Investigating Two-dimensional Horizontal Mesh Grid Effects on the Eulerian Atmospheric Transport Model Using Artificial Neural Network

Abstract

The complexity of monitoring is compounded by the environmental and health impacts linked to air pollution. The elevated expenses and intricate execution involved in measurements prompt the integration of modeling as a complementary approach alongside monitoring and surveillance efforts. Transport chemistry models like CHIMERE operate deterministically, utilizing meteorological factors, emissions data, boundary conditions, and various physical processes such as transport and Horizontal Mesh-Grid to influence inputs and outputs. The findings are validated using monitoring data over different periods of 2010, 2016, and 2021 and compared with results from prior research. The initial aspect reveals: (1) Enhanced resolution increases the probability of accurate forecasts, particularly for ozone, with PM10 displaying less distinct patterns. (2) Spatial resolution has minimal impact on temperature and wind speed. (3) Planetary Boundary Layer Height (PBLH) exhibits higher sensitivity, influencing Land Use and Land Cover (LULC), primarily due to emissions, advocating for higher resolution. The second aspect demonstrates: (4) CHIMERE-Artificial Neural Network (CHIMERE-ANN) demonstrates high accuracy in predicting ozone levels for Agadir and Casablanca, achieving improved correlation coefficients of 80% and 94%, respectively, accompanied by a notable decrease in Root Mean Square Error (RMSE) to 7.5 µg m–3 and 7.4 µg m–3. (5) Implementing CHIMERE-ANN with high spatial resolution concentrations (RA3 and RC3) enhances the accuracy of pollutant concentration forecasts. The proposed model enables rapid and detailed simulation of air pollution scenarios alongside flexibility for continuous updates.