Climate and streamflow characteristics for selected streamgages in eastern South Dakota, water years 1945–2013

Abstract

Upward trends in precipitation and streamflow have been observed in the northeastern Missouri River Basin during the past century, including the area of eastern South Dakota. Some of the identified upward trends were anomalously large relative to surrounding parts of the northern Great Plains. Forcing factors for streamflow trends in eastern South Dakota are not well understood, and it is not known whether streamflow trends are driven primarily by climatic changes or various land-use changes. Understanding the effects that climate (specifically precipitation and temperature) has on streamflow characteristics within a region will help to better understand additional factors such as land-use alterations that may affect the hydrology of the region. To aid in this understanding, a study was completed by the U.S. Geological Survey, in cooperation with the East Dakota Water Development District and James River Water Development District, to assess trends in climate and streamflow characteristics at 10 selected streamgages in eastern South Dakota for water years (WYs) 1945–2013 (69 years) and WYs 1980–2013 (34 years). A WY is the 12-month period, October 1 through September 30, and is designated by the calendar year in which it ends. One streamgage is on the Whetstone River, a tributary to the Minnesota River, and the other streamgages are in the James, Big Sioux, and Vermillion River Basins. The watersheds for two of the James River streamgages extend into North Dakota, and parts of the watersheds for two of the Big Sioux River streamgages extend into Minnesota and Iowa. The objectives of this study were to document trends in streamflow and precipitation in these watersheds, and characterize the residual streamflow variability that might be attributed to factors other than precipitation. Residuals were computed as the departure from a locally-weighted scatterplot smoothing (LOWESS) model. Significance of trends was based on the Mann-Kendall nonparametric test at a 0.10 significance level. Of the 10 streamgages selected, only the Elm River at Westport (in the upper part of James River Basin) did not have a significant upward trend in annual mean streamflow for WYs 1945–2013, whereas only one-half of the streamgages had significant upward trends in annual mean streamflow for WYs 1980–2013. Mean and 7-day minimum streamflows also had upward trends for the spring runoff period (March–May) for most of the streamgages during WYs 1945–2013 and for one streamgage during WYs 1980–2013. Magnitudes of increases in streamflow were as great as 30 cubic feet per second per year for the streamgage on the James River near Scotland during WYs 1980–2013. Precipitation trends for WYs 1945–2013 were not necessarily significant for the watersheds of streamgages with a significant streamflow trend. Annual total precipitation had a significant upward trend for the watersheds of 4 of the 10 streamgages during WYs 1945–2013 and no significant trends for WYs 1980–2013. The most widespread precipitation increase was for September–November, with significant upward trends for the watersheds of 8 of the 10 streamgages during WYs 1945–2013; however, no trends in September– November precipitation were significant for WYs 1980–2013. The greatest magnitude of increase in precipitation was for the December–May season during WYs 1980–2013, which had a mean increase of 0.106 inch per year in the watersheds of streamgages with significant trends. The correlation between streamflow and precipitation metrics was low as indicated by the mean coefficient of determination (R 2 ) of 0.18 for all pairs considered. The highest locally-weighed scatterplot smoothing (LOWESS) correlation was between annual precipitation (by water year) and annual mean streamflow (by water year), which had a mean R 2 of 0.47 for all streamgages and was as high as 0.72 for one streamgage. The correlation between annual precipitation and March–May mean streamflow had a mean R 2 of 0.33 for all streamgages and was as high as 0.52 for one streamgage. Other metrics had R 2 values for LOWESS correlations that were less than 0.3 and were not further considered for analyses of residuals. For annual precipitation as a predictor of annual mean flow, precipitation-removed streamflow had significant upward trends during WYs 1945–2013 for one-half of the streamgages. Upward trends in residual annual mean streamflow were indicated for the Whetstone River and lower part of the Big Sioux River Basin, indicating that other factors are contributors to streamflow variability during WYs 1945–2013. In contrast, most of the streamgages in the James and Vermillion River Basins had no trends in residual annual mean streamflow, indicating that streamflow trends can be explained primarily by precipitation. Precipitation-removed streamflow had fewer trends during the more recent analysis period of WYs 1980–2013 than WYs 1945–2013 for all streamgages in eastern South Dakota. Upward trends in residuals for March– May mean streamflow were indicated for Skunk Creek at Sioux Falls and the Big Sioux River at Akron, but trends in residuals were not significant at the remaining streamgages. For the streamgages with significant trends in residual streamflow (such as the streamgage on the Whetstone River and streamgages in the Big Sioux River Basin), land-use changes likely are minor factors, with the main factors probably being changes in the timing and frequency of large precipitation events and persistently wetter antecedent conditions. Changes in the relation between precipitation and streamflow since 1945 were evident when considering the runoff efficiency of the watershed. For example, the streamflow response to annual precipitation of 25 inches for the James River near Scotland increased from approximately 1,000 cubic feet per second for WYs 1945–1990 to about 2,500 cubic feet per second for WYs 1991–2013. The importance of antecedent conditions on annual mean streamflow also was indicated by the significance of the multiple linear regression coefficients of annual mean streamflow and precipitation from preceding water years for all but one streamgage. In addition, rising groundwater levels are present in wells in eastern South Dakota, particularly since the 1980s.