Sensitivity of a GCM Simulation to Inclusion of Inland Water Surfaces

A land surface model that includes a subgrid parameterization for inland water (lake, swamp, marsh) was coupled to a modified version of the NCAR CCM2. The coupled model was run for 5 yr with and without inland water subgrid points to determine the importance of inland water for global climate simulation. In July, the inclusion of these water bodies resulted in a spatially consistent signal in which high inland water regions were 2°–3°C cooler, had increased latent heat flux (10–45 W m−2), and decreased sensible heat flux (5–30 W m−2) compared to the simulation without these water bodies. These changes were statistically significant in the lake region of northwest Canada, the Great Lakes region of North America, the swamp and marsh region of the Siberian lowlands, and the lake region of East Africa, but were not significantly different in the swamp and marsh region of Finland and northwest Russia. The effect on Northern Hemisphere January air temperature was difficult to interpret due to large interannual variability. In tropical lake regions (East Africa), the response to lakes was less in the rainy season (January) than in the dry season (July). Precipitation was unchanged in both months except for the Great Lakes region where precipitation increased in January. These changes in temperature, precipitation, and surface fluxes were consistent with mesoscale modeling studies of the effects of lakes on climate and tended to bring the model closer to observations. In particular, the summer cooling in North America helped reduce a large warm temperature bias in the model, but did not eliminate the bias. The lakes had little effect on atmospheric moisture, radiation, or zonal circulation. These results show that subgrid-scale inland water bodies can be successfully added to global land surface models for use with GCMS.