Document Type : Research Paper

Authors

• Sadaf Piri ¹
• Mohamad Ansari ghojghar ²

¹ PhD candidate, Department of Reclamation of Arid and Mountainous Regions Engineering, Faculty of Natural Resources, University of Tehran, Karaj, Iran

² Assistant Professor, Department of Reclamation of Arid and Mountainous Regions Engineering, Faculty of Natural Resources, University of Tehran,, Karaj, Iran

Abstract

Introduction
Drought, as an abnormal and dangerous phenomenon, seriously damages water resources, agriculture, economic sectors, and the environment. Within the framework of comprehensive watershed management, accurate and timely drought prediction is very important. This is a necessity, especially in sensitive and vulnerable areas such as Khuzestan Province, for the use of water resources, management of consumption and increasing the resilience of natural and human ecosystems. If a drought period coincides with a vegetation growth period, it disrupts the ecological balance, leading to changes such as reduced soil moisture, changes in ground surface temperature, and even impacts on evaporation and transpiration processes. AI metaheuristic algorithms are able to predict water demand by examining historical data and environmental factors. These predictions allow managers to make better plans for water supply and prevent waste of resources. Considering the uniqueness of Khuzestan Province in terms of its geographical location and water conflicts in recent years, examining the power and efficiency of artificial intelligence algorithms in predicting and identifying climate change can fill the research gap in this field and, through scientific innovation, have a significant impact on environmental protection and the balance of water resources in the face of drought conditions.
 
Materials and methods
In this research, in order to monitor the drought areas of the stations located in the Khuzestan Province, the Precipitation data during the statistical period (1989-2020), and using the values Standardized Precipitation Index (SPI) were calculated to separate dry and wet years. In the following, the inverse distance weighted (IDW) method was used to interpolate the data obtained from SPI. The FCMR model was used to predict meteorological drought. The FCMR fuzzy regression model is a hybrid method that uses linear regression and fuzzy clustering to model data. The GOW and ACO_R algorithms were used to build the hybrid model.
 
Results and discussion
According to the results obtained from the goodness of fit assessment criteria at eight study stations, the 12-month and 6-month SPIs showed relatively better and more accurate results than the 3-month and 1-month SPIs. In the comparison of the 12-month and 6-month SPIs, the 12-month SPI also showed better performance. The RMSE, R, NS and MAE values ​​decreased, increased, and decreased after combining the GOW catalyst and the FCMR model compared to the individual FCMR model, respectively. The combination of the ACOR catalyst and the FCMR model also increased, decreased, decreased and increased in the RMSE, R, NS, and MAE values ​​compared to the individual FCMR model, respectively. Accordingly, it can be concluded that combining the gray wolf with the FCMR model has improved performance compared to using the individual FCMR model. Combining the ant colony catalyst with the FCMR model can also be used with reduced accuracy and lower performance compared to using the individual FCM model.
 
 
 
Conclusions
In the present study, the accuracy and performance of the individual FCMR model were compared and analyzed with the dual hybrid FCMR-GOW and FCMR-ACOR models at eight synoptic stations in Khuzestan province. According to Tables 2 to 4, the GOW catalyst improved the FCMR model and the ACOR catalyst reduced the accuracy of the FCMR model. At all eight stations, the dual hybrid FCMR-GOW model ranked first with the highest accuracy in predicting SPI. Also, the long-term SPI time windows had higher accuracy than the short-term time windows. Furthermore, there is no significant gap in terms of accuracy and precision between the individual FCMR model and the dual hybrid FCMR-GOW model. Therefore, it can be concluded that considering the increasing costs of the aforementioned dual hybrid models, using the individual FCMR model seems more logical. In general, it can be said that combining individual models with meta-heuristic algorithms does not necessarily mean increasing the accuracy of SPI index modeling.
 

Keywords

• Climate change
• Dry year
• Metasearch algorithm
• SPI
• Watershed management