ARTIFICIAL NEURAL NETWORK BASED RAINFALL RUNOFF MODEL FOR FLOOD PLAIN MAPPING USING SATELLITE IMAGES

Abstract

Flood hazard is one of the most severe problems in Himalyan basins. Floods are responsible for the formation of flood plains, which are built of alluvium, the sediments carried, deposited and reworked by the river. These flood plains are very productive agriculturally and are densely populated. Altough flood are essentially hydrological phenomenon, the uneven distribution of floods in the river basin highlights the control of geomorphological and geological factors. Remote sensing data is of immerse value in evaluating the geomorphological and geological features during flooding event. Landsat imagery has been found a prolific source of remotely sensed data for the mapping of flood plains due to its broad synoptic coverage and availability in multiband and multidate, while the conventional ground survey and mapping methods are costly and time consuming. Rainfall-runoff is one of the most important hydrological variable used in most of the water resource applications. Runoff information is needed in dealing with many watershed development and management problem. Remote sensing technology combined with the geographic information system (GIS) can augment the conventional methods to a great extent in rainfall-runoff studies. The remote sensing data provides spatial information on land cover, soil as input to SCS model. In the present study, the area of interest is Rapti river basin of U.P for which an Artificial neural network (ANN) model has been developed for estimating runoff but, before that Curve number (CN) was estimated as this is an important factor in SCS model for runoff estimation. Finally a small area has been selected for mapping the flood plains of the river. Remotely-sensed data in the form of satellite images and toposheets at scale 1:50,000 were used in conjunction with ancillary data to study flood related features of River Rapti in the state of Uttar Pradesh, India. The flood plain areas have been delineated by considering the presence of suitable flood susceptibility indicators such as land use boundaries, soil type, cropping pattern, abundant channels etc.