Artificial Neural Networks (ANNs) and their Application in Soil and Water Resources Engineering

Artificial neural networks (ANNs) are a type of artificial intelligence (AI) widely used in the simulation, prediction, estimation and forecasting of time series for many complex hydrologic and hydraulic problems in recent years. The chapter is designed and intended to serve as a basic conceptual note for beginning soft computing water resource specialists. Soft computing tools made a remarkable contribution, especially in the decision support system of soil and water resources in the current trends of speed and more accurate results. The chapter discusses more on ANN and a little on fuzzy logic, and how these are applied in hydrologic time series and their conceptual understanding in the application process. Further, the discussion also progresses to analyze time series problems such as rainfall-runoff; runoff-sediment transportation; river stage-discharge; groundwater table elevation fluctuations; and spring discharge interaction with rainfall, temperature and evaporation in particular. The applicability of various statistical parameters is discussed as they play a key role in the performance evaluation of the simulation and modeling processes. The chapter is mainly aimed at describing ANN applicability, commonly used algorithms and guidelines for applicability in water science and engineering. In view of the above, this chapter is designed and organized step by step with an ANN description: inputting and pre-processing the time series data, training the network and validating the network to see the performance evaluation of the methodology. An attempt has been made to brief various applied research results on sub-topics. Mathematically, ANN application is viewed as a universal approximator; hence, simple applied mathematical equations have been employed in the chapter with lucid, understandable interpretations. Input vector and architectural design of ANN will be discussed in detail in the following explanatory notes for different design problems as it involves a clear physical understanding of hydrologic as well as hydraulic processes. Further, data pre-processing, selection of input and output variables for a number of designed problems in the chapter are explained.