Smooth Dynamical Downscaling of the Wind with the Model for Prediction Across Scales (MPAS)

Abstract

In the wind industry, wind time series for the past years are commonly generated using an atmospheric model to dynamically downscale large-scale reanalysis to local wind flow. Instead of relying on a nesting strategy like the Weather Research and Forecasting (WRF) model, the Model for Prediction Across Scales (MPAS) runs on variable resolution meshes that allow for smooth transitions. The goal of this study is to design MPAS meshes that are robust, accurate and computationally efficient for wind resource assessment applications. We have designed a benchmark validation of one-year simulations in wind-energy-relevant locations representing different geographies and flow complexity scenarios. To focus on identifying real differences between modeling wind time series using regional MPAS meshes compared to using WRF nested domains, both models share the same settings whenever possible and the post-processing is analogous for MPAS and WRF output. Besides, WRF simulations were run with and without nudging (the assimilation of reanalysis data on all points of the outer WRF domain), which is an option known to improve the accuracy of the results, and it is not implemented in MPAS yet. The real-data comparison shows that MPAS improves all wind speed metrics with respect to WRF simulations without nudging, but it has generally worse accuracy than the WRF simulations that do have nudging. This is a strong indication that the MPAS model indeed benefits from smooth transitions between scales, and that further developments on MPAS may turn it into a standard for wind resource applications.