The worthiness of using information on land-use–land-cover in watershed models for Western Ghats: A case study

The variable source area (VSA) theory of runoff generation mechanisms has been proved to hold good in many wet mountainous areas, decades ago. According to this theory, infiltration-excess overland flow is limited to very small areas in mountainous and forested catchments. But, the perception that the land surface characteristics, including land-use–land-cover (LULC), form the major factors influencing the response of the catchment to rainfall has dominated the thought in hydrology to such an extent that models based on the overland flow theory continue to be used even in such areas. The present study was taken up in order to understand the worthiness of using parameters, including the curve number (CN), that are based on the physiographic characteristics of the catchment in a watershed model designed to estimate runoff in the wet mountainous areas of the Western Ghats in southern India, where the VSA theory has been proved to hold good. The study has been accomplished by applying the NITK model developed for estimating runoff using daily rainfall data. This model is believed to estimate reliably the streamflow in the region using parameter values that can be computed from catchment characteristics. In the present study, it is applied on three gauged streams in the region of Western Ghats in Karnataka. Initially, the performance of the model has been studied with the parameters fixed using the catchment characteristics. Later, the model has been used as a tool to test hypotheses concerning the catchment response, by varying the parameter values, adopting a trial and error procedure. Initial results showed that the model performance is poor as the coefficients of efficiency vary between –66.9 and 82%. The sensitivity analysis carried out subsequently showed that the model parameters are required to be altered greatly for good performance and that the model simulations are not sensitive to the parameter CN. Further, the performance of this model was compared with that of a VSA model, known to suit the region well. This showed that even after all the changes in the model parameters, the model results are not highly reliable. Hence, in order to understand the reasons for the poor performance of the model, a technique was developed to compute the CN values that would be actually necessary to simulate daily direct runoff (DRO) reliably in this method, the daily values of CN are computed by applying backwards the expression for runoff on the DRO estimated by the VSA model. The variations in the values of CN computed using this method are then studied. It is found that the variations in daily CN are high and highly random too, whereas the NITK model uses only three fixed values of CN. It is thus concluded that factors other than those on which the CN is popularly believed to depend control the runoff generation in the region and that influence of LULC on runoff is not discernible at all from the kind of data that is commonly available.