Air quality simulation with WRF-Chem over southeastern Brazil, part I: Model description and evaluation using ground-based and satellite data

Abstract

A comprehensive Weather Research and Forecasting with Chemistry (WRF-Chem) model evaluation is conducted using ground-based and total column observational data from air quality stations and satellite retrievals. Fine particles (PM2.5; ≤ 2.5 μm in aerodynamic diameter), nitrogen oxides (NOx, NO + NO2), carbon monoxide (CO), tropospheric ozone (O3) concentrations and AOD values over southeastern Brazil were analyzed to assess the model's capability in reproducing atmospheric observation. The model simulations were performed over simple one domain at grid resolution of 10 km over southeastern Brazil. This spatial resolution was chosen due to a previous evaluation between five MODIS AOD products with AERONET estimates, resulting in Dark Target at 10 km of spatial resolution the best product to represent the AOD values over our study domain. Model input emissions comprise vehicular emissions derived from a bottom-up emission model, as well as on-line calculations of biogenic and fire emission rates. Given that the atmospheric state affects air pollution dispersion, a model evaluation on the meteorological conditions was carried out to better evaluate the model performance in reproducing the pollutant concentrations. Good agreement between model simulations and observations for air temperature and relative humidity at 2 m height was found, with correlation coefficients higher than 0.85 in most periods. Expected benchmarks for wind speed and direction at 10 m height were also found in this analysis, though with larger uncertainties. Underestimation occurred for daily accumulated precipitation due to the limitations of the cloud microphysics scheme or cumulus parameterization. Model simulations of PM2.5, NOx, CO and O3 agreed well with ground-based observations in terms of temporal variations and trends, with model-observation discrepancies due to uncertainties in the emission inventories. O3 was the better simulated pollutant in terms of temporal variability, with the characteristic large and small amplitudes observed over urban and rural areas being well represented by the model. High O3 concentrations were observed at the Botucatu station, due to transport of pollutants generated in the Metropolitan Area of São Paulo, and were also represented by the model, indicating the need of more active air quality monitoring stations over inland regions in southeastern Brazil. Moderate and high correlation coefficients (ranging 0.46–0.81) were found for tropospheric NO2 VCD and CO column, and AOD at 550 nm due to uncertainties in the emission inventories and aerosol model simplifications. Both the model and satellite captured higher values in similar regions over our study domain. This work represents a first effort, in southeastern Brazil, that combines numerical modeling, remote sensing and ground-based stations to analyze and understand the impact exerted by the emissions of urban pollution over surrounding areas. A more in-depth analysis of the impact of emissions transport to inland regions from urban areas in southeastern Brazil will be discussed in the second part of this work.