The objective of this subproject is to evaluate the ability of GCMs in simulating the global hydrologic cycle and to explore means of validation of GCM precipitation and hydrologic processes with space- and ground-based observations. We have completed the inter-comparison of the precipitation (P) evaporation (E) and surface hydrologic forcing (P-E) for 23 AMIP GCMs and relevant observations over a variety of spatial and temporal scales. These include global and hemispheric means, latitudinal profiles, selected areal means for the tropics and extratropics, ocean and land respectively. In addition, we have computed pattern correlations among models and observations for different seasons, harmonic analysis for annual and semi-annual cycles and rain-rate frequency distributions. We also compare the joint influence of temperature and precipitation on local climate using the Koeppen climate classification. A summary of the findings is as follows.
The global mean surface temperature and precipitation show large variation among models. The models have as much as 6-7°C differences in global mean temperature and up to about 2.00 mm/day differences in global precipitation. The global and hemispheric mean and the annual precipitation in the models are consistent to within approximately 20-30%. The differences among model precipitation estimates over the ocean are as large as those among climatological estimates of Jaeger, Legates and Willmott, and satellite estimates from the Microwave Sounding Unit (MSU). The discrepancies among models are largest over the land region. These may be due to the diverse land surface schemes used in the different models. All models seem to underestimate global rainfall variability compared with climatological estimates. Except for a few, most models maintain a global water balance (E-P) to within 5-10% of the total precipitation.
Another noteworthy result is in the intercomparison of ensemble mean
rainfall frequency distribution. A common feature shared by the models
is that they all under-estimate the precipitation in the light rain (0-1
mm/day) category. The ratio between the lowest rainrate and the next (1-2
mm/day) category for the model ensemble average is approximately 1 to 0.8.
In contrast, the same ratio from the observation is approximately 1 to
0.4. This discrepancy is most severe over land. In the higher rainrate,
the models are consistent with observation. The above findings imply that
while there is a reasonable degree of realism in the model parameterization
of convective rain, there may be a fundamental problem regarding the production
of light rain. This may be due to the improper treatment of shallow convection,
in particular boundary layer stratus in most GCMs. Comparing the global
rainfall distribution of the 23 AMIP GCMs and the climatological estimate
shows that the overall large-scale rainfall pattern is reasonably well
simulated. The largest difference between model and observation is in the
Asian Monsoon region during the boreal summer. In particular, most models
do not reproduce the rainband related to the Mei-yu over East Asia. Large
discrepancies are also found in the rainfall pattern over SPCZ and ITCZ
of the eastern Pacific. These are also the regions where the satellite
and ground-based rainfall estimates differ the most.