Analysis of WRF Model Wind Estimate Sensitivity to Physics Parameterization Choice and Terrain Representation in Andalusia (Southern Spain)

Abstract

This paper reports on an evaluation of the relative roles of choice of parameterization scheme and terrain representation in the Weather Research and Forecasting (WRF) mesoscale model, in the context of a regional wind resource assessment. As a first step, 32 configurations using two different schemes for microphysics, cumulus, planetary boundary layer (PBL), or shortwave and longwave radiation were evaluated. In a second step, wind estimates that were obtained from various experiments with different spatial resolution (1, 3, and 9 km) were assessed. Estimates were tested against data from four stations, located in southern Spain, that provided hourly wind speed and direction data at 40 m above ground level. Results from the first analysis showed that wind speed standard deviation (STD) and bias values were mainly sensitive to the PBL parameterization selection, with STD differences up to 10% and bias differences between −15% and 10%. The second analysis showed a weak influence of spatial resolution on the STD values. On the other hand, the bias was found to be highly sensitive to model spatial resolution. The sign of the bias depended on terrain morphology and the spatial resolution, but absolute values tended to be much higher with coarser spatial resolution. Physical configuration was found to have little impact on wind direction distribution estimates. In addition, these estimates proved to be more sensitive to the ability of WRF to represent the terrain morphology around the station than to the model spatial resolution itself.

Sensitivity of the WRF model to PBL parametrisations and nesting techniques: evaluation of wind storms over complex terrain

“…Hence, the number of vertical levels has been set to 40 in every domain. This number is similar to the number in the recent literature, which ranges between 30 and 46 (Miguez-Macho et al, 2004;Lo et al, 2008;Kwun et al, 2009;Santos-Alamillos et al, 2013;Etienne et al, 2013). The vertical resolution ensures that several eta levels lie below the PBL height at any time.…”

“…It employs an Eulerian mass-coordinate solver with a nonhydrostatic approach, and a terrain-following eta-coordinate system in the vertical. It is a state-of-the-art mesoscale model used in a variety of studies also for hindcast simulations (Kwun et al, 2009;Jiménez et al, 2010Awan et al, 2011;García-Díez et al, 2013;Jiménez and Dudhia, 2013;Santos-Alamillos et al, 2013;Menendez et al, 2014, among others). A first decision in regional climate modelling concerns the selection of the domain to be simulated.…”

“…The PBL plays a major role in simulating surface winds (Stensrud, 2007;Kwun et al, 2009;Santos-Alamillos et al, 2013;Menendez et al, 2014). Nowadays, there are many different approximations to account for the relevant subgrid processes that lead to different PBL formations.…”

“…More recently, García-Díez et al (2013) focused on the role of different PBL schemes. Other studies have investigated the role of the PBL schemes in the ability to simulate the surface wind of typhoons (Kwun et al, 2009) along the coast of the Mediterranean Sea (Menendez et al, 2014) or in southern Spain (Santos-Alamillos et al, 2013). Recently, García-Díez et al (2015) investigated the added value of dynamical downscaling compared to the driving data set, and found that over the Iberian Peninsula, 9 km resolved simulations are barely able to add value compared to the driving data set.…”

Australian National Electricity Market Model - Version 1.10

“…WRF physics scheme configurations were chosen based on those found optimal for wind modelling (Deppe, Gallus & Takle 2012;Santos-Alamillos et al 2013;Yang et al 2013;Zhang, Pu & Zhang 2013) and suitable for Australian environmental conditions (Evans, Ekström & Ji 2012). The main physics scheme selections being: Each month model run included a two day prior spin-up time to equalise model dynamics and a model run time-step of 45 seconds was set.…”