Drought and its effect on streamflow are important to understand because of the potential to adversely affect water supply, agricultural production, and ecological conditions. The Red River of the North Basin in north-central United States and central Canada is susceptible to dry conditions. During an extended drought, streamflow conditions in the Red River of the North may become inadequate to support existing water supply needs in the basin for agriculture, industry, human use, and aquatic life. To understand potential future low-streamflow conditions in the Red River of the North Basin, the U.S. Geological Survey, in cooperation with the International Joint Commission, North Dakota Department of Water Resources, Red River Joint Water Resource District, and Red River Watershed Management Board, developed a water-balance model of the Red River of the North Basin upstream from Emerson, Manitoba, Canada, and coupled the model with stochastic weather inputs to simulate possible future low-streamflow conditions.

Historical changes in low-streamflow conditions were characterized across the Red River of the North Basin using multiple change-point analysis for 12 streamgages. Across these stations, significant change-point years in 1943 and 1994 marked increases in the magnitude of low-streamflow conditions. During 1920–2015, conversion of primary land (not affected by human use) to agricultural and secondary land was followed by a conversion from smalls grains to corn and soybeans as the dominant crop type. From land-use analysis, 1940–2000 was determined to have relatively stable land use and therefore was used as the calibration period for the water-balance model.

A deterministic water-balance model was developed for the Red River of the North Basin upstream from Emerson, Manitoba. The water-balance model was calibrated with data from 37 U.S. Geological Survey streamgages for 1940–2000 and verified using data for 2001–15. The calibrated water-balance model simulated streamflow distributions that mirrored the seasonal patterns of the observed mean monthly streamflow and the standard deviation of the monthly streamflow data, especially during the fall and winter months when streamflow was lowest. For the verification period, during the low-streamflow months of December through January, the difference between simulated and observed data was similar to the calibration comparison and successfully reproduced seasonal trends in the distribution of streamflow, even when using weather data that were outside the calibration period.

To determine the future risk of low-streamflow conditions in the Red River of the North Basin, a block-bootstrap method was used to generate multiple possible future climates. These stochastically generated weather time series were then input to a water-balance model to simulate a distribution of possible streamflows. Three sets of experiments were performed, with each experiment containing a set of scenarios. The first set of experiments from the stochastic streamflow model were designed to investigate how changes in reservoir management would affect the distribution of low streamflow. Relative to scenario 1 (present-day [2023] reservoir operation), scenario 2 (no reservoir operation) shifted the low-streamflow frequency curves downward, reducing the annual minimum monthly streamflow for the Emerson subbasin. Subbasins were defined by the contributing area upstream from a selected streamgage station. Relative to scenario 1, scenario 3 (regulated streamflow with an increased reservoir capacity of 10 percent) shifted the low-streamflow frequency curves upward for the Emerson subbasin. The magnitude of this upward shift, caused by increased reservoir capacity, was lower than the magnitude of the shift caused by the absence of the reservoirs, which indicates that the streamflow was most affected when the reservoirs were first constructed.

The second set of experiments from the stochastic streamflow model included two scenarios that were performed to better understand how the Red River of the North Basin responds to long periods of low or high precipitation. The results indicate that the model consistently overestimated streamflow, but the relative change between a wet and dry climate state of simulated streamflow distribution reasonably matched the relative change of historical streamflow. Across the subbasins, the model was most accurate for low-streamflow conditions associated with nonexceedance probabilities between 20 and 40 percent.

The third set of experiments from the stochastic streamflow model were done to investigate low-streamflow response across the basin to several drought events. Low-end streamflow was reduced when the basin was exposed to a drought, and the magnitude of the reduction increased with longer or more intense droughts. Compared to the low-intensity drought scenarios, the range of percent reductions (as indicated by the interquartile range) was larger for the high-intensity drought scenarios for all subbasins, and the subbasins of Grand Forks and Emerson had a smaller range of reductions compared to the other three subbasins. The larger drainage area—combined with the large contribution of the Red Lake River and several other Minnesota tributaries that generally experience wetter climate conditions—upstream from the Emerson and Grand Forks subbasins may contribute to the smaller range in reductions under the high intensity scenarios. Comparison of the percent reduction in low-end streamflow among subbasins also indicated that the effects of drought duration and intensity could be cumulative. Combining factors of time and intensity produced a larger reduction in streamflow than when each effect was isolated. The array of drought scenarios can be used to determine how a subbasin would respond to multiple possible future conditions. Based on climate predictions, the drought scenario that best matches a future anticipated drought scenario can be used to estimate a low streamflow response for a given subbasin.