Two-Dimensional Simulations of Drainage Winds and Diffusion Compared to Observations

Abstract

A vertically integrated dynamical drainage flow model is developed from conservation equations for momentum and mass in a terrain-following coordinate system. Wind fields from the dynamical model drive a Monte Carlo transport and diffusion model. The model needs only topographic data, an Eulerian or Lagrangian time scale and a surface drag coefficient for input data, and can be started with a motionless atmosphere. Model wind and diffusion predictions are compared to observations from the rugged Geysers, California area. Model winds generally agree with observed surface winds, and in some cases may give better estimates of area-averaged flow than point observations. Tracer gas concentration contours agree qualitatively with observed contours, and point predictions of maximum concentrations were correctly predicted to within factors of 2 to 10. Standard statistical tests of model skill showed that the accuracy of the predictions varied significantly from canyon to canyon in the Geysers area. Model wind predictions are also compared to observations from the South Carolina Savannah River Plant, which has gently rolling terrain. The model correctly simulated the slower development of drainage winds and slower deepening of the drainage layer in the Savannah River Valley, relative to the Geysers, California simulations. The South Carolina simulations and observations suggest that drainage winds are more frequent in the southeastern United States than is generally recognized. They may be responsible for some of the errors in air pollution concentration predictions made by Gaussian models, which assume homogeneous winds and turbulence.