Influence of resolution and parameterization of the WRF model on the verifiability of weather forecast

Numerical weather prediction (NWP) models are frequently used tools in operational weather forecasting. The NWP bases on current weather observations and processing of this data using computational models to forecast possible weather conditions. The aim of the study was to determine the optimal configuration of the Weather Research and Forecasting (WRF) model , version 4.2 (Skamarock et al. 2008), for more effective weather forecasting for the area of Poland. For model evaluation, we used observations from the IMWM-NRI network (above 50 meteorological stations). Numerical simulations were run using GFS model data was obtained from NOAA's NCEP servers. The WRF model was configured for a 3 km horizontal resolution grid, using unique parameterization settings for this model. Validation of forecast data was performed using statistical measures recommended by the WMO, e.g. mean error, mean absolute error, mean squared error, showing the values of forecast error. In this study, the model settings were configured based of other papers for Europe (Stergiou et al. 2017, Mooney et al. 2013, Kioutsioukis et al. 2016, Garcia-Diez et al. 2015, Carvalho et al. 2014, Santos Alamillos 2013), especially from its central part (Wałaszek et al. 2014, Kryza et al. 2017). The results of the work present statistical summaries of optimal model parameterization schemes, depending on their verifiability. Model configuration characterized by the best performance will be further examined over a longer time period (in the study, the average MAE for air temperature was 0.8°C). The research was funded by National Science Center (project number: 2017/27/N/ST10/00565)