WRF evaluation exercise using open sea in situ measurements

Abstract

The quality of the prediction of the atmospheric transport and dispersion of toxic, hazardous materials and pollutants strongly depends on the atmospheric forcing and in particular on the wind field. Especially in coastal regions, another important factor for correctly predicting the pollutant levels is the precise quantification of the sea aerosol particles formed predominantly by the action of the wind on the ocean surface. Furthermore, acting as a cloud condensation nuclei to form cloud drops, exchanging gases with the atmosphere and engaging other reactions, scattering lights and exchange moisture with the atmosphere, sea aerosol plays a fundamental role in the Earth's weather and climate regulation. Despite its importance, no continuous and long term monitoring of sea aerosol emissions is available and the measurements performed during dedicated campaigns are extremely scarce and spotted. As a consequence, sea aerosol trends are mainly predicted by simulations whose skill and precision rely on the atmospheric forcing fields used to drive models. The "Research on Energy Systems" (RSE S.p.A.) implemented the model system WRF-SMOKE-CAMx in order to simulate both concentration and deposition of pollutants in Italy. In particular, the WRF model was used for modeling the meteorological fields, the SMOKE model was selected to process the emission inventory and the CAMx was selected as the pollutant dispersion and chemical model to simulate different scenarios. The paper addresses an analysis of the performance of two WRF model versions (WRF-ARW 3.0 and WRF-ARW 3.2.1) in the Ligurian sea area carried out by comparing model outputs against observations at five different monitoring sites, four along the coast and one in the open ocean. The new parameterizations introduced in the 3.2.1 version of the WRF model contribute to improve the overall model performance although, for both the model releases, reproduction of the wind field shows criticisms due to the particularly complex orography of the Ligurian basin area. Nonetheless, the analysis evidences the feasibility of using the WRF model outputs as input meteorological fields to model the sea aerosol emissions.

Evaluation of modeled wind field for dispersion modeling

“…Our results, as well as previous results reporting better agreement with observations of the MM5 using 12 km resolution over a finer resolution (URS Corporation, 2008;Zhang et al, 2011) and differences between models in prediction of the surface wind direction (Balzarini et al, 2012), reveal that a detailed and careful assessment is required when applying meteorological models to study pollutant dispersion as part of environmental impact surveys (EIS) or for regulatory purposes. Table 4 Comparison over the whole year of 2011 of surface wind predictions of the MM5 and WRF meteorological models with observations.…”

“…In contrast, overestimated annual average wind speed is obtained for the complex terrain conditions. These results are in agreement with the findings of Balzarini et al (2012), who reported no systematic tendency of WRF towards underestimation or overestimation of the surface wind speed. Our results do not support, however, Liu and Warner (2004)'s conclusions that MM5 predictions of surface wind speed and direction are more accurate than those of WRF.…”

“…whereas other studies (e.g. Balzarini et al, 2012) found no systematic tendency of WRF towards Atmospheric Research 166 (2015) 150-156 underestimation or overestimation of the surface wind speed. Moreover, to our best knowledge, a quantitative evaluation of the predicted wind direction over a long period has never been reported.…”

“…Interestingly, MM5 performed better when using the 36 km or the 12 km grid resolutions than with the 4 km grid (URS Corporation, 2008;Zhang et al, 2011). Likewise, running WRF simulations with a 5 km grid resolution did not improve, in general, the model performance when compared to using a 15 km grid (Balzarini et al, 2012). Yet, most studies that examined different grids (from 36 km to 3 km) did not evaluate systematically the accuracy of the results vs. observations (Durante et al, 2012;NPS-CIRA, 2004), and in other cases the comparisons were inconclusive.…”

“…for estimating pollutant dispersion, the wind data are obtained by defining a pseudometeorological station at the location where the meteorological records are missing. Several evaluation studies of the MM5 and WRF mesoscale meteorological models against point meteorological measurements have been reported (Durante et al, 2012;Balzarini et al, 2012;Carvalho et al, 2006;NPS-CIRA, 2004). Specifically, Liu and Warner (2004) reported that WRF forecasts the wind speed and direction in the upper troposphere better than MM5 but that MM5 generates more accurate forecasts at the surface and the lower troposphere.…”