We compared simulations of soil moisture by the AMIP climate models with observations taken in the former Soviet Union (FSU) for 1979-1985, the People's Republic of China for 1981-1988 (which extend to 1991 in our collection), and the state of Illinois in the United States for 1983-1988 (which extend to the present). The mean annual cycles and interannual variations were compared to plant available soil moisture in the upper 1 m of soil using data from more than 100 stations from the FSU, more than 100 from China, and 17 from Illinois.
In general, the model output was quite different from the data. The reasons for these discrepancies were investigated. Models with 15-cm field capacities did not capture the large high latitude values of soil moisture. This has implications for their ability to also properly simulate evapotranspiration, sensible heat flux, and runoff. The observed interannual variations of soil moisture in Russia were not captured by any of the AMIP models. This implies that interannual variations of precipitation in this region are not driven by global sea surface temperature variations. Interannual variations in China and the United States were also investigated. It was also found that several models have large soil moisture trends during the first year or two of the AMIP simulations, with potentially large impacts on global hydrological cycle trends and on other climate elements. This is because the simulations were begun without spinning up the soil moisture to the model climatology, and soil moisture has a long enough memory to influence climate for this time scale.
Humidity and moisture fluxes
We have also examined aspects of tropospheric humidity by comparing total column precipitable water (W) and zonally-averaged specific humidity (q) fields from AMIP models with various observations. Our best verification data for the AMIP period are radiosonde observations over the continent of North America. There we find that the majority of the models simulate the annual cycle, including its asymmetry, and the meridional profile of W reasonably well, although there are a few far outliers. Simulations of interannual variations in W are very diverse and generally quite poor.
To expand the study to tropical (10S-10N) and midlatitude (30N-50N) bands, we used ECMWF analyses as verification data. The AMIP models underestimate zonally averaged q below 500 hPa and slightly overestimate it above. Relative humidity simulations show a very broad range of values at all levels for both regions. The models capture the seasonal shift over the Northern Hemisphere oceans, where the maximum in W shifts eastward from summer to winter. However, the models appear to overestimate W substantially in the western Pacific during JJA.
On the global scale, we used analyses of radiosonde data by A. Oort to examine W and the net meridional flux of water vapor. Decadal- and global-mean values of W tend to be lower for the AMIP models than estimates from the observations, and differences of up to 15% exist among the model values. As over N. America, the seasonal cycle in global W appears to be generally well simulated, although there is a larger range of model values during northern summer than at other times. Interannual variations in the models' global W are marked by an ENSO signal (which is not prominent in the N. American simulations), but the reality of this signal is difficult to confirm with available observations. The simulated net poleward fluxes of water vapor across midlatitudes, inferred from the models' hydrological fields, appear to be too large, with possibly important consequences for the energetics of these models.
For most parameters, we find that a consensus (the mean or the median)
of the models agrees with the observations better than any individual model
does. We have also developed preliminary skill scores for quantitatively
aggregating our intercomparison diagnostic results for water vapor.