Abstract
To quantify the influence of agricultural water use on streams, we developed a hydrological model incorporating diversion and return flow processes. To rigorously validate the model, we applied it to the Kinu River watershed with intensive observation of streamflow on both the main and branch channels, all of which experience substantial human influences. We sequentially applied the information related to agricultural water use with different details, trying to identify the details in the operational rules necessary for modelling watersheds. The operational and management rules were divided into eight steps, including the initial model, two for reservoir operation, four for water diversion, and one for water allocation in the irrigation areas. The accuracies of the simulated streamflow along the main stream improved by introducing the first rule on reservoir operation related to the seasonal variation of maximum water levels, whereas detailed information related to the integrated operation of multiple reservoirs was not as effective at improving accuracy. Each rule on water diversion from the main stream improved the accuracies and water balances, suggesting the diversion volume in the initial model was overestimated. The sequential screening also revealed the issues regarding the structural errors in the current model and their potential direction for improvement.
