A New Approach for Flood Routing Modelling using a Neurofuzzy System ‘For a Contribution to the Protection of Agricultural Land and Crops’

Monitoring the phenomenon of flood routing is of great techno-economic interest and a preventive factor for agricultural land and crops protection. Particularly, modelling the phenomenon constitute an essential component. This work presents the development and evaluation of a numerical model for simulating flood routing by optimising the temporal variation of downstream flow in river reaches. The proposed approach employs an Adaptive Neuro-fuzzy Inference System (ANFIS) as a stochastic, data-driven tool.

Two model structures were tested; A three-input version ANFIS (3 Var) based on inflow, outflow and the discharge derivative (∂Q/∂t) at time t. And an extended four-input version ANFIS (4 Var) that additionally incorporates the derivative at t+Δt. The models were applied to two real-world cases. The first case involved a non-smooth hydrograph from the river wye in the United Kingdom and the second case concerned the Karun River hydrograph in Iran.

In the first case, the four-input model reduced maximum and mean outflow errors at 7.7% and 1.6%, outperforming both the three-input version and the Grey Wolf Optimizer (GWO) benchmark. For the second case, the four-input model achieved mean outflow error of 3%, again exhibiting a distinct and consistent error-reduction pattern. In both cases, simulated and observed hydrographs were nearly superimposed with an average deviation less than 1.8%, confirming the model’s accuracy and robustness. These findings demonstrate that ANFIS (4var) can offers strong predictive capability, high reliability and practical applicability for forecasting and decision-making even for other rivers.