Satellite-Based Application For Air Quality Monitoring and Management at National Scale (SAANS)

Why PM2.5?

PM2.5 is fine particulate matter of diameter smaller than 2.5 µm. To put into perspective, average human hair is 70 µm thick. Some of the particles are emitted directly from a source, while some are formed by complex chemical reactions in the atmosphere. Owing to their fine sizes, they tend to stay in the air longer. When we inhale air, these fine particles penetrate deep into our lungs and may even enter circulatory system causing various diseases. In India, PM2.5 is the main criteria pollutant that needs continuous monitoring. According to Disease Burden India study led by the Indian Council of Medical Research and Public Health Foundation of India, 670,000 (95% UI 550,000-790,000) premature deaths are attributable to ambient PM2.5 exposure in India in 2017 [Balakrishnan et al., 2019].


Why do we need satellites to monitor air pollution in India?

In India, PM2.5 monitoring started in 2008-2009 by Central Pollution Control Board; however, the network mainly expanded in the recent years (Fig 1). Currently, 230+ continuous monitoring sites are operational in the country that monitors PM2.5. In addition, 700+ manual monitoring sites are operational. However, these sites are confined to the urban centres and almost 45% of the districts in India do not have any monitoring sites.

Satellites can fill this spatial gap in sampling. Since the satellites have been measuring particulate matter from the year 2000, they provide continuous, long-term and consistent data to examine the spatial as well as temporal pattern in air pollution. This allows us to monitor air pollution at national scale, track the background pollution level, differentiate urban vs. rural areas, and generate pollution data at various administrative levels (e.g. state, district, municipal corporation). Using satellite data, air pollution can be mapped at very high resolution. This enables us to identify local hotspots and guide the expansion of ground-based network, especially in the regions without any prior information. To summarize, satellites provide valuable information complimentary to ground-based monitoring and are a great tool for air quality management in country like India where the traditional monitoring is not adequate. Such hybrid approach is need of the hour in India [Brauer et al., 2019].


How do satellites retrieve PM2.5?

Satellites retrieve a parameter called aerosol optical depth (AOD). AOD represents columnar concentration of particulate matters in terms of light extinction. Over the years, satellite-derived PM2.5 estimation has evolved. In the early days, PM2.5/AOD ratio (or the scaling factor) was estimated for the grids that has a ground-based site. This ratio was then modelled using additional information of meteorology. The algorithm got matured over the years. The scaling factors can be derived from chemical transport model constrained by observed vertical distribution and AOD [see Dey et al., 2012; Chowdhury et al., 2019 for more details] or from reanalysis data (in this project). The scaling factors that vary spatially and temporally are used to convert columnar AOD to surface PM2.5.


Resolution Sensor R2 Application

≥10 km

MODIS, MISR 0.75-0.78 District level
1 km MODIS MAIAC 0.88 City level

Validation of Satellite AOD


modis

Figure 1. AERONET vs MODIS AOD


merra2

Figure 2. MERRA-2 vs AERONET AOD


daily2

Figure 3. 24-hour Daily Validation


annual2

Figure 4. 24-hour Annual Validation


Algorithm of this project

In this project, we upgraded the research algorithm discussed in the literature (Dey et al., 2012; Chowdhury et al., 2019). The entire flow chart is shown in Figure 1. We derived the scaling factor using MERRA-2 reanalysis data that is calibrated against measurements from the CPCB continuous monitoring sites. We did not use sites that only measure PM10 to avoid any additional uncertainty due to conversion of PM10 to PM2.5. The grids where no ground-based monitoring sites are available, we used a nearest neighbour algorithm to adjust the bias. We processed 20+ years of AOD data from MODIS-MAIAC at 1 km resolution (>3 Tb of data). The variance across the swaths is minimised using the Savitzky-Golay filter. The MODIS and AERONET AOD show very good correlation (Figure 2), so does the MERRA-2 and AERONET AOD (Figure 3). MERRA-2 scaling factors are interpolated to match the satellite resolution and AOD is converted to PM2.5 for every satellite overpass from Feb 26, 2000 onwards. The instantaneous satellite-derived PM2.5 is calibrated using a percentile based regression approach. Since the satellites cross the Indian region between 10 am to 2 pm, we further convert the instantaneous PM2.5 to 24-hr average using diurnal scaling factors from MERRA-2. Our final product is daily (24-hr average) and annual PM2.5 products. These are validated (see the regression statistics in Figures 4 and 5). We report <5% error in satellite-derived estimates for PM2.5 up to 200 µgm-3. The median error remains <10% for PM2.5 up to 500µgm-3 , beyond which the error further increases.


flowchart1

Figure 1. The flow of the entire process with the MERRA-2 calibration steps shown in the blue colored font,satellite PM2.5 evaluation in red colored font and generation of the final products that is disseminated trough SAANS portal in bold black font.


References

Balakrishnan, K., S. Dey et al. (2019), The impact of air pollution on deaths, disease burden, and life expectancy across the states of India: the Global Burden of Disease study 2017, The Lancet Planetary Health, Vol. 3(1), 26-39.

Chowdhury, S. (2019), Ambient PM2.5 exposure and associated premature mortality burden for India in present and future climate, PhD Thesis, IIT Delhi.

Chowdhury, S., S. Dey, L. Di Girolamo, K. R. Smith, A. Pillarisetti and A. Lyapustin (2019), Tracking ambient PM2.5 build-up in Delhi national capital region during the dry season over 15 years using a high-resolution (1 km) satellite aerosol dataset, Atmospheric Environment, Vol. 204, 142-150.

Dey, S., L. Di Girolamo, A. van Donkelaar, S. N. Tripathi, T. Gupta and M. Mohan (2012), Decadal exposure to fine particulate matter (PM2.5) in the Indian subcontinent using remote sensing data, Remote Sensing of Environment, Vol. 127, 153-161.

Brauer, M., S. Guttikunda, K. A. Nishad, S. Dey, S. N. Tripathi, C. Weagle, R. V. Martin, “Examination of monitoring approaches for ambient air pollution: A case study for India article", Atmospheric Environment, 216, 116940, 2019 (pdf). (Times Cited 3, Impact Factor = 4.012)