Abstract
Using an atmospheric general circulation model(JMA89), basinwide water balances of 27 major rivers were estimated and compared with actual observations. Two different deep cumulus convection schemes were used in the 10-year integrations from 1979 to 1988 that were forced by the observed global sea surface temperature boundary conditions. Results show that the Prognostic Arakawa-Shubert(PAS) scheme can simulate more realistic global distributions of annual mean precipitation than the Kuo scheme. In both cases, simulated annual mean precipitation, and simulated runoff coincides with the observations fairly well. The differences in precipitation or runoff between model simulations and observations are caused by the land surface hydrology and other parts of the model, and these two factors cannot be distinguished from each other. To check the reliability of the parameterizations in land surface hydrology in a simple biosphere(SiB) model, simulated runoff ratio(runoff/precipitation)is compared with the observations for each river basin. There are considerable difference between simulations and observations for runoff ratios. Differences in runoff ratios(observations minus simulations)for each river basin are significantly negative because when compared to the observations simulated runoff has been considerably overestimated. Simulated runoff has been greatly overestimated in most river basins where the observed runoff is less than 200 mm/year and the runoff ratio is less than 0.2. It is demonstrated that the parameterizations of land surface hydrology in the SiB model tend to overestimate the runoff or to underestimate the evapotranspiration in dry regions.
