Introduction
Rising temperatures, altered precipitation patterns, and extreme weather events not only affect water resources and food security but can also significantly transform river dynamics, including the transport of suspended sediments. Riverine suspended sediments, as a vital indicator of watershed erosion and aquatic ecosystem health, play a crucial role in water quality, reservoir capacity, and the stability of water infrastructure. Therefore, understanding and accurately predicting future climate changes and their subsequent effects on sediment dynamics are of high importance for optimal water resource management and sustainable development planning. Climate predictions at various scales, including General Circulation Models (GCMs), provide powerful tools for simulating future climatic conditions under different greenhouse gas emission scenarios. After generating future climate scenarios, the next step is to analyze the consequences of these changes on hydrological and sediment processes, which requires more specialized and precise modeling. In summary, considering recent research, General Circulation Models from the sixth report and Artificial Neural Networks are effective tools for estimating climatic parameters and river suspended sediments. In this study, hybrid models of Artificial Neural Network-Wavelet, Artificial Neural Network-Chicken Swarm Optimization, and Artificial Neural Network-Particle Swarm Optimization were used to estimate the suspended sediments of the Keshkan River located in Lorestan Province.


Materials and Methods:
The methodology of this research was designed and implemented in two main stages. In the first stage, historical meteorological data, including precipitation, minimum temperature, and maximum temperature, as well as hydrological data related to daily discharge and suspended sediment load of the Keshkan River, were collected and refined for a long-term statistical period. Subsequently, using the output of a General Circulation Model under two representative scenarios with different levels of greenhouse gas emissions, future climate data for the study area were downscaled. The performance of the climate model was evaluated during the baseline period using various statistical indices. In the second stage, by combining climate and hydrological data, intelligent hybrid models including Artificial Neural Network-Wavelet Transform, Artificial Neural Network-Particle Swarm Optimization (Correction: the previous text mentioned “Chicken Swarm Optimization”, but in the new text “Particle Swarm Optimization” is mentioned. I will stick to the new information provided), and Artificial Neural Network-Particle Swarm Optimization (Correction: it seems there is a repetition, I will assume it refers to “Artificial Neural Network-Chicken Swarm Optimization” as in the original text, or another swarm-based algorithm if specified) were developed to predict suspended sediment load. The performance of these models was evaluated using error metrics, and the best model was selected for predicting suspended sediment load under climate change conditions. Finally, the trend of suspended sediment load changes under different climate scenarios in the future period was analyzed.

Results and Discussion:
The statistical analysis of the model’s simulation for rainfall and temperature parameters during the historical period revealed that rainfall simulation error was higher than other parameters. The capability of the LARS-WG model in simulating meteorological parameters was confirmed, but the model showed less accuracy in rainfall simulation. The performance of the Sixth Assessment Report scenarios in forecasting future fluctuations compared to the baseline period indicated that both models successfully reproduced the seasonal rainfall pattern, which includes a maximum in winter and early spring and a minimum in summer. However, there are differences in absolute rainfall values between the models and scenarios. A reduction in rainfall during specific months, such as summer and early autumn, was predicted by some models in the SSP5-8.5 scenario, which could lead to an intensification of seasonal drought during these periods. The temperature graphs clearly showed a general warming trend in Pol-e Dokhtar county in the coming decades, with predicted temperatures in both SSP5-8.5 and SSP1-2.6 scenarios being significantly higher than observed values throughout the year. For modeling sediment load, a Support Vector Artificial Neural Network model was used with Wavelet, Chicken Swarm, and Particle Swarm Optimization algorithms. According to the evaluations, hybrid structures had less error compared to individual structures. Therefore, the results of the model evaluation showed that the Artificial Neural Network-Wavelet hybrid model demonstrated better performance in the validation stage, with a correlation coefficient of 0.965, the lowest root mean square error of 0.067, the lowest mean absolute error of 0.034, and the highest Nash-Sutcliffe efficiency coefficient of 0.970.

Conclusions
In general, the findings of this research indicate that climate change, as an influential factor, will alter the hydrological pattern of the Keshkan River basin in Lorestan Province. The results of temperature changes indicated that during the study period, the county is affected by global warming, with temperature changes showing an increase in temperature from 2020 to 2050 in the SSP126 and SSP585 scenarios, respectively. The results from predicting rainfall and temperature fluctuations showed that the BCC-CSM2-MR model predicts much higher rainfall in the months of June to October, while the SSP585 scenario generally leads to higher temperatures than SSP126. The results from the performance of the hybrid models showed that the examined models in a combined structure, including all input parameters, performed better due to increased memory. The Artificial Neural Network-Wavelet model exhibited greater accuracy and less error compared to the other models investigated. The results from predicting sediment load in the coming years indicate a 23% increase in river sediment. This highlights the necessity for serious attention to comprehensive watershed management and the implementation of erosion control and desilting programs.