Hooshang Seifi; Ismail Gorbani
Abstract
Introduction Presently, population growth, urban development, the importance of agriculture in economic development, the need for supplying water demands of this sector and improving public health have multiplied water consumption as compared to the past. For appropriate and optimal use of water resources, ...
Read More
Introduction Presently, population growth, urban development, the importance of agriculture in economic development, the need for supplying water demands of this sector and improving public health have multiplied water consumption as compared to the past. For appropriate and optimal use of water resources, it is necessary to know the amount of available water in the area, its temporal and spatial changes, and the exact planning for maintenance and utilization of the available water. Accordingly, studying and measuring changes in snow levels, as an important source of water in mountainous areas, is very important. Snow cover is one of the important parameters involved in the amount of snowmelt. Due to the difficulty of monitoring and measuring snow cover level in mountainous basins, satellite images are used as alternatives to monitoring and ground operations in the preparation process of snow cover map. In this regard, the use of satellite imagery and remote sensing, due to low cost, up-to-date and extensive coverage, is a major breakthrough which can be used to identify snowy areas and evaluate changes in that method. Detailed analysis of snow-related issues requires a set of snow measurements and observations. OLI and TIRS sensors with various advantages like appropriate number of bands, referable spatial resolution, and sequential time series are considered to be an appropriate tool for this purpose. The main objective of this research is to estimate snow coverage of Sahand Mountain using satellite images received from OLI and TIRS sensors and by object-oriented classification method. Materials and Methods The present study use images received from OLI and TIRS sensors, and Landsat 8 satellite on 08/02/2017 (Pass and Row no. 34-168), as well as Digital Elevation Model based on data received from Aster Sensor and Terra satellite with a resolution of 28.5 meters to produce snow coverage map. Geo TIFF satellite data were originally requested from the American Geological organization and received from USGS site. Envi 5.3, eCognition 9.1, and ArcMap 10.4.1 software were used for processing and preparation of images, as well as classification and extraction of the final maps. In order to classify and extract snow cover surface with high precision, NDSI, NDVI, LST, and Brightness algorithms were used along with fuzzy algorithms. Results and Discussion Classification of satellite digital images is one of the most important methods for extracting applied information, which is currently performed by two general methods of pixel-based processing and object-based processing. The former method is based on the classification of numerical values of images, and the latter use not only numerical values, but also content, texture, and background information in the image classification process. Recent researches have processed image pixels and have only applied NDSI algorithm to estimate snow cover level. Therefore, pixels recognized as snow in such researches may contain snow cover and other land uses, which reduces the precision of snow cover extraction and makes the process of extracting all snow covers difficult. Extraction of snow cover using MODIS images is one of such researches. Due to low spatial resolution of these satellite imageries, extraction of mountainous valleys snow cover, as well as the separation of snow cover from the cloud cover is done with very low accuracy. Therefore, due to higher accuracy of object-oriented classification as compared to pixel-based classification, object-based techniques were used to classify and estimate snow cover. In object-oriented method, pixels are classified based on shape, texture and gray tone of the image. Thereby, pixels change into image objects and resolves the pixel blend problem. Therefore, by assigning each object to a specific land use, classification accuracy increases. Also, using complementary algorithms such as Brightness and NDVI along with the NDSI algorithm will improve the accuracy of findings as compared to other recent research. Therefore, using Landsat 8 satellite images and the new method of image classification, the present study extract snow cover from different domains of the study area. The snow cover in valleys was also extracted with appropriate and acceptable accuracy using different algorithms. Using LST algorithm in object-oriented processing method, detecting and separating snow cover from cloud cover was made possible. In this way, a satisfactory result was obtained from the snow cover. Finally, snow cover for Sahand Mountain Range was calculated to be 1882.88 km2. The results can be used as an alternative to snow measurement stations. Conclusion Based on research findings, using Landsat 8 satellite imagery and object-oriented processing methods for image classification have the necessary efficiency in extracting snow cover in mountainous regions. Given the precise estimation of the snow surface and the low cost of using this type of satellite imagery, it is possible to use this type of images and check the snow cover with great confidence. While ground observations are expensive due to impassibility of mountainous regions, and also they are not sufficiently precise.
