Lake level rise modeling with Artificial Neural Networks (ANN) and Hydrodynamic model

Abstract

The water level in a lake increases as a result of changes in the hydrologic processes in the lake catchment that can cause a flood in the lake's surrounding areas. The lake system needs a comprehensive investigation to alleviate an increase in the water level and consequently flooding. Some different methods and approaches can be applied to predict the lake level increase. A water balance study, hydrodynamic model application, and ANN simulation applications are some of the methods. Each method has merit and complexity in terms of data requirement, the reflection of the physical processes, and ease of modeling application. A simple water balance model application is relatively good in considering all the physical processes. However, model outputs are not good during the model performance evaluation. The hydrodynamic models are simulation models which base on the physical processes of the lake system, with the same modification. Somehow it considers all the real processes with simplification. A good understanding of the application provides a favorable condition to provide all the input/factors contributing/ indirectly to a system modeling. The simulation through a hydrodynamic model is within a good range on performance indicators. Artificial Neural networks (ANN) are recent applications that focus on data science rather than physical processes. The ANN provides an opportunity to include all the contributing factors in the modeling processes. The output from ANN modeling processes is in a very good range. The choice of the method of prediction needs to understand the actual event in a lake system and the model's background for good prediction.