Abstract
There is a global abundance of non-perennial rivers and streams, of which are predicted to increase due to environmental change and anthropogenic influences. However, most modeled representations of streamflow have been constructed with perennial systems in mind, leaving a gap in our understanding and representation of non-perennial systems. To adapt to future challenges, there is a need to determine what modeled representations of low- and no-flow in non-perennial rivers and streams do well and where uncertainties may lie in the internal representations of hydrologic processes. Here we compare four publicly available process-based hydrologic models: Variable Infiltration Capacity, Precipitation Runoff Modeling System, and National Water Model versions 2.1 and 3.0, in their ability to represent non-perennial streamflow regimes across 156 streamgages that experience non-perennial streamflow behavior in the Pacific Northwest. Our results show that process-based models are largely unable to capture non-perennial streamflow behavior, and that simulation skill decreases as a function of increasing aridity of a streamgage location. Most simulations underestimate the number of no- and low-flow days a streamgage experiences and overestimates the magnitude of low-flows. The ability to accurately model non-perennial systems is paramount to draw inferences about the connections between hydrologic characteristics of low- and no-flow and the potential ecological, biogeochemical, and societal implications of these important systems. Our findings suggest that improving our predictive understanding of non-perennial streamflow of rivers and streams within the Pacific Northwest will fill critical gaps and better target the timing and location of future research, management, and conservation efforts as well as improve the usability of these models for a wider audience of practitioners across fields.