Random Forest approach to air quality assessment for port emissions in a coastal area in the Gulf of Mexico.

This study analyzed the impact of shipping emissions on air quality in the Port of Veracruz using the Random Forest model for meteorological normalization. Air quality data for SO2, PM10, and PM2.5 (2017-2020) were processed with the rmweather package, evaluating different model configurations. The default setup (300 trees, 1000 predictions) provided optimal R2 and MSE. Meteorological variables such as temperature, wind speed/direction, and humidity significantly influenced pollutant concentrations. SO2 was strongly linked to temperature and wind direction, while PM10 and PM2.5 correlated with wind speed and humidity. Since the monitoring station captures multiple pollution sources, a broader emissions inventory was needed, but only shipping data was available. The study assessed SO2 emissions under different sulfur content scenarios, showing that the 3.5 % sulfur case overestimated concentrations, while 0.5 % and 0.05 % underestimated them. This aligns with the CALPUFF model, considering only shipping emissions.