Evaluating the use of satellite observations to supplement ground-level air quality data in selected cities in low- and middle-income countries

Using data from ground-level measurements, this work evaluates the performance of one chemical transport model (CTM)-based approach using MERRA2 model output and one statistical approach using a generalized additive model to translate remotely-sensed Aerosol Optical Depth (AOD) measurements from the MODIS combined Deep Blue and Dark Target algorithm to surface-level PM₂.₅ concentrations for nine cities in low and middle-income countries that include a range of environmental conditions (e.g., mountainous, dusty, or coastal conditions) and only have between 1 and 10 ground level monitoring sites available. This evaluation shows that the CTM-based and statistical approaches tested here generally had a low correlation with the true daily-average PM₂.₅ values within these nine cities, in contrast to previous studies that had showed stronger correlations over other regions. In addition, the uncertainty in the satellite-based estimates of the daily-average PM₂.₅ concentration at a given location in a city tended to be very large (21–77% for the statistical methods, and 48–85% for the CTM-based methods). Many cities also had significant limitations in the availability of satellite observations of aerosols throughout the year due to their coastal location, persistent clouds, or persistent seasonal snow cover. The satellite-based methods tested in this work appear to work best for low altitude, inland cities like Hanoi and Delhi, but still have significant errors (43–60%) in predictions of daily-average PM₂.₅ concentrations at sites within these cities. The CTM-based satellite approach tested here tended to underestimate PM₂.₅ in high-altitude cities (except for Addis Ababa, Ethiopia, where they overestimate). However, this work suggests that under some conditions, adding satellite data to ground-level monitoring (GLM) network data via co-kriging may reduce the number of GLM sites needed to characterize PM₂.₅ concentrations within a city, but this needs to be determined on a case-by-case basis. This evaluation is then used to make recommendations to countries with sparse GLM networks on how to incorporate the use of satellite observations in their PM₂.₅ monitoring and under what conditions satellite approaches are likely to be unsuccessful.