| Abstract | This study explored the potential of satellite remote sensing for the monitoring of particulate matter (PM) using synoptic meteorological patterns with the aim to not only account for the local and regional meteorological variables but also the vertical distribution of PM typical in each pattern. Satellite remote sensing by Moderate Resolution Imaging Spectroradiometer (MODIS) and Multi-Angle Imaging Spectroradiometer (MISR), both measuring Aerosol Optical Depth (AOD),has emerged as a potential breakthrough in large scale PM monitoring which can partially substitute the expensive ground monitoring network, especially in areas where the data are scarce. Apart from other factors, the PM-AOD relationship is affected by meteorology that determines PM optical properties, its dispersion, accumulation and vertical distribution. This study presents a novel approach to analyze PM-AOD relationship by considering the totality of meteorological factors expressed as synoptic patterns. The main objectives of the study are; (1) To determine the synoptic meteorological patterns governing Bangkok Metropolitan Region (BMR) during the dry season (November-April). (2) To analyze PM-AOD relationship with and without meteorological patterns consideration, and (3) To analyze the potential of AOD to be used for PM composition (SO4-2, EC/BC, NO3-) monitoring in BMR. Meteorological dataset of 18 variables, recorded at 07:00 LST at nine regional surface weather stations during the dry seasons between2000 and2010 was subjected to the principal component analysis(PCA) which reduced it to six principal components (PCs). Two stage clustering was applied to the six PCs which produced four meteorologically distinct patterns, representing four commonly observed weather patterns in BMR. The number of PCs retained and the statistical significance of the difference among the 4patterns were statistically evaluated and confirmed by using parallel analysis (PA) and multi-variate analysis of variance (MANOVA) respectively. Pattern 1 represents the highest pressure and the lowest temperature situation which was associated with the highest average 24hPM10concentrations (83 ± 27 μg/m 3) across 22 monitoring stations in BMR during the study period. Pattern 2 which is characterized by the highest temperature and the lowest pressure is found to have the lowest average 24hPM10(53 ± 17 μg/m3). Pattern 3 represents the second highest temperature and the second lowest pressure with negative pressure changes during previous 24 hours and intermediate 24hPM10(73 ± 24 μg/m3). Pattern 4, with similar pressure and temperature conditions as pattern 3, described maximum moisture, lowest visibility and wind speed prevailing but with positive pressure tendencies and 24hPM10of 80 ± 28 μg/m3.Lidar aerosol extinction as well as backscatter coefficient vertical profiles were used to test the potential of synoptic patterns to account for the PM vertical distribution. Both, the typical extinction and back scatter profiles for each pattern were found consistent with their respective meteorological conditions. MODIS and MISR AOD, extracted from their respective level 2 aerosol data products for the study period were validated against AERONET AOD for the Pimai remote site, 200 km from BMR, where the AERONET data were available. The results showed that they were highly comparable. First, the relationship betweenPM10and AOD was analyzed without consideration of meteorological patterns (lumpcase) which showed that the correlation coefficient (R values) varied among stations from 0.28to 0.55forMODISTerra, from 0.27to 0.58for MODIS Aqua, and from 0.12 to 0.54 for MISR AODs. When considering the synoptic patterns, R values between PM10and MODIS AOD, averaged across all stations, improved for all patterns except for pattern 2. For the case of MISR AOD, R values noticeably improved only for pattern 1. Pattern 1 had the maximum R iv values ranging between 0.38–0.68(averaged at 0.46)for MODIS Terra, 0.33-0.68(averaged at0.38)for MODIS Aqua and 0.13-0.73 (averaged at 0.42) for MISR. The higher R values found for pattern 1 maybe explained by the presence of low temperature and high pressure with typical low mixing height shown in the aerosol Lidar vertical profiles. The R values were minimum and unexpectedly negative for some stations for pattern 2which is characterized by the low pressure with high mixing height, shown in the aerosol extinction/backscatter profiles, hence thePM10measured at the ground level in this case may be the least representative for the overall column aerosol loading (AOD) in this pattern. R values for pattern 3 and 4 were intermediate and mostly lower than those of pattern 1.The number of MISR observations matched up with ground PM observations was still small and much less than that of MODIS. Hence more data are required for a better analysis of the PM10-MISR AOD relationships. Eight MISR component AODs were extracted from MISR level 2 data which were then used in stepwise regression as independent variables against PM components (SO4-2, BC/EC, NO3-).Available hourly and daily EC and daily SO4-2and NO3-measurements, made at AIT, were used in the analysis. The results showed that MISR component AODs were better correlated with PM components (R=0.69for hourly BC,R=0.72for daily SO4-2inboth,PM2.5and PM10and R= 0.53 and 0.54 for daily NO3- in PM10 and PM2.5respectively) as compared to total AOD (R=0.49for hourly BC,R=0.40 and 0.45 for daily SO4-2 and R=0.22 and 0-.12 for NO3-in PM10and PM2.5respectively). The correlation coefficient between the MISR AOD components and the PM components was comparable to that between MISR and MODIS total AODs with the PM mass concentrations. The improved PM-AOD correlations produced an be used to semi-qualitatively evaluate the PM air quality in the remote areas and the locations where no ground monitoring data are available. Because of the sparse monitoring networks in the developing countries the results would be particularly useful in this aspect. In particular, the PM-AOD correlations obtained in this study can be used to evaluate the regional modeling results for the locations in the study region where no monitoring is available. It is also suggested to further improve the PM-AOD relationship in the region by calculating near ground AOD using AOD vertical profiles produced by 3D Chemical Transport Models. |