Extended Abstract
Introduction
Snow is a key component of the hydrological cycle in mountainous regions because it temporarily stores cold-season precipitation and gradually releases water during the melting period. This seasonal storage contributes to streamflow, groundwater recharge, spring discharge, and agricultural water supply. Reliable assessment of snow resources is therefore essential in mountainous and data-scarce regions, particularly where water availability depends strongly on cold-season precipitation. Snow-covered area alone cannot adequately represent the amount of water stored in a snowpack. Snow depth indicates the vertical thickness of the snow layer, whereas snow water equivalent (SWE) represents the depth of water produced by the complete melting of the snowpack. Because SWE is also influenced by snow density, moisture content, and compaction, the simultaneous analysis of snow depth and SWE provides a more comprehensive assessment of snow storage. Ground-based measurements of these variables are limited in many mountainous regions because of sparse monitoring networks, difficult access, high field-survey costs, and pronounced spatial variability. Gridded land-surface model products can partly overcome these limitations by providing spatially continuous and temporally consistent estimates. However, their outputs must be interpreted cautiously because of their relatively coarse spatial resolution and model-related uncertainties. This study investigated the monthly and between-period variations in snow depth and SWE, evaluated the performance of two satellite precipitation products, and examined the relationship between snow storage and topographic factors in Marivan County, western Iran, during three four-month cold periods extending from December to March in 2021–2022, 2022–2023, and 2023–2024.
Materials and Methods
Marivan County covers approximately 2,337 km² in western Kurdistan Province near the Iran–Iraq border. The county is located in the western Zagros Mountains and is characterized by complex topography, with elevations ranging from approximately 1,191 to 3,153 m above sea level. Monthly snow depth and SWE data were obtained from the FLDAS-Noah land-surface modeling system at a spatial resolution of 0.1°. Daily precipitation estimates were acquired from CHIRPS at a spatial resolution of 0.05° and PERSIANN-CDR at 0.25°. The daily precipitation data were aggregated into monthly totals and compared with monthly observations from the Marivan meteorological station. Elevation, slope, and aspect were derived from the 30 m Shuttle Radar Topography Mission digital elevation model. Because the spatial resolution of SRTM was considerably finer than that of FLDAS-Noah, the topographic data were aggregated and analyzed at a scale compatible with the FLDAS grid. Consequently, the topographic results were interpreted as regional patterns rather than direct snow estimates at the 30 m pixel scale. Data preparation and extraction were performed in Google Earth Engine, whereas ArcGIS, Excel, and Python were used for spatial processing, statistical analysis, and graphical presentation. Descriptive statistics, including the mean, median, standard deviation, coefficient of variation, minimum, maximum, and quartiles, were calculated for snow depth and SWE. The relationship between monthly snow depth and SWE was examined using Pearson’s correlation coefficient and the coefficient of determination. The performance of CHIRPS and PERSIANN-CDR was evaluated using R², root mean square error, mean absolute error, bias, and the Nash–Sutcliffe efficiency coefficient. Area-weighted mean snow depth and SWE were also calculated for elevation, slope, and aspect classes.
Results and Discussion
The results revealed substantial temporal variability in snow storage. Over the 12-month study period, the mean snow depth and SWE were 2.563 cm and 4.569 mm, respectively, with coefficients of variation of 107.45% and 96.44%. The highest values occurred in February 2023, when snow depth reached 10.120 cm and SWE reached 16.362 mm, followed by a sharp decline in March 2023. Overall, the 2022–2023 four-month cold period had the greatest snow storage, whereas the 2023–2024 period had the lowest, despite recording the highest station precipitation. This finding indicates that total precipitation alone does not determine snow storage; precipitation phase, air temperature, timing of precipitation, and intermittent melting during the cold period are also important. Snow depth and SWE were strongly and significantly correlated (r=0.985, R2=0.971), indicating close temporal agreement between the two FLDAS-Noah variables. Nevertheless, SWE also depends on snow density and compaction and therefore cannot be fully represented by snow depth alone. Both satellite precipitation products substantially underestimated observed precipitation. The 12-month totals were 1,696.93 mm at the Marivan station, 980.72 mm for CHIRPS, and 731.45 mm for PERSIANN-CDR, corresponding to underestimations of 42.2% and 56.9%, respectively. CHIRPS produced lower RMSE, MAE, and absolute bias values and a less negative NSE than PERSIANN-CDR; however, neither product performed sufficiently well to replace station observations without local evaluation and bias correction. Elevation showed the clearest relationship with snow storage, with the highest mean snow depth and SWE recorded in the 2,624.1–3,153 m elevation class. Snow storage increased up to the 35–40° slope class and then declined on steeper terrain. Differences among aspect classes were comparatively limited, although southeastern slopes showed the highest mean values.
Conclusion
Snow storage in Marivan County varied considerably among months and across the three four-month cold periods. The 2022–2023 period, particularly February 2023, had the greatest snow depth and SWE, whereas the 2023–2024 period had the lowest snow storage despite high observed precipitation. Snow depth and SWE were strongly correlated, although SWE remained the more direct indicator of the amount of water stored in the snowpack. Both CHIRPS and PERSIANN-CDR substantially underestimated station precipitation, with CHIRPS showing relatively better overall performance. Elevation was the topographic factor most consistently associated with increasing snow storage, whereas slope showed a nonlinear relationship and differences among aspect classes were limited. The integration of FLDAS-Noah data, satellite precipitation products, station observations, and topographic information can support regional snow monitoring in data-scarce mountainous areas. However, the relatively coarse model resolution, spatial-scale mismatch among datasets, reliance on a single meteorological station, and absence of ground-based snow observations should be considered when interpreting the results.