Big data analyses for determining the spatio-temporal trends of air pollution due to wildfires in California using Google Earth Engine

Abstract

California has a long history of wildfires that release substantial amounts of smoke, contributing to regional air pollution. This study employs advancements in Earth observation satellites, cloud computing, and the Google Earth Engine to analyze MODIS MCD19A2 Version 6 level 2, Sentinel-5P NRTI AER AI, and Sentinel-5P NRTI NO2, evaluating the impact of wildfires from 2010 to 2022 on air quality in California. Historical fire events are cross-validated using the MODIS 1-km MYD14A1 V6 dataset, and the MODIS MAIAC-derived Aerosol Optical Depth (AOD) values are validated against AERONET AOD measurements from ground-based sun photometers. To assess the analysis's uncertainty, metrics such as Mean Absolute Error, Relative Mean Bias, and Root Mean Square Error are applied. Additionally, linear regression is used to verify the correlation between satellite and ground measurements. Results show that the average monthly MODIS MAIAC AOD at 470 nm and 550 nm tends to overestimate AOD by 24% and 6%, respectively, compared to AERONET AOD values. The correlation coefficient and adjusted R-squared value range from 0.78 to 0.80 and from 0.61 to 0.65, respectively, for AOD values at 470 nm and 550 nm. During the fire season, correlations exhibit closer associations, with a coefficient above 0.78 and an adjusted R-squared value above 0.63 are observed for both wavelengths. Sentinel-5P data reveal a significant increase in NO2 concentration in the surrounding areas during the 2020 and 2021 wildfires. This study identifies historical spatiotemporal trends of air pollution attributable to wildfires in California from 2010 to 2022, informing decision-making in the development of effective prevention and mitigation programs.