Influence of Initial Conditions on Summer Precipitation Simulations over the United States and Mexico

Abstract

Abstract The impact of initial conditions on summer precipitation over North America for July–September was examined by comparing long-term simulations of the Atmospheric Model Intercomparison Project (AMIP) runs with the ensemble simulations (SIMUs) initialized at the end of June each year. Both types of simulations use the observed SSTs as boundary conditions, and hence, the differences are a result of the initial conditions. Experiments were performed using the NCEP Global Forecast System (GFS) T126L28 model with 28 vertical levels. Over the monsoon core region, the model has the correct relationship between evaporation (E) and soil moisture and the T126 model has enough horizontal resolution to simulate the moisture transport from the Gulf of California. Over the Great Plains, the model has dry and warm biases. These biases are larger in the AMIP runs in comparison with the SIMU. Two major model errors contribute to the biases: 1) the deep soil layer (10–200 cm) is far too dry and does not have enough variability; and 2) the model does not have the correct relationship between E and soil moisture at the top level 10 cm (SM10cm). The linear E–SM10cm relationship in the model has a large slope in spring and early summer. Evaporation tends to drop too sharply and too quickly during the model dry periods; it does increase during rainfall periods, but the model still has an overall deficit in E. The SIMU runs have better estimated E values supplied by the initial conditions and the surface conditions deteriorate slow enough to have better seasonal mean summer E and precipitation than the AMIP runs. The results here may be model dependent. For the GFS model, the initial conditions supply better estimates of soil moisture and surface fluxes at the beginning of a simulation, which compensates for model errors.