A new multivariate stochastic model of hourly precipitation, air temperature, dewpoint, directional wind velocity components, and solar radiation has been developed, primarily for use in predicting wind erosion with wind erosion models such as the Wind Erosion Prediction System (WEPS) that predict surface soil moisture based on a physical model of soil water processes. Most previous stochastic models of wind have operated at the daily time scale or ignored correlations with other weather variables. A four-component Markov model of air temperature, dewpoint, and the two components of the wind velocity is forced by an hourly stochastic precipitation model that accounts for the diurnal cycle of the occurrence and depth of precipitation. Solar radiation is simulated as a third step; its simulation is conditioned on the output of the precipitation model and a Markov process for cloud cover. Parameters are computed separately for each month, and have been computed for three sites in the Great Plains of the United States. The effect on evapotranspiration, surface soil moisture, and wind erosion predictions of including the more complex aspects of the model, i.e., the auto- and cross-correlations between variables and the inclusion of the diurnal cycle, is being tested.