نوع مقاله : مقاله پژوهشی
1 گروه مهندسی آب-دانشکده کشاورزی- دانشگاه فردوسی مشهد-ایران
2 استاد گروه علوم و مهندسی آب- دانشکده کشاورزی- دانشگاه فردوسی مشهد
3 استادیار، گروه جغرافیا- دانشگاه فردوسی مشهد
عنوان مقاله [English]
Introduction: The lack/inadequacy of climate observational data in different regions can lead to lack of awareness in different climates and low management power in the fields of meteorology, hydrology and agriculture. Today, the development of centers for forecasting and modeling of climatic data has provided access to almost real-time data. Reanalysis data is used in conjunction with static data, or in non-data locations, for instances, The European Centre for Medium-Range Weather Forecasts (ECMWF). The results of MCEVOY et al. (2014) studies on evaluation of 4 network databases with observations showed that the type of climate variable and the spatial resolution of network data has an impact on the results of statistical comparisons. Another Study have shown that ERA-Interim data has a very high accuracy in predicting rainfall in many parts of Iran and its error rate is low and can be neglected in over 70% of the stations studied. The result of the temporal-spatial accuracy of the rainfall data base of the ECMWF on Iran showed that not only the time synchronization but also the amount of similarity was very similar between the estimated values of the ECMWF data base and observed rainfall (Darand & Karimi, 2015). Regarding the advantage of using grid data and examining the possibility of using them, this study evaluates the accuracy of the ECMWF (ERA-Interim) reanalysis data on rainfall estimation and temperature variables in regions with diverse climates in Iran.
Materials and Methods:
2.1. Study area
Five areas of climate diversity (Rasht, Mashhad, Birjand, Bushehr, Kermanshah) are considered in the country. These are located in North, North-East, East, South and west East exponents of Iran.
Daily reanalysis data from precipitation, average temperature, maximum and minimum temperature and dew point temperature from the ERA-Interim database with synoptic stations data at corresponding geographic locations for the statistical years 2015 to 2017 and with The spatial resolution of 0.5 * 0.5 degrees were used. In order to compare, the correlation coefficients (R2), RMSE and NRMSE, the mean bias error (MBE), Nash–Sutcliffe model efficiency coefficient (N_S) and residual coefficient (CRM) were used.
Results and Discussion: The cumulative rainfall in the corresponding coordinates of each station from the ERA-interim model shows a similar trend with the measured data and shows good rainfall fluctuations. Checking the linear regression models fitted between the studied variables (observational and predictive values of the ERA-interim model) indicates that these models are meaningful. The range of R2 for the maximum daily temperature and average temperature variables is above 90%. On average, the lowest and highest standard errors are estimated for precipitation in Birjand and Rasht stations, respectively. In addition, the best result of correlation of Tdew was found in Mashhad station and over/under estimation was shown in Birjand and Rasht respectively. It seems that model has less error for estimating average temperature and maximum temperature. The RMSE range of precipitation is at the selected stations (1.3-7.25) mm. NRMSE with the observed distance indicates that model accuracy is high in predicting rainfall values at all stations, and the error rate varies from 5 to 10%. The range of N-S efficiency at stations and various variables is between -0.3 and -0.9. The distribution of points around the line 1:1 for variables indicates a good relationship between observational data and the ERA-interim model. Pearson correlation coefficients vary from 0.7 to 0.99 and indicate a strong positive linear relationship between observational values and the ERA-interim model in various variables. The slope of the regression equation is between 0.95 and 1.05, which indicates a good relationship between the two data and the high ability of the model. High values of r coefficient show a strong correlation between two time series of data. Regarding the spatial distribution of selected stations, there is a high correlation between the daily time series and observational values and model predictions in all variables that have been investigated by Karimi (1394), Ratii and Sotoudeh (1396), and Sharifi et al 2016) and MCEVOY et al (2014).
Conclusion: In general, the model has been able to simulate the process of time variation of different variables at selected stations, and the accuracy of the model is acceptable. According to the results of this research, the data of this model can be used along with the station data. This study highlights the importance of conducting spatial analysis of observations and potential measurement errors in order to obtain an understanding of the potential deviations of network data before being used in hydro-climatic applications.