Several studies have found rising ambient particulate matter (PM 2.5 ) concentrations in urban areas across developing countries. For setting mitigation policies source-contribution is needed, which is calculated mostly through computationally intensive chemical transport models or manpower intensive source apportionment studies. Data based approach that use remote sensing datasets can help reduce this challenge, specially in developing countries which lack spatially and temporally dense air quality monitoring networks. Our objective was identifying relative contribution of urban emission sources to monthly PM 2.5 ambient concentrations and assessing whether urban expansion can explain rise of PM 2.5 ambient concentration from 2001 to 2015 in 15 Indian cities. We adapted the Intergovernmental Panel on Climate Change’s (IPCC) emission framework in a land use regression (LUR) model to estimate concentrations by statistically modeling the impact of urban growth on aerosol concentrations with the help of remote sensing datasets. Contribution to concentration from six key sources (residential, industrial, commercial, crop fires, brick kiln and vehicles) was estimated by inverse distance weighting of their emissions in the land-use regression model. A hierarchical Bayesian approach was used to account for the random effects due to the heterogeneous emitting sources in the 15 cities. Long-term ambient PM 2.5 concentration from 2001 to 2015, was represented by a indicator R (varying from 0 to 100), decomposed from MODIS (Moderate Resolution Imaging Spectroradiometer) derived AOD (aerosol optical depth) and angstrom exponent datasets. The model was trained on annual-level spatial land-use distribution and technological advancement data and the monthly-level emission activity of 2001 and 2011 over each location to predict monthly R. The results suggest that above the central portion of a city, concentration due to primary PM 2.5 emission is contributed mostly by residential areas (35.0 ± 11.9%), brick kilns (11.7 ± 5.2%) and industries (4.2 ± 2.8%). The model performed moderately for most cities (median correlation for out of time validation was 0.52), especially when assumed changes in seasonal emissions for each source reflected actual seasonal changes in emissions. The results suggest the need for policies focusing on emissions from residential regions and brick kilns. The relative order of the contributions estimated by this study is consistent with other recent studies and a contribution of up to 42.8 ± 14.1% is attributed to the formation of secondary aerosol, long-range transport and unaccounted sources in surrounding regions. The strength of this approach is to be able to estimate the contribution of urban growth to primary aerosols statistically with a relatively low computation cost compared to the more accurate but computationally expensive chemical transport based models. This remote sensing based approach is especially useful in locations without emission inventory.
Impact of Urban Growth on Air Quality in Indian Cities Using Hierarchical Bayesian Approach