The status report reviews some of the important applications of artificial neural networks (ANNs) in surface water hydrology, highlighting their advantages and limitations. The review also covers the basic aspects of ANNs, various ANN architectures and learning algorithms.
An ANN is a computational method inspired by studies of the brain and nervous systems in biological organisms. ANN represents highly idealized mathematical models of our present understanding of such complex systems. One of the characteristics of the neural networks is their ability to learn.
A neural network is not programmed like a conventional computer program, but is presented with examples of the patterns, observations and concepts, or any type of data it is supposed to learn. Through the process of learning the neural network organizes itself to develop an internal set of features that it uses to classify information or data.
Due to its massively parallel processing architecture the ANN is capable of efficiently handling complex computations, thus making it the most preferred technique today for high speed processing of huge data. These characteristics render ANNs to be very suitable tools for handling various hydrological modeling problems.
The possible applications of ANNs in the field of surface water hydrology are numerous. They can be used for rainfall-runoff modeling, calibration of rainfall-runoff models, precipitation estimation from remotely sensed information, estimation of rainfall rate from satellite infrared imagery, retrieval of snow parameters from microwave measurements, flood forecasting, drought estimation, multivariate modeling of water resources time series, modeling of water retention curves, deriving general operating policy for reservoirs and for prediction of water quality parameters.
Download the report here:
/articles/applications-artificial-neural-networks-surface-water-hydrology-research-report-national