Snow is an important water resource. In the mountainous regions, particularly the arid ones, snowmelt water is the primary source of many rivers. Moreover, the global radiant energy balance is highly affected by the albedo and low thermal conductivity of snow cover (Liu, C et al 2020). Also, in such high-latitude areas, the snow cover is a significant driver in many ecological, climatic and hydrological processes (Gascoin, S et al 2019). Since snow can be considered as the water storage with a short delay during the seasonal run-off, it constitutes a crucial component of the hydrological cycle (Muhammad, S. and Thapa, A 2020). Thus, the seasonal snowpack has considerable control over the hydrology and economy in many mountainous and cold regions globally. Similarly, the snow variability affects various ecological procedures, such as the species composition, distribution, and phenology (Alonso-González, E et al 2018). Therefore, achieving a deeper understanding of the present and future climate, water cycle, and ecological changes requires an accurate assessment of the seasonal snow cover. The climatological, hydrological, and ecological significance of snow cover is linked to its energy storage, high reflectance, good insulating properties, remarkable thermal capacity, being a substantial water storage resource, with the eventual release during the melting season (Czyzowska-Wisniewski, E et al 2015).
Monitoring the snow cover is an important means of studying its spatial and temporal changes, as well as the distribution analysis of the regional precipitation. This climatic phenomenon can be assessed using the measuring stations, modeling, remote sensing technology, and applied programs. Although accurate information regarding the measurement location can be provided through the ground stations, these stations are still limited in terms of the spatial scale; that is because providing sufficient information to produce long-term data about snow (on a spatial scale) is not achievable in many parts of the world when the information is obtained through a scattered network of meteorological stations. Nevertheless, the spatial and temporal characteristics of snow can still be monitored through modeling. Yet, the accuracy of such modeling results has been proven to be low due to the lack of information regarding the initial conditions (Azizi, G et al 2017). Moreover, due to the limited number of meteorological stations and the pointwise measurements thereof, these stations are not suitable alternatives for studying snow as a continuous phenomenon. Also, the snowfield measurement and sampling are timely procedures and not cost efficient. Alternatively, using remote sensing technology not only makes it possible to access high-altitude sites, but also it is generally less expensive than the formerly mentioned methods. Furthermore, satellites are proper imagery tools for measuring snow cover due to the snow reflectance and the visible contrast between snowfields and most surfaces (Raispour, K 2016). Remote sensing data can provide better estimates of snow cover ranges compared to the traditional surveying methods. Thus, nowadays, the use of remote sensing data with more accurate information on snow cover is an operational method of efficient water resource management (Mirmousavi, S. H. and Saboor, L 2014).
The image processing methods of remote sensing can be divided into two general categories. The first category with a single-pixel processing unit is called the pixel-based method. The processing units in the second category include image objects or a group of pixels; in other words, since a homogeneous group of pixels or the object image is the main processing unit, the image is processed in the object space and not in the pixel one; this makes it possible to define additional properties other than the spectral one, such as the shape, size, texture, and neighborhood (Momeni, M., Khosravi, I. and Mostaejeran, B 2013). There are many research studies worldwide on measuring the snow cover level and the trend of its changes using remote sensing. For instance, Lopez et al. (2008), after monitoring the images from the period of 2000–2006 based on the NDSI index, examined the amount of snow cover and its changes in northern Patagonia. The results of this study marked the minimum snow cover with an area of 3600 km² in March 2007 and the maximum snow cover with an area of 11323 km² in August 2001. Boi (2009) presented a snow cover monitoring technique for Italy and the Alpine regions using visible, near infrared and infrared MSG data. Accordingly, the monthly and annual maps regarding the snow cover frequency has been estimated (Boi, P 2010). In another study, Mölg et al. (2010) examined and controlled the snow cover data of MODIS multi-temporal imagery at high altitudes of Italy. In this study, snow cover estimation in time-series images from 2002 to 2008 was conducted; the output maps were derived from combining Aqua and Terra snow cover maps, thereby reducing the cloudless and value-free pixels. Additionally, the snow cover maps, obtained from Landsat E.T.M. + satellite images, were utilized to validate the results. Moreover, this study confirmed the classification improvement by a combination of Aqua and Terra images. Finally, using the object-oriented fuzzy classification and Landsat satellite data, Farhan et al. (2018) estimated the changes in the seasonal snow cover level in the Astore River Basin (western Himalayan part of Pakistan). Subsequent to the segmentation of the satellite images, the degree of fuzzy membership was determined, and the area’s snow cover level was estimated. As such, López-Moreno et al. (2020) considered the long-term trends of snow cover duration and depth from December to April of the years from 1958 to 2017 in the Pyrenees. The Mann–Kendall test illustrated that snow cover duration and its average depth decreased during the research time scope; moreover, the persistent warming was proved to be a major factor for the snow cover decrease in the Pyrenees.
At last, the present study was performed to compare the performance of the support vector machine (SVM) kernel functions and object-oriented fuzzy operators in estimating the snow cover amount in Alvand Mountain (Hamadan Province, Iran) using Sentinel-2B satellite images.