This report describes an assessment of streamflow and water-quality data from a long-term water-quality network in the basin, extending from the Ark Granite site to the Ark Coolidge site. Specifically, this report establishes baseline water-quality conditions based on data from October 1989 through September 2022 (water years [WY] 1990−2022) that could be used to help identify changes in water quality that may result from changes in land and water use or water management. A WY is the 12-month period from October 1 through September 30 of the following year and is designated by the calendar year in which it ends. The report (1) describes methods used to establish a relation between daily mean SC at a recently discontinued (2023) water-quality site USGS 07099970 Arkansas River at Moffat Street (Ark Moffat) and daily mean SC at a newly established (2021) water-quality site USGS 07099971 Arkansas River below Moffat Street (Ark blw Moffat), (2) describes methods used to establish relations between discrete SC and Ca, Mg, Na, SO₄, and TDS values across 13 sites using data collected from October 1989 through December 2023, (3) describes the spatial and temporal variation in TDS composition in the Arkansas River from Portland, Colorado, downstream to Coolidge, Kansas, using data collected during calendar year 2023, (4) describes seasonal and temporal trends in streamflow and TDS concentrations and loads from October 1994 through September 2022, and (5) identifies areas with higher TDS using data collected from October 1989 through December 2023. The concentrations documented in this report are from samples that are representative of the raw, untreated ambient conditions of the surface water in the Basin. The purpose of drinking water standards presented in this report for TDS are for reference only and to provide environmentally relevant context to the results.

This report uses surface-water data collected by the USGS and the Colorado Division of Water Resources (DWR) from October 1989 through December 2023. Hydrologic and water-quality data collected by the USGS are publicly available through the USGS National Water Information System (NWIS) database (USGS, 2024) using the USGS site number provided in table 1. Hydrologic data collected by the DWR are publicly available through Colorado’s Decision Support Systems database (DWR, 2024) using the DWR site abbreviation shown in table 1.

The Arkansas River originates in the high mountains of the Continental Divide in central Colorado and flows onto the eastern plains into Kansas (fig. 1). Upstream from Cañon City, Colo., the Arkansas River is characterized by steep gradient, high velocity flows that are confined to a narrow rock and cobble stream channel. Downstream from Cañon City, the river gradient decreases as the river flows out of the mountains onto the plains. Colorado Springs, the largest city in the study area, is in the Fountain Creek Basin in El Paso County (figs. 1 and 2). Fountain Creek, a major tributary in the study area, flows into the Arkansas River within the city limits of Pueblo. Downstream from Fountain Creek, the river channel is a shifting sand channel that meanders along the alluvial floodplain (Miller and others, 2010).

Pueblo Reservoir is the first main stem reservoir on the Arkansas River; is about 6 miles west of Pueblo, Colorado (fig. 1); and provides water storage for agricultural, municipal, and industrial uses; recreation; fish and wildlife; and flood control (Bureau of Reclamation, 2022). Pueblo Reservoir was constructed between 1970 and 1975 as part of the Bureau of Reclamation’s Fryingpan-Arkansas Project (Bureau of Reclamation, 2022). The Fryingpan-Arkansas Project provides approximately 57,400 acre-feet of nonnative (transmountain) water annually from the Fryingpan River Basin (fig. 1) on the western slope of the Continental Divide to the eastern slope by means of the Arkansas River (Southeastern Colorado Water Conservancy District, 2025). During WYs 1990−2022, Fryingpan-Arkansas Project water imports varied from 13,960 acre-feet in WY 2012, to 99,810 acre-feet in WY 2011 (Southeastern Colorado Water Conservancy District, 2025; DWR, 2024). Much of the water from the project is routed to Twin Lakes Reservoir upstream from Ark Granite (fig. 1) and then released to the Arkansas River to water users upstream from Pueblo Reservoir or stored in Pueblo Reservoir (Southeastern Colorado Water Conservancy District, 2025). Native or nonnative water belonging to Colorado Springs or Aurora (fig. 1), can be directly diverted from Twin Lakes Reservoir into a pipeline for delivery to the municipal water-supply systems of the two cities (Lewis, 1999).

Under Colorado law, nonnative water imported from outside a watershed can be used and reused to the point of extinction if the amount of water can be quantified at its points of intended use (Radosevich and others, 1976). Colorado Springs, like other cities in eastern Colorado, supplements its water supply with nonnative water from outside of the Basin (Colorado Springs Utilities, 2026). This nonnative water is used to supply water for municipal, commercial, industrial, and irrigation purposes and then much of it is returned to Monument Creek and Fountain Creek (fig. 1), mostly in the form of treated wastewater effluent, to potentially be reused. Since April 1989, a physically based transit loss accounting model has been used to track native and nonnative water and administer water rights along Fountain Creek on a daily basis (Colarullo and Miller, 2019). The total volume of nonnative water flowing into Monument Creek and Fountain Creek varies from year to year but has generally increased over time as the population in El Paso County (fig. 2) has grown (Colarullo and Miller, 2019; Hennessy and others, 2024). Bern and others (2024) found significant (p-value less than [<] 0.05) increases in nonnative return flows at sites on Monument Creek and Fountain Creek for WYs 1999−2022. The annual percentage of nonnative reusable return flow ranged from 7 percent to 54 percent of the total streamflow measured near the mouth of Fountain Creek at USGS site 07106500 Fountain Creek at Pueblo, Colo., (Fnt Pueblo; fig. 1; table 1) for WYs 1999−2022 (Bern and others, 2024; Hennessy and others, 2024).

Streamflow in the Arkansas River shows considerable seasonal variability due to the timing and magnitude of snowmelt and rainfall runoff; reservoir operations; and municipal, agricultural, and industrial diversions and releases. Reservoir operations, such as the Winter Water Storage Program (WWSP), greatly affect the quantity and timing of streamflow in the Arkansas River downstream from Pueblo Reservoir. Water is stored in Pueblo Reservoir from November 15 of each year to March 14 of the next year under the WWSP and released in the spring or late summer when natural streamflow in the river may not be sufficient for irrigation needs (Ortiz, 2013; Southeastern Colorado Water Conservancy District, 2025). For nearly 120 miles downstream from Pueblo Reservoir, water from the river is diverted through a network of irrigation canals and is applied and reapplied to crops in the valley before flowing into John Martin Reservoir downstream from Las Animas (Ortiz, 2013). The Arkansas River flows into Kansas about 58 miles downstream from John Martin Reservoir (fig. 1).

From the headwaters to near Cañon City in Fremont County, the Arkansas River traverses mountainous terrain underlain by Cambrian and Proterozoic age igneous and metamorphic rocks (figs. 1 and 2). East of Cañon City, the river traverses flatter terrain underlain by Cretaceous marine shales and limestones (figs. 1 and 2) as it flows eastward through the plains toward the Colorado–Kansas State line (Green, 1992; Zielinski and others, 1995). Soils on the plains east of Cañon City originate from locally weathered marine shale and limestone and mixtures of rock fragments and minerals (Zielinski and others, 1995). Miller and others (2010) found that TDS concentrations in groundwater from Quaternary alluvial, glacial drift, and wind-laid deposits (fig. 2) increased downgradient with median values of about 220 milligrams per liter (mg/L) in the Basin headwaters to Pueblo Reservoir to about 3,400 mg/L from John Martin Reservoir to near the Colorado–Kansas State line. The sedimentary rocks, particularly the Upper Cretaceous rocks (fig. 2), in the Basin downstream from Pueblo Reservoir and in the Fountain Creek Basin were a primary source of TDS in groundwater and surface water in the Basin (Miller and others, 2010). Bern and Stogner (2017) found that inflows originating from areas affected by the Cretaceous Niobrara Formation (fig. 3) between Ark Cañon City and Ark Portland had substantially higher concentrations of TDS, particularly during the nonirrigation season when geologic effects were likely greater, than inflows where surface exposures of Niobrara Formation were absent. Ivahnenko and others (2013) found that Ca and HCO₃ were the dominant ions in samples at Ark Granite and Ark Cañon City (fig. 1) where igneous and metamorphic (crystalline) rocks are present (fig. 2). Inflows affected by the Niobrara generally had Na-SO₄ compositions, whereas inflows not affected by the Niobrara Formation had Ca-HCO₃ compositions (Bern and Stogner, 2017).

From January 2022 through December 2023, water-quality samples were collected and analyzed for TDS at all 13 study sites. In 2023, selected ions (carbonate [CO₃], HCO₃, Ca, Mg, Na, K, Cl, SO₄, fluoride [Fl], Si, iron [Fe], manganese [Mn], alkalinity, and bromide [Br]) were added to the sampling effort to determine the composition of TDS spatially and temporally at 11 of the 13 sites, including 9 sites in the Arkansas River from Portland, Colorado, downstream to Coolidge, Kansas, and 2 sites in Fountain Creek, USGS 07106000 Fountain Creek near Fountain, Colo., (Fnt nr Fnt) and USGS 07106500 Fountain Creek at Pueblo, Colo., (Fnt Pueblo; fig. 1; table 1). All water-quality samples were collected and processed using standard USGS techniques and procedures described by the USGS “National Field Manual for the Collection of Water-Quality Data” (USGS, variously dated). Isokinetic, depth-integrating, discharge-weighted samples were collected using intervals of equal width when site and flow conditions allowed. When site or flow condition did not allow for collection of equal-width interval samples, a single point sample or multiple point (verticals) samples were collected using a DH–81 sampler (USGS, variously dated). Sample processing was done primarily in a mobile USGS laboratory at the sampling site. Sampling equipment was cleaned using part-per-billion protocols (USGS, variously dated). Specific conductance, pH, and water temperature measurements were made using protocols described in the National Field Manual (USGS, variously dated). Water-quality constituents (CO₃, HCO₃, Ca, Mg, Na, K, Cl, SO₄, Fl, Si, Fe, Mn, alkalinity, TDS, and Br) were analyzed at the USGS National Water Quality Laboratory (NWQL) in Lakewood, Colorado (Fishman and Friedman, 1989; Fishman, 1993; Struzeski and others, 1996; Clesceri and others, 1999).

Quality control and quality assurance (QAQC) samples (10 field blank and 13 replicate samples) accounted for about 10 percent of the total number of samples collected at the 13 study sites (fig. 1; table 1) from January 2022 through December 2023. The QAQC samples were analyzed for TDS in 2022. In 2023, QAQC samples were also analyzed for major ions (CO₃, HCO₃, Ca, Mg, Na, K, Cl, SO₄, Fl, Si, Fe, Mn, alkalinity, TDS, and Br) at selected sites. Procedures from the National Field Manual for equipment cleaning, sample collection, and field processing were followed to minimize bias and contamination (USGS, variously dated). Field blank and replicate samples were collected to evaluate potential contamination and variability with the sampling process and analysis (Mueller and others, 2015) and analyzed by the USGS NWQL.

During this study, 10 field blank samples were collected at 8 sites (1 field blank at 6 sites and 2 field blanks at 2  sites). A field blank consisted of processing inorganic-free blank water at a field site using the same equipment and methods used in the environmental sample processing. Inorganic-free blank water was obtained from the NWQL. Field blanks were used to evaluate the potential for sample contamination from several sources, including from equipment, ambient contaminants at the site, and laboratory procedures (Mueller and others, 2015). An analysis of field blank data showed that all cleaning and processing protocols were mostly effective in minimizing contamination of water-quality samples throughout the study.

Only two constituents, filtered Cl and filtered SO₄, were detected in field blanks. Filtered Cl was detected at concentrations ranging from 0.05 to 0.06 mg/L, slightly greater than the NWQL laboratory reporting limit (RL) of 0.04 mg/L, at four sites (USGS, 2024). Concentrations of filtered Cl in environmental samples, collected from the 11 sites where major ion data were collected as part of this study, ranged from 3.37 to 143 mg/L (USGS, 2024), more than 50 times higher than concentrations detected in field blanks. Filtered SO₄ was detected at a concentration of 0.023 mg/L in a field blank at one site, USGS 07099400 Arkansas River above Pueblo, Colo., (Ark abv Pueblo). Concentrations of filtered SO₄ in environmental samples collected from the same 11 sites ranged from 34.8 to 2,210 mg/L (USGS, 2024), more than 1,500 times higher than the concentration detected in the field blank. Consequently, concentrations of filtered Cl and filtered SO₄ in field blanks likely did not affect the interpretation of environmental data.

During this study, 13 field replicate samples were collected at 11 sites (1 replicate at 9 sites and 2 replicates at 2 sites). A replicate sample involves processing two environmental samples at the same site during the same visit. Three types of replicates were collected during this study: concurrent, sequential, and split (Mueller and others, 2015). Concurrent and sequential replicates are used to target the variability in constituent concentrations resulting from sample collection, processing, and analysis. Split replicates target the source of variability resulting from sample processing and analysis and exclude variability resulting from sample collection (Mueller and others, 2015). For a concurrent sample, two volumes of water were collected in separate churns. This was accomplished by collecting two samples at each stream vertical (that is, each sample point in the cross section) using a DH-81 sampler. One sample was poured into a churn, and the other sample was poured into a different churn, resulting in two composited samples from the same cross section at about the same time. For a sequential sample, one volume of water was collected using a DH-81 sampler at multiple verticals and poured into a churn, and then a second volume of water was collected using a different DH-81 sampler along the same channel cross section and poured into a different churn. Both samples were then processed. For a split replicate, 1 volume of water was collected, and 2 sets of samples were processed out of the same churn. The environmental sample was processed first, then the replicate sample was processed.

Of the 119 constituent replicate pairs, 112 had detections greater than the RLs for the environmental and replicate sample. The relative percent difference (RPD) between the environmental and replicate concentrations was calculated for the 112 pairs using the following equation (Mueller and others 2015; eq. 1):,(1)where

Seven of the 112 constituent replicate pairs were flagged for having an RPD greater than (>) 10 percent (table 2). The constituents for these seven pairs included carbonate (CO₃) as CO₃ processed in the field from a filtered sample, filtered SO_4, filtered Fe, filtered Br, and acid neutralizing capacity (ANC) as calcium carbonate (CaCO₃) processed in the laboratory from an unfiltered sample (table 2). All the flagged pairs were from split replicates.

Concentrations of filtered CO₃ and Br in the flagged environmental and replicate pairs were low (<1 mg/L; table 2); as a result, small changes in concentrations produced large changes in RPD values. Concentrations of filtered Fe were near or less than the RL of 10 micrograms per liter (μg/L; table 2) in the flagged pair. The flagged pair for filtered Fe had an absolute difference (that is, the difference between the environmental and replicate values) of 2.6 μg/L (table 2). An analysis of the QAQC data from the NWQL from October 2017 through September 2022 found that there was a mean absolute difference for Fe at surface water sites of 17.354 μg/L (Bexfield and others, 2023), so the difference found in this replicate pair is likely due to variance within the NWQL analysis method with a higher RPD because of the low Fe concentration. Possible sources of variability during the processing of the sample could include clogging of the 0.45-micron filter between the environmental and replicate samples or dilution of one of the samples in a pair by not fully purging the filter of deionized water. For the purposes of this study, the variability in filtered CO₃, Fe, and Br at the low concentrations shown in table 2, did not affect data interpretation because these constituents made up a very small part of the overall TDS. Sulfate concentrations in the environmental and replicate sample fell within the historical range of values for SO₄ at Fnt nr Fnt (USGS, 2024). Because the primary purpose of sampling for major ions for this study was to determine the composition of TDS at each site, the difference in SO₄ concentrations did not affect the data interpretation.

Acid neutralizing capacity and alkalinity are measures of the ability of a sample to neutralize a strong acid (Hem, 1985). Acid neutralizing capacity is determined using an unfiltered sample, whereas alkalinity is determined using a filtered sample. The ANC replicate pair at Ark blw JMR had the largest RPD (43.6 percent) of all samples (table 2). The replicate ANC value (218 mg/L) was substantially higher than the environmental value (140 mg/L; table 2), likely because the replicate sample was more concentrated and had more particulates than the environmental sample. The alkalinity values, determined using filtered samples, only varied by about 1 percent in the replicate (150 mg/L) and environmental (153 mg/L) samples (USGS, 2024). As a split replicate, the replicate sample was processed using the water remaining in the churn that possibly was not well mixed after the environmental sample had been processed. Consequently, the variability in ANC was likely due to how the split replicate was processed and not to sampling techniques or analysis.

Statistical analyses used in this report include (1) linear regression techniques; (2) computation of baseline period mean, 10th percentile, 25th percentile, 75th percentile, and 90th percentile daily streamflow, SC, and TDS values for each day of the year; (3) computation of flow adjusted TDS concentrations; (4) the locally estimated scatterplot smoothing (LOESS) regression technique; (5) the Seasonal Kendall test; and (6) the Mann-Kendall test.

Two different laboratory methods were used to analyze TDS through time at study sites: (1) residue on evaporation at 180 degrees Fahrenheit and (2) the sum of constituents. Only TDS determined by residue on evaporation was used in the analyses presented in this report. These water-quality data are publicly available through the USGS NWIS database (USGS, 2024) using the USGS site number provided in table 1.

Prior to using regression techniques to develop relations between SC and Ca, Mg, Na, SO_4, and TDS values, scatter plots of the discrete paired data were visually inspected to determine whether linear relations were present and consistent compared with the range of measured values. Simple linear regression was used to determine the best straight-line fit between paired discrete SC (independent variable) and Ca, Mg, Na, SO₄, and TDS (dependent variables) values measured from October 1989 through December 2023, to obtain predictive equations for each constituent. The coefficient of determination, r-squared (r²), is a statistical measure of how well the predictive equation (regression line) explains variation in the actual data, varies between 0 and 1, and indicates the amount of variability in the dependent variable described by the independent variable. An r² value of 0 indicates that there is no relation between the dependent and independent variables, and an r² value of 1 indicates that there is a perfect relation between the dependent and independent variables (Helsel and others, 2020). The null hypothesis is that there is no relation between the dependent and independent variables (that is, the slope of the regression line is equal to zero). Small p-values indicate that the null hypothesis is unlikely to be true. A p-value of <0.05 was used in this report to indicate if a statistical test was significant.

Water years 1990−2022 were chosen as the baseline period to represent current conditions. Mean, 10th percentile, 25th percentile, 75th percentile, and 90th percentile daily streamflow and SC values were computed for each day of the calendar year at each site using all available daily mean data during the 34-year baseline period. For example, the mean, 10th percentile, 25th percentile, 75th percentile, and 90th percentile for all January 1st streamflow and SC values were calculated for January 1, 1990, through January 1, 2022. Baseline TDS concentrations were computed for each calendar day using the relations developed between SC and TDS (table 3) and the baseline daily SC values. Baseline TDS loads (tons per day [tons/d]) were computed by multiplying the baseline TDS concentrations (mg/L) by the baseline streamflow values (cubic feet per second [ft³/s]; Liebermann and others, 1987) and a conversion factor of 0.002697. In the same way, instantaneous TDS loads were computed for each sample by multiplying the discrete TDS concentration by the instantaneous streamflow value and the conversion factor. The conversion factor is the product of the unit conversions required to go from mg/L per ft³/s to tons/day. The NWIS surface-water daily statistics function (USGS, 2024) was used to calculate the baseline period, mean, 10th percentile, 25th percentile, 75th percentile, and 90th percentile daily values for SC at all sites and for streamflow at sites operated by the USGS (table 1). For streamflow gages operated by DWR (table 1), the R statistical software version 4.4.0 (R Core Team, 2024) was used to compute baseline streamflow values.

Values of SC and TDS were flow-adjusted using the residual method to remove the effect of streamflow (Cohn and others, 1992; Ortiz, 2004). First, multiple linear regression was used to determine the best fit relation between daily mean streamflow and instantaneous SC values for WYs 1990−2022, to obtain a predictive equation 2 for SC due to streamflow (Cohn and others, 1992). Second, these predictive equations (table 4) were used to compute the daily mean SC attributed to streamflow at each site. Finally, the daily mean SC attributed to streamflow was subtracted from the daily mean SC (USGS, 2024) to obtain the flow-adjusted daily mean SC. The daily mean flow-adjusted TDS concentration was then computed by using the flow-adjusted daily mean SC with the linear regression relations developed between paired discrete SC and TDS values (table 3).ln(SC)=β₀ + β_1* ln(Q⁄Qbar)+β₂ [ ln(Q⁄Qbar)]² +β₃ sin(2πT)+ β₄ cos(2πT)+ε(2)where

The equation requires the estimation of five parameters: β₀ is a constant, β₁ and β₂ describe the relation between SC value and streamflow, and β₃ and β₄ describe the seasonal variability in the measured SC data. The error, denoted ε, is assumed to be independent and normally distributed with zero mean and variance (Ortiz, 2004). Qbar is the normalized daily streamflow for each day of the year for WYs 1990−2022.

The LOESS smoothing technique (Cleveland and others, 1992) was used to visually evaluate temporal trends in mean monthly streamflow and median monthly SC and TDS values. The LOESS smoothing technique allows the shape of the relation between the dependent variables (streamflow, SC, and TDS) and the independent variable (time) to be dictated by the data and creates a visual representation of the dominant pattern in the data (Helsel and others, 2020). To evaluate temporal monotonic trends in streamflow, SC, and TDS, two nonparametric tests were used: a Seasonal Kendall test (Hirsch and others, 1982) and a Mann-Kendall test (Mann, 1945; Kendall, 1975). The Seasonal Kendall test is a nonparametric trend test that is used to test for trends in environmental data where data from the same season is compared from year to year (Hirsch and Slack, 1984; Helsel and others, 2020). For this study, months were used as the seasons. The Seasonal Kendall test computes the Mann-Kendall test on each of the seasons, then combines the results, yielding an overall statistic (Helsel and others, 2020). The Mann-Kendall test was used to evaluate the overall change for each month during the baseline period. The slope estimates for both tests were calculated using the Theil-Sen estimator (Sen, 1968; Theil, 1992). The LOESS analysis, Seasonal Kendall test, and Mann-Kendall test were conducted using the R statistical software version 4.4.0 (R Core Team, 2024). The “EnvStats” package (Millard, 2013) was used to conduct the Seasonal Kendall test and the Mann-Kendall test.

Specific conductance is highly correlated with Ca, Mg, Na, SO₄, and TDS concentrations in the Basin (Cain, 1987; Ortiz and others, 1998). Simple linear regression was used to determine the best straight-line fit between paired discrete SC (independent variable) and Ca, Mg, Na, SO₄, and TDS (dependent variables) values measured from October 1989 through December 2023, at each study site, to obtain predictive equations for each constituent. The metadata, input, output, and script files used to develop these equations are shown in Miller and Eng (2025). Table 3 shows the regression model coefficients and the r² value for each equation. Statistically significant (p-value <0.001) linear relations were present between SC and Ca, Mg, Na, SO₄, and TDS at all sites (table 3). Values of r² for these relations varied from 0.720 to 0.993 (table 3). However, the relation between SC and Mg, SO₄, and TDS at Fnt Pueblo was not the same across the full range of SC values; a slope change in the linear relation was evident for each constituent. This change in slope likely indicates a change in source water composition during certain flow regimes (Clean Water Team, 2004). As a result, SC specific linear regression equations dependent on SC magnitude were developed to estimate Mg, SO₄, and TDS at Fnt Pueblo (table 3). For example, a slight break in the relation between SC and Mg, SO₄, and TDS occurred when SC values were greater than about 1,250 microsiemens per centimeter (μS/cm) at Fnt Pueblo. The single relation for that site, developed using all data, generally underestimated Mg, SO₄, and TDS concentrations when SC values were <1,000 μS/cm and >1,500 μS/cm and slightly overestimated TDS values when SC values were between about 1,200 and 1,300 μS/cm. The SC specific relations (less than or equal to 1,250 μS/cm and greater than 1,250 μS/cm) had lower standard errors and fit the measured data better than a single relation at Fnt Pueblo (table 3). For relations developed between SC and TDS at study sites, r² values varied from 0.894 at Ark Granite to 0.992 at Ark Rocky Ford (table 3).

Daily mean TDS concentrations were computed using the new relations developed between SC and TDS during this study (table 3) and previously developed relations by Ortiz (2004) and Miller and others (2010) and daily mean SC values for the baseline period (WYs 1990–2022) at 12 of the 13 sites. Ark Rocky Ford was not included in this analysis because no previous relation between SC and TDS was published for the site. Daily mean TDS concentrations, based on the new and old relations, were compared to determine whether relations between SC and TDS had changed through time. Absolute percentage differences between these TDS concentrations varied from 0.09 to 1.75 percent; TDS concentrations at most sites varied by less than 1.0 percent. These results indicate that the relations between SC and TDS have remained consistent through time at the compared sites.

Mean, 10th percentile, 25th percentile, 75th percentile, and 90th percentile daily streamflow and TDS concentrations and loads were computed for each day of the calendar year at each site using all available daily mean data during the 34-year baseline period as described in the “Statistical Analyses” section of the report. Comparisons can be made between these baseline values and current or future values to help identify changes in water quality that may result from changes in land or water use or timing of runoff events. Figure 7 illustrates how baseline values can be used to identify changes in water quality over time and shows a comparison of daily mean TDS concentrations and loads and the computed mean, 10th percentile, and 90th percentile daily baseline values at Ark Avondale. Daily mean TDS concentrations during June 2024, were below the mean daily baseline TDS concentrations (fig. 7A). TDS loads (fig. 7B) were near the mean daily baseline values at the beginning of June, increased rapidly starting about June 5th, peaked on June 10th when the highest streamflow of the calendar year occurred because of snowmelt runoff, and then decreased below the mean daily baseline values starting around June 19th (USGS, 2024).

As shown in the “Relations between Specific Conductance and Total Dissolved Solids” section, SC values can act as a surrogate for TDS concentrations, and relations between streamflow and SC can indicate likely effects on TDS concentrations at the sites. Streamflow in the Arkansas River shows considerable seasonal variability due to the timing and magnitude of snowmelt and rainfall runoff; reservoir operations; and municipal, agricultural, and industrial diversions and releases (Lewis and Brendle, 1998). The SC values and streamflow are generally inversely related in the study area, where higher values of SC generally occur with lower streamflow and lower values of SC with higher streamflow, as shown in figure 8A, B. Figure 8A, B illustrates how water operations and seasonal variability in streamflow sources can affect SC values. Distinct seasonal differences in median streamflow and SC values were apparent during the winter and summer months at Ark Portland and the USGS 07124000 Arkansas River at Las Animas, Colo., (Ark Las Animas; fig. 8A, B) site. During the winter, water is stored primarily in Pueblo Reservoir starting on November  15 and ending March 14 of the following year as part of the WWSP (Southeastern Colorado Water Conservancy District, 2025). Abrupt changes in median streamflow and SC values were evident in the Arkansas River downstream from Pueblo Reservoir at Ark Las Animas (figs. 1 and 8B) around the beginning and ending of the WWSP because water is stored in the winter and released and diverted by canals upstream from the site during the spring and summer. Ark Portland is upstream from Pueblo Reservoir and upstream from most major canals, so these abrupt changes in streamflow and SC values were not apparent at the beginning and end of the WWSP and irrigation season, which generally goes from March through October (fig. 8A). Snowmelt runoff from the mountainous areas of the Basin, generally in May and June, produces the highest streamflow and lowest SC values of the year at most sites as illustrated by Ark Portland and Ark Las Animas (fig. 8A, B). During low flow periods, especially during the winter months, SC and TDS values are higher than when the dominant source of flow to the river originates from snowmelt or storm runoff. Inflows to the river during low flow periods likely originate from groundwater sources, return flows, or other releases (Burns, 1985) that have had more interaction with the land surface, resulting in higher SC and TDS.

Much of the variance in SC and thus TDS can be attributed to the variance in streamflow, as shown in figure 8. Specific conductance and TDS were flow-adjusted to remove the effect of streamflow to evaluate changes in SC and TDS not due to changes in streamflow (Cohn and others, 1992; Ortiz, 2004) using relations developed between discrete SC and streamflow values (Miller and Eng, 2025; table 4). For these relations between SC and streamflow, r² values varied from 0.355 at Ark abv Pueblo to 0.854 at Ark Catlin Dam (table 4; fig. 1). Ark Avondale, Ark Catlin Dam, and Ark Rocky Ford relations had the highest r² values and Ark abv Pueblo, Ark Cañon City, and Ark Portland relations had the lowest r² values (table 4). Low r² values indicate poor relations between SC and streamflow. Reservoir operations, water exchanges, diversions, and return flows from wastewater and irrigation affect the quantity of streamflow in the Arkansas River and Fountain Creek and may also affect SC values (Lewis, 1999) and the relations between SC and streamflow.

Temporal changes in TDS concentrations and loads can be used to identify changing water-quality conditions. Changes in TDS concentrations could be attributed to a change in streamflow, a change in the source water to the system, or a combination of both. The mixing of nonnative water and native water may affect water quality in the Arkansas River and Fountain Creek. However, the quality of nonnative water was not evaluated as part of this study. To accurately assess trends in water-quality constituents, such as TDS, it is important to have a dataset that accurately represents the seasonal variability in source waters and year-to-year variability in streamflow (Hirsch and others, 2010). Mean monthly streamflow and median monthly SC values, TDS concentrations, and TDS loads used in the trend analyses were computed from daily mean streamflow and SC values recorded at the sites. The period of records varied across sites for SC, and most sites have complete records starting in WY 1995 (table 1). Therefore, trends in SC and TDS were evaluated for WYs 1995 through 2022 at 9 of the 13 sites. Water-quality monitors were installed at Ark Coolidge in 1999, at Ark Rocky Ford in 2012, and Ark blw Moffat in 2021 (table 1). Trends in SC and TDS were evaluated using data for WYs 2000 through 2022 at Ark Coolidge and for WYs 2013 through 2022 at Ark Rocky Ford. Trends in streamflow were evaluated using data for WYs 1995 through 2022 at both Ark Coolidge and Ark Rocky Ford. Trends in streamflow were not evaluated at Ark Catlin Dam because the streamflow gage operated only seasonally from 2019 through 2022. Trend analyses were not done at Ark blw Moffat.

Higher than mean streamflow occurred in the mid to late 1990s at the start of the trend analysis period (October 1994) and below mean streamflow occurred in 2020–22 at the end of the trend analysis period (September 2022) at sites downstream from Pueblo Reservoir; very low flows occurred in 2002 and 2012 (fig. 9). Streamflow values and TDS loads are generally directly correlated, and streamflow values and TDS concentrations are generally inversely correlated in the study area as illustrated in figures 7 and 8. For this reason, downward trends in streamflow and TDS loads and upward trends in TDS concentrations would be expected during the WYs 1995−2022 analysis period. It is likely that trend results could turn out differently if the analysis period began in the early 2000s or mid to late 1970s, when streamflows were lowest at Ark Las Animas (fig. 9).

Seasonal Kendall tests (WYs 1995−2022) showed no significant trends in streamflow at the two most upstream sites on the Arkansas River, Ark Granite and Ark Cañon City (table 5). However, significant downward trends in streamflow were detected at Arkansas River sites from Ark Portland to Ark Coolidge and at Fountain Creek sites (table 5). Significant upward trends in TDS concentrations (table 6) and loads (table 7) were detected at Ark Granite. No significant trends in TDS concentrations were detected at Ark Cañon City or Ark Moffat (table 6), but significant downward trends in TDS loads were detected at the sites (table 7). Significant downward trends in TDS concentrations and loads were detected at the Fountain Creek sites, Ark Portland, Ark abv Pueblo, and at all Arkansas River sites from Ark Avondale to Ark Coolidge (fig. 1; tables 6 and 7). At Ark Rocky Ford, analyses for WYs 2013–2022 showed a significant downward trend in streamflow and a significant upward trend in TDS concentrations (table 6) but no significant trend in TDS loads (table 7).

Results from the Mann-Kendall test (WYs 1995−2022) on monthly values showed fewer significant trends than the Seasonal Kendall test, likely because seasonal differences were present in the data (figures 8A, B). The Mann-Kendall test is used to test for consistent upward or downward trends in continuous data, not to detect seasonal patterns (Helsel and others, 2020). No significant trends in monthly mean streamflow were detected at the three sites (Ark Granite, Ark Cañon City, and Ark Portland) upstream from Pueblo Reservoir (fig. 1; app. 1; table 1.1) using the Mann-Kendall test. Significant upward trends in monthly median SC values were detected at Ark Granite from June through September (table 1.2; fig. 10A). No significant trends were detected in SC values at Ark Cañon City or Ark Portland (table 1.2).

Immediately downstream from Pueblo Reservoir at Ark abv Pueblo, significant downward trends in monthly mean streamflow occurred from October through March and May (table 1.1); significant downward trends occurred in monthly median SC values from October through May at the site (table 1.2). Along Fountain Creek, significant downward trends in monthly median SC values occurred during November and December at Fnt nr Fnt and October through December at Fnt Pueblo (fig. 1; table 1.2), but significant downward trends in monthly mean streamflow occurred only during February (table 1.1). Significant downward trends in monthly mean streamflow occurred for one or more months at all sites downstream from Pueblo Reservoir (fig. 1; table 1.1). Significant downward trends in monthly median SC values occurred for one or more months at all sites downstream from Pueblo Reservoir (fig. 1; table 1.2). These downward trends in monthly median SC values occurred more frequently during the WWSP months, particularly during December (table 1.2). At Ark Rocky Ford, analyses for WYs 2013–2022 showed no significant trends in monthly median SC values (table 1.2), but a significant downward trend in monthly mean streamflow was detected for April. When evaluated over the longer WYs 1995–2022 period, significant downward trends in monthly mean streamflow were detected in January, April, November, and December at Ark Rocky Ford (table 1.1).

Upward trends in TDS concentrations and loads and SC values at Ark Granite could have resulted from remedial actions that took place to minimize erosion of fluvial mine waste deposits along the Arkansas River upstream from the site during the analysis period. As part of the remedial process, selected irrigated meadows and tailings deposits were revegetated. Initial treatment consisted of lime amendment or lime and organic matter amendment, deep tilling, seeding, and mulching (Christner, 2013). Lime application reduces soil acidity and increases Ca and possibly Mg in the soil (Ritchey and others, 2025). Bicarbonate, Ca, Mg, Na, and SO₄ made up most of the TDS in the Arkansas River samples (fig. 6A, B). Runoff from areas treated with lime during the analysis period could have had higher than background concentrations of Ca and Mg and, consequently, increased concentrations of TDS in the Arkansas River at Ark Granite.

Downward trends in streamflow and TDS loads were expected at sites downstream from Pueblo Reservoir because streamflow was higher than mean at the beginning of the trend analysis period and lower than mean at the ending of the trend analysis period (fig. 9). However, downward trends in TDS concentrations and SC values were not expected because higher TDS concentrations and SC values generally occur during lower streamflow in the study area (fig. 8A, B). Because downward trends in streamflow and SC (surrogate for TDS) were more prevalent during winter base flow months (tables 1.1 and 1.2; fig. 10B), this trend may indicate changes in groundwater quantity, and possibly groundwater quality, potentially affecting the river water-quality condition. From 1994 to 2022, irrigated farmlands in Colorado declined by more than 23 percent (U.S. Department of Agriculture, 2025). Along the Arkansas River between Pueblo Reservoir and Las Animas (fig. 1), land in farms declined by 7 percent, 23 percent, and 40 percent in Pueblo, Otero, and Crowley counties (fig. 2), respectively, from 2017 to 2022 (U.S. Department of Agriculture, 2025). Seasonal Kendall tests (WYs 1995−2022) showed mean annual decreases in TDS concentrations of 1.33 mg/L per year (mg/L/yr) at Ark Avondale, 1.65 mg/L/yr at Ark Catlin Dam, and 19.4 mg/L/yr at Ark Las Animas, respectively (table 6; fig. 3). Declines in irrigated farmlands, changes in farming practices, and transfer of water to municipalities may have reduced the amount of water being applied to the land surface through time, which in turn could reduce groundwater return flows and the effects of water recycling on water quality during WYs 1995–2022 (CWCB, 2022; DWR, 2025; U.S. Department of Agriculture, 2025). In addition, mixing of importated Fryingpan-Arkansas Project water and native streamflow may also act affect TDS concentrations and SC values in the Arkansas River and Fountain Creek. Lewis and Brendle (1998) found that imported water released to the river during the winter tended to dilute TDS concentrations and lower SC values upstream from Pueblo Reservoir.