Objective: Reanalysis products can be an alternative to characterize air quality in regions where in situ monitoring is insufficient. Through advanced data assimilation methods, reanalysis combines data from atmospheric transport models with observations from around the world, both remote sensed or in situ. Before reanalysis data can be used to assess air quality at local and regional scales, its accuracy must be evaluated by comparing it with temporally and spatially co-located in situ data to assess correlations and biases. The objective of this project is to validate Copernicus Atmosphere Monitoring Service (CAMS) and Modern-Era Retrospective Analysis for Research and Applications (MERRA-2) reanalysis data near surface against in situ data collected in the Metropolitan Area of São Paulo (MASP) between 2015 and 2019, focusing on the pollutants ozone (O3), nitrogen dioxide (NO2) and particulate matter (PM10 and PM2.5).

Methods: Eight air quality monitoring stations operated by the Sao Paulo State Environmental Agency (CETESB) were selected to represent MASP in terms of air pollutant concentrations. The choice of ground based stations was based on data availability, location and spatial representativity, favoring background and regional air quality stations. Daily averages of air pollutant concentrations were calculated from the hourly in-situ and MERRA-2 data and from the 3-hourly CAMS data. For MERRA-2, PM10 and PM2.5 concentrations were calculated from organic and inorganic aerosol components, following the recommendations of the data provider. Six grid points from CAMS and six grid points from MERRA-2 intersecting the metropolitan area were selected, with spatial resolutions of 0.75°x 0.75° and 0.5°x 0.625°, respectively. In order to select the most representative CAMS and MERRA-2 grid points, correlation analysis between the MASP in situ data and the concentrations in each grid point was performed for each of the four air pollutants under analysis. Metrics to evaluate the performance of reanalysis data in representing in situ observations included: Pearson correlation coefficients, slopes and offsets, mean bias and standard deviations. Taylor diagrams were used to support the choice of the best CAMS and MERRA-2 grid points to represent air quality conditions in the MASP, allowing for simultaneous comparison of different metrics. Additionally, the interannual and the seasonal variability of reanalysis products were compared against in situ observations.

Findings: The best grid points to represent air quality conditions in the MASP was CAMS’ grid point #4, intersecting the MASP southwestern corner, and MERRA-2’s grid point #3, intersecting the central and western parts of the MASP. For the grid point selected, the Pearson coefficient ranged from 0.63 to 0.73 when comparing CAMS reanalysis and in situ data, depending on the air pollutant. CAMS-PM10 showed the highest correlation coefficients, followed by CAMS-PM2.5 , CAMS-NO2 and CAMS-O3. MERRA-2 showed lower correlation coefficients with in situ data, ranging from -0.21 to 0.31, since PM10 showed a negative correlation. Particularly, MERRA-2 failed to represent the temporal variability of PM2.5 and PM10 concentrations, possibly due to an overestimation of the sea salt component in the austral spring and summer. CAMS data overestimated the concentrations of O3, PM2.5 and PM10, respectively by 136, 50 and 16%. CAMS underestimated NO2 by 45%, suggesting that the magnitude of anthropogenic emission sources in the MASP could be underestimated in the CAMS’ atmospheric transport model. MERRA-2 overestimated O3 by 171% and underestimated PM2.5 and PM10 by 20 and 11%. The ratio between reanalysis and in situ data did not show a significant interannual variability in the study period (2015-2019), nor a seasonal variability in the case of CAMS. The O3 seasonal behavior was well represented by CAMS and MERRA-2 reanalysis data, with higher concentrations in the austral spring and summer, when there is typically a greater incidence of solar radiation at the surface in Sao Paulo. CAMS reproduced the expected seasonal variability for NO2, PM10 and PM2.5, with higher concentrations in the austral winter, when mild cold and dry conditions disfavor the dispersion of primary air pollutants. MERRA-2 failed to reproduce the seasonal behavior of PM in the MASP, showing greater concentrations in the austral spring and summer. The possible reasons are under investigation, with the analysis of the temporal variability of MERRA-2’s individual aerosol components. Overall, we conclude that CAMS reanalysis data was better suited to represent ground based air quality conditions in the MASP, although biases should be accounted for.

Keywords: Air quality,tropospheric ozone, particulate matter, CAMS, MERRA-2.