Abstract

Introduction:



Flooding is a natural phenomenon that can have devastating effects on communities and ecosystems, making it a significant concern for disaster preparedness and management. It can cause significant damage to the environment and human life, resulting in damage to property and infrastructure, and can occur gradually or suddenly, resulting in flash floods. Various factors such as global warming, land use and land cover change, and urbanization can exacerbate the impact and frequency of flood events. One of the important aspects of understanding and managing floods is capturing the dynamics of runoff, which is one of the main factors in flood events. Accurate flood risk assessment relies on accurate estimates of peak runoff, which are determined through rainfall-runoff simulations. Accurate discharge prediction is a critical factor in flood control and reducing damage to the environment and infrastructure. In recent years, due to the nonlinear and complex nature of hydrological problems, models based on artificial intelligence approaches have been used. These models are inspired by the nature of living organisms and are capable of solving problems of great complexity and scope.Therefore, in this study, optimization algorithms were used with the aim of combining with the support vector regression model to estimate flood discharge.

Methodology:



The study area is the Kakarezha station located in Lorestan province. This station is located in a river called Kakarezha in Lorestan province, which is one of the permanent rivers of Lorestan province and originates from the southeastern mountains of Al-Ashtar city and Chaghlondi district (Herud) and is known as Kakarezha within the Al-Ashtar city. This river is located between 15°48″ to 49°48″ east longitude and 22°32″ to 52°33″ north latitude and is located in Lorestan province and east of Khorramabad city and forms part of the headwaters of the Karkheh River in the Zagros. In this study, a support vector regression model with wavelet, bat and firefly algorithms was used to model the flood discharge of the Kakarezha River located in Lorestan province. The precipitation parameter corresponding to the flood discharge was used as the input of the model and the flood discharge parameter was used as the output of the model in the daily time period, 2012-2022.





Results and Discussion:



In order to model the flood discharge of the Kakarreza River located in Lorestan Province, a support vector regression model with wavelet, bat and firefly algorithms was used. Also, in the support vector regression model, driving functions called kernels were used. These functions include radial, polygonal and linear basis functions, which were investigated in this study. For this purpose, the precipitation parameter values of the Kakarreza hydrometric station are normalized and then entered into the support vector regression model. In recent years, because the values of the kernel function adjustment parameters are randomly selected in the support vector regression model, optimization algorithms have been used to increase the accuracy and reduce the model error. As is clear in Table 1, all models have better accuracy in the radial basis kernel function. The results of the models under study are shown in Table 1. As is clear from the table, the support vector-wavelet regression model with the highest correlation coefficient of 0.980, the lowest root mean square (m3/s) of 0.168, the lowest mean absolute error (m3/s) of 0.088, and the highest Nash-Sutcliffe coefficient of 0.985 has shown better performance in the validation stage. Therefore, the support vector-wavelet regression model has better performance than the other models under study. The superiority of this model is due to the wavelet transform, which divides the received signals into two high-pass and low-pass categories, and in the high-pass category, the resolution is increased, which causes the maximum signal values to be analyzed with desired accuracy.

Conclusions:

Flood discharge estimation using hybrid models based on support vector regression is an efficient tool in designing hydrological systems. In the present study, a case study was conducted to evaluate the performance of the hybrid meta-heuristic model of support vector regression to estimate flood discharge in the Kakarreza watershed located in Lorestan province. The results of the evaluation criteria showed that the wavelet-support vector regression model has high accuracy and negligible error. Also, according to the graphs examined, the wavelet-support vector regression model has estimated flood discharge values close to their actual values. In summary, the results of this study show that the use of artificial intelligence models based on the support vector regression model approach can be used in the field of flood discharge estimation for other regions of the country and a step towards making appropriate management decisions.