Evaluation of the performance of ERA5, ERA5-Land and MERRA-2 reanalysis to estimate snow depth over a mountainous semi-arid region in Iran

Study region

Mountainous semi-arid region, Iran.

Study focus

Snow is a critical component of the cryosphere, with significant seasonal and annual variability that impacts global water circulation and energy balance. While ground-based observations provide the most reliable snow depth (SND) data, their sparse distribution in remote regions necessitates the use of alternative datasets for monitoring snow depth. This study evaluates the ability of three reanalysis datasets—ECMWF's ERA5, ERA5-Land and the Modern-Era Retrospective Analysis (MERRA-2)—for estimating snow depth across Iran from 1980 to 2020. A comparison was conducted using SND data from synoptic stations within the study area. The evaluation was performed on both temporal and spatial scales, employing statistical indicators such as correlation coefficients, bias, and root mean square error (RMSE).

New hydrological insights for the region

This study provides critical new insights into the hydrology of the region, particularly in understanding the limitations of existing datasets in mountainous areas. Our findings indicate that all datasets can approximate observations, although their performance varies considerably across different regions. All datasets report maximum snow depth in the mountainous regions of Iran, particularly in the Alborz and Zagros Mountain ranges. Despite the higher correlation and lower RMSE of ERA5 and ERA5-Land compared to MERRA-2, all datasets exhibit common weaknesses in accurately estimating SND in complex terrains. The superior performance of ERA5-Land in this study can be attributed to its fine horizontal resolution, advanced data assimilation techniques and improved physical modeling, which enhance its ability to capture snow dynamics accurately. Additionally, the study highlights the challenges MERRA-2 faces in capturing snow depth in mountainous regions. Future research could benefit from integrating additional datasets and employing machine learning algorithms to improve snow depth assessments, as these approaches may reduce estimation uncertainties and enhance the understanding of snow dynamics across various regions, ultimately contributing to more reliable hydrological assessments.