عنوان مقاله [English]
General Circulation Models (GCMs) are indispensable tools for understanding future climate change. However, the utilization of GCM output data requires an evaluation of its accuracy in comparison to observational data. The aim of this study is to compare the outputs of 13 CMIP6 models with observations using two bias correction methods called Quantile Mapping and Linear Scaling, along with a Bayesian model averaging approach. To explore the future changes in temperature and precipitation variables in the Khazar basin, climate model simulations in two scenarios, SSP245 and SSP585, were chosen. The results show that the quantile mapping method, despite its better performance in modifying climate models, is inefficient at the stations where the bias is unstable over time and the correction is not done properly. Also, using the Multi Models Ensemble (MME) provides a higher level of accuracy than individual models. Although averaging models after bias correction reduces the extreme fluctuations of the time series, it has better accuracy and performance than the second method (averaging models before correction and then applying bias correction). The findings show that the average precipitation in the Khazar basin for the two scenarios, SSP585 and SSP245, is not statistically different from the observational periods and ranges from 0% to 6% for both the near-future periods (2050–2021) and the far-future periods (2080–2051). The average temperature will rise 1.67 and 3.14 (1.8 and 3.65) ℃ more than it did during the observational periods for the near-future (2050-2021) and far-future (2051-2080) periods, respectively, under the SSP245 (SSP585) scenario.