The climate version of the GEOS 2 GCM at NASA Goddard Space Flight Center did not simulate a realistic 1988 summertime drought in the central United States. Despite several new upgrades to the model's parameterizations including cloud dynamics and microphysics (McRAS) and land-surface processes, hydrology, and snowmelt physics (HYSSiB), the model did not succeed in better simulating the summer 1988 drought. Even the model's horizontal grid spacing, changed from 4x5 degrees to 2x2.5 degrees, did not make much difference.

In an attempt to understand the influence of winds and soil moisture on the model's climate drift, the following investigation was performed. An ensemble set of four simulations with the GEOS 2 GCM started from 1 January to 31 August each for 1987 and 1988 served as control. Three companion sets of four simulations in which wind speed values (at all levels) and/or soil moisture values from best available data were inserted into the model during the simulations. The first set replaced the model's wind values every 6 hours with re-analysis wind-vector data from the NASA Goddard Data Assimilation Office. The second set of simulations replaced the model's soil moisture values every day with soil moisture values from an offline analysis of soil moisture produced with HYSSiB using ISLSCP Initiative I forcing data for 1987 and 1988. The third set replaced both wind-vectors and soil moisture values.

The results show that replacing the wind vectors had a strong effect on the resultant moisture convergence and precipitation in the mid-western U.S. The drought in June was well-simulated as compared to observations of precipitation (from the GPCP estimates). The low-level jet in 1988 was weak as a consequence of using analyzed winds. The control case produced too strong of a low-level jet that brought in too much moisture producing very similar precipitation for both years in this region. The case with replaced soil moisture values also simulated an improved 1988 drought, although to a lesser degree. The circulation wind pattern was somewhat improved; however, the low-level jet was still stronger than analyzed and able to bring in more moisture. The case with both re-analysis winds and soil moistures simulated the most realistic precipitation for 1987 and 1988 as compared to GPCP. This study shows that drifts in the wind patterns were largely responsible for the realism of simulated rainfall. Although physical parameterizations also affect climate drifts, eliminating wind drifts allows the divergent circulation to develop realistically.