Post-process correction improves the accuracy of satellite PM_2.5 retrievals

<p>Estimates of PM<span class="inline-formula">_2.5</span> levels are crucial for monitoring air quality and studying the epidemiological impact of air quality on the population. Currently, the most precise measurements of PM<span class="inline-formula">_2.5</span> are obtained from ground stations, resulting in limited spatial coverage. In this study, we consider satellite-based PM<span class="inline-formula">_2.5</span> retrieval, which involves conversion of high-resolution satellite retrieval of aerosol optical depth (AOD) into high-resolution PM<span class="inline-formula">_2.5</span> retrieval. To improve the accuracy of the AOD-to-PM<span class="inline-formula">_2.5</span> conversion, we employ the machine-learning-based post-process correction to correct the AOD-to-PM conversion ratio derived from Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2) reanalysis model data. The post-process-correction approach utilizes a fusion and downscaling of satellite observation and retrieval data, MERRA-2 reanalysis data, various high-resolution geographical indicators, meteorological data, and ground station observations for learning a predictor for the approximation error in the AOD-to-PM<span class="inline-formula">_2.5</span> conversion ratio. The corrected conversion ratio is then applied to estimate PM<span class="inline-formula">_2.5</span> levels given the high-resolution satellite AOD retrieval data derived from Sentinel-3 observations. The region of study is central Europe during the year 2019. Our model produces PM<span class="inline-formula">_2.5</span> estimates with a spatial resolution of 100 m at satellite overpass times with <span class="inline-formula"><i>R</i>²</span> <span class="inline-formula">=</span> 0.55 and RMSE <span class="inline-formula">=</span> 6.2 <span class="inline-formula">µ</span>g m<span class="inline-formula">⁻³</span>. The corresponding metrics for monthly averages are <span class="inline-formula"><i>R</i>²</span> <span class="inline-formula">=</span> 0.72 and RMSE <span class="inline-formula">=</span> 3.7 <span class="inline-formula">µ</span>g m<span class="inline-formula">⁻³</span>. Additionally, we have incorporated an ensemble of neural networks to provide error envelopes for machine-learning-related uncertainty in the PM<span class="inline-formula">_2.5</span> estimates. The proposed approach can produce accurate high-resolution PM<span class="inline-formula">_2.5</span> data that can be very useful for air quality monitoring, emission regulation, and epidemiological studies.</p>