Results of an intercomparison study under the Atmospheric Model Intercomparison Project (AMIP) to assess the abilities of 29 global climate models (GCMs) in simulating various aspects of regional and hydrologic processes in response to observed sea surface temperature and sea ice boundary forcings are presented. The authors find that the models generally portray an earthlike climate to approximately 10%–20% of the global land surface temperature (= 14.8°C) and global precipitation (= 2.3 mm day−1) While a majority of the models have a reasonable global water budget, about a quarter of the models show significant errors in the total global water balance.
While the model frequency distributions of heavy precipitation associated with deep convection are in reasonable agreement with observations, a systematic underestimate of the frequency of occurrence of light precipitation events (< 1 mm day−1) is present in almost all the AMIP models, especially over continental desert regions and over tropical and subtropical oceanic regions contiguous to the west coasts of continents where low-level stratocumulus clouds tend to occur. This discrepancy is presumably related to the crude treatment of moist processes, especially those related to low clouds and nonconvective precipitation in the models. Another common problem in the global rainfall distribution is the presence of spectral rain or spurious gridpoint-scale heavy rain. The artificial anchoring of rainfall to topographic features in the Maritime Continent appears to be a generic problem in many GCMs. Models differ substantially in the magnitude of the rainfall amount over the eastern Pacific ITCZ for all seasons. The simulated boreal summer rainfall distributions have large variability over the Indian subcontinent and the Bay of Bengal. The northward migration of the monsoon convective zones are not well simulated. In particular, the East Asian monsoon rainband over the subtropical western Pacific is ill-defined or absent in all models.
On the interannual timescale, the models show reasonable skills in simulating the fluctuations of the Southern Oscillation and the eastward migration of the major equatorial precipitation zone during ENSO. Most models show useful rainfall prediction skill in the Tropics associated with ENSO-related SST forcing. However, the models do not show any useful skill for extratropical rainfall prediction from specified anomalous global SST forcing. Overall, the models depict a reasonably realistic annual cycle of water balance over regions where long-term local moisture balance is maintained—that is, (P–E) ≈ 0—over large interior land regions in the extratropics. In regions of strong dynamic control—that is, (P–E) >>0—such as the tropical western Pacific, monsoon regions, and the ITCZ, the intermodel variability is very large.
The simulated water balance over large river basins has been validated against hydrographic river discharge data using a river-routing model. Results show that while the model ensemble mean runoffs are consistent with the climatological observed river discharge for the Amazon and Mississippi, the intermodel variability is substantial. The models yield even more divergent results over other world river basins. These results suggest that while some GCMs may have moderate capability in capturing some aspects of the climatological variation of runoff, it is premature to use them for climate studies related to continental-scale water balance. A ranking of the AMIP models and some possible implications based on the above performance are also presented.
*Additional affiliation: Applied Research Corporation, Landover, Maryland.