Intercomparisons of ECMWF ERA and TOGA data with observations for the 1993 Great Flood

Abstract

The 1993 summer rainfall over the Midwestern United States caused major flooding in the Upper Mississippi and Lower Missouri River basins. In this paper, a monthly comparative study of the analysis products from the European Centre for Medium‐Range Weather Forecasts (ECMWF) with observations is conducted for the period September 1992 to August 1993. The model products include the ECMWF analysis for the Tropical Ocean and Global Atmosphere (TOGA) project and ECMWF reanalysis (ERA) project. Overall, the large‐scale hydrological components (precipitation, evaporation, precipitable water, surface temperature, and relative humidity) appear partially improved from TOGA to ERA. Some internal inconsistencies exist in the assimilation of the ERA data from October 1992 to March 1993, in association with the dramatic prevalence of large‐scale moisture divergence at 0000 UT. The separation of moisture flux into meridional and zonal components indicates that the correct description of the meridional moisture flow is of crucial importance for simulation of anomalously wet conditions. While the spatially averaged monthly fields in ERA and TOGA are reasonable (in particular, the models maintain very good mean flow), the fine‐scale extremes are less accurate. The models do not reproduce observed mesoscale convective complexes due to limitations on resolution and the parameterization of precipitation processes. The new soil hydrology and boundary layer schemes of ERA work better in warm months but with no consistent improvement over TOGA. The terrestrial water budget in ERA is unrealistic, with accumulated model runoff being largely deviated from the observed streamflow. In addition to modeling internal water balance, the lack of the routing procedure and the low spatial resolution of surface hydrology are the other major reasons for this imbalance.