A hybrid inverse-modeling technique to estimate circulation in a steady wind driven open channel flow

• An easy to calibrate high resolution hydrodynamic model for inverse modeling is developed.
• Geostatistical interpolation technique kriging was introduced to model velocity profile in the system.
• Methodology to apply neural network for inverse modeling is described.
• Experimental results validating the inverse modeling technique is reported.

A pollutant transport model for small water bodies is proposed in this paper. This novel model uses kriging and a simple neural network in combination with limited physics to model transport of pollutants in water body instead of trying to model all natural hydrodynamic processes. This technique helps to keep the unknown parameters in the model to fewer numbers compared to models with similar spatial resolution. As such, the model is suitable for ‘inverse modeling’ and predicting the pollutant distribution in water bodies. The model is validated with extensive simulated data and lab scale experiments.