Flood routing in rivers by artificial neural networks.

Commonly used hydrological methods for flood routing in rivers have restrictions in the analysis of complex problems, as for example in the case of sequential flows, lateral flows or river junctions, mainly in cases without hydrometrical information in the whole hydrological network. The characteristics of artificial neural networks make them a possibility for their application to stream flood routing, because they have several advantages with respect to the traditional hydrological methods. Application of artificial neural network to different sample cases, with enough information and selecting appropriate topology, shows that it is possible to obtain results with a similar precision to the hydraulic and hydrological methods, with usually available data in hydrometrical records that are scarce for the application of such methods. Application of artificial neural networks of simple architecture in isolated stream flood routing cases and sequential flows in the hydrological region 30 in Mexico, as well as an annual hydrometrical record in the junction of the Manso and Cajones streams flowing into Tesechoacan river, shows clearly their advantages.