Optimisation technique for improving wind downscaling results by estimating roughness parameters

Abstract

The characterisation of the aerodynamic roughness length (z(0)) and the displacement height (d) is critical when modelling the wind field using the log vertical profile. It is known that the values of these parameters depend on land coverage and weather conditions. Thus, many authors have studied their relationship, providing typical values for each land cover. In this paper, we have performed a comprehensive literature review to collect the intervals of z(0) and d values for each land coverage. Using these intervals, we estimate their values using an optimisation technique that improves the results of a downscaling wind model. The downscaling model is a 3D adaptive, mass-consistent finite element model (Wind3D) that takes values from the HARMONIE-AROME or ECMWF mesoscale numerical weather prediction models. The optimisation is carried out by a memetic algorithm that combines the Differential Evolution method, a rebirth operator and the L-BFGS-B algorithm. The fitness function to be minimised is the root mean square error (RMSE) against observed wind data. This fast procedure allows updating the aerodynamic parameters for any weather condition. Numerical experiments have been carried out to show the performance of the methodology.