Impact of higher-resolved satellite-based land cover classificatio on near surface wind speed forecasts

Within the project "LandCover4Wind", we investigate the potential for improving 24 h mesoscale wind speed forecasts at altitudes of interest for wind energy applications by using different high-resolution satellitebased land cover maps such as the Global Land Cover Characterization (GLCC) from the US Geological The Advanced Weather Research and Forecasting (WRF) model version 3.9 was run at three different grid resolutions covering a wide range of weather conditions during July and November 2015. We evaluate results by tower measurements at 10, 40 and 98 m altitude for the rural station Falkenberg in the eastern part of Germany characterized by f at and open terrain. We test different conf gurations for model domain horizontal resolution, land cover classif cation (LCC) data, and land surface and boundary layer physics. We discuss the need of transforming native LCC spatial resolution to lower WRF domain grid resolutions and transforming different thematic resolutions in LCC datasets into the less-resolving WRF classes. In general, comparisons show better forecast performance for the stormy November period compared to hot-summer July conditions. Especially, correlations are signif cantly higher in November. All experiments show a positive wind speed bias which changes with altitude, time of day and model conf guration. With respect to LCC, CLC shows the smallest wind speed errors while GLCC performs worst. Two different land surface models NOAH LSM and 5-layer thermal diffusion scheme (5LD) are tested in combination with Mellor-Yamada-Janjic (MYJ) and Yonsei University (YSU) boundary layer modules. In general NOAH-MYJ performs best with CLC for November conditions, while it gives worst results with MODIS LCC for July conditions. For any given LCC, results depend signif cantly on the WRF land surface and boundary layer physics conf guration and the performance on a daily base varies considerably depending on the test metric and model conf guration. Results show only a weak dependency on domain resolution with increasing biases with increasing resolution. In summary, using a high resolution LCC dataset as the CLC proves to be valuable for wind speed forecasts.