Accuracy Assessment of WRF Model in the Context of Air Quality Modeling in Complex Terrain

Abstract

Output data from the Weather Research and Forecasting (WRF) model are frequently used in air quality modeling for scientific, practical and regulatory purposes. Therefore, it is crucial to determine whether the accuracy of WRF predictions is suitable for application in air quality models on a local scale (<50 km) and in complex terrain. The presented research is unique because, to assess the accuracy of the WRF model, data from experimental data sets for the assessment of air quality models were used, which contained information about the actual conditions of selected meteorological parameters along the vertical profile of the atmosphere. The aim of the study was to conduct an evaluation of the WRF model using data derived from three field experiments designated to conduct air quality model evaluation studies for models such as AERMOD, ADMS or CALPUFF. Accuracy evaluation was carried out in relation to the grid resolution, station location (on-site and weather airport) and vertical profile of the atmosphere. Obtained results of the evaluation for temperature, wind speed and direction were analyzed with regard to the possibilities of application in air quality modeling systems. It was stated that the use of a grid with a resolution of 1 km generally resulted in statistically significantly lower values of errors for wind speed compared to a 4 km resolution. The outcomes of simulations for temperature and wind speed were sensitive with regard to the location. In on-site locations (complex terrain) significantly higher values of prediction errors (MB, MGE, RMSE) were obtained compared to the standard weather station locations (airport). In addition, wind speed predictions in on-site locations were generally biased (overestimated). Along the vertical profile of the atmosphere, up to the altitude of 100 m a.g.l., statistically significantly different outcomes of accuracy evaluation were achieved for wind speed and direction. Considering the above, caution should be exercised when using data from meteorological simulations in air quality modeling.