Streamflow is an important source of water for industrial and domestic uses. Changing trends in streamflow pose a challenge to water management and may lead to unmet water demand. Our current understanding of the relative importance of the drivers of streamflow changes is not comprehensive. This project investigated the dominant drivers of changes in streamflow over space and time and predicted future streamflow across the contiguous United States. Specifically, this study identified the time-varying dominant drivers of streamflow change at mean annual and seasonal timescales from 1950 to 2009. The hydrological drivers are precipitation, potential evapotranspiration, and other factors represented by an omega parameter in the Budyko framework. The omega parameter can represent factors such as climate seasonality, agricultural drainage, and urbanization. This project also investigated the spatial heterogeneity the surface water balance by considering anthropogenic factors. Future changes in streamflow were predicted under climate and land-use change from 2040 to 2069.

Results show that: 1) Dominant drivers of streamflow in the United States vary over space and time. Precipitation is the dominant driver in wet gets drier watersheds, while other factors are the dominant driver in dry gets drier watersheds. 2) Climatic, physiographic, and anthropogenic controls of the water balance are better understood as spatially heterogeneous. Snowfall and forest coverage are important predictors of spatial-varying Budyko omega parameter in the reference watersheds without human disturbance. Relative cumulative moisture surplus, dam storage, and developed land in riparian areas are important predictors in non-reference watersheds with human disturbance. 3) Random forest models outperformed multiple linear regression and gradient boosting in simulating the temporal variability of controls of water balance. Streamflow is projected to decrease in the central, southwestern, and southeastern United States and increase in the northeast. 4) During the dry season, other factors represented by ω is the dominant driver of streamflow change in most watersheds in the western U.S. In the wet season, effective precipitation is the dominant driver of changes in streamflow in most watersheds after the 1980s.

Intellectual Merit: This study improved our physical understanding of water cycle variability. First, the time-varying drivers demonstrate that factors leading to non-stationary patterns of streamflow can also change. Identification of the time-varying drivers is an important development for understanding time variability and change in hydrology. It suggests that the current labeling of a driver as statically dominant may not fully capture the causes of the non-stationary patterns of streamflow. Second, this study found that climate is a major control of the importance of forest coverage. The spatially varying results help address an unsolved problem in hydrology (i.e., What causes the spatial variability of runoff sensitivity?). Third, research results helped improve the understanding of the physical mechanisms that lead to temporal changes in water availability. The results are derived from the application of machine learning algorithms, which answers the call proposed in Shen et al. (2021) that "the hydrology community is poised to fully explore the power in the vast amount of data using machine learning in various subdomains of hydrology".

Broader Impacts: This project provides implications for water resources management. Hydrological drivers identified as dominant in this project can inform watershed managers of the processes that need to pay close attention to. The spatially varying importance of environmental and anthropogenic predictors is useful for local water resources management. The nation-wide integrated watershed management strategies can be adapted to the local scale by referring to the spatially varying relationship. Prediction of future changes in water availability highlights regions where there are decreases in water availability, providing insights for water and food security. Findings were disseminated to stakeholders, research scientists, and watershed managers in seven academic conferences. Two of the presentations won a best paper award. Results were also disseminated as peer-reviewed journal articles that have already been published (two papers) and under revision after review (one paper).

Last Modified: 03/14/2022
Modified by: Steven Quiring