Parameter estimation in a diagnostic wind model over complex terrain

Abstract

The aerodynamic roughness length (z0) and the displacement height (d) are critical for wind modelling based on the log vertical profile. It is well known that the values of these parameters depend on weather conditions and land coverage. Thus, many authors have studied its relationship, providing typical ranges for each land coverage. In this work, a comprehensive literature review is performed to collect the ranges of z0 and d for each surface type. In particular, we have focused on the coverages present in the Information System of Land Cover of Spain (SIOSE) [1]. Using these ranges, we propose a procedure to construct z0 and d maps through a downscaling wind model. Results from the HARMONIE-AROME and ECMWF mesoscale numerical weather prediction models are downscaled using a 3D diagnostic wind model with adaptive finite element method [2, 3]. The values of z0 and d are estimated with a memetic algorithm that combines the Differential Evolution method [4], a rebirth operator and the L-BFGS-B algorithm [5]. So, the root mean square error (RMSE) of the wind model is minimised against the observed wind data. This fast procedure allows updating the roughness parameters for any weather condition. Some numerical experiments are presented to show the performance of this methodology. Although we work with the SIOSE database and the Wind3D model, the method can be used in conjunction with other databases and downscaling models.