investigating the impact of climatic components on daily rainfall simulation (case study: khorramabad station)

This study investigates the performance of three models random forest, gaussian process regression and contemporaneous autoregressive moving average in simulating rainfall values at a rain gauge station, khorramadad, iran base on canesm2 predictors. the models were evaluated using root mean square error and nash-sutcliffe efficiency statistics to determine their predictive accuracy and efficiency. in the training phase, rf model exhibited an rmse of 3.98 mm and an nse of 0.32, indicating moderate predictive accuracy and efficiency. gpr showed improved performance with an rmse of 2.55 mm and an nse of 0.67, reflecting better predictive accuracy and a higher level of efficiency than rf. carma model demonstrated the best performance, achieving an rmse of 1.2 mm and an nse of 0.94, signifying high predictive accuracy and excellent efficiency. in the testing phase; the progressive improvement in rmse values from 4.8 mm (gpr) and 4.1 mm (rf) to 1.3 mm (carma) across the models highlights the increasing accuracy in rainfall simulation. similarly, the nse values, ranging from 0.15 (gpr) and 0.2 (rf) to 0.93 (carma), underscore the enhanced efficiency of the models. the results of a graphical examination of different models in rainfall simulating values at the studied station also showed that the values simulated by the carma model are much more similar in terms of dispersion to the observed values. among the three, carma model stands out as the most reliable and effective model for simulating rainfall values, making it a valuable tool for hydrological studies and water resource management.