ارزیابی شبیه سازی دما و بارشِ مدل های اقلیمی Cmip5 در مطالعات منطقه ای تغییر اقلیم (مطالعه موردی: مناطق عمده تولید گندم دیم در ایران)

ارائه راهکارهای مناسب سازگاری و کاهش اثرات تغییر اقلیم در هر منطقه ملزم به پیش بینی های صحیح تغییرات اقلیمی در آن منطقه است. که خود این پیش بینی ها به شدت متکی به خروجی مدل های gcm است. اما وجود تعداد زیادی از این مدل ها و خروجی های متفاوت آنها در هر منطقه، باعث سردرگمی محققان در انتخاب آنها می شود. در این راستا عملکرد 21 مدل به روز gcm از cmip5 بر اساس نمره مهارت ارزیابی شد و در ادامه پیش نگری از تغییرات دما و بارش طی سال های 2045-2065 و 2080-2100 تحت سناریوهای انتشار جدید rcp2.6 و rcp8.5 ارائه شد. از روش نگاشت هم فاصله ی تابع توزیع تجمعی، برای تصحیح خطای شبیه سازی مدل ها استفاده شد. ارزیابی منطقه ای شبیه سازی دما و بارشِ مدل های اقلیمی cmip5 نشان داد، شبیه سازی ها با خطا همراه است و می بایستی قبل از استفاده تصحیح شوند. اگرچه تصحیح خطای سبب کاهش خطای غیرسیستماتیک شد اما خطای سیستماتیک در شبیه سازی مدل ها همچنان قابل توجه است. به دو طریق می توان نتایج شبیه سازی مدل ها را بهبود بخشید، اولاً، لحاظ کردن تمام گروه های یک مدل در تحلیل ها و دوماً یافتن ترکیبی بهینه از مدل ها متناسب با منطقه. از این رو ترکیبی بهینه از مدل ها متناسب با منطقه انتخاب شد (مدل های انتخابی). بالاترین متوسط مقدار نمره مهارت برای شبیه سازی متوسط سالانه بارش و دما به ترتیب 0.04 و 38/. مربوط به مدل های انتخابی، در بین ایستگاه ها بود. عدم قطعیت در پیش نگری های تغییرات دما و بارش در قرن حاضر، تحت تاثیر انتخاب سناریوی انتشار، دوره زمانی و مدل ها است.

Evaluation of Simulated Precipitation and Temperature from CMIP5 Climate Models in Regional Climate Change Studies (Case Study: Major Rainfed WheatProduction Areas in Iran)

Introduction: Achieving sustainable practices of mitigation and adaptation to climate change requires accurate projections of climate change in each region. In this regard, Coupled Model Intercomparison Project (CMIP) over the past 20 years has shown a good performance. Therefore, new CMIP5 climate models are expected to be bases for many climate change studies. These models use a new set of emission scenarios called Representative Concentration Pathway (RCP) to project climate change. Climate change is expected to impact wheat production and food security in Iran. So far, no study has not been conducted to regionally project climate change based on new CMIP5 models and RCP scenarios over the major wheatproducing areas in Iran. Our objective was to evaluate the performance of CMIP5 climate models in simulating temperature and precipitation in these areas. In addition, different combinations of climate models were evaluated to select appropriate models in these areas. Materials and Methods: According to the latest data, nearly 60% of rainfed wheat is produced within our study area. The mean monthly temperature and precipitation data were provided by Meteorological Organization of Iran for synoptic stations. Period of 19752005 was considered as a historical period (baseline period). We evaluated outputs from 21 GCMs from CMIP5 climate models for monthly values of total precipitation and mean surface air temperature. One in ten ensembles of each GCM model was evaluated as available. We used model outputs for two emission scenarios i.e. RCP2.6 and RCP8.5, for the future periods of 2045–2065 and 20802100 to project temperature and precipitation changes. We assigned the models into two groups, high resolution (models less than 2° latitude/longitude, highre; 11 models) and low resolution (models greater than 2° latitude/longitude, lowre, 10 models). Output GCM models were used for a grid in which recorded data are available. We applied the equidistant quintilebased mapping method (EDCDF) to correct bias of monthly precipitation and temperature simulated by models in the historical period (19752005) and, then in the future periods. We also used the root mean square error (RMSE), the coefficient of correlation and the skill scores (SS) to evaluate the model performance. Result and Discussion: Average of all ensembles of an individual model outperformed the other ensembles in simulating the historical climate. This superiority is largely caused by the cancellation of offsetting errors in individual ensembles of a model, and also reduces the effects of natural internal climate variability in simulations. Taylor diagram showed, contrary to a simulation of temperature, simulations of precipitation have great variability than observations and the standard deviation of simulated precipitation values was less than that of observations for most used models. The models simulated temperature much better than precipitation across the region. Contrary to precipitation, the simulated temperature did not show a significant difference among the models. Several combinations of models resulted in an improvement in precipitation and temperature simulations. Therefore, a combination of models can be used in regional climate change assessment studies. The models performance for simulating the historical climate was evaluated based on skill score (SS) and Δ (the Euclidian distance from perfect skill, point (1, 1, 1, . . . , 1)). Many different combinations of 21 GCM models were evaluated, which combination of 7 models as selected models yielded a lower Δ and higher skill scores. For multimodal ensemble (MME) mean (All, highre, lowre and Selected, models) Δ value was less than that for individual models. SS values in the simulation of precipitation were more than 3 for 75% of models during the high precipitation months. Uncertainty in the simulation of precipitation during the low precipitation months was more than that of high precipitation months and it was even much more in southern areas (especially in August and September). Uncertainties in temperature and precipitation changes projections were affected by the scenario, the time period and models selected. All models showed biases indicating the fact that direct use of such models in climate change studies (without bias correction) is not recommendable. Although the use of statistical methods for bias correction resulted in a significant reduction of nonsystematic biases, systematic biases were not considerably influenced. Precipitation will increase in northern areas toward the end of the century and a higher reduction in precipitation is anticipated in the southern areas. The average, longterm (2080–2100) temperature increase was 5.5°C under RCP8.5. Further, temperature increase will be greater in the southern regions. Conclusion: Performance of 21 GCMs from CMIP5 climate models were evaluated in major rainfed wheatproduction areas in Iran and temperature and precipitation changes were projected under RCP2.6 and RCP8.5. Taking into account all GCM’s initial conditions (if they are available) leads to a better performance. Simulations of models exhibited biases, so models output must be corrected before they can be used in regional climate change assessment studies. Although bias correction resulted in a significant reduction of nonsystematic biases, systematic biases were not significantly affected. The MME (All, highre, lowre and Selected, models) consistently outperformed individual models for both precipitation and temperature suggesting that a smaller group of models can be used in regional climate change assessment. We recognized a subset of 21 models (7 selected models) based on performance that combination of them can provide the best performance and plausible future projections.