The Rio Grande is the fifth longest river in the U.S. and among the top twenty of the world. The Rio Grande originates in south-central Colorado and its main water source is from snowmelt in the Rocky Mountains (Figure 1; Ellis et al., 1993; Moore et al., 2008). The Rio Grande river water is currently distributed to three states (Colorado, New Mexico, Texas) and two countries (U.S. and Mexico; Water 2025: Preventing Crisis and Conflict in the West, 2005; Alley, 2013). Rapid population growth in this region has led to increasing demand for freshwater (Wong et al., 2012; Sheng, 2013). The Rio Grande river may soon reach tipping points with respect to both freshwater quality and availability under current scenarios of climate changes and population growth (Swetnam and Betancourt, 1998; Phillips et al., 2003, 2011; Seager et al., 2007; Gutzler and Robbins, 2010). Consequently, the U.S. Department of Interior has identified the Rio Grande region as having the highest potential for conflict and crisis among any U.S. river systems.

The studied stretch of the Rio Grande is located in southern New Mexico and western Texas, flowing through the Rio Grande Valley along a series of rift basins filled with alluvial, fluvial, playa and lacustrine sediments (Figure 1; Phillips et al., 2003; Szynkiewicz et al., 2011). The climate in the study area is semi-arid to arid, with hot summers (mean air temperature in June ~28°C) and mild winters (mean air temperature in January ~8°C). For Elephant Butte reservoir, New Mexico and El Paso, Texas (Figure 2), the annual rainfall averages ~250 and 300 mm, respectively, with most of the rainfall occurring during the late summer months as monsoons (Miyamoto et al., 1995).

In the study area, water from the Rio Grande is mainly used to irrigate ~193,000 acres of agricultural land (~800 km²) within the Hatch and Mesilla Valley irrigation districts (Garfin et al., 2013) (Figure 1A). Downstream from El Paso, TX, Rio Grande water is used to irrigate additional agricultural land in the Hueco Basin in Texas and Mexico. Water for irrigation is delivered through an interconnected system of dams, reservoirs, canals, and drains (commonly known as the Rio Grande project). In the study area, the Rio Grande river flows are regulated by Elephant Butte and Caballo Reservoirs, which generally store river water during fall and winter months (November through February, or non-irrigation seasons) and release water to downstream users during spring and summer (March through October, or irrigation seasons). Consequently, Rio Grande river flows vary significantly between irrigation and non-irrigation seasons as a result of considerable human impacts on the local hydrological cycle (Moyer et al., 2013). The river flows at downstream locations are high in spring and summer months but significantly reduced in fall and winter months. At some locations the riverbed can be completely dry for several months. Most of the low flows during non-irrigation seasons have been attributed to seepage from reservoirs, groundwater base flows, irrigation return flows, and effluents from wastewater treatment plants near large cities and population centers (Moore et al., 2008).

Previous investigations have identified multiple salinity sources to the Rio Grande but there is a general disagreement about which source is dominant. Upwelling of deep groundwater has been suggested as important salinity source at distal ends of alluvial basins where groundwater is inferred to flow upward due to uplifted basement blocks (Figure 1B; Moore and Anderholm, 2002; Phillips et al., 2003; Hogan et al., 2007; Moore et al., 2008; Williams et al., 2013). However, in the semi-arid region, the Rio Grande is a losing stream in most of southern New Mexico and western Texas (Driscoll and Sherson, 2016), and in dry years the Rio Grande channel completely dries out for several months during non-irrigation seasons. In addition, intense pumping of the groundwater has led to significant decreases of the groundwater table. These observations raise the question of the importance of the natural upwelling of saline groundwater in these alluvial basins as a dominant mechanism of direct input into the Rio Grande.

Other studies have suggested possible sources of salinity related to agricultural activities (e.g., evapotranspiration from agricultural field and return flows, displacement of shallow groundwater, irrigation use of groundwater, and application of fertilizers), urban/industrial inputs (e.g., point sources of treated or non-treated waste effluents), and shallow geological sources (e.g., dissolution of secondary evaporites in soil zones, and shallow brackish groundwater) (Figure 1B; Lippincott, 1939; Haney and Bendixen, 1953; Wilcox, 1957; Ellis et al., 1993; Moore and Anderholm, 2002; Phillips et al., 2003; Witcher et al., 2004; Hogan et al., 2007; Szynkiewicz et al., 2011; Moyer et al., 2013; Driscoll and Sherson, 2016).

Groundwater is another important source of irrigation water in this region particularly in drought years to supplement shortages of Rio Grande surface water. While elevated salinity values in some local groundwater is usually due to long residence time in sedimentary and crystalline bedrock in this region, evapotranspiration associated with agricultural fields has been suggested to increase salinity in irrigation return flows that in turn recharge to shallow groundwater underneath agricultural areas (Walton et al., 1999). It is unclear, however, whether there are additional important salinity inputs to shallow groundwater from other sources such as application of fertilizers on agricultural fields, and urban effluents discharging directly to the Rio Grande or infiltrating to shallow aquifers (Figure 1B).

We selected fifteen locations (RG-1 to RG-15) along a ~200 km-stretch of Rio Grande, from Elephant Butte Reservoir in southern New Mexico to State Highway 273 in the city of El Paso, Texas for river sample collection (Figure 2; Appendix Table 1). The water in the Elephant Butte reservoir (RG-1) is the source of river water in this region and represents a “baseline” for salinity investigations. Locations RG-2 (Williamsburg) and RG-3 (Percha Dam) have some agricultural activities but less than further downstream locations. RG-2 and 3 are located in an area with many reported natural occurrences of hot springs and geothermal activity (e.g., Williams et al., 2013; Pepin et al., 2015). However, local spa resorts near Location RG-2 may also lead to human-induced discharge of highly saline, geothermal groundwater sourced from artesian wells (Szynkiewicz et al., 2015b) and there is also effluent inflow into the Rio Grande at RG-2 from a nearby wastewater treatment plant. Locations RG-4 through RG-7 are along the Hatch and Mesilla Valley irrigation districts in the southern Palomas Basin and Mesilla Basin. Locations RG-8 through RG-11 are in close proximity to the city of Las Cruces, a relatively large urban area in southern New Mexico (population ~100,000) with one WWTP (UB-1) located below RG-8. Locations RG-12 through RG-15 are located between Anthony, New Mexico and El Paso, Texas (population ~800,000) near the distal end of the Mesilla Basin where plausible upwelling of highly saline groundwater to the Rio Grande has been previously suggested (e.g., Phillips et al., 2003; Hogan et al., 2007). However, there may be also inflow of saline water originating from shallow groundwater of the salt flat intersected by Montoya Drain near El Paso, which conveys return flows to the Rio Grande from nearby irrigation districts (Szynkiewicz et al., 2015b). At these 15 sampling sites, a total of ~200 river samples were collected monthly between 2014 and 2016 for elemental and isotopic analyses.

In this study, we also sampled 79 groundwater wells (GW-1–GW-79), 7 waste effluents from the cities of Las Cruces (UB-1), Sunland Park (UB-4), and El Paso (UB-2, 3, 5, 6, 7), and 35 irrigation canals/drains (DR-1–DR-35) along the studied stretch of the Rio Grande (Figure 2; Appendix Table 1). It is noted that due to the nature of population distribution in this region, many of the groundwater, urban, and agricultural samples were from the Messila Basin in the south while samples from the north study area were limited. Available archived samples from Szynkiewicz et al. (2015b) were also included to determine signatures of an agricultural drain near Tornillo, Texas which combines return flows from downstream irrigation districts located below the city of El Paso (Figure 2). Two archived samples (Szynkiewicz et al., 2011, 2015b) of shallow saline groundwater from monitoring piezometers near Fabens, Texas, one archived sample from an artesian well with geothermal groundwater near Truth or Consequences, New Mexico, and two shallow groundwater samples from private wells near Anthony, New Mexico were also included for isotope analysis. Finally, several archived samples of commonly used fertilizers in the region were also analyzed for U, B, and Sr isotope composition (Szynkiewicz et al., 2015b).

Water samples were collected into 1L acid washed HDPE Nalgene bottles. Field measurements of pH, temperature, and electric conductivity were taken in situ during sample collection using a YSI Professional Plus multimeter, which was calibrated prior to sampling. The samples were stored in a cooler for ~5–6 h before arrival to the laboratory. In the laboratory, ~400 mL of each water sample was filtered using a 0.45 μm cellulose acetate filter to remove suspended sediment and particulates, and the filtered water was placed in two 250 mL acid washed HDPE Nalgene bottles. One bottle was acidified with 3 drops of ultrapure concentrated nitric acid for cation and isotope analysis, and the second bottle was archived without acidification for immediate anion analysis. The samples were stored at 4°C in the refrigerator before analysis.

For major cation concentrations (Na, Ca, Si, K, and Mg), the acidified water sample was analyzed on a Perkin Elmer 5300DV Optical Emission Spectrometer (OES) at University of Texas at El Paso (UTEP). Two water standards (USGS M-210 and NIST 1640a) were analyzed at least 3–5 times during each analytical session to assess analytical precision of cations. The analytical precision was estimated to be better than 10% for major cations.

For major anion concentrations (Cl, SO₄, and NO₃), the non-acidified filtered sample was analyzed using a Dionex ICS-2100 at UTEP. An in-house water standard was measured at least twice during each analytical session to ensure accuracy. In general, the analytical precision of anions for the standards used was better than 12%. A selected number of samples were measured for alkalinity by the titration method. For the rest of samples, their alkalinity values were calculated based on the mass charge balance of the analyzed major chemical species.

A subset of samples was analyzed for trace element concentrations (U, B, and Sr) with a Thermo Fisher Scientific X Series 2 ICP-MS at the Pennsylvania State University Laboratory for Isotopes and Metals in the Environment (LIME). The NIST water standard (NIST 1640a) was used to assess accuracy. Analytical error was between 1 and 11%. In this study, only U, B, and Sr concentrations are reported and discussed. More details of the analytical methods for major and trace elements are reported in Nyachoti (2016) and Garcia (2017).

Uranium isotopic ratios, ²³⁴U/²³⁸U, were measured for a subset of 116 river and other types of water samples at UTEP. A minimum of 50 ng of U was used to carry out U isotope analysis. The volume needed to obtain 50 ng of U was calculated using U concentrations [U] with trace elements analysis carried out at LIME. Samples were evaporated overnight at 90°C in a class-100 clean room. U column chemistry followed a procedure similar to Chabaux et al. (1995).

Purified U samples were analyzed on a Nu Plasma HR MC-ICP-MS at UTEP to determine ²³⁴U/²³⁸U isotopic ratios with the uranium standard (NBL145B) for standard-sample-standard bracketing. The estimated errors (2SE) of the isotope ratios were better than 0.5%. From the measured isotope ratios of ²³⁴U/²³⁸U, we calculated (²³⁴U/²³⁸U) activity ratios (here the parenthesis specifies the activity ratio) using decay half-lives of ²³⁴U and ²³⁸U (Cheng et al., 2000). The USGS rock standard BCR-2 was processed along with column chemistry and measured multiple times to ensure accuracy of measurements: average measured (²³⁴U/²³⁸U) activity ratio is 1.004 ± 0.002 (2SE, n = 10), consistent with the expected (²³⁴U/²³⁸U) activity ratio at equilibrium (1.0). The procedure blank for U was ~30 pg and negligible.

A total of 45 representative water samples (river and other types of water samples) were selected for B isotope analysis at the Institut de Physique du Globe de Paris (IPGP) in France. The measurements were made with the procedure described in Louvat et al. (2010, 2014). The procedure required ~300 ng of boron and the final solution to be at 200 ng/mL of B for isotope analysis. For samples with high salinity, the sample was diluted with 5 mL of 18 MΩ water to prevent clogging of the column.

The purified B samples were analyzed on a Neptune MC-ICP-MS at IPGP using a direct injection high efficiency nebulization (d-DIHEN) method developed by Louvat et al. (2014) to measure boron isotopic ratios. Each sample was measured three times. The standard reference material SRM 951 was used at the sample concentration for standard-sample bracketing. The boric acid reference material AE121 was measured 8 times to assess accuracy and precision (19.44 ± 0.07%, 2 SD). The North Atlantic Surface Seawater (NASS-5) standard was also processed along with the samples (39.55 ± 0.15%, N = 2; in agreement with Louvat et al., 2014).

A total of 106 river and other types of water samples were selected for ⁸⁷Sr/⁸⁶Sr isotope analysis at UTEP. Around 25 ml of water samples were evaporated to dryness, the dried samples were re-dissolved in 3.5 N HNO₃ then separated and purified through Sr-Spec resin. The purified samples were measured for ⁸⁷Sr /⁸⁶Sr ratios on MC-ICP-MS using the standard-sample bracketing method with NIST SRM 987 as the standard bracketing solution (Konter and Storm, 2014). About 200 mg of rock standard BCR2 was acid-digested in HNO₃-HF and HCl-H₃BO₃ then separated through Sr-Spec resin. ⁸⁷Sr/⁸⁶Sr ratios in BCR2 reported an average value of 0.70502 ± 0.00001 (2σ; N = 5) consistent with values reported in the literature (0.70502; Raczek et al., 2003). Blanks for Sr analysis (~82 pg) are negligible.

In general, salinity values in the Rio Grande river in our study area vary both spatially and temporally. Between May 2014 and May 2016, the measured TDS values of the Elephant Butte Reservoir water (EBR: RG-1) showed a narrow range of 450 to 630 mg/L (average: 550 ± 60 mg/L; n = 15; Figure 3). In contrast, TDS values of the Rio Grande at downstream locations showed much greater variability. More specifically, under high flow conditions from April to July, Rio Grande water (RG-2 to RG-15) had TDS values ranging from ~450 to 840 mg/L (Figure 3). Higher TDS values (e.g., 840 mg/L) were generally observed in close proximity to agricultural areas and large urban centers (e.g., between RG-4 and RG-15). Under low flow conditions from August to March, Rio Grande water below EBR showed a large variability of TDS values from ~370 to 5000 mg/L (Figure 3). The highest TDS values (up to ~5,000 mg/L) was observed in close proximity to both Truth or Consequences (RG-2) and west El Paso (RG-15). The lowest TDS values (e.g., 370 mg/L) were observed near urban areas (RG-5 and RG-11).

Major elemental concentrations for the Rio Grande river samples between May 2014 and May 2016 (http://dx.doi.org/10.1594/IEDA/111231) were used to calculate water chemistry types and saturation indices (SI) by Nyachoti (2016) and Garcia (2017). The main results are summarized here. The EBR (RG-1) water was largely the Ca-Na-HCO₃-SO₄ type during both irrigation and non-irrigation seasons. The downstream Rio Grande sites during the irrigation seasons were also dominated by the Ca-Na-HCO₃-SO₄ type. During the non-irrigation seasons, the water types at the downstream locations were more variable, including Ca-Na-HCO₃-SO₄ (RG-3, 4, 6), Na-Ca-SO₄-HCO₃ (RG-5, 7, 8), Na-Cl-SO₄ (RG-9, 14, 15), and Na-Cl types (RG-2). These variable water types at downstream locations are mostly related to the contributions of additional Na, Ca, Cl, and SO₄ sources that lead to changes of the Ca-Na-HCO₃-SO₄ type from upstream. According to the calculated SI values (Nyachoti, 2016; Garcia, 2017), all river water samples between May 2014 and May 2016 were saturated with respect to calcite and dolomite and under-saturated with respect to gypsum, halite, and thenardite. The degree of saturation (SI values) for calcite and dolomite increases from RG-1 to RG-15 while the SI values for gypsum and halite show higher values for selected locations (RG-2, 6, 7, and 15).

The salinity of Rio Grande river at downstream locations to RG-1 is impacted by possible salinity end members from geological, agricultural, and urban sources. The groundwater, agricultural canal/drain water, and treated wastewater samples from this study showed highly variable TDS values: 320–16,900, 550–4,300 mg/L, 780–1,554 mg/L, respectively (Figure 3). The water types of these end-member samples were also highly variable, mostly overlapping with the water types of the Rio Grande. The groundwater and agricultural canal/ drains had mainly Ca-HCO₃-SO₄, Ca-Na-HCO₃-SO₄, Na-SO₄, and Na-Cl water types, and the dominant water type in the urban wastewater was Ca-Na-Cl-SO₄ (Nyachoti, 2016; Garcia, 2017). It is noted that as a limitation of the study's approach to assess the end member salinity sources, the sample locations for the groundwater, agricultural canal/drain, and wastewater treatment plants are largely restricted to the nature of the distribution of population and agricultural centers that is centered along the river in the southern study area and only with a limited number of sample locations from the northern study area.

Generally, the measured U concentrations in the Rio Grande water samples (~0.6 to 10 μg/L) are considered to be higher than U concentrations of other major rivers (Chabaux et al., 2003) and the average U concentration of ocean water (~3 μg/L; Drever, 1997). Similar to the observations of the TDS values, the Rio Grande river water had less variable U concentrations during the irrigation seasons (~2 to 4 μg/L) and large variability during the non-irrigation seasons (~0.6 to 10 μg/L) (Appendix Table 2). There is a positive correlation between U and HCO₃ concentrations in Rio Grande river samples (Figure 4). Noticeably, elevated U concentrations were observed during non-irrigation seasons mostly near agricultural areas and urban areas in Palomas and Hueco Basin (RG-3, 4, 15; Figure 5A). The lowest U concentrations in the Rio Grande were observed during non-irrigation season below the inflows of urban effluents (RG-9 and RG-14) as well as in several agricultural areas in Mesilla Basin (RG-7) (Figure 5). This observation is consistent with lower U concentrations measured in the urban effluents in this study (<0.1 to 2 μg/L; Figure 5A). By contrast, agricultural drains in this study generally have elevated U concentrations (~2 to 12 μg/L). The groundwater samples showed the most variations in U concentrations among the end-member samples, between <0.1 to 95 μg/L.

During irrigation seasons, the measured (²³⁴U/²³⁸U) ratios of the Rio Grande river water showed a narrow range between 1.7 and 1.8, similar to the Elephant Butte reservoir water (Figure 5B). During non-irrigation season the (²³⁴U/²³⁸U) ratios of Rio Grande showed much greater variations, with higher ratios (2.0 to 2.5) in locations RG-2, RG-3 and RG-4, in proximity to Truth or Consequences and Caballo Reservoir, and lower ratios (1.5 to 1.6) in locations RG-6, RG-9, RG-14, and RG-15, mostly observed near large agricultural and urban areas (Figure 5B). Noticeably, agricultural canals and drains showed characteristically lower (²³⁴U/²³⁸U) ratios (~1.1 to 1.6) compared to the Rio Grande. The urban wastewater had slightly higher (²³⁴U/²³⁸U) ratios (1.5 to 1.9) and mostly similar to the Rio Grande river water. The studied groundwater samples showed a large range of (²³⁴U/²³⁸U) ratios (~1.2 to 2.7). Significantly higher (²³⁴U/²³⁸U) ratios in groundwater were observed from areas with hot spring and hydrothermal activities such as from Truth or Consequences, NM or Fabens, TX, compared to relatively low ratios generally observed in shallow irrigation wells near agricultural areas along the Rio Grande (Appendix Table 2).

Both temporal and spatial variability of B concentrations in Rio Grande water samples follows similar trends as observed in U concentrations. There is also a positive correlation between B and Na concentrations (Figure 4). More specifically, the Rio Grande river water had limited variation of B concentration during the irrigation seasons (~90 to 200 μg/L) compared to much larger variations (~180 μg/L to 1,000 μg/L) during the non-irrigation season (Figure 6A). The groundwater, drain and urban water samples all had higher B concentrations (up to ~3,700 μg/L) compared to the Rio Grande river samples (Figure 6A).

The Rio Grande river samples also showed similar spatial and temporal trends of δ¹¹B values to the measured (²³⁴U/²³⁸U) ratios, with smaller variation during irrigation seasons (+8% to +11%) and larger variation during non-irrigation seasons (+3% to +16%; Figure 6B). Urban wastewater had uniquely lower δ¹¹B values (+4% to +11%) compared to higher values in drains (+9% to +15%). The highest δ¹¹B values were measured in groundwater (+11% to +30%), with one exception for the geothermal groundwater sample in Truth or Consequences with a significantly lower δ¹¹B of +6% (Figure 6B).

The Rio Grande river samples showed low Sr concentrations during the irrigation seasons (~0.5 to 1.2 mg/L) and higher concentrations during the non-irrigation seasons (~1 to 3 mg/L) (Figure 7A). There is a positive correlation between Sr and Ca concentrations in these samples as well (Figure 4). The highest Sr concentrations in the Rio Grande water (~2 to 3 mg/L) were observed during non-irrigation season in Locations RG-2, 6, 14, and 15. Noticeably higher Sr concentrations were observed in agricultural drains (~2 to 4 mg/L). The groundwater samples showed the largest variation of Sr concentration, from ~1 to 16 mg/L. The urban effluent samples had Sr concentrations similar to the Rio Grande water (~1 mg/L).

During irrigation seasons, the measured ⁸⁷Sr/⁸⁶Sr ratios of the Rio Grande showed a narrow range of values (0.710 to 0.711), similar to Elephant Butte reservoir water (Figure 7B). In contrast, the ⁸⁷Sr/⁸⁶Sr ratios of Rio Grande during non-irrigation seasons showed much greater variation, with lower ratios in locations RG-9 and RG-15 (0.709 to 0.711) and higher ratios in locations RG-2 and RG-6 (0.712 to 0.715). Slightly high ⁸⁷Sr/⁸⁶Sr ratios (0.711–0.712) were observed in agricultural drains in Mesilla Basin, while lower ratios (~0.710) were observed in drains from the Hueco Basin (Figure 7B). The urban effluent had similar or slightly lower (⁸⁷Sr/⁸⁶Sr) ratios (~0.709 to 0.710). The studied groundwater samples showed the most diverse groups of values: much higher (⁸⁷Sr/⁸⁶Sr) ratios in Palomas Basin (~0.720) compared to low ratios in downstream locations in Mesilla and Hueco Basin (~0.708 to 0.710).

Previous studies have identified multiple salinity sources of geological, agricultural, and urban origins in the semi-arid Rio Grande watershed of southern New Mexico and western Texas (Figure 1B; e.g., Ellis et al., 1993; Phillips et al., 2003; Moyer et al., 2013). Total dissolved solid contents, major element chemistry, and several light stable isotope ratios (δ¹⁸O, δH, and δ³⁴S) have been used to attempt to identify and quantify contributions from each possible salinity end-members in the Rio Grande watershed. Several specific processes that have been previously identified, including: (1) natural upwelling of deep saline groundwater at the distal ends of alluvial basins, (2) mixing with saline groundwater in shallow aquifers via faults, (3) agricultural return flows from irrigation canals and drains, and (4) urban streams and discharges of treated or untreated wastewater from urban areas. Indeed, the groundwater, agricultural canals/drains, and urban wastewater samples in this study all show variable but elevated TDS (Figure 3) and are justified as potential salinity sources to the Rio Grande. However, accurate identification and quantification for contributions of these salinity end members are difficult due to the general overlapping values of TDS, major element concentrations and elemental ratios, as well as light stable isotope ratios (δ¹⁸O, δD, and δ³⁴S) (e.g., Szynkiewicz et al., 2015b). In addition, several major elements (Ca, K, SO₄ or NO₃) or stable isotope ratios (δ¹⁸O, δD, and δ³⁴S) may not behave conservatively in the Rio Grande watershed (e.g., Phillips et al., 2003; Szynkiewicz et al., 2015b), undermining their use as salinity tracers to identify the original sources. It is expected that U and Sr isotope ratios may behave conservatively in rivers under oxic conditions, while B isotope ratios may experience modification due to its intrinsic nature as a light stable isotope system. To apply these isotope ratios of trace elements to identify major salinity sources, it is generally expected that these trace elements (U, Sr, and B) mimic the changes of major elements (TDS, or Ca, Mg, K, Na, Cl, SO₄, NO₃) in water. Such an assumption is not easily to validate due to the different geochemical and physical behavior of these different elements. However, the Sr, B, and U concentrations of Rio Grande water samples in this study show a positive trend with respect to the changes of Ca, Na, and HCO₃ concentrations, respectively (especially for Sr and B, Figure 4), strongly suggesting that the salinity changes were accompanied by the changes of Sr, B, and U concentrations. Such positive correlations of Sr, B, and U with TDS values support the use of their isotope tracers to identify sources of major elements and salinity in this study.

Here, we first assess using the U, Sr, and B isotope ratios of groundwater, agricultural, and urban samples to characterize possible salinity end members in the Rio Grande watershed in this study area.

We aim to assessing the spatial and temporal variability of the Rio Grande river salinity for the study period with the above identified U, B, and Sr isotope and elemental signatures of the agricultural, urban, and geological sources. However, during the irrigation (high flow) seasons, the concentrations and isotopic compositions of U, B and Sr in the Rio Grande were very similar to the Elephant Butte reservoir water (Figures 5–7). This is in agreement with that during high river flows, the river chemistry mainly reflects the chemistry of upstream reservoirs. In contrast, the Rio Grande showed highly variable U, B and Sr concentrations and isotope ratios during the non-irrigation (low flow) seasons and we hence focus on the Rio Grande during non-irrigation (low flow) seasons from Elephant Butte reservoir to El Paso, TX.

The multi-tracer dataset of U, B, and Sr isotope ratios highlights the presence of multiple salinity sources in the Rio Grande watershed and that their contributions to Rio Grande river var both spatially and temporally (Figure 8). As discussed in previous sections, despite that no single isotope tracers of U, B, and Sr can effectively distinguish the multiple salinity sources, the combination of the isotope tracers and elemental concentrations of U, B, and Sr shows potentials in a multi-tracer approach. Here, we use a mass balance model to quantify the contributions of individual salinity source, as summarized below.

For a given sampling location, the isotopic composition of Rio Grande water (R_RioGrande) and elemental concentration (C_RioGrande) for an element (U, Sr, or B) can be calculated as:

Here, the f values in the above equations are the relative water mass contribution of salinity sources: Elephant Butte reservoir water (EBR), deep groundwater, shallow groundwater, agricultural water, and urban water. For each salinity source, the characteristic values for U, Sr, and B isotope ratios and concentrations have been discussed and are summarized in Table 1. The f values from each salinity source can be constrained by using U, Sr, and B isotope ratios and concentrations in the measured river water sample with a mathematical inverse procedure (e.g., Roy et al., 1999; Chetelat and Gaillardet, 2005; and Engel et al., 2016). Briefly, the above set of equations is over determined for a given water sample because of the number of constraining equations (7 in this case: 3 from U, Sr, B concentrations, 3 from isotope ratios, and 1 from Equation 3) is greater than the number of the unknowns (5: f values for EBR, deep_GW, shallow_GW, agricultural, and urban). In the inversion technique, we can solve the 5 unknowns (f values) by using a Monte Carlo procedure to minimize the difference between the observed and the calculated values in the 7 equations by iteration, considering the uncertainties of measurements R and C (Engel et al., 2016). If the algorithm converges to the preset difference level (e.g., 5%), a solution is found. For this study, the above mass balance approach is applied at 5 selected locations downstream to EBR to El Paso, TX for both irrigation and non-irrigation seasons (Table 2). The 5 selected locations are characterized by various geological, agricultural, and urban settings, respectively. The above algorithm was solved 100 times and the average f values from the 100 solutions are presented in Table 2 and Figure 9. It is noted that although the f values represent the contribution calculated from U, Sr, and B mass balance, the positive correlation between U, Sr, and B concentrations and the TDS values in these samples supports that the f values also represent the salinity contribution from each end member.

The mass balance quantification shows that during the irrigation (high flow) seasons, salinity at the 5 sampling locations downstream to Elephant Butte reservoir is dominated (>85%) by EBR release water. Only a small fraction of urban signature (~13%) is observed at RG-15 location next to the city of El Paso (Table 2). Small amounts of agricultural, groundwater, and urban (<2%) can be observed at the RB-2 site, which is at the Truth or Consequences area with known upwelling of deep groundwater, small agricultural and urban activities. The Elephant Butte reservoir stores the water coming from the headwater region of the Rio Grande watershed in southern Colorado and Northern New Mexico where climate is temperate and with more precipitation than the southern New Mexico and West Texas. The chemical solutes in EBR are mainly from chemical weathering processes within the Critical Zone in the headwater regions (e.g., Szynkiewicz et al., 2015a).

During the non-irrigation (low flow) seasons, salinity in the Rio Grande is still largely contributed by an EBR type of water (~40–80%), most likely as the residual EBR flows of the previous irrigation seasons. However, due to the decreased river flow, other salinity signatures can be readily observed at many locations (Figure 9), with a significant component of urban signature (8–58%) and a large contribution from shallow groundwater (up to 43%). Agricultural water contribution is low (~2–4%) and deep groundwater contribution is also low except at the RB-2 location where upwelling of groundwater is documented (~23%).

Indeed, the above salinity quantification is consistent with the previous notation that upwelling of deep saline groundwater and mixing with groundwater in shallow aquifers may be responsible for salinity increases in certain locations such as at the distal end of the Mesilla Basin, between Locations RG-12 and 15 (Phillips et al., 2003; Witcher et al., 2004; Hogan et al., 2007; Hiebing et al., 2018). Upwelling of geothermal groundwater to shallow aquifers has been also suggested in areas with uplifted basin basements such as in the hot spring district of Truth or Consequences, near Location RG-2 and 3 (Williams et al., 2013).

However, there is no clear evidence for direct connection between the Rio Grande and the deep groundwater via natural upwelling in the most part of the Mesilla Basin. Below Elephant Butte and Caballo Reservoir, the Rio Grande is a losing stream with significant infiltration of surface water to the alluvial aquifers (e.g., Szynkiewicz et al., 2015a,b; Driscoll and Sherson, 2016). During the irrigation season, the Rio Grande surface water is diverted to the large network of irrigation canals. The end points of main agricultural drains are directly connected to the Rio Grande and return the excess irrigation water from the agricultural fields. Therefore, several studies have proposed that agricultural return flows may contribute to river salinity either through use of groundwater with elevated salinity for irrigation, flushing of secondary salts precipitating in soil during dry seasons, and/or high evapotranspiration rates in the agricultural fields (e.g., Ellis et al., 1993; Szynkiewicz et al., 2015b). Our mass balance quantification suggests that the agricultural contribution is present but not significant (at 2–4%). However, a large contribution from shallow groundwater (up to 43%) is documented in this region that may represent the indirect contribution of agricultural water to Rio Grande via infiltration underneath agricultural fields and shallow groundwater flows back to Rio Grande (Walton et al., 1999).

In addition, urban effluents from the cities could be additional sources of salinity since they significantly increase stream flows during the non-irrigation seasons when there is no water release from Elephant Butte and Caballo reservoirs. Indeed, high urban contributions are observed at locations near large cities and towns in the study area (RG-15 is near El Paso, TX as discussed above; RG-2 is near T or C, NM). Note that one WWTP (not sampled in this study) is located ~3 km upstream from RG-2 and highest NO₃ was mainly observed during non-irrigation season (up to 15 mg/L) when there was no water releases from upstream Elephant Butte reservoir. Additional sources of salinity might be spa resorts from Truth or Consequences that discharge geothermal water directly to the Rio Grande. Indeed, both local groundwater used for municipal supplies and artesian geothermal groundwater from spa resorts have elevated TDS (~800–4,000 mg/L) due to high concentrations of Na and Cl (~200–1,200 mg/L).

It is also noted that during non-irrigation seasons, the percentage contribution of EBR is expected to decrease with increasing distance from EBR at downstream locations, due to the addition of other non-EBR types of water sources. However, the calculated EBR contribution in this study increases from RG-2 (~40%), to RG-4 (~65%), and RG-6 (~80%; Figure 9). Such an observation may point to the presence of an additional source (or sources) that has similar isotopic composition to the EBR type water but such a source was included in the EBR contribution by the Monte Carlo mass balance model. For example, the seepage of EBR and Caballo reservoirs to shallow aquifers may lead to increased contribution of a shallow groundwater end-member that has “EBR”-like isotope signatures to the Rio Grande at downstream locations (such as at RG-4 and RG-6). The shallow groundwater contribution calculated by the Monte Carlo mass balance model is in fact unusually low at these locations (Figure 9), especially at RG-6, which is located at the distal end of the Palomas Basin and expected to receive high groundwater contribution (e.g., at RG-15 location; Figure 2). Such inconsistencies of the calculated EBR and shallow groundwater contributions at RG-4 and RG-6 locations point to several limits of the current Monte Carlo mass balance model: (1) the current model cannot solve effectively the end-members with similar or overlapping signatures; (2) the current model does not take into account of river flow discharge that may provide additional constraints on the mass balance model; (3) the current model does not offer sensibility and uncertainty analysis that may help to assess the model results with uncertainties. Future model calculations that may include river discharges, more accurate characterization of salinity end-member and uncertainties will be more useful to address the salinity issues for the complex surface-shallow-deep groundwater system in the Rio Grande watershed.