, despite the fact that land cover change maps are the same. Our analyses suggest that the primary reason for those discrepancies is how LSMs parameterize albedo. Another reason, of secondary importance, results from differences in their simulated snow extent. Our methodology is a useful tool not only to infer observations-based historical changes in land surface variables impacted by LCC, but also to point out deficiencies of the models. We therefore suggest that it could be more widely developed and used in conjunction with other tools in order to evaluate LSMs.