A New Diffusion Scheme for Numerical Models Based on Full Irreversibility

Abstract

In this work, the forecast accuracy of a numerical weather prediction model is improved by emulating physical dissipation as suggested by the second law of thermodynamics, which controls the irreversible evolutionary direction of a many-body system like the atmosphere. The ability of the new physics-based scheme to improve model accuracy is demonstrated via the case of the one-dimensional viscous Burgers equation and the one-dimensional diffusion equation, as well as a series of numerical simulations of the well-known 1998 successive torrential rains along the Yangtze River valley and 365 continuous 24-h simulations during 2005?06 with decreased root-mean-square errors and improved forecasts in all of the simulations.