A Comparison of Soil-Moisture Sensitivity in Two Global Climate Models

Abstract

Area-averaged surface hydrological processes from two global spectral general circulation climate models coupled to simple slab-ocean mixed layers are compared for the climates simulated with present-day (control) and increased atmospheric carbon dioxide (CO2). The models were developed at the National Center for Atmospheric Research (NCAR) and the Geophysical Fluid Dynamics Laboratory (GFDL). Both models use highly parameterized surface processes and the so-called ?bucket? (15-cm field capacity) soil-moisture method. For increased CO2 compared to the control, both models simulate more snowmelt in winter and early spring and less in late spring. more runoff in early spring and less in late spring, and increased precipitation, evaporation, and latent-heat flux, particularly in sprint However, the models differ in several key respects for the increased CO2 compared to the control. In spite of a qualitatively similar annual cycle of soil moisture, the NCAR model simulates soil-moisture amounts much less than saturation, while the GFDL model shows soil moisture near saturation in winter and spring. Therefore, the runoff is significantly less in the NCAR model than in the GFDL model. Accordingly, the increased winter and spring precipitation with increased CO2, (similar in both models) mostly runs off in the GFDL model, but is retained in the soil-moisture ?byckets? at most gridpoints in the NCAR model. The increased springtime evaporation rate (also similar in both models) results in a soil-moisture deficit in the GFDL model almost as soon as the warm season begins, while in the NCAR model the excess soil moisture retained in the ?buckets? during winter and spring must first be evaported before a soil-moisture deficit can occur. This delays or even prohibits a ?summer drying? which is noted in the GFDL model. In addition, a soil-moisture-cloud-precipitation feedback occurs which either prolongs the soil-moisture surplus well into the warm season (NCAR) or further accentuates the summer dryness (GFDL). The critical factor determining the magnitude of the summer drying is the soil-moisture amount in the control case, particularly during spring. The highly parameterized hydrology in the models, lack of appropriate observed data, and complexity of hydrological processes in the real world prohibit accurate verification and/or Calibration of the parameterizations in the models. Observed estimates of surface runoff based on streamflow suggest that the NCAR model is simulating too little soil moisture and the GFDL model too much. This would imply that, when both models use the highly parameterized bucket method, the NCAR model is underestimating and the GFDL overestimating summer dryness due to of an increase of CO2.