Validation of temperature and precipitation variables of CMIP5 models in Iran under CORDEX-WAS and NEX-GDDP projects

ABSTRACT
No study has so far evaluated the NEX-GDDP and CORDEX-WAS downscaling methods to validate the output of CMIP5 models in Iran for temperature and precipitation parameters. Therefore, this study is the first in Iran to compare the MPI-ESM-LR model performance from the CMIP5 model series for temperature and precipitation variables with the combined approach of dynamic and statistical downscaling methods for the historical period of 1980-2005. The verification was performed using the MBE, RMSE, and R statistics. The slope of the data trend in the time series is estimated using Sen’s non-parametric method. The findings revealed that degrees of bias equal to -0.34 and -0.46 C were recorded in CORDEX and NEX-GDDP projects, respectively, indicating the better performance of the MPI-ESM-LR model under the CORDEX dynamic downscaling project than the NEX-GDDP statistical project in temperature simulation. In both projects, the maximum and minimum temperatures were simulated in Iran's southern coasts and north-western heights. The MBE index shows a decreased bias in the NEX-GDDP project (-2.60 mm) compared to the CORDEX project (-8.21 mm), suggesting the better performance of the MPI-ESM-LR model in the NEX-GDDP project than the CORDEX project in precipitation simulation. Both projects' maximum and minimum precipitations were simulated in the Zagros highlands and the southeast of Iran, respectively.
Extended abstract
Introduction
Currently, most of the research on climate change relies on GCMs. These models are an important tool for simulating and predicting past and future climate changes in various research fields. To develop a climate change-resistant strategic plan, policymakers and decision-makers should be informed of the potential changes in the forecasted future climate. To this aim, there is a need to emphasize an in-depth study of uncertainty using CMIP5 models with CORDEX-WAS dynamic downscaling and NEX-GDDP statistical methods on a regional scale. No study has so far evaluated the NEX-GDDP and CORDEX-WAS downscaling methods to validate the output of CMIP5 models in Iran for temperature and precipitation parameters. Therefore, this study is the first in Iran to compare the MPI-ESM-LR model performance from the CMIP5 model series for temperature and precipitation variables with the combined approach of dynamic and statistical downscaling methods for the historical period of 1980-2005.
 
Methodology
To verify the accuracy of the air temperature and precipitation data extracted from the MPI-ESM-LR model from the CMIP5 model series, 49 synoptic stations were selected in Iran during the statistical period of 1980-2005 (according to the historical data of CORDEX-WAS and NEX-GDDP projects). The verification was performed using the statistics of mean bias error (MBE), root mean square error (RMSE), and Pearson correlation coefficient (r). The simulated temperature and precipitation data were evaluated by the selected model with station data (observed data). This research uses CORDEX-WAS range data with a spatial resolution of 0.44 arc degrees, the RCA4 model for RCM, and the r1i1p1 ensemble. The output is also obtained from the NEX-GDDP downscaling project for Iran according to what is implemented for CORDEX-WAS. The slope of the data trend in the time series is estimated using Sen’s non-parametric method.
 
Results and discussion
In the temperature variable, the MPI-ESM-LR model shows a correlation coefficient 0.99 in both projects. The RMSE indexes are equal to 0.78 and 0.51 °C in CORDEX and NEX-GDDP projects, respectively. Degrees of bias equal to -0.34 and -0.46 C were recorded in CORDEX and NEX-GDDP projects, respectively, indicating the better performance of the MPI-ESM-LR model under the CORDEX dynamic downscaling project than the NEX-GDDP statistical project in temperature simulation. The temperature downtrend slopes in each decade were calculated at -0.848 °C in the synoptic station, -1.191 °C in the CORDEX project, and -1.075 °C in the NEX-GDDP project. In both projects, the maximum and minimum temperatures were simulated in Iran's southern coasts and north-western heights. In the precipitation variable of the NEX-GDDP project, correlation coefficients of 0.85 and 0.65 were obtained in the CORDEX and NEX-GDDP projects, respectively. The RMSE index shows error values of 9.53 mm in the NEX-GDDP project and 6.52 mm in the CORDEX project. The MBE index shows a decreased bias in the NEX-GDDP project (-2.60 mm) compared to the CORDEX project (-8.21 mm), suggesting the better performance of the MPI-ESM-LR model in the NEX-GDDP project than the CORDEX project in precipitation simulation. Except for a precipitation downtrend of -11.766 mm per decade in the synoptic station, the uptrend precipitation slops of 8.513 mm and 12.524 mm were simulated in CORDEX and NEX-GDDP projects, respectively, in each decade. Both projects' maximum and minimum precipitations were simulated in the Zagros highlands and the southeast of Iran, respectively.
In the temperature variable in the CORDEX project, the high correlation coefficients in the majority of regions in Iran indicate the high accuracy of the model in temperature simulation. Under this project, the highest error in the RMSE index was observed in Chabahar and Jask in the southeast of Iran, and the lowest error was noticed in the extreme western slopes of Zagros. In the MBE index, the performance of the model in temperature simulation under the mentioned project shows a temperature overestimation or a positive bias in the southern coasts and western highlands and a negative bias in the Alborz highlands and low-altitude interior regions. In the NEX-GDDP project, a correlation of 0.99 exists between observed and simulated data in the whole country. In the RMSE index, the maximum error is visible on the west coast of the Caspian Sea and Ardabil, and the minimum error is seen in Jask, Chabahar, Iranshahr, and Zahedan (the southeast of Iran). In the MBE index, a positive bias was recorded in the same regions and on the eastern coasts of the Caspian Sea, versus a negative bias recorded in other regions of Iran.
In the precipitation variable under the CORDEX project, the correlation coefficient statistics in the heights of Binalud and Aladagh range from 0.99 to 0.92 in the north-east of the country and the heights of the western Zagros and up to 5. 0 in the Caspian Sea coasts and the coastal areas of Oman (southeast of Iran). In the CORDEX project, the maximum negative bias in the MBE index was observed in the western shores of the Caspian while the maximum positive bias belonged to Tabriz and Khorramabad. In the RMSE index, the minimum error was seen in Zabul and Birjand in eastern Iran. In the NEX-GDDP project, a good correlation coefficient of > 89% was obtained in 81.5% of the country, indicating that the simulated data is close to the real data. In the RMSE index, the maximum error was recorded on the country's northern coasts. The minimum error is between 6.8 and 1.4 mm in Alborz heights and pitfalls in the central and eastern parts of the country. Caspian coasts show the highest negative bias or underestimation in the MBE index. The maximum positive bias was estimated in the Zagros highlands, central pitfalls, and the northeastern highlands of the country.
 
Conclusion
Climate change-driven disasters can be prevented to a large extent, provided that warnings of adverse weather conditions are taken seriously. The need to pay attention to risk management and increasing resilience in climate change conditions caused by global warming can account for a road map in this context. Given Iran's mostly dry climate, temperature and precipitation changes necessitate an integrated management plan for water resources and a long-term vision of the relevant managers and officials in the country. This is because climate change leads to challenges in various environmental, agricultural, food security, social, economic, cultural, political, and international fields.
 
Funding
There is no funding support.
 
Authors’ Contribution
All of the authors approved the content of the manuscript and agreed on all aspects of the work.
 
Conflict of Interest
Authors declared no conflict of interest.
 
Acknowledgments
We are grateful to all the scientific consultants of this paper.