Spatio-temporal variations in the estimation of PM₁₀ from MODIS-derived aerosol optical depth for the urban areas in the Central Indo-Gangetic Plain

Abstract

Particulate air pollution poses a serious health problem to the urban centers in the central Indo-Gangetic plain (IGP) in northern India. Health management planning is constrained by the lack of availability of continuous dataset of particulate matter (PM) at a regional scale. Recently, researchers have established the strength of regression models for estimating PM from satellite-derived aerosol optical depth (AOD) and meteorological factors. The present study is focused on three cities, namely, Agra, Kanpur and Varanasi located in the central IGP. The study envisages four approaches of multi-linear regression modeling to estimate PM₁₀ (particulates smaller than 10 µm) from AOD and the meteorological parameters. The first approach consists of four regional models, three of which estimate regional mean PM₁₀ and the fourth one estimates the distributed PM₁₀. These models have a weak-to-moderate coefficient of determination (R ² = 0.37–0.63). Spatial and temporal variations in the estimators are separately addressed by the second modeling approach, i.e., city models (CMs) and the third modeling approach, i.e., seasonal models (SMs), respectively. R ² of these models varies from 0.40 to 0.68. Finally, the spatio-temporal variability of the estimators are addressed by the fourth modeling approach, i.e., city-wise seasonal models (CSMs) which exhibited better results (R ² = 0.49–0.88). Remarkable variations in the regression estimators of the CSMs are observed both spatially and temporally. The model adequacy checks and the validation studies also support CSMs for more reliable estimation of PM₁₀ in the central IGP. The proposed methodology can, therefore, be reliably used in generating the regional PM₁₀ concentration maps for health impact studies.