Document Type : Research Article
Authors
1
Professor in Climatology, Department of Physical Geography, Faculty of Social Sciences, University of Mohaghegh Ardabili, Ardabil, Iran.
2
Ph.D. in Climatology, Department of Physical Geography, Faculty of Social Sciences, University of Mohaghegh Ardabili, Ardabil, Iran
Abstract
The purpose of the study was to assess the accuracy of snow depth estimates from two reanalysis stations, ERA5-Land and ERA2, in the Northwestern region of Iran, for a period of twenty years (2004-23). For this purpose, 10 selected synoptic stations were used as proxy study sites. Ground station data were obtained from the Iranian Meteorological Organization and NASA data for MERRA-2 and Copernicus data for ERA5-Land were obtained for both reanalysis stations. The Quantile mapping method has been used to downscale the output of both databases. The accuracy of the two reanalysis stations was evaluated by comparing the estimated snow depth with the ground-synoptic weather station over a monthly time scale using the KGE, Taylor, Pearson correlation coefficients and RMSE. ERA5 Land Reanalysis is able to determine snow depths in the northwest of Iran in winter with more accuracy than ERA 2. The results show that ERA5 Land Reanalysis database performed best at the Mahabad, Khoy, Tabriz, Ahar, Urmia and Marghheh stations. The Kling-Gupta index was very close to 1 and the best database performance in the period was in January. According to this study, the maximum difference in bias for snow depth assessment between two reanalysis databases over a 20-year period was 542 in December. The largest spatial differences between the two databases were observed at the Ahar Station in December and February and at the Tabriz and Maragheh Stations in January. The results also showed that the uncertainty of snow depths data from ERA5 Land Reanalysis increases as one moves eastwards from the western Iranian region.
Introduction
Snow is an important event in mountainous areas and at high latitudes. Since snow causes transport and agricultural problems due to the temperature drop, it should be investigated. Snow evaluation without reliable data cannot produce desired and effective results. Due to the constraints of mountainous areas in the construction and maintenance of ground stations and the lack of synoptic stations, there is also a lack of coverage of data. One suitable alternative is to use data from reanalysis databases, which can be a good option for researchers if they have a suitable horizontal resolution and are compatible with the site being investigated. The accuracy and effectiveness of these data shall be evaluated before their use.
Material and Methods
After obtaining data from both reanalysis databases on a monthly time scale, the snow depth parameter has been converted from metres to centimeters for consistency. The accuracy of each set of data from both databases was evaluated using a Taylor plot and compared with observation data. For this purpose, 30 percent of the study stations (three stations) were selected and evaluated on the basis of their respective dispersion, different climatic conditions, distance between them (in km) and different altitudes. Due to poor snow depth data results from both databases in the Taylor diagram, the raw output of the two reanalysis databases was scaled down using the quartile mapping method. In the next step, the downscaled results of the three selected stations were compared with the raw snow depth values of all three stations. The calculations showed that the downscaled results were better and further steps were taken with downscaled station data (10 stations). Due to the geographical location of the study area and the mountainous topography, as well as favourable conditions for snowfall in the winter period, the snow depth measurements were performed in January, February and December in the northern part of Iran. Then, using statistical indices Pearson correlation, Root Mean Square Error and Kling-Gupta, seasonal and spatial snow depths were evaluated using ERA5 and MERRA2 reanalysis databases and R software. For each station, the Percentage Bias Index (PBIAS) was calculated separately from the output of the two reanalysis databases for each month of the winter period and the difference between the two databases was averaged over each station. The calculated bias difference was zoned and plotted separately in the GIS for each period of three months. Finally, to provide a good visual overview and to separate the results of the two reanalysis databases for determining the snow depths in the northwest of Iran at 10 stations, a box plot was drawn in R software.
Results and Discussion
According to Taylor's statistical measures, before using snow depth data in both databases, a downscaling operation is required. In this study, the ERA5 land reanalysis data performed better than MERRA2, both in terms of spatial (station) and temporal (seasonal) aspects. Based on the results of the Taylor diagram, the reduced data from both reanalysis databases were used to assess snow depths. A bar graph was obtained for each synoptic station for winter over a 20-year period. Among the 10 synoptic stations, Tabriz, Khalkhal, Meshkinhahr and Urmia stations (from different parts of the study area according to altitude, geographical location and climatic conditions) were selected to represent the winter season. In January, the correlation coefficient, RMSE and KGE showed that the accuracy of the ERA5 land reanalysis snow depth estimates is higher than the MERRA2 snow depth estimates. Compared to ERA5 land data, the Root Mean Square Error was lower and the Kling-Gupta and correlation coefficients were higher than those of MERRAs 2. The maximum accuracy of the ERA5 Land database was observed at Tabriz synoptic station. In December, the best performance of the mentioned database was recorded by the Urmia station in December with a Kling-Gupta index of 0.9 for ERA5 land output. The seasonal distribution of the two databases, based on the difference in bias, showed that the maximum spatial difference in snow depth measurements was in January at Tabriz and Maragheh stations (in the central part of the study area), and in February and December at Ahar station (in the northern part of the study area).
Conclusions
The results showed that the raw data from the two reanalysis databases evaluated in this study cannot determine snow depths in the northwest of Iran and the output of the reanalysis data needs to be downscaled before it can be used. The results also demonstrated the reliability of the Quartile Mapping (QM) bias correction method for the variable snow. According to the study, ERA5 land reanalysis is very accurate in comparison with ERA2. The results of this study showed that the ERA5 Land reanalysis database is performing best at the Mahabad, Khoy, Ahar, Rahimabad and Maragheh stations.
Keywords
Main Subjects
)
Send comment about this article