اريا سروي

This research aimed to investigate the impacts of climate change on temperature an‎d precipitation in the Caspian region an‎d the runoff of the Shafarood River. Emphasis was placed on comparing downscaling methods an‎d eva‎luating global datasets, presenting a three-stage framework: reference data selec‎tion, GCM selec‎tion, an‎d comparison of downscaling methods with runoff simulation. In the first stage the accuracy of three datasets (CPC, APHRODITE, an‎d ERA5) was eva‎luated directly using statistical indices an‎d indirectly through runoff simulation using the SWAT model in the Shafarood watershed. In the second stage, climate models were ranked from 265 variants belonging to 27 CMIP6 models, using 9 indices an‎d the TOPSIS method, based on synoptic an‎d ERA5 data at daily an‎d monthly scales. Subsequently, a set of downscaling an‎d error reduction methods, including machine learning models (SVM, RF, an‎d XGBoost) for classification an‎d regression, the statistical model SDSM, an‎d distribution-based methods (QM an‎d EDCDF improved), were eva‎luated individually an‎d in combination. The proposed hybrid approach included combinations of classification-regression-distribution matching, as well as regression-quantile mapping an‎d classification-regression. In these models, the K-means++ algorithm was first used to improve class balance, an‎d two indices, ALCE an‎d ALCEₜ, were introduced to address the bias in the Accuracy metric in the presence of no-rain days. In the third step, the selec‎ted GCM outputs, adjusted using the optimal downscaling method, were input into the SWAT model. The model was calibrated using SWAT CUP an‎d the SUFI-2 algorithm during the calibration period (2000–2010) an‎d showed acceptable performance during the validation period (2010–2014) with R² an‎d NSE values of 0.77 an‎d 0.76, respectively. According to the results of this study, ERA5 data were selec‎ted as the reference dataset, an‎d the INM-CM4-8 model was identified as the most suitable GCM. Furthermore, for downscaling an‎d error reduction, the models XGB-QM-SVC an‎d RF-QM-SVC performed best for precipitation, while XGB-QM, SVR-QM, an‎d RF-QM performed best for temperature. The assessment of climate change effects under the SSP2-4.5 scenario showed that in the near future (2020–2040) compared to the baseline period (1990–2014), temperature will increase across the region, an‎d a decrease in precipitation is predicted in part of the Caspian region. Hydrological assessments in the Shafarood watershed indicated a decrease in average runoff. Under the SSP5-8.5 scenario, the average minimum an‎d maximum temperatures across the region are projected to increase, while average precipitation an‎d runoff are expected to decrease

ريز‌مقياس‌نمايي , كاهش خطاي آماري , تطبيق‌توزيع بهبود‌يافته , كلاسه‌بندي , تغيير اقليم , مدل‌هاي گردش‌ عمومي جو , داده‌هاي اقليمي جهاني , حوضه‌ي آبريز درياي خزر , شفارود

Downscaling , Statistical error reduction , Improved quantile mapping , Classification , Climate change , GCMs , Global climate data , Caspian Sea watershed , Shafarood

Aria Sarvi

Hossein Afzalimehr