Estimation of simulated precipitation uncertainty by using statistical models (Case study: Northwest of Iran)

In this research, the precipitation data needed for the stations of the study area during the statistical period of 1981-2015 was obtained from the Meteorological Statistics Center. To predict and produce statistical data from 2026 to 2065, the LARS-WG and CLIMGEN models were used. In this study, bootstrap method was used to assess precipitation uncertainty. The correlation between the observed and simulated monthly rainfall data shows that the CLIMGEN model simulates the synthetic rainfall data more accurately. The lowest error of the RMSE and MAE in both of these models is at Jolfa Station And the highest error in both models was measured at Sardasht station. Estimation of precipitation output (mean precipitation) by LARS-WG model with bootstrap method indicates higher uncertainty of LARS-WG model than CLIMGEN model. The variance of precipitation observations also indicates a significant change in confidence intervals in the spring and autumn months. Therefore, with regard to the above, it can be said that in the case of rainfall, the LARS-WG model shows more uncertainty than the CLIMGEN model in most of the studied stations in the study area. The absolute magnitude of annual precipitation error with the output of the CLIMGEN model is less than the magnitude of the error with the output of the LARS-WG model. This indicates a lower error value of the CLIMGEN model in the simulated precipitation spatial uncertainty than the LARS-WG model in the northwestern region of Iran.