Investigation of the Effect of Atmospheric Correction (DOS) On Water-Vegetations Indexes Mapping (Case Study: South Wetlands of Lake Urmia)

Wetlands comprise roughly 6–9 percent of the Earth’s surface. The role of wetlands in maintaining environmental quality includes the storage of global terrestrial carbon. In addition, they influence many aspects of ecology, economy and human welfare. Furthermore, wetlands act as an oasis in an urban area which is important in the reduction of surrounding surface air temperature. Changes in the spatial distribution of wetlands (croplands, forests, water bodies and rivers), either by natural factors or anthropogenic activities could significantly affect the ecosystem. Satellite imagery enables us to monitor short-and long-term changes in wetlands and its vegetation density. Due to the fact that the electromagnetic waves reaching the satellites pass through the Earth's atmosphere, the reflections recorded by the sensor of these satellites do not really reflect the phenomena of the Earth's surface. By applying some corrections on the images, it is possible to convert them to Top Of Atmosphere reflectance values (TOA) and the Earth's surface reflections (BOA) ones. In this paper, we have reviewed and compared the results of extraction of wetlands and water body using raw and atmospheric corrected images of Sentinel 2 in south of Lake Urmia. The study area includes wetland and agricultural lands of southern Urmia Lake. Due to the existence of two large Zarrineh and Simineh rivers in the region as well as its very fertile soil, agriculture has expanded rapidly. The main data of this study is satellite images of Sentinel 2 (spring 2019). The dark object subtraction (DOS) is one of the methods of atmospheric correction of satellite images, which with a partial fraction of the dark object's reflection of the whole image, it makes an atmospheric correction of the satellite image. In this paper, this method has been used for atmospheric correction. In fact, a copy of the raw satellite image of the study area was made and the atmospheric corrections were applied. Then the results were compared with the raw images. In order to compare two raw and corrected images, it was attempted to separate the wetlands (rivers, ponds, wetlands) from non-moisturizing lands, so that the effect of atmospheric correction on the ground reflection could be observed. For this purpose, NDWI2 MNDWI, NDTI and SAVI indexes have been used in this paper. To compare the effect of atmospheric correction on Sentinel satellite images, the image of the study area was first provided and entered into the QGIS software for atmospheric correction. Then, using a combination of short infrared, near-infrared and green bands, the extraction and classification of wetlands, water bodies and vegetation cover density was made by SNAP software. Due to the lack of ground control points, the images were sampled by the ArcMap software and verified by using google earth images. Three precision coefficients were used to check and compare the accuracy of raw data with the atmospheric corrected data. In order to compare the accuracy of the outputs, Kappa coefficients, users’ accuracy and producers’ accuracy were calculated using the ArcMap software. The MNDWI and NDTI indices are the best indicators for raw images and for corrected images, to extract wetlands and water bodies. Kappa values of these indicators are above 0.9 and also users’ accuracy and producers’ accuracy are above 96%. Among the four above-mentioned indicators, the NDWI2 index has the lowest accuracy as well as the minimum Kappa coefficient. The results show that corrected images have high accuracy in extracting and displaying wetlands and water bodies. The area of the wetlands and water bodies to be redirected from corrected images is closer to actual areas. The actual area in the SAVI index is 25.15 square kilometers and the redistributed area of raw and corrected images are 25.71 and 25.38 km2, respectively. The actual area in the NDWI2 index is 180 km2, with a corrected area of 178.93 square kilometers. For other indicators, NDTI and MNDWI, the actual areas are 10.37 and 29.5 km2, respectively. In general, it can be concluded that atmospheric corrected images using the DOS method show better results in showing wetlands and water bodies areas. The results of this paper show that the application of atmospheric corrections to the Sentinel 2 images can increase the accuracy of the extraction of wetlands and water bodies areas and even other landcovers. Considering the four indicators and extracting the zones from raw and atmospheric corrected images, it was determined that, firstly, MNDWI and NDTI indices are the best indicators for extracting wetlands and water bodies in the south of Lake Urmia. Secondly, among these two indicators, the data from the corrected atmospheric images have high precision coefficients than raw images. Therefore, it can be said that in estimating the wetlands and water bodies using Sentinel images 2, these images must be corrected using different methods to minimize their error of representations.