Modeling of Hydrological Processes Using Pattern Recognition and Soft-Computing Techniques.
Most of the hydrological processes vary both spatially and temporally and embedded with nonlinearity in spatial and temporal scales. The mechanistic models used to model such processes would require large amounts of high-quality data and a good understanding of the underlying physics to model the nonlinear relationships. Also, many mechanistic hydrologic models ignore patterns and thresholds, and assume that physical relationships hold over the entire range of hydrological conditions. They also ignore scaling issues - formulae developed at a local scale are used in watershed models at various spatial and temporal scales. The result is a complicated and ad-hoc model calibration process that accounts implicitly for the above-noted shortcomings.
A viable alternative to the mechanistic modeling approach is to make use of soft-computing techniques, such as Clustering Techniques (CT), Artificial Neural Networks (ANNs), Wavelet Networks (WNs), Wavelet Analysis (WA), Genetic Algorithms (GAs), and Genetic Programming (GP) to model the hydrological processes. These techniques map the inputs to outputs without the need to identify the physics a priori. The aim of this research project is to investigate the possibilities of employing pattern recognition and soft-computing techniques to build robust and parsimonious hydrologic models.