Publications

Paul, B; Mishra, MK; Das, AK (2022). Spatial heterogeneity and estimation of PM10 concentration over Brahmaputra Valley using geographic weighted regression model assimilating surface, MODIS, and ERA-interim reanalysis data. AIR QUALITY ATMOSPHERE AND HEALTH, 15(3), 425-435.

Abstract
The geographic weighted regression (GWR) model is utilized to estimate PM10 from a 3-year (2016 to 2018) collocated monthly mean of predictive variables, aerosol optical depth (AOD 550 nm), planetary boundary layer (PBL), and air temperature (AT) over the intermontane Brahmaputra River Valley (BRV). The coefficient of determination (R-2 = 0.62) and the root mean square error (RMSE = 22.74 mu g/m(3)) obtained from the GWR model fitting decreased by 0.09 (R-2 = 0.53) and 0.14 mu g/m(3) (RMSE = 22.60) in cross-validation analysis. Regression coefficient (beta) values computed by GWR reflect the heterogeneity of relationship among the variables. The PM10-AOD 550-nm relationship showed asymmetry (strong positive to negative beta values) at spatial scale from the west (beta(1) = 5.94 to 23.91), central (beta(1) = -2.92 to -17.68), to the east (beta(1) = -4.19 to -27.27) valleys of the BRV due to the effect of local meteorological conditions. Results of monthly mean value of PM10 and the predictive variables indicate surface PM10 and AOD 550-nm mean values vary strongly in sites over the valley. This variation is dependent on the proximity to sources and dispersal mechanisms within the PBL. Further in this study, the PM10 (mu g/m(3)) is predicted at valley site scale for 2016 based on annual mean of AT, PBL, and AOD 550 nm at higher spatial resolution. The results show interesting spatial contrast of AOD 550 nm and PM10 with PBL.

DOI:
10.1007/s11869-022-01160-9

ISSN:
1873-9326