Constrained Stochastic Climate Simulation

Abstract

A stochastic, multivariate, hydrometeorological data generation algorithm is presented. Hourly values of precipitation, cloud cover, shortwave radiation, longwave radiation, temperature, dewpoint, wind speed, and wind direction are jointly generated for the two-meter level. The procedure is designed to provide coherent sets of input data for models of various land surface processes. The model's flexibility and economy allow the study of land surface responses to different atmospheric forcings. Generated data plots, model output statistics, and generated mean diurnal curves are compared to observations for the months of January and July at two sites, Boston, Massachusetts and Dodge City, Kansas. Data representing three "climates", normal, wet, and temperature-biased were generated and applied to a detailed model of the land surface. The resulting energy fluxes across the land-atmosphere interface are reviewed and the differences are noted.