Purpose and Scope

This report presents a basin-wide approach to trend analysis, which provides insight into how water quality is changing spatially and temporally across the Souris River Basin. Water-quality data and streamflow or reservoir inflow data for 34 sites (30 stream sites and four reservoir sites) and 23 constituents with established WQOs were compiled from 1970 to 2020. Not all sites had water-quality data for all constituents during the entire period. Nine constituents were selected for trend analysis, and data requirements for the trend analysis method combined with data availability for a site-constituent pair determined the number of site-constituent pairs analyzed for trends. To describe spatial patterns of concentrations across the basin, descriptive statistics were computed for all site-constituent pairs having 10 or more observations between 1970 and 2020. To understand short-term temporal water-quality changes in streams (15 years or less) and allow for comparison of trends among the most sites for the same period, a recent trend period from 2009 to 2019 was used to evaluate trends in TDS, selected ions, total phosphorus, and selected trace metals for 12 sites. Likewise for reservoir sites, short-term trends were evaluated for four reservoir sites for TDS, sulfate, sodium, total phosphorus, and total iron from 2000 to 2015. To understand longer term temporal changes in streams and allow for comparison of trends among sites for the same periods, a historical trend period from 1976 to 2019 was used to evaluate trends, and depending on the constituent, consisted of one, two, or three piecewise monotonic trends. Three piecewise monotonic trend periods from 1976 to 1988, 1988 to 2005, and 2005 to 2019 were used to evaluate trends at 10 sites for TDS, sulfate, and sodium, and at nine sites for chloride; four piecewise monotonic trends, 1976–88, 1988–2000, 2000–09, and 2009–19, were used to evaluate trends in total phosphorus for six sites, and two piecewise monotonic trends, 1999–09 and 2009–19, were used to evaluate trends for total iron for five sites. Additionally, a detailed evaluation of the probability of exceedances for constituents consistently exceeding WQOs (TDS, sulfate, sodium, total phosphorus, and total iron) at the two binational sites is presented.

The trend-analysis method used in this report removes natural or hydroclimatically induced variability in constituent concentration because of interannual variability in streamflow. As such, trends are assumed to be related to drivers other than interannual natural hydroclimatic variability, such as fertilizer application, land-use change, changes in agricultural practices, livestock production, and urban or industrial development; however, trends may also be related to longer-term hydroclimatic variability that is not captured by the trend analysis method. Potential factors affecting observed trends are discussed, but a full interpretation of the causation of trends is beyond the scope of this study.

Description of Study Area

The Souris River Basin is a 24,600-square-mile basin in the Provinces of Saskatchewan and Manitoba, Canada, and the State of North Dakota, United States (Vecchia, 2000, fig. 1). The basin topography is characterized by the presence of shallow wetlands or potholes nestled among rolling prairie hills, grasslands, and agricultural fields (International Souris River Study Board, 2021b). The Souris River originates near Weyburn, Saskatchewan; flows southeasterly across the international border near Sherwood, N. Dak.; flows past Minot, N. Dak.; and forms a loop and turns northeast through Verendrye, N. Dak. The river continues to flow northwesterly, crossing back into Canada near Westhope, N. Dak. The Souris River eventually empties into the Assiniboine River, which flows to the Red River of the North at Winnipeg, Manitoba (not shown). The total length of the river is about 729 miles with 358 miles in North Dakota (International Souris River Study Board, 2021a).

The primary land use in the basin is agriculture, which accounts for more than 72 percent of the total basin area, and most of the agricultural use can be attributed to row crops (Falcone, 2018; International Souris River Study Board, 2021b). The primary land use has been agricultural since 1970 and the overall land use has not changed, but changes have been made in the types of crops planted, best management practices applied, and alterations to drainage to improve crop production. Other major land cover includes grasslands (about 12 percent), forests (about 4 percent) and wetlands (nearly 3 percent). Artificial drainage of wetlands has increased with time, but the extent is difficult to quantify owing to lack of complete and comparable datasets across jurisdictions (International Souris River Study Board, 2021b). Urbanization is limited and population is sparse in the basin, with the total population of the Souris Basin, including both countries, estimated to be 157,000 (International Souris River Study Board, 2021b), but land-use modifications to accommodate urban and rural infrastructure have been made with time.

The Souris River Basin is in the prairie pothole region, which has a unique and complex hydrology, sometimes referred to as “fill-and-spill hydrology.” Streamflow varies seasonally, with the highest streamflow generally in the spring because of snowmelt or rainfall on partially frozen soils. During spring, wetlands and potholes are “filled,” but under nonflood conditions the potholes store the water and much of the watershed does not contribute to streamflow in the Souris River (International Souris River Study Board, 2021b). In the summer months, streamflow recedes as runoff diminishes and evaporation increases. Storage in potholes combined with high evaporation rates results in a small percentage, generally less than 1 percent, of the precipitation that falls on the basin and ultimately flows out of the basin (runoff ratio; International Souris River Study Board, 2021b). In fall and winter months, streamflow is low and dominated by groundwater or reservoir discharge. Interannual variability in streamflow is considerable. For example, the annual mean streamflow in 2011 and 2012 for Sherwood was 2,270 cubic feet per second (ft³/s) and 122 ft³/s, respectively. Long-term climate persistence in which precipitation alternates between a wet climate state and a dry climate state, with each state lasting for multiple decades, ultimately affects the streamflow (Vance and others, 1992; Shapley and others, 2005, Vecchia, 2008; Kolars and others; 201624, Ryberg and others, 2016). Under continued wetter conditions, shallow potholes “spill,” increasing contributing drainage area and creating increased connectivity between the landscape and streamflow.

Several studies have shown increases in streamflow in the basin, although the timing of the increase varies depending on the methods used in the study. Using annual maximum 10-day mean streamflow, Kolars and others (2016) identified two distinct equilibrium frequency distributions, one for a dry climate state (1912–69) and one for a wet climate state (1970–2020), for four sites on the Souris River, providing evidence that the wet climate state has increased streamflow for the entire main-stem Souris River upstream from Westhope (see figures 26–28 in Kolars and others, 2016). Hulley and others (2019) and International Souris River Study Board (2021b) detected an increase in annual mean streamflow during 75 years or more of record for tributaries and main-stem sites in the Souris River Basin, but they did not identify an abrupt shift in response to changing climate states. Ryberg and others (2016) reported that precipitation during summer and fall months trend towards higher precipitation starting in 1980, and the International Souris River Study Board (2021b) reported increases in streamflow during the spring, summer, and fall, but not specifically since 1980. In the Red River of the North Basin, which is the adjacent drainage basin bordering the Souris River Basin on the east, an abrupt increase in streamflow around 1993 has been attributed to a shift from a dry climate state to a wet climate state 13 years earlier, in or around 1980 (Vecchia, 2003, 2008; Kolars and others, 2016; Ryberg and others, 2016). A long time lag between the onset of a new climate state and the eventual onset of a new streamflow equilibrium can be caused by soil moisture and surface-water storage (Vecchia, 2008; Kolars and others, 2016). From the studies presented here, a clear abrupt increase in streamflow in the Souris River around 1993 was not evident. Some of the features of the Souris River Basin hydrology, such as prairie potholes and dam control, may make it more difficult to detect an abrupt shift, or it may be that the changes are more gradual or may be more distinct at the lower end of the streamflow distribution. Changes in the amount and timing of runoff from year to year and decade to decade can have a large effect on the relative amount of natural and anthropogenic sources of dissolved ions and nutrients that are transported to streams, which in turn, can result in large year-to-year and decade-to-decade changes in concentrations and loads.

Streamflow in the Souris River Basin is controlled and managed by three reservoirs in Canada and one reservoir in the United States that are operated as a system as outlined in Annexes A and B of United Nations (1989). Reservoir operations for high flows are also described in Appendix A and Annex B of United Nations (1989). Construction of Lake Darling on the Souris River (United States) was completed in 1936 and its primary purpose was to provide water to support fish and waterfowl habitat at the J. Clark Salyer National Wildlife Refuge, which is located 110 miles downstream (fig. 1). Construction of Boundary Dam (Canada) on Long Creek was completed in 1958 and its primary purpose was to provide cooling to the Boundary Dam Power Station. A series of floods during the 1970s and growing energy development in Saskatchewan spurred improvement of existing structures and the construction of new water management structures known as the Souris River Project. Rafferty Reservoir, Boundary Reservoir, Grant Devine Lake, and Lake Darling collectively constitute the Souris River Project (fig. 1, International Souris River Study Board, 2021a). Rafferty Reservoir is the largest reservoir in the system, was constructed in 1991 to provide flood control and water supply benefits, and was filled to its full supply level in 1997. In 1993, modifications were made to Boundary Reservoir to provide water to Rafferty Reservoir. Grant Devine Lake on Moose Mountain Creek started filling in 1992 and because of litigation was not filled to its full supply level until 1997. The primary usage of reservoir releases from Rafferty Reservoir is cooling of the boundary dam power station. Additional releases are made from Rafferty Reservoir during the winter when reservoir levels are drawn down. Releases are made from Rafferty Reservoir and Grant Devine Lake after the spring runoff forecasts are made. Between 1994 and 1998, Lake Darling underwent a major rehabilitation that altered its flood capacity. Lake Darling Dam was raised 0.5 foot and a gated spillway was installed to replace the uncontrolled spillway and emergency spillway. Changes in flow management along with timing and volume of reservoir releases may contribute to changes in water-quality conditions in the Souris River.

Site and Constituent Selection

Thirty-four sites with 10 or more years of water-quality data during 1970–2020 and 23 constituents with established WQOs were selected for analysis (tables 1 and 2). For sites with at least 10 observations for a given constituent, descriptive statistics were computed. Nine constituents were selected for trend analysis (table 2) and data requirements for the trend analysis method combined with data availability for a site-constituent pair determined the number of site-constituent pairs analyzed for trends.

Water-Quality Data Compilation

Water-quality data compiled from multiple agencies collected during multiple decades introduce the potential for inconsistencies between data for the same constituent. Inconsistencies in data caused by field-collection and laboratory-analytical methods were primarily addressed through identifying known or documented changes in field-collection or laboratory-analytical methods, recensoring to a common censoring level, and normalizing data based upon paired statistical testing. Other inconsistencies not addressed by the aforementioned methods were addressed using tools in the software package used for trend analysis and will be discussed in the “Trend Analysis” section. Addressing data inconsistencies as completely as possible ensured that detected trends reflect real environmental changes.

For the two binational sites, an understanding of the history of data collection and statistical testing of paired datasets was required before combining data collected and analyzed by different agencies. At each binational site, water-quality samples were collected and analyzed at varying sampling frequencies by multiple agencies for varying purposes and during different periods and locations; some paired data have been collected with an intended purpose to test comparability of results (table 3). Between 1970 and 1991, Westhope samples were regularly collected by two agencies at different locations: (1) by the USGS at the boundary crossing near Westhope, N. Dak.; and (2) by ECCC 5.1 miles downstream from Westhope, N. Dak., near Coulter, Manitoba. Since 1992, Westhope samples have been primarily collected and analyzed by ECCC at the boundary crossing near Westhope, N. Dak., with occasional samples collected at Coulter, Manitoba. There are likely differences owing to location in some constituent concentrations from samples collected at Westhope and Coulter, and although 10 paired samples were available from ECCC, this was too few paired samples for testing comparability.

Since about 1992, samples have been collected concurrently by the USGS and ECCC once or twice a year at Sherwood and Westhope to test data comparability between laboratory-analytical methods, and these paired samples will be referred to as “USGS–ECCC Sherwood” and “USGS–ECCC Westhope,” respectively (table 3). The field-collection method of a sample can cause differences in sample results, particularly when a lot of sand-sized sediment particles are present or when the sample cross-section is not well-mixed. Primary samples are collected by the USGS at Sherwood using an isokinetic or multiple vertical field-collection method (U.S. Geological Survey, variously dated) and samples are analyzed by the USGS National Water Quality Laboratory (NWQL). Primary samples are collected by ECCC at Westhope using a grab or single vertical field-collection method and samples are analyzed by an ECCC laboratory (Jennifer Bradley, ECCC, written commun., September 2022). For paired samples, the same field-collection method is used as the primary samples, but samples are analyzed by both laboratories; that is, paired samples at Westhope are collected using the grab field-collection method, but samples are analyzed by the NWQL, and paired samples at Sherwood are collected using isokinetic or multiple vertical field-collection method, but samples are analyzed by ECCC laboratory. Because the same field-collection method is used for paired samples, any differences in the paired USGS–ECCC Sherwood and USGS–ECCC Westhope samples reflect laboratory-analytical differences.

Since 1972, Sherwood samples have regularly been collected by the USGS using an isokinetic or multiple vertical field-collection method and analyzed by the NWQL. Starting in July 2018, the same field-collection method has been used but analyzed by two different laboratories: (1) North Dakota Department of Environmental Quality Laboratory (NDDEQL) and (2) NWQL. These paired samples will be referred to as “NWQL–NDDEQL Sherwood” and differences only reflect laboratory-analytical differences. Occasional samples have been collected at both binational sites by the North Dakota Department of Environmental Quality (NDDEQ) using a grab or single vertical field-collection method and analyzed by the NDDEQL. Statistical testing was performed for paired datasets and is described in the “Statistical Testing of Paired Datasets for Binational Sites” section.

At the U.S. stream sites, samples were collected by the USGS or NDDEQ. Most samples at the U.S. stream sites were collected by the USGS, but depending on the site, they were analyzed by different laboratories. At four of the 11 U.S. stream sites (Sherwood, Souris River above Minot, N. Dak. [USGS station 05117500; hereafter referred to as "site 11"] and Souris River near Verendrye, N. Dak. [USGS station 05112000; hereafter referred to as "site 12”], and Westhope), samples were collected by the USGS and analyzed at NWQL, and at the other seven sites, samples were collected by the USGS and analyzed at NDDEQL. Revisions to the North Dakota statewide network sampling design in October 2012 resulted in discontinued water-quality data collection at two U.S. sites (Souris River near Foxholm, N. Dak. [USGS station 05116000] and Souris River near Bantry, N. Dak. [USGS station 05122000; hereafter referred to as "site 14”]; table 1) and increased frequency of collection and addition of constituents at other sites (Long Creek near Noonan, N. Dak. [USGS station 05122000; hereafter referred to as "site 3”], Willow Creek near Willow City, N. Dak. [USGS station 05123400; hereafter referred to as "site 15”], and Deep River near Upham, N. Dak. [USGS station 05123510; hereafter referred to as "site 16”]; table 1; Galloway and others, 2012). For constituents primarily in dissolved form (sulfate, chloride, and nitrate plus nitrite), results from unfiltered and filtered samples were combined. Most results for these constituents were for a filtered sample but results from unfiltered samples were used to supplement the dataset when the filtered result was not available. Filtered and unfiltered samples were used for sodium. All compiled data for ammonia were filtered. Phosphorus data were separated into filtered and unfiltered and were labeled as dissolved phosphorus and total phosphorus, respectively. Samples identified as “supernate” by NDDEQ were grouped with total phosphorus (Nustad and Vecchia, 2020). For TDS, the most complete record was used for each site, which was a measured concentration analyzed by NWQL for Sherwood and sites 11 and 12 and a calculated concentration from the NDDEQL for all other sites. For boron and trace metals, only data from 1999 through 2020 were considered for this study because of substantial changes to USGS sample collection techniques and analysis methods at the NWQL in the mid- to late 1990s (U.S. Geological Survey, 1992, 1993; Hoffman and others, 1996). For boron and trace metals, only unfiltered samples were used and were classified as “total.” All U.S. data were recensored as needed to a common censoring limit to the U.S. and Canadian data (table 2).

For Manitoba sites, multiple analytical laboratories were used, and alternating periods of filtered and unfiltered sample results for the same constituent were identified. Because there were alternating periods of filtered and unfiltered results within MARD major ion data, filtered and unfiltered sample results were combined into one dataset. For MARD total phosphorus results, a laboratory-analytical method change was previously identified as causing a high bias of total phosphorus results during April 2001 to March 2009 (Nustad and Vecchia, 2020). While developing trend models, this method change was tested for statistical significance at each site. For MARD data, only filtered samples were used for boron and trace metals data. As with the U.S. data, Manitoba data were recensored to a common censoring limit as needed for each constituent (table 2).

For the Saskatchewan sites, water-quality data were provided for compilation by the WSA. As with the Manitoba sites, although multiple analytical laboratories were used, WSA provided data for major ions, nutrients, and total suspended solids (TSS) that were comparable and could be combined from the same constituent (John-Mark Davies, Water Security Agency Saskatchewan, written commun., 2022). Data for boron and trace metals were separated between “total” and “dissolved” based upon the description of the sample in the data provided. Samples that were listed as “total” in the Saskatchewan data were assumed to be unfiltered, and “dissolved” samples were assumed to be filtered samples.

Censored values, or values less than the method detection limit for which an exact value is not known (Foreman and others, 2021), need to be considered during trend analysis (Helsel and others, 2020). Although the software package R–QWTREND will estimate censored values, it is recommended that no more than 25 percent of the dataset be censored values (Vecchia and Nustad, 2020). Boron, total ammonia, nitrate plus nitrite, arsenic, beryllium, cadmium, chromium, cobalt, copper, lead, molybdenum, nickel, selenium, zinc, and TSS all had censored values with multiple censoring levels and were recensored to a common censoring level (table 2). Although trend analysis was originally intended for nitrate plus nitrite, total ammonia, total arsenic, and TSS, more than 50 percent of the data were censored for nearly all site-constituent pairs after recensoring to a common censoring level, excluding them from trend analysis. Although 30 percent of the data were censored for molybdenum at Westhope, trends were evaluated but should be interpreted with caution.

Descriptive Statistics

Descriptive statistics were computed for all sites in the Souris River Basin with at least 10 samples collected between 1970 and 2020 to describe the spatial variability of concentrations in the basin. Statistics for boron and trace metals were computed using data between 1999 and 2020 owing to laboratory analysis and sample collection changes (U.S. Geological Survey, 1992 and 1993; Hoffman and others, 1996). Descriptive statistics were calculated on the raw recensored data. Although 10 or more samples were collected at the sites, the number of values for specific constituents varied by site because samples were collected by various agencies or groups for different purposes. The distribution of the data also varied with time. Although a site may have data starting in 1970 and ending in 2020, there could be periods of data missing for many or a few years in between. Statistics included minimum; maximum; and the 10th, 25th, 50th (median), 75th, and 90th percentiles of values for individual constituents at each site. Median concentrations for selected constituents were plotted on a map of the Souris River Basin to show spatial patterns in concentration across the basin (fig. 2).

Trend Analysis

Water-quality trends were evaluated for this study using R–QWTREND, a publicly available software package developed by USGS for analyzing trends in stream water quality (Vecchia and Nustad, 2020). The methodology of the time-series model was originally developed and applied to the two binational sites of the Souris River Basin (Vecchia, 2000). The time-series model was modified in subsequent water-quality studies by the USGS and has been applied to other basins near the Souris River Basin (Jones and Armstrong, 2001; Vecchia, 2003, 2005; Galloway and others, 2012; Nustad and Vecchia, 2020; Vecchia and Nustad, 2020; Tatge and others, 2022), as well as other basins across the United States (Risch and others, 2014; Sando and others, 2014a, 2014b, 2015; Giorgino and others, 2018; Barr and Kalkhoff, 2021). The complex hydrology of the Souris River Basin combined with a multidecadal, multiagency, multiconstituent water-quality dataset required a trend method that could account for many of these complexities. R–QWTREND was used because it has the capability to address many of the complexities of the Souris River Basin water-quality dataset, including the ability to remove variability in concentrations owing to interannual flow-related variability and seasonal variation; remove serial correlation, which addresses variability in sampling frequency; correctly handle censored values (as much as 25 percent of the data); and test for step trends caused by nonenvironmental factors. Examples of nonenvironmental factors include differences in parts of the sample analyzed (filtered or unfiltered), collection method, or laboratory-analytical method. If statistically significant step trends caused by non-environmental factors are detected, data can be corrected prior to analyzing for piecewise monotonic trends.

Flow-Averaged Exceedance Probability

The ISRB WQO, along with other water-quality standards and objectives established by other jurisdictions in the Souris River Basin, were used to evaluate the flow-averaged exceedance probability (EP) for TDS, sulfate, sodium, total phosphorus, and total iron for the binational sites. Two measures of exceedance probabilities were evaluated to describe the probability of exceeding a specified concentration threshold: flow-averaged EP and annual mean flow-averaged EP (Vecchia and Nustad, 2020). The annual mean flow-averaged EP is a measure of the proportion of time during the year concentrations are expected to exceed the concentration threshold, assuming average flow conditions. For example, if the annual mean flow-averaged EP for a given year is 0.25 (or one-fourth of the year), it is expected that the WQO would be exceeded about 25 percent of the time during that year (about 90 days), assuming normal flow conditions. The flow-averaged EP is computed for each 5-day time interval in the period of record and is interpreted as the chance of exceeding the concentration threshold during that time interval, assuming flow conditions were the same year after year. For example, if the flow-averaged EP is 0.5 for June 1–5 of a specified year, there is an equal chance of exceeding the WQO during that time interval, assuming normal flow conditions for that time of year. Concentration thresholds are set by jurisdictions for different purposes and typically depend on the designated use. The most restrictive concentration threshold was listed for each jurisdiction regardless of designated use (table 5). For each constituent, the ISRB WQO was evaluated along with additional concentration thresholds from other jurisdictions. Because the TDS concentration threshold from all other jurisdictions was 500 milligrams per liter (mg/L), an additional concentration threshold of 750 mg/L (approximately the median concentration for Sherwood and Westhope) was used for comparison. For total iron, because concentration thresholds for the ISRB and other jurisdictions were all 300 micrograms per liter (µg/L), the median concentration for Sherwood (550 µg/L) and 75th percentile concentration for Westhope (800 µg/L) were used as additional concentration thresholds (table 1.3; Tatge and Nustad, 2023). Three figures of exceedance probability for TDS, sulfate, sodium, total phosphorus, and total iron at Sherwood and Westhope are presented: (1) the annual mean flow-averaged EP and flow-averaged EP of the WQO during 1976–2019; (2) the flow-averaged EP of the WQO for 3 separate years that represent the start or end of a trend period for most constituents (1988, 2005, and 2019); and (3) the annual mean flow-averaged EP of the WQO during 1976–2019 compared with other concentration thresholds for the same period.

Recent Water-Quality Trends

Trends were evaluated for the recent period (2009–19) for 12 stream sites and nine constituents. The WQOs are only applicable to Sherwood and Westhope, and in practice, raw measured concentrations are compared against the WQOs. To provide basin-wide perspective, it is helpful to use the WQO as a reference to compare the flow-averaged GMC for all sites. For example, if the flow-averaged GMC of tributaries is less than the WQO, these tributaries are likely not contributing to exceedances of the WQO at Sherwood and Westhope.

Total Dissolved Solids and Selected Ions

Trends in TDS concentrations are affected by the same factors that affect the major dissolved ions that constitute TDS. It is possible for one constituent to drive the TDS trend or for individual constituents to have trends in opposite directions, cancelling out the trend in TDS. For this reason, trends in TDS can be more difficult to interpret. Depending on the abundance of any one of the major dissolved ions, a single constituent can dominate the TDS concentration and the dominant constituent can vary by location, local geology, soils, and other factors. In the Souris River Basin, sulfur is abundant in the soils and can account for a substantial fraction of the TDS, as much as about 50 percent for some sites. Also, sodium is another substantial fraction of the TDS and sodium-sulfate evaporites are known to be present in large quantities (Keller and others, 1986).

The annual flow-averaged GMC of TDS increased during the recent period (2009–19) at all sites evaluated except for Sherwood, and by 2019 one-half of the sites had an annual flow-averaged GMC greater than the TDS WQO of 1,000 mg/L (fig. 5; table 6). Most of the significant or mildly significant increasing concentrations for TDS were at sites in the upper basin (fig. 5). Given that sulfate and sodium constitute about one-half of the TDS for most sites, these ions were likely driving the TDS trends. Using the TDS WQO of 1,000 mg/L as a reference, six of the sites evaluated for TDS trends in the recent period had a flow-averaged GMC in 2019 greater than 1,000 mg/L, but only two of these sites (sites 3 and 11) started with concentrations less than 1,000 mg/L in 2009 (fig. 5; table 6). Site 15, Pipestone Creek Bridge at Kola (NE18-10-29W) [MARD station MB05NGS085, referred to hereafter as “site 30”], and Souris River at Provincial Road 530 near Treesbank, Manitoba (MARD station MB05NGS085, referred to hereafter as “site 34”) were the only three sites with increasing flow-averaged concentrations that remained slightly less than the WQO (988, 999, and 929 mg/L, respectively) for the entire recent period (table 6; fig. 5).

Trends in sulfate concentrations may be affected from atmospheric deposition or land-use and climate changes, which may increase or decrease the exposure of naturally occurring sulfur in runoff. Since 1970, precipitation has increased in the Souris River Basin, resulting in increased runoff (Ryberg and others, 2016). Some studies have linked increasing sulfate concentrations to urbanization (Kaushal and others, 2018), but urbanization in the sparsely populated Souris River Basin is minimal. Sulfur is naturally and abundantly present in soils in the Souris River Basin and across North Dakota (Galloway and others, 2012). Sulfur can be reduced and oxidized to produce sulfate ions, which are highly soluble (Hem, 1985). Keller and others (1986) determined that saline soils within North Dakota are dominated by sodium-sulfate salts.

Annual flow-averaged GMC of sulfate increased during the recent period at all sites, with most sites having mildly significant or significant increases and five sites with increases greater than 100 mg/L (table 6; fig. 6). Of the 12 sites evaluated, nine sites had mildly significant or significant increases. The largest increase was at site 3 (263-mg/L increase; 57 percent). Site 4 in Saskatchewan had the smallest increase (nonsignificant increase of about 41 mg/L). Of the nine sites evaluated in North Dakota, five had increases greater than 100 mg/L and the remaining four had increases between 50 and 80 mg/L (table 6). The two sites evaluated in Manitoba (sites 30 and 34) had mildly significant and significant trends, respectively, and the increase was less than 100 mg/L between 2009 and 2019 (table 6). Using the WQO of 450 mg/L as a reference, by 2019, the annual flow-averaged GMC was greater than 450 mg/L for five sites (sites 3, 10, 11, 12, and Westhope) (table 6; fig. 6). Three main-stem U.S. sites (site 11, 12, and Westhope) had significant or mildly significant increases in concentrations going from less than the WQO to greater than the WQO between 2009 and 2019.

Table 6.    

Summary of trend results for the recent period 2009–19 for total dissolved solids and selected ions at selected sites in the Souris River Basin.

[p-value, probability value; GMC, geometric mean concentration]

Table 6.    Summary of trend results for the recent period 2009–19 for total dissolved solids and selected ions at selected sites in the Souris River Basin.

Site location (fig. 1)	Site name	Trend period	p-value	Significance level	Flow-averaged GMC for first year in trend period	Flow-averaged GMC for last year in trend period	Change in flow-averaged GMC between first and last year	Change, in percent from first year to last year
Total dissolved solids, in milligrams per liter
3	Long Creek near Noonan, North Dakota	2009–19	0.0048	Significant increase	916	1,317	401	44
4	Souris River at Highway 39 near Roche Percee, Saskatchewan	2009–19	0.0937	Nonsignificant increase	1,025	1,220	194	19
7	Souris River near Sherwood, North Dakota	2009–19	0.7961	Nonsignificant decrease	861	846	−16	−2
10	Des Lacs River at Foxholm, North Dakota	2009–19	0.0019	Significant increase	1,215	1,448	233	19
11	Souris River above Minot, North Dakota	2009–19	0.2295	Nonsignificant increase	988	1,061	73	7
12	Souris River near Verendrye, North Dakota	2009–19	0.0174	Mildly significant increase	1,051	1,181	130	13
13	Wintering River near Karlsruhe, North Dakota	2009–19	0.0428	Mildly significant increase	685	796	111	16
15	Willow Creek near Willow City, North Dakota	2009–19	0.1030	Nonsignificant increase	841	988	147	18
16	Deep River near Upham, North Dakota	2009–19	0.0656	Nonsignificant increase	680	823	143	21
18	Souris River near Westhope, North Dakota	2009–19	0.4157	Nonsignificant increase	1,030	1,062	32	3
30	Pipestone Creek Bridge at Kola, Manitoba	2009–19	0.2096	Nonsignificant increase	894	999	106	12
34	Souris River at Provincial Road 530 near Treesbank, Manitoba	2009–19	0.0314	Mildly significant increase	863	929	66	8
Sulfate, in milligrams per liter
3	Long Creek near Noonan, North Dakota	2009–19	0.0011	Significant increase	465	728	263	57
4	Souris River at Highway 39 near Roche Percee, Saskatchewan	2009–19	0.5086	Nonsignificant increase	352	392	41	12
7	Souris River near Sherwood, North Dakota	2009–19	0.0794	Nonsignificant increase	282	342	60	22
10	Des Lacs River at Foxholm, North Dakota	2009–19	0.0027	Significant increase	551	659	107	20
11	Souris River above Minot, North Dakota	2009–19	0.0014	Significant increase	362	477	114	32
12	Souris River near Verendrye, North Dakota	2009–19	0.0001	Significant increase	407	522	116	29
13	Wintering River near Karlsruhe, North Dakota	2009–19	0.0012	Mildly significant increase	256	324	68	27
15	Willow Creek near Willow City, North Dakota	2009–19	0.2078	Nonsignificant increase	280	357	77	28
16	Deep River near Upham, North Dakota	2009–19	0.0036	Significant increase	209	329	120	58
18	Souris River near Westhope, North Dakota	2009–19	0.0185	Mildly significant increase	420	475	55	13
30	Pipestone Creek Bridge at Kola, Manitoba	2009–19	0.0357	Mildly significant increase	378	421	43	12
34	Souris River at Provincial Road 530 near Treesbank, Manitoba	2009–19	0.0004	Significant increase	328	410	82	25
Sodium, in milligrams per liter
3	Long Creek near Noonan, North Dakota	2009–19	0.0347	Mildly significant increase	167	228	61	37
4	Souris River at Highway 39 near Roche Percee, Saskatchewan	2009–19	0.8687	Nonsignificant decrease	198	194	−4	−2
7	Souris River near Sherwood, North Dakota	2009–19	0.6449	Nonsignificant decrease	134	128	−6	−4
10	Des Lacs River at Foxholm, North Dakota	2009–19	0.8842	Nonsignificant increase	218	219	2	1
11	Souris River above Minot, North Dakota	2009–19	0.0397	Mildly significant decrease	170	142	−28	−17
12	Souris River near Verendrye, North Dakota	2009–19	0.2947	Nonsignificant decrease	171	161	−10	−6
13	Wintering River near Karlsruhe, North Dakota	2009–19	0.4184	Nonsignificant increase	128	137	8	7
15	Willow Creek near Willow City, North Dakota	2009–19	0.3722	Nonsignificant increase	91	104	13	15
16	Deep River near Upham, North Dakota	2009–19	0.0373	Mildly significant increase	54	67	13	24
18	Souris River near Westhope, North Dakota	2009–19	0.7833	Nonsignificant decrease	158	155	−3	−1
30	Pipestone Creek Bridge at Kola, Manitoba	2009–19	0.0470	Mildly significant increase	69	81	12	18
34	Souris River at Provincial Road 530 near Treesbank, Manitoba	2009–19	0.2978	Nonsignificant increase	110	116	7	6
Chloride, in milligrams per liter
3	Long Creek near Noonan, North Dakota	2009–19	0.0089	Significant increase	19	29	10	52
7	Souris River near Sherwood, North Dakota	2009–19	0.350	Nonsignificant decrease	36	34	−2	−6
10	Des Lacs River at Foxholm, North Dakota	2009–19	0.0523	Nonsignificant increase	30	33	3	9
11	Souris River above Minot, North Dakota	2009–19	0.0060	Significant decrease	36	29	−7	−21
12	Souris River near Verendrye, North Dakota	2009–19	0.0148	Mildly significant decrease	45	38	−7	−16
13	Wintering River near Karlsruhe, North Dakota	2009–19	0.1103	Nonsignificant increase	17	20	3	18
15	Willow Creek near Willow City, North Dakota	2009–19	0.9004	Nonsignificant decrease	29	29	−1	−2
16	Deep River near Upham, North Dakota	2009–19	0.2344	Nonsignificant decrease	44	38	−6	−13
18	Souris River near Westhope, North Dakota	2009–19	0.0568	Nonsignificant decrease	43	38	−5	−12
34	Souris River at Provincial Road 530 near Treesbank, Manitoba	2009–19	0.7469	Nonsignificant increase	32	33	0	1
Total boron, in micrograms per liter
7	Souris River near Sherwood, North Dakota	2009–19	0.4962	Nonsignificant increase	165	179	14	8
11	Souris River above Minot, North Dakota	2009–19	0.0173	Mildly significant decrease	172	139	−33	−19
12	Souris River near Verendrye, North Dakota	2009–19	0.0064	Significant decrease	217	168	−49	−23
18	Souris River near Westhope, North Dakota	2009–19	0.0373	Mildly significant decrease	190	167	−23	−12
30	Pipestone Creek Bridge at Kola, Manitoba	2009–19	0.0489	Mildly significant increase	111	135	24	21
34	Souris River at Provincial Road 530 near Treesbank, Manitoba	2009–19	0.0353	Mildly significant decrease	153	136	−17	−11

Trends in sodium concentrations can be affected by anthropogenic and natural processes such as urbanization, road deicing, energy development, soil evaporites, and groundwater (Pettyjohn, 1967; Keller and others, 198623; U.S. Energy Information Administration, 2013; Granato and others, 2015). In the sparsely populated and agricultural Souris River Basin, urbanization and road deicing are not widespread human processes. Oil and gas development is present in the Souris River Basin, as some of the subbasins contained more than 1,000 oil and gas development wells by 2011 (Susong and others, 2012; U.S. Energy Information Administration, 2013). Sodium is naturally available in many of the aquifers in the basin (Pettyjohn, 1967) and from sodium-sulfate evaporites in the soils (Keller and others, 1986).

Trends in annual flow-averaged GMC of sodium generally were small and nonsignificant, and most of the sites had concentrations greater than the sodium WQO of 100 mg/L (table 6; fig. 7). The largest mildly significant increase in sodium concentration was at site 3, which had a 61-mg/L, or 37-percent, increase (table 6; fig. 7). The largest and only mildly significant decrease was on the site 11, which had a 28-mg/L, or 17-percent, (table 6; fig. 7). Other than site 3, the change in flow-averaged GMC for all sites was less than 30 mg/L in either increasing or decreasing direction. Nine of the 12 sites evaluated had flow-averaged GMC greater than the WQO of 100 mg/L for sodium in 2009, and by 2019, 10 sites had flow-averaged GMCs greater than the WQO (table 6; fig. 7).

Changes in chloride concentrations have been linked to human activities such as road deicing, dust control of unpaved roads, agriculture, and energy development (Granato and others, 2015). Agricultural sources of chloride can come from fertilizer in the form of sylvite, animal waste from livestock holding areas, or irrigation practices can increase soil salinity (Granato and others, 2015). Chloride is naturally present in many aquifers in the basin (Pettyjohn, 1967).

Changes in the annual flow-averaged GMC of chloride were generally small and nonsignificant with about one-half of the sites decreasing and one-half increasing, but, unlike sodium, all sites had concentrations much less than the chloride WQO of 100 mg/L (fig. 8; table 6). Of the 10 sites evaluated, six had decreasing concentrations and four had increasing concentrations, but seven sites had small nonsignificant changes, indicating there was little change in chloride concentrations between 2009 and 2019. The largest increase was at site 3, which was significant and increased by 10 mg/L or 52 percent (fig. 8; table 6). The largest decreases were significant or mildly significant at sites 11 and 12, with a 7-mg/L decrease for both sites, or 21 and 16 percent, respectively (fig. 8; table 6). All sites had flow-averaged GMCs between 20 and 40 mg/L in 2019, which was much less than the chloride WQO of 100 mg/L (site 13; table 6; fig. 8).

Boron concentrations are mostly affected by the weathering of local geologic formations (Hem, 1985; Canadian Council of Ministers of the Environment, 2009). The release of boron through the natural weathering process is slow and the concentrations released are generally low (Canadian Council of Ministers of the Environment, 2009). Because the main source of boron is likely geological, gradual changes are likely from climatic changes and abrupt changes may be indicative of an additional source such as wastewater effluent, mining activities, or industrial activities (Canadian Council of Ministers of the Environment, 2009).

The annual flow-averaged GMC of total boron decreased for four of six sites during the recent period, but all sites had concentrations much less than the 500-µg/L WQO (fig. 9; table 6). The largest decrease was a significant decrease of 49 µg/L or 23 percent at site 12. The largest increase was a mildly significant increase of 24 µg/L or 21 percent at site 30. By 2019, all sites had a flow-averaged GMC between 100 and 200 µg/L, which was much less than the WQO of 500 µg/L for total boron.

Total Phosphorus

Anthropogenic sources of phosphorus can be related to crop management, livestock, fertilizers, land-use changes, and industrial or municipal effluents (Hem, 1985; Tornes and Brigham, 1993; Dubrovsky and others, 2010). Crop and livestock management practices can affect phosphorus concentrations through the application of fertilizer and runoff from feed lots (Hem, 1985; Tornes and Brigham, 1993). Overabundance of nutrients in lakes and rivers can cause excessive algal growth in receiving waterbodies, harming aquatic ecosystems (Paerl and others, 2001).

Annual flow-averaged GMC of total phosphorus decreased for nearly all sites across the Souris River Basin during the recent period including the binational sites, but all sites had concentrations greater than the total phosphorus WQO of 0.1 mg/L for the entire period (fig. 10; table 7). The largest decrease was at the site 25, where the flow-averaged GMC decreased by 0.16 mg/L or 45 percent (fig. 10; table 7). The only site with increasing concentrations was site 11, but the increase was small and nonsignificant (fig. 10; table 7). Mildly significant or significant decreases in total phosphorus were detected at Westhope, and sites 4, 25, and 30 (fig. 10; table 7). None of the sites had a flow-averaged GMC during the entire period that was less than the WQO for total phosphorus of 0.1 mg/L (fig. 10; table 7).

Table 7.    

Summary of trend results for the recent period 2009–19 for total phosphorus at selected sites in the Souris River Basin.

[p-value, probability value; FAGMC, flow-averaged geometric mean concentration; <, less than]

Table 7.    Summary of trend results for the recent period 2009–19 for total phosphorus at selected sites in the Souris River Basin.

Site location (fig. 1)	Site name	Trend period	p-value	Significance level	FAGMC for first year in trend period	FAGMC for last year in trend period	Change in FAGMC between first and last year	Change, in percent from first year to last year
Total phosphorus, in milligrams per liter
4	Souris River at Highway 39 near Roche Percee, Saskatchewan	2009–19	<0.0001	Significant decrease	0.24	0.18	−0.06	−26
7	Souris River near Sherwood, North Dakota	2009–19	0.0668	Nonsignificant decrease	0.22	0.16	−0.06	−28
10	Des Lacs River at Foxholm, North Dakota	2009–19	0.2438	Nonsignificant decrease	0.2	0.17	−0.04	−18
11	Souris River above Minot, North Dakota	2009–19	0.5485	Nonsignificant increase	0.23	0.25	0.02	8
12	Souris River near Verendrye, North Dakota	2009–19	0.3555	Nonsignificant decrease	0.22	0.2	−0.02	−7
18	Souris River near Westhope, North Dakota	2009–19	0.0076	Significant decrease	0.29	0.22	−0.07	−25
25	Souris River east of Melita, Manitoba on Highway 3	2009–19	0.0010	Significant decrease	0.36	0.2	−0.16	−45
30	Pipestone Creek Bridge at Kola, Manitoba	2009–19	0.0208	Mildly significant decrease	0.16	0.11	−0.05	−31
34	Souris River at Provincial Road 530 near Treesbank, Manitoba	2009–19	0.1517	Nonsignificant decrease	0.21	0.18	−0.03	−13

Trace Metals

Concentrations of trace metals are complex and can be related to different processes in the basin. Natural sources of iron, barium, and molybdenum are likely from the weathering of geologic materials in the basin (Hem, 1985; Canadian Council of Ministers of the Environment, 1999, 2009) and potential anthropogenic sources could be from atmospheric deposition from increased energy production in North Dakota (Hem, 1985; U.S. Energy Information Administration, 2013). Increased levels of trace metals can be toxic for aquatic and nonaquatic wildlife as well as humans (Canadian Council of Ministers of the Environment, 1999, 2009; Centers for Disease Control, 2022).

Small and nonsignificant changes in annual flow-averaged GMC of total iron were detected at all sites, except Sherwood, and by 2019 all sites other than Sherwood had concentrations greater than the total iron WQO of 300 µg/L (table 8; fig. 11). At Sherwood, a large significant decrease of 73 percent or 623 µg/L was detected (table 8; fig. 11). All other sites had nonsignificant and small changes in total iron concentrations, with most of the nonsignificant increases upstream from Westhope (table 8; fig. 11). The measurement of total iron concentrations is more variable than many other constituents, likely due to a high degree of variability in the amount of iron attached to sediment particles in each sample, which can make it difficult to identify changes in concentrations over time.

Table 8.    

Summary of trend results for the recent period 2009–19 for trace metals at selected sites in the Souris River Basin.

[p-value, probability value; FAGMC, flow-averaged geometric mean concentration; <, less than]

Table 8.    Summary of trend results for the recent period 2009–19 for trace metals at selected sites in the Souris River Basin.

Site location (fig. 1)	Site name	Trend period	p-value	Significance level	FAGMC for first year in trend period	FAGMC for last year in trend period	Change in FAGMC between first and last year	Change, in percent from first year to last year
Total iron, in micrograms per liter
4	Souris River at Highway 39 near Roche Percee, Saskatchewan	2009–19	0.3858	Nonsignificant increase	407	512	104	26
7	Souris River near Sherwood, North Dakota	2009–19	<0.0001	Significant decrease	854	231	−623	−73
11	Souris River above Minot, North Dakota	2009–19	0.5507	Nonsignificant increase	273	309	36	14
12	Souris River near Verendrye, North Dakota	2009–19	0.4555	Nonsignificant increase	669	759	90	14
18	Souris River near Westhope, North Dakota	2009–19	0.7507	Nonsignificant decrease	482	447	−36	−8
30	Pipestone Creek Bridge at Kola, Manitoba	2009–19	0.1121	Nonsignificant decrease	468	311	−158	−34
34	Souris River at Provincial Road 530 near Treesbank, Manitoba	2009–19	0.3952	Nonsignificant increase	469	524	55	12
Total barium, in micrograms per liter
4	Souris River at Highway 39 near Roche Percee, Saskatchewan	2009–19	0.1124	Nonsignificant decrease	91	74	−17	−19
7	Souris River near Sherwood, North Dakota	2009–19	0.0184	Mildly significant decrease	78	62	−17	−21
11	Souris River above Minot, North Dakota	2009–19	0.0965	Nonsignificant decrease	83	70	−13	−15
12	Souris River near Verendrye, North Dakota	2009–19	0.9038	Nonsignificant increase	81	82	1	1
18	Souris River near Westhope, North Dakota	2009–19	0.1210	Nonsignificant decrease	112	97	−14	−13
30	Pipestone Creek Bridge at Kola, Manitoba	2009–19	0.0893	Nonsignificant decrease	68	59	−9	−14
34	Souris River at Provincial Road 530 near Treesbank, Manitoba	2009–19	0.5086	Nonsignificant decrease	90	88	−2	−2
Total molybdenum, in micrograms per liter
4	Souris River at Highway 39 near Roche Percee, Saskatchewan	2009–19	0.2063	Nonsignificant decrease	4.3	3.4	−0.9	−22
7	Souris River near Sherwood, North Dakota	2009–19	0.1062	Nonsignificant increase	2.9	3.7	0.8	28
11	Souris River above Minot, North Dakota	2009–19	0.6854	Nonsignificant increase	3.6	3.7	0.2	5
12	Souris River near Verendrye, North Dakota	2009–19	0.0097	Significant decrease	4.2	3.3	−0.9	−21
18	Souris River near Westhope, North Dakota	2009–19	0.2965	Nonsignificant decrease	3.5	2.7	−0.8	−22
30	Pipestone Creek Bridge at Kola, Manitoba	2009–19	0.0002	Significant increase	2.7	3.8	1.1	40
34	Souris River at Provincial Road 530 near Treesbank, Manitoba	2009–19	0.4026	Nonsignificant decrease	3.0	2.8	−0.2	−7

Nonsignificant small decreases in annual flow-averaged GMC of total barium were detected at six of the seven sites evaluated during the recent trend period. The largest and only mildly significant decrease in total barium flow-averaged GMC was detected at Sherwood, which decreased by 17 µg/L or 21 percent (fig. 12; table 8). The only increase was at site 12, which was a nonsignificant increase of 1 µg/L or 1 percent. All sites had flow-averaged GMCs less than 100 µg/L, or an order of magnitude less than the total barium WQO of 1,000 µg/L (fig. 12; table 8). Median barium concentrations for other sites in the basin were less than 120 µg/L (table 1.3).

Overall, trends in total molybdenum concentrations in the Souris River Basin were generally small and nonsignificant, except at sites 12 and 30, which were significant (table 8; fig. 13). The only two significant trends were a decrease of 0.9 µg/L or 21 percent at site 12, and the largest increase of 1.1 µg/L or 40 percent at site 30 (table 8). All seven sites evaluated had flow-averaged GMCs less than the WQO of 10 µg/L throughout the trend period (table 8; fig. 13).

Reservoir Trends

Four reservoirs were analyzed for trends in TDS, sulfate, sodium, and total phosphorus for a period of commonly available data (2000–15). Total iron was analyzed for Lake Darling and J. Clark Salyer Pool.

For TDS, sulfate, and sodium during 2000–15, concentrations increased in the reservoirs and most were significant (table 9). During 2000–15, Grant Devine Lake and Lake Darling reservoirs had the largest increases in flow-averaged GMC for TDS, sulfate, and sodium with all increases significant or mildly significant. J. Clark Salyer Pool had the smallest concentration increases for TDS and sulfate, and a nonsignificant increase of 3 percent was detected for sodium. Among TDS, sulfate, and sodium, increases in flow-averaged GMC concentration of sulfate were the largest, ranging from 19 percent at J. Clark Salyer Pool to 82 percent at Lake Darling. Despite increases, TDS, sulfate, and sodium concentrations in the upstream reservoirs were below the WQOs established for the binational sites during the trend period. TDS and sulfate concentrations in Lake Darling also remained below the WQO for TDS (1,000 mg/L) and sulfate (450 mg/L), and only exceeded the sodium WQO (100 mg/L) at the end of the trend period. J. Clark Salyer Pool exceeded the WQO for TDS and sodium during the trend period.

Reservoir changes in total phosphorus were variable with a nonsignificant increase in Grant Devine Lake and mildly significant or significant increasing and decreasing trends in the other reservoirs. Mildly significant and significant increases in flow-averaged GMC of total phosphorus were detected for Rafferty Reservoir and Lake Darling, respectively. J. Clark Salyer Pool had a significant decrease in total phosphorus concentration during the trend period. Relative to the total phosphorus WQO of 0.1 mg/L, all sites had a flow-averaged GMC greater than 0.1 mg/L by the end of the trend period. The largest increase in flow-averaged GMC of total phosphorus was 168 percent at Rafferty Reservoir. Because concentrations in Rafferty Reservoir are small (median of 0.16 mg/L; table 1.2), a small change in concentration can result in a large percentage change (table 9). The largest significant decrease in total phosphorus flow-averaged GMC was 28 percent at J. Clark Salyer Pool (table 9).

Nonsignificant increases in flow-averaged GMC of total iron were observed at the two reservoir sites evaluated for trends. The largest increase in total iron concentration was at J. Clark Salyer Pool, which was 49 percent and nonsignificant. Lake Darling had a smaller nonsignificant increase of 14 percent in total iron concentrations. Relative to the total iron WQO of 300 µg/L for the binational sites, J. Clark Salyer Pool had a flow-averaged GMC greater than the 300 µg/L during the trend period.

There are some limitations in the trend results for the reservoirs because of the trend method used and the period selected for trend analysis. Reservoir water quality is affected to some degree by changes in reservoir volume and season, both of which are accounted for in R–QWTREND through flow anomalies. For the reservoir trend models, lagged reservoir inflows were assumed to be related to water-quality concentrations at the downstream end of the reservoir weeks to months later. This assumption does not account for changes in concentration owing to in-reservoir processes such as dilution, evaporation, settling, nutrient cycling, or oxidation-reduction reactions. The effect of reservoir processes on concentrations varies with constituent and other properties of the reservoir (for example, residence time and reservoir morphometry). For constituents like TDS, sulfate, and sodium, dilution and evaporation are major reservoir processes affecting concentration and are directly related to changes in reservoir volume, but other processes such as groundwater inputs or ice formation can also affect concentrations. For total phosphorus and total iron, reservoir processes such as settling, nutrient cycling, or oxidation-reduction reactions cause changes in concentration between the upstream and downstream ends of the reservoir. For the reservoir trend models of TDS, sodium, and sulfate, the long-term (annual) flow anomaly was significant and negative, meaning that concentrations decrease as inflow increases (Tatge and Nustad, 2023). A significant and negative long-term flow anomaly is an indication that variability in inflows explains some of the variability in concentrations on an annual time scale. For the reservoir trend models of total phosphorus and total iron, at least one seasonality term was significant for all models, and for Lake Darling and J. Clark Salyer Pool the long-term flow anomaly was significant (Tatge and Nustad, 2023). A significant seasonal term is an indication that variability in inflows explains some of the seasonal variability and annual variability in total phosphorus and total iron concentrations. Although a one-period trend model is used, multiyear patterns were evident in TDS, sulfate, and sodium concentrations for Rafferty Reservoir and Grant Devine Lake, and in response to the flood of 2011, reservoir concentrations of TDS, sulfate, and sodium decreased substantially for all reservoirs except J. Clark Salyer Pool. For the upstream reservoirs, multiple piecewise monotonic trend models may have provided a better fit, but a longer one-period trend model was used to assess overall change in concentration using a trend period most like the stream site trend periods. Trend results presented in this report are specific to the period chosen for analysis, and results can differ if the trend period is redefined. R–QWTREND was not specifically designed for reservoir trend analysis and other methods may be available, but an approach that considers change of inflow volume in some manner is important. Reservoir dynamics are complex and a better understanding of the effects of flow, season, and reservoir processes on concentrations could be gained from a detailed mass balance study, but that was beyond the scope of this work.

Historical Trends

Historical trends in TDS, sulfate, sodium, chloride, total phosphorus, and total iron were evaluated for 10 sites during 1976–2019 using water-quality and streamflow data from 1970 to 2020. The number of sites analyzed for each constituent depended on data availability: 10 sites were analyzed for TDS, sulfate, and sodium; 9 sites were analyzed for chloride; 6 sites were analyzed for total phosphorus; and 5 sites were analyzed for total iron. For TDS, sulfate, sodium, and chloride, a three-period trend model was used for all sites and consisted of three piecewise monotonic trends: 1976–88, 1988–2005, and 2005–19. For total phosphorus, a four-period trend model, consisting of four piecewise monotonic trends (1976–1988 and 1988–2000, 2000–09, 2009–19), was used for sites on Sherwood, site 12, Westhope and site 34. Because total phosphorus data were not available for the early period, only the latter two periods were used for sites 10 and 11. For total iron, a two-period trend model, consisting of two piecewise monotonic trends (2000–09 and 2009–19), was used. Trend results are reported in table 10 in terms of percentage change and magnitude of change in the annual flow-averaged GMC from the first year to the last year of each trend period, and the fitted trend in annual flow-averaged GMC over time is shown in figures 14–19.

During the historical period, increases in TDS and sulfate have been large and consistent since the late 1980s, with the largest increases and the most sites with mildly significant or significant increases generally during the middle period (1988–2005; table 10 and figs. 14–15). During the early period (1976–88), changes were a mix of increasing and decreasing concentrations, but nearly all changes were nonsignificant and were generally small (less than 20 percent for TDS and 25 percent or less for sulfate). During the middle period (1988–2005), large (greater than about 20 percent for TDS and greater than 40 percent for sulfate) and significant or mildly significant increases in concentrations in TDS and sulfate were detected with the smallest increase at site 3 (19 percent for TDS, 41 percent for sulfate) and the largest increase at site 13 (74 percent for TDS, 172 percent for sulfate). During the late period (2005–19), other than sulfate at site 15, TDS and sulfate continued to increase, but increases generally were smaller and more sites had nonsignificant trends than the middle period. During the late period, concentrations increased by 3 to 29 percent for TDS and about 9 to 49 percent for sulfate. At the start of the historical period (1976), flow-averaged GMCs at nine sites were less than the WQOs of 1,000 mg/L for TDS and 450 mg/L for sulfate. By the end of the historical period (2019), flow-averaged GMCs for one-half of the sites increased to concentrations greater than the WQOs for TDS and sulfate. For all sites, the increasing sulfate concentrations (table 10) were more than one-half of the magnitude of the TDS increases, indicating that sulfate accounted for at least one-half of the increase in TDS for those sites.