Search for: All records

Large-scale hydrologic models are increasingly being developed for operational use in the forecasting and planning of water resources. However, the predictive strength of such models depends on how well they resolve various functions of catchment hydrology, which are influenced by gradients in climate, topography, soils, and land use. Most assessments of hydrologic model uncertainty have been limited to traditional statistical methods. Here, we present a proof-of-concept approach that uses interpretable machine learning techniques to provide post hoc assessment of model sensitivity and process deficiency in hydrologic models. We train a random forest model to predict the Kling–Gupta efficiency (KGE) of National Water Model (NWM) and National Hydrologic Model (NHM) streamflow predictions for 4383 stream gauges in the conterminous United States. Thereafter, we explain the local and global controls that 48 catchment attributes exert on KGE prediction using interpretable Shapley values. Overall, we find that soil water content is the most impactful feature controlling successful model performance, suggesting that soil water storage is difficult for hydrologic models to resolve, particularly for arid locations. We identify nonlinear thresholds beyond which predictive performance decreases for NWM and NHM. For example, soil water content less than 210 mm, precipitation less than 900 mm yr−1, road density greater than 5 km km−2, and lake area percent greater than 10 % contributed to lower KGE values. These results suggest that improvements in how these influential processes are represented could result in the largest increases in NWM and NHM predictive performance. This study demonstrates the utility of interrogating process-based models using data-driven techniques, which has broad applicability and potential for improving the next generation of large-scale hydrologic models. 

The protection of headwater streams faces increasing challenges, exemplified by limited global recognition of headwater contributions to watershed resiliency and a recent US Supreme Court decision limiting federal safeguards. Despite accounting for ~77% of global river networks, the lack of adequate headwaters protections is caused, in part, by limited information on their extent and functions—in particular, their flow regimes, which form the foundation for decision-making regarding their protection. Yet, headwater streamflow is challenging to comprehensively measure and model; it is highly variable and sensitive to changes in land use, management and climate. Modelling headwater streamflow to quantify its cumulative contributions to downstream river networks requires an integrative understanding across local hillslope and channel (that is, watershed) processes. Here we begin to address this challenge by proposing a consistent definition for headwater systems and streams, evaluating how headwater streamflow is characterized and advocating for closing gaps in headwater streamflow data collection, modelling and synthesis. 

Abstract Non-perennial streams are widespread, critical to ecosystems and society, and the subject of ongoing policy debate. Prior large-scale research on stream intermittency has been based on long-term averages, generally using annually aggregated data to characterize a highly variable process. As a result, it is not well understood if, how, or why the hydrology of non-perennial streams is changing. Here, we investigate trends and drivers of three intermittency signatures that describe the duration, timing, and dry-down period of stream intermittency across the continental United States (CONUS). Half of gages exhibited a significant trend through time in at least one of the three intermittency signatures, and changes in no-flow duration were most pervasive (41% of gages). Changes in intermittency were substantial for many streams, and 7% of gages exhibited changes in annual no-flow duration exceeding 100 days during the study period. Distinct regional patterns of change were evident, with widespread drying in southern CONUS and wetting in northern CONUS. These patterns are correlated with changes in aridity, though drivers of spatiotemporal variability were diverse across the three intermittency signatures. While the no-flow timing and duration were strongly related to climate, dry-down period was most strongly related to watershed land use and physiography. Our results indicate that non-perennial conditions are increasing in prevalence over much of CONUS and binary classifications of ‘perennial’ and ‘non-perennial’ are not an accurate reflection of this change. Water management and policy should reflect the changing nature and diverse drivers of changing intermittency both today and in the future.