RS and Geographical Information System–Based Evaluation of Distributed and Composite Curve Number Techniques

Abstract

This paper compares the composite and distributed curve number (CN) techniques in computation of runoff from a hypothetical watershed of 100×100 cells and from a natural medium-sized agricultural Tarafeni watershed located in West Bengal, India. To this end, remote sensing (RS) satellite digital data from 1989 and 2000 were used for land use and land cover classification using a maximum likelihood classifier algorithm of supervised classification for CN-generation. Runoff was estimated for each grid by using a distributed-CN approach and averaged for the whole watershed by using a weighted-runoff approach. The estimated runoff values were compared with those from the traditional composite-CN technique. In both cases of initial abstraction (Ia) taken as 0.2 S and 0.3 S (where S is the potential maximum retention), the estimated runoff caused by distributed CN technique was more than that caused by the composite-CN approach. The difference in runoff values was more for Ia=0.3S than that for Ia=0.2S; runoff caused by distributed CN with Ia=0.2S matched more closely with the observed. Furthermore, the difference was very high for small events, moderate for medium, and low for high rainfall events; it was very high for the watershed exhibiting greater CN variation.