Abstract: In Latin America, limited PM2.5 monitoring stations hinder air quality assessment and understanding health impacts. This study utilizes the Aerosol Optical Depth (AOD) measured by satellite platforms, particularly focusing on the MAIAC product with its high spatial resolution (1 km), often used as a PM proxy. Our aim is to evaluate the uncertainty of the MAIAC 1km product in six densely populated cities: Sao Paulo, Santiago, Buenos Aires, Medellín, La Paz, and México, over 2015-2022. We employed typical global metrics (R2, RMSE, bias) and the detailed Relative Expanded Uncertainty (REU) for a comprehensive assessment. Unlike common metrics, REU provides insights into the error structure across the AOD’s dynamic range. Our findings indicate relatively low AOD values (< 0.2) with moderate correlation (R < 0.6) with AERONET sun photometers. In the AOD range of 0.05 to 0.15, the relative uncertainty exceeded 50% for all locations. Additionally, we compared MAIAC with the MODIS DT 3km product, finding consistently lower uncertainties (REU~20-30%) for MODIS in the same dynamic range. Despite MAIAC showing slightly better RMSE and R2 (~0.08 and ~0.55) than MODIS (~0.1 and ~0.49), regression analysis revealed a greater bias in MAIAC. We conclude that at lower aerosol loads, the MAIAC algorithm exhibits higher uncertainty and bias compared to MODIS DT, highlighting the importance of considering uncertainties when using AOD products for PM estimation and spatial PM2.5 distribution characterization.

Keywords: MAIAC 1km, AOD, Uncertainty, Latin American cities