标题：Atmospheric Smog Modelling, Using EOS Satellite ASTER Image Sensor, with Feature Extraction for Pattern Recognition Techniques and its Analysis of Variance with In-Situ Ground Sensor Data
期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：2008
卷号：XXXVII-B8
页码：467-472
出版社：Copernicus Publications
摘要：Atmospheric pollution was previously considered as a 'Brown Cloud' phenomenon restricted to industrialized urban regions. Studies in field stations and satellite observations made since the last decade revealed that it now spans continents and ocean basins world wide. Anthropogenic activities are considered to be the primary cause of pollution in the atmosphere. The appearance of a smog layer with more absorption and scattering of solar radiations, particularly long-wave infrared radiations, decreases the atmospheric transmission factor, significantly perturbing the atmospheric absorption of solar radiation. The objective of this research is to create a geo-spatial model based on probability density of common atmospheric pollutants in the troposphere by using feature extraction and pattern recognition technique with high-spectral and spatial resolution Earth Observation System (EOS) satellite imaging sensor Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) data, US Environmental Protection Agency (EPA) ground sensor data and "Analysis of Variance" of satellite pixel value with EPA ground sensor observations. This research investigated three locations: 1. San Francisco Bay area, in California has a unique land feature having maritime climate surrounded by coastal ranges. 2. Los Angeles, in California (a maritime climate) and, 3. Charleston, in West Virginia which has humid continental climate in the mountain ranges in Kanawha Valley. Polluting industries data and Tapered Element Oscillating Microbalance (TEOM) ground sensor data are collected from the US-EPA. Spectral signatures of common atmospheric pollutants are collected from Jet Propulsion Laboratory's ASTER spectral Library, HITRAN Database, and USGS spectral library. Climatic data are collected from National Oceanic and Atmospheric Administration - National Climatic Data Center (NOAA-NCDC). All spatial data are stored as point features in shapefiles, with pollutants concentration as attributes in ArcGIS 9.2 software and imported into ER Mapper vector files for locating ground sensors and pollutants sources in ASTER imagery. ASTER L1B data are georegistered and geocorrected by image-to-image registration with georegistered Digital Orthophoto Quarter-Quadrangles (DOQQ's). Principal Component Analysis, Density Slicing and Band Ratioing techniques are applied to extract features in the ASTER datasets. Spectral signatures in graphical form of the atmospheric features are obtained in ER-Mapper 7.1 geospatial software and compared both in short wave infra-red (SWIR) and thermal infra-red (TIR) bands. Correlation between ground sensor pollution level and ASTER image pixel digital numbers are investigated in the study of Analysis of Variance, by creating a general linear model in SAS software It is observed that there is significant city effect, suggesting different kinds of atmospheric pollutants in different cities under investigation. Despite the broader bandwidth of ASTER as compared to hyperspectral satellite systems, it is observed that TIR band 14 is highly correlated with EPA monitored concentration in NOx and PM 10 and SWIR band 7 is moderately correlated with EPA monitored CO concentration data in all the three areas, (i.e. San Francisco Bay area and Los Angeles, in California, and in Charleston in West Virginia). Future investigation is envisioned to study the subtle differences in spectral signatures of air pollutants by using hyperspectral satellite data and advanced sensors
关键词：Smog; Absorption Bands; Satellite-Imagery; Spectral Signature; Sensors; and ASTER
