Drought Prediction Using Compatible Adaptive Neuro Fuzzy Inference System with the Ant Colony Optimization Algorithm in Zabol Watershed

The use of simulations combined with the drought index enhances forecast accuracy. The exploratory approach provides the scope for drought warning and the opportunity for financial support and insurance mechanisms for local communities. The most important part related to the reduction of the effects of drought is the improvement of the precision of forecasting systems.

In this research, the precision of the new Anfis Simulation combined with the meta-heuristics of ant colonies in drought prediction is compared to the normal ANFIS Simulation. Simulation performance is estimated using the Mean Squared Error, Mean Absolute Error and Coefficient of Determination.

Validation of the simulation in the three month profile shows that the Absolute Error Value and the Root Mean Square Error in the Zabol Station are lower by Anfis Simulation. This article examines the Compatible Adaptive Neuro Fuzzy Inference System (ANFIS) and ANFIS models with Ant Colony Optimization Algorithm (ANFIS-ACOR) for drought forecasting. Drought forecasting was done using monthly precipitation data from synoptic stations Zabol (1983-2020) and Zahak (1994-2020). The results showed (ANFIS-ACOR) improved the performance of the Adaptive Neuro Fuzzy Inference System. The model predictive correlation test (ANFIS-ACOR) values in the 3, 6, 9 and 12-month intervals are equal to 0.738, 0.854, 0.801 and 0.898 at the Zabol Station. As well 0.792, 0.804, 0.759 and 0.887 at Zahak Station, respectively. In addition, Anfis Simulation-ant colony has a higher coefficient of determination. Validation of the simulation related to the three-month profile of Anfis-ant Colony Simulation showed that the absolute error and the Root Mean Square Error are lower and the coefficient of determination is higher. In the Zahak Station, the Anfis Simulation is more accurate in the 3-month formation, but in the validation section, the Anfis-ant Colony Simulation is superior. Overall, in the three-month SPI profile, Aanfis Simulation training is more accurate, but Anfis Colony Simulation is superior in validation.

Conclusion and Suggestions :

Ant colony optimization as a function of the data population progressively brings the proposed solutions closer to the overall optimal solution. This problem increased the efficiency of the numerical calculation of the ant colony compared with the Anfis in predicting drought. Anfis Simulation in combination with ant colony provides efficiency for large-scale problems in a short time by auto-tuning. These features reduce the cost of calculating natural data with irregular geometric pattern or data recorded in high volume. Simulations which optimize will fail at the local optimum level. Anfis-Ant Colony Simulation evaluates simulation space for predicting droughts on a global scale. This research proposes to compare the Anfis Simulation with other numerical calculations and to choose the best simulation for the Zabol Watershed.