Monitoring the spatial changes of the snow cover of Central Alborz using the SVM algorithm and landsat images

The snows of the Alborz mountain range play an important role in providing underground and surface water for the settlements around it and the densely populated Caspian Plain, and the permanent rivers of the northern and southern slopes of Alborz are fed by the snows of the high places of Alborz. Global warming has caused changes in the atmospheric-climatic parameters of Iran; in such a way that temperature and evaporation have increased and precipitation, especially snow, has decreased. These changes, which are considered as independent variables, will cause changes in the dependent variable, which is the snow cover of the high points of the Alborz Mountains. Its consequences can be accelerating the melting of snow, increasing the process of flooding of rivers in the catchment areas of the study area, and the destruction of habitats and settlements downstream. Therefore, the monitoring of Alborz snow area can be used in formulating water management strategies and sustainable development.

In this study, the temporal-spatial variations of the Central Alborz snow cover on a seasonal scale for the years 1985 to 2020 were monitored using Landsat TM, ETM+ and OLI for 1985, 1995, 2005, 2015 and 2020. For each season of the year, an image was prepared that was a combination of 4 satellite images and radiometric and geometric corrections were made on it. SVM algorithm was used to extract the snow cover. The area of snow cover was obtained by transferring the classified map to the ArcGIS.

The results showed that the kappa coefficient for the classified maps was more than 0.91. The average snow covers for winter, autumn, spring and summer were 1.19, 0.47, 0.14 and 0.004 million hectares, respectively. Snow covers have been declining from 1985 to 2020, reaching 1.98 million hectares in 2020 from 1.68 in 1985 to 1.68 in winter. In the autumn, it increased from 0.84 in 1985 to 0.15 million hectares in 2020. In winter and autumn, snow cover in the eastern part of central Alborz has decreased sharply compared to the western part. In the winter of 1985, snow started at an altitude of 1,500 meters, but by 2020 it reached 2,500 meters. In the summer, snow was more than 3,900 meters high in 1985, but peaks more than 4,200 meters in 2015 and 2020. The area of snow cover in Central Alborz has a decreasing trend, which has the highest rate in winter. In order to increase the spatial accuracy of snow, the present research used Landsat series images with 30 meters pixels, and its results are more favorable than the output of MODIS images.

Therefore, the results of this study showed that the accuracy of the support vector machine algorithm is more than 0.91% in the classification of Landsat images, this method can be used to extract snow patches; in such a way that it separated the shadow snow and cloud from the snow and identified the accumulation of snow in the valleys. It can be concluded that using images with spatial resolution of 30 meters and applying classification algorithms for snow extraction is better than using NDSI index and MODIS images. On the other hand, the process of snow cover in central Alborz has been such that during 25 years, the area of snow has decreased from about 0.7 million hectares and a large amount of fresh water storage in Alborz has been lost.