Runoff sediment process modeling is highly variable and nonlinear in nature. The kashinagar watershed of Vamsadhara river basin (Orissa) has been taken under the study. The linear and nonlinear daily sediment yield prediction models based on concept of determining and assigning the varying weightages to the antecedents events for rainfall-runoff-sediment process were developed for watershed. Only active daily runoff data along with antecedent runoff index and antecedent sediment index were used as input for development of the first data set (1984–86) and the second data set (1992–94) models. The daily linear dynamic models were not found suitable for study area due to low value of R2 (less than 50%). However, the value of R2 for the first and second dataset, non-linear sediment yield prediction models were found equal to 77% and 67%, respectively. It was found that the Kashinagar watershed fluvial system exhibits a strong memory on daily basis. The prediction of sediment yield is very much important in order to adopt the suitable soil conservation measures in the watershed for minimizing the sediment load in the reservoir to increase the life of structure.
Absolute prediction error, Antecedent runoff index, Antecedent sediment index, Daily dynamic model, Sediment yield
