Review of Soft Computing in Water Resources Engineering: Artificial Neural Networks, Fuzzy Logic and Genetic Algorithms by Gökmen Tayfur

Application of soft computing techniques to hydrologic and water resources problems is becoming increasingly popular, especially in areas in which it is difficult to formulate a physics-based model, or as the complexity of the problem at hand increases and the mathematical solution becomes unwieldy. Soft computing techniques can be a convenient method for exploring difficult problems in water resources engineering and may help acquire better empirical insights into complex multiparameter problems and hence reduce the range of plausible solutions or the possible solution field. The volume of literature published using such techniques has grown exponentially in recent decades, and courses introducing them have become almost standard in any graduate-level water resources program.

Soft Computing in Water Resources Engineering is, therefore, a welcome addition to the literature. The book has twelve chapters divided, unequally, into three parts: “artificial neural networks” (ANN), “fuzzy logic algorithm” (FL), and “genetic algorithms” (GA). Several of the chapters are very short. Each section traces the beginnings and evolution of the technique and early applications, followed by chapter(s) discussing the specifics of the technique and concludes with one chapter illustrating its application in the field of hydrology and water resources.

Artificial neural networks (Part I) are discussed in the first five chapters. Chapter 1 gives an overview of how ANN has been inspired by biological nervous system, and then presents different types of ANN and how they compare with other models. The artificial neuron is discussed in detail in Chapter 2 along with an explanation of the methods for computing net information and the different activation functions available, including the ones commonly used in water resources. Chapter 3 deals with network training, starting with data standardization procedures for two activation functions: sigmoid and hyperbolic tangent. Five supervised network training algorithms are presented, followed by learning rules and parameters. Chapter 4 discusses model testing procedures, performance evaluation, and reasons for over-training and how it can be avoided. The last chapter in this section presents a wide range of applications under five categories. The first category is prediction, which includes total suspended sediment, seepage, dispersion coefficient, sheet sediment, runoff at plot scale, runoff at watershed scale, and flood hydrograph at basin scale. This is followed by the classification problem. The third category of application includes flood hydrograph forecasting at the basin scale. The fourth application is extrapolation, and the last category of application is filling gaps in time series.

Fuzzy logic (Part II) is covered in Chapters 6 to 9. Chapter 6 gives a brief history of FL and introduces basic concepts. Chapter 7 discusses the theory of fuzzy set operations, and Chapter 8 deals with FL model construction. Chapter 9 presents applications of FL models in water resources engineering, such as prediction of total suspended and sheet sediment, runoff hydrograph simulation, hydrograph simulation at watershed scale, and dispersion coefficient.

The last 3 chapters of the book are devoted to genetic algorithms. Chapter 10 gives a brief introduction and the theory underlying genetic algorithms. In Chapter 11, a most recently developed variant of GA, incorporating trait-based heterogeneous populations, is presented. Chapter 12 concludes with three case studies of GA applications in hydrology and river hydraulics, including longitudinal dispersion coefficient prediction, hydrograph simulation, and prediction of event-based hydrographs and mean and bankfull discharge.

The book is written in a clear and easy-to-understand manner and will be of great value to those who would want to either employ soft computing techniques for addressing practical problems or investigate these techniques further. Examples presented in the book are a major strength and reflect the author’s experience and knowledge. Although the current edition of the book will be a good addition to the library of any graduate student or faculty member interested in soft computing, the book’s value would have greatly increased if examples at the end of each chapter were provided. Further, because many of the soft computing techniques presented require a certain level of computer programming experience, it would have been beneficial to see sample algorithms or codes written either in MATLAB or R, two commonly used programming languages, that students may use to get started in their application.

The book draws extensively from the author’s work, which covers a wide range of hydrological and water resources problems. However, these techniques have also been applied to a range of other problems, such as reservoir operation, flood frequency analysis, evaporation and evapotranspiration, crop modeling, irrigation scheduling, and soil moisture. It would have been desirable to also discuss applications other than those the author has investigated. Overall, Soft Computing in Water Resources Engineering offers a good introduction to the application of ANN, FL, and GA to hydrological and water resources engineering problems.