Introduction
As an important type of precipitation, snow is especially important in the hydrological cycle. This importance can be examined and analyzed from several aspects such as water supply in other seasons. The most important aspect is the possibility of creating hazards for human beings and human infrastructure (snow avalanches, floods during seasonsof snowmelt). Therefore, it is necessary to study the snow phenomenon and its covered surfaces in winter. Monitoring the changes in this important climatic phenomenon has always been considered important by researchers and planners. Remote sensing methods have revolutionized the field of natural environment monitoring since their inception. Snow depth is an example of what can be monitored and evaluated by remotely sensed data and techniques.
 
Materials & Methods
The present study seeks to evaluate the efficiency of several important remote sensing indices in monitoring snow depth, andalso to introduce and evaluate a proposed spectral index. To reach this aim, satellite images of Landsat 8 and Sentinel 2 have been used. These images were received from the relevant portal and used to calculate snow indicesafterinitial corrections. Four spectral indices were usedto extract snow covered surfaces. These indices include: NDSI - S3 - NDSII - SWI. These indices are based on reflection from snow covered surfaces in light reflection and absorption spectra of snow covered surfaces.Light reflection from snow covered surfaces in the visible spectra and absorption in the short infrared spectrum allow automatic detection and extraction of snow covered surfacesin remote sensing multispectral images. The above mentioned indices have the ability to extract snow, but they fail to differentiatebetween snow and other related phenomena such as water (in the absorption band) and light-color salt marshes (in the reflection band) and thus, similarity of the spectra occurs. This spectral mixing which occurs due to the similarity of the reflections, cannot be eliminated even when threshold limits are defined. Thus, the extracted snow cover includes not only snow, but also other similar zones. To solve this problem and extract snow covered surfaces correctly,a new index is presented in this paper based on principal component analysis (PCA) and the first component of the set, and short wave infrared (SWIR) spectrum reflection.Using the first component of the set with the highest variance makes the difference between reflectance of snow and similar phenomena visible and thus, solves the issue of spectral mixing to a very large extent. The proposed new index called PCSWIRI is also evaluated and validated along with 4 other indices in the present paper.
 
Results & Discussion
Spectral indices introduced in the previous section were examined and evaluatedusing 7 sets of images (4 Landsat images and 3 sentinel 2images) captured in different days of winter from the main study area (Lake Urmia in the northwest) and two other study areas. The results indicate efficiency of the proposed index in the extractionof snow covered surfaces. The proposed index has improved the accuracy of snow cover extractionin the whole collection of images. This increased accuracy has been confirmed withstatistical evaluation criteria, such as kappa coefficient, overall accuracy and in the visual review of indices(comparing to the composition of the original image). The main study area includes Lake Urmia, an important geographic feature containing water and salt and a mixture of the two, which makes its spectrum similar to snow. This lake is incorrectly identified by other indices as a snow covered surface. Like the main study area, the first study and assessment area contains salt covered zones (salt lake). Despite the spectral similarity between snow and salt,the proposed index has been able to distinguish between this phenomena (in both regions) and snow and to extract only realsnow covered surfaces. In addition, visual review of existing water bodies (Dam Lake) and 5 evaluated indicesindicates higher accuracy of the proposed index. In order to automate the process of calculation in the proposed spectral indices, a software was also providedbased on MatLAB.
 
Conclusion
The findings of the present study indicates higher accuracy and efficiency of the proposed index (PCSWIRI) for snow cover extraction. Snow cover maps are very useful in various hydrological, climatic, precipitation-runoff modeling studies, and etc. Therefore, increasing the accuracy of snow cover maps is of great importance and results inimprovedaccuracy and reliability of modeling processes.