A diagnostic approach to modeling watersheds with human interference

Most watersheds have human impacts that modify hydrological responses differently over a range of timescales. However, these impacts are not accounted for in most hydrological models. Human impacts in watersheds are diverse and case specific, unlike natural hydrologic processes. Incorporating all plausible human impacts comes at a high data acquisition and modeling cost. This raises the question, which human impacts do we need to incorporate to represent observed streamflow patterns at different timescales? To answer this question, we develop a diagnostic approach to modeling watersheds with human interference. This mixed methods approach is informed by the case history and builds on the top-down hydrological modeling approach where process complexity is incrementally added with changing timescales to identify and respond to changing dominant hydrological processes in a given watershed. Here we implement this modeling approach in the East Fork watershed in California, USA for which data on changes in water imports, withdrawals, irrigation and agriculture land cover is available from the early 1940 ' s, making it an ideal demonstration case. In the East Fork watershed, we find that incorporation of water imports and rights are sufficient to replicate annual patterns of runoff variability, and that adding crop water demand and irrigation enables replication of monthly and daily patterns, while incorporation of groundwater pumping results in negligible improvements. To demonstrate the capabilities of the diagnostic approach in and beyond this case we conducted two computational experiments: checking for needed model structural change and exploring a counterfactual scenario of intensified agriculture.