Estimation of ground level PM₂.₅ concentration using aerosol optical thickness from modis images in Peninsular Malaysia

Fine particulate matter is particulate matter lower in diameter than 2.5 μm (PM2.5). It affects the public health, economic development, and the regional climate. Governments worldwide have been concerned about the levels of the PM2.5 in the atmosphere and long ago began monitoring their levels continuously using air quality monitoring stations. The Malaysian government invested much in building ground monitoring stations. Most of these stations have been located in urban areas. In this context, use of remote sensing (RS) techniques and the geographic information system (GIS) in estimating the levels of the ambient PM2.5 become more widespread. The Moderate Resolution Imaging Spectroradiometer (MODIS) Aerosol Optical thickness (AOT) product of the Terra Satellite can be used to estimate the PM2.5 levels with an accuracy that depends on the statistical relationship between PM2.5 and AOT. This study aimed at estimating the PM2.5 mass concentration in Peninsular Malaysia by linking the MODIS sensor data with the measured PM2.5 concentrations and meteorological parameters in the year 2013. The first objective of this study was to validate MODIS-AOT retrievals with ground AERONET AOT data after extracting the AOT from MODIS images. The second objective of the study was to correlate the ground PM2.5 levels with the validated MODIS AOT after identifying the spatial and temporal AOT distributions by using Multiple Linear Regression Analysis (MLRA) and Geographically-Weighted Regression models (GWRMs) for Peninsular Malaysia. The third objective of the study was to accurately evaluate the accuracy of estimates of the relationship between the PM2.5 levels and the corresponding MODIS AOT data in different pixel size groups. The geographic domain of this study was Peninsular Malaysia, which has an area of about 131,598 km2, covering 40% of the land area of Malaysia and hosting approximately 80% of its population and economic activities. The methodology of the study consisted of three stages. First, the values of AOT were extracted from MODIS images and the AOT retrievals were validated with ground AERONET- AOT data. Second, the MLRA and GWR modeling were attempted to spatially and temporally correlate the reported ground PM2.5 levels and meteorological data with the validated MODIS AOT data after identification of the spatial and temporal AOT distributions in Peninsular Malaysia in the year 2013. Subsequently, a comparison of strengths and weaknesses was held between the various generated models. Lastly, an assessment of the accuracy of PM2.5 estimation has been conducted on the MODIS spatial models. The results showed that the MODIS AOT retrievals have a good correlation with the ground observations derived from AERONET as indicated by the values of the coefficient of determination (R2) for the linear regression models, which were 0.87 and 0.78 for the daily average and the hourly average (± 30 min) data, respectively. The map of distribution of AOT indicated that the AOT concentrated in the western coast of Peninsular Malaysia. Mostly, the spatial and temporal AOT values in the southwest monsoon were higher than in the northeast monsoon throughout the study period. The MLRA and GWRM both gave almost identical estimates of the PM2.5 concentrations. The analysis outcomes revealed that the R2 value for the hourly PM2.5 regression model (0.66) was higher than that for the daily PM2.5 (R2 = 0.53). Comparison with the literature uncovers that results of estimation of PM2.5 using AOT from MODIS for Peninsular Malaysia are similar to the results of other studies in other parts of the world. Furthermore, assessment of the accuracies of the hourly and daily estimates of PM2.5 disclosed that the 5 x 5 pixel size model had the lowest values of the mean-squared error (MSE), root mean-squared error (RMSE), and relative root mean-squared error (rRMSE). The relatively low error values associated with the 5 x 5 pixel size model indicate the accuracy of this model in Peninsular Malaysia.

Text FPAS 2020 10 IR.pdf Download (1MB)

Actions (login required)

View Item