Dissolved solids can negatively impact livestock drinking water. To characterize dissolved solid concentrations in the northern Great Plains, we studied 45 water sources over 11 yr. Ca, F, Mg, Na, and total solids sometimes exceeded recommended levels but rarely reached levels empirically shown to negatively impact livestock. Fe often attained concentrations that reduce water consumption, which can reduce feed intake and weight gain. Excessive Fe was likeliest in creeks and reservoirs. SO4 sometimes reached concentrations that reduce weight gain but rarely reached concentrations causing health problems. Excessive SO4 was least likely in reservoirs. A safeguard against water quality problems is ensuring livestock can access multiple water source types (e.g., wells and reservoirs) so that they can use taste and smell to choose acceptable water. Where this is impossible, water quality monitoring is of greater importance. Inexpensive electrical conductivity meters roughly estimate water quality and can identify watering locations requiring more detailed monitoring.
Excessive dissolved solids in livestock drinking water reduce feed efficiency and growth, cause health problems (e.g., scours, tooth decay), and can be fatal (Weeth et al. 1960; Van Leeuwen 1999; Burgess et al. 2010; Ollivett and McGuirk 2013; Choubisa and Choubisa 2016; Sharma et al. 2017). Safe limits for dissolved solids are unclear. Several publications recommend maximum levels (e.g., Socha et al. 2003; Olkowski 2009), but studies often exceed recommended maximums without apparent consequence. For example, a recommended maximum for total dissolved solids (TDSs) is 3000 ppm (e.g., Olkowski 2009), but in dairy cattle, 3570 ppm for 200 d caused no detected harmful effects and instead prolonged milk production (Bahman et al. 1993).
Given safe levels are uncertain, how can we determine where dissolved solids pose risks? This question is particularly relevant in rangelands where uncertainty about what concentrations exist compounds uncertainty about what levels are safe. We continued work by Petersen et al. (2015) to quantify dissolved solids in the northern Great Plains. Petersen et al. (2015) measured solids at 45 water sources for 5 yr, and we measured 6 more yr to quantify interannual variation. We hypothesized accounting for water source type (reservoirs spring, well, creek), precipitation, and other variation would improve dissolved solid predictions. If so, this should help managers anticipate water quality problems.
This study occurred near Miles City, Montana at the US Department of Agriculture–Agricultural Research Services Fort Keogh Livestock and Range Research Laboratory, which has a mean elevation of 780 m. Vegetation is mixed-grass prairie and includes grasses like western wheatgrass (Pascopyrum smithii [Rydb.] Á. Löve), shrubs such as big sage (Artemisia tridentata Nuttall), and forbs such as American vetch (e.g., Vicia americana Muhl. ex Willd). The uppermost geologic layer is the Fort Union formation, which is common in Montana, Wyoming, North Dakota, South Dakota, and Saskatchewan. This formation contains beds of fine- and medium-grained sandstone, siltstone, mudstone, coal, and clinker (Smith et al. 2000).
Study locations were 12 reservoirs, 10 springs, 12 wells, 10 creeks, and the Yellowstone River. These locations were scattered about a roughly 15 × 15-km area (for a map of locations, see Petersen et al. 2015). The Yellowstone River is included with creeks for analysis and presentation. Well depth ranged from 30 m to 110 m, and reservoir surface area ranged from 4 ha to 87 ha. From 2009 to 2019, water sources were sampled between 5 May and 10 June and then between 3 September and 21 October, except 28%, 1%, 17%, and 13% of creek, well, reservoir, and spring samples are missing due to sources being dry. For wells, water was sampled from 950-L fiberglass drinking troughs filled with pumped water. For other sources, samples were gathered ≤ 3 m from the water's edge near areas cattle congregated to drink. Water was sampled at the surface by filling 250-mL plastic bottles (Thermo Scientific, Waltham, MA). Samples were iced in coolers and then shipped for analysis in cooled shipping containers within 24 h of collection. Concentrations of Ca, Cl, F, Fe, Na, Mg, and SO4 were measured, and TDS was estimated from electrical conductivity (Midwest Laboratories, Omaha, NE).
We analyzed log-transformed concentration data with a multivariate mixed model having fixed effects for source (creek, well, reservoir, spring); random effects for year and location; and co-variates for Julian day of measurement and total precipitation the 365 d preceding measurement. Julian day × source and precipitation × source were also included. Precipitation data were gathered at the Frank Wiley Field Airport near Miles City, Montana ( http://www.noaa.gov). Values were missing when locations were dry. Had these locations been measured just before all water evaporated, concentrations would have been relatively high. Therefore, the data are not “missing at random,” and ignoring missing values would generate biased estimates (Rubin 1976). Thus, we replaced missing observations with the observation having the greatest mean (averaged over dissolved solids) for the location. We assumed noninformative prior distributions for all parameters (Huang and Wand 2013; Gelman et al. 2014) and fit the model using a Gibbs sampler coded in Fortran (Intel Corporation 2013). In quantifying differences between early and late sampling periods, we use average sampling dates (21 May and 19 September) as predictor values.
Dissolved solids except Cl varied widely among locations (Fig. 1). Averaged over years, F exceeded the maximum recommended by Socha et al. (2003) in some wells. This was also true for Fe in all sources except wells and Na and SO4 in all sources except reservoirs (see Fig. 1). Even where mean concentrations were safe, individual measurements sometimes exceeded safety thresholds. For example, one creek exceeded the TDS threshold twice (data not shown), even though mean TDS for that creek was well below the threshold (see Fig. 1). Likewise, 15% of reservoir samples exceeded the Na threshold, even though mean Na concentrations were below the threshold in all reservoirs (see Fig. 1).
Unexplained interannual variation was miniscule for all dissolved solids except Fe. Mean Fe ranged from ∼6 ppm in 2009 to ∼18 ppm in 2016. Several dissolved solids decreased with increasing precipitation in reservoirs, and Ca and Mg increased with increasing precipitation in creeks and wells (Fig. 2). Precipitation ranged from 150 to 490 mm over the study period, and long-term (30-yr) average annual precipitation was 315 mm.
Fig. 1 suggests mean concentrations varied considerably among the four water sources, and Fig. 3 confirms this for all but SO4. Creeks and reservoirs were highest in Fe, creeks and springs were highest in Ca and Mg, and wells were highest in F, Cl, Na, and TDS (see Fig. 3). Time of measurement (May vs. September) was a minor source of variation (see Fig. 3).
Ca, Cl, F, Mg, Na, and TDS were generally safe
Mean TDS and Cl were unproblematic, though two TDS measurements (5910, 9490) exceeded the 5000-ppm limit of Socha et al. (2003). The Ca mean at one source exceeded the 150-ppm limit of Socha et al. (2003), but no measurements exceeded greater recommended limits (e.g., 1000 ppm Carson 2000). Most guidelines ignore Ca as a water toxin (Soltanpour and Raley 1993; Carson 2000; Higgins et al. 2008). Some wells had mean F exceeding the 2.0-ppm limit of Socha et al. (2003). However, our maximum concentration (8 ppm) equaled the maximum recommended by Olkowski (2009). Prolonged F exposure at 8 ppm has modestly changed cattle teeth and bones without otherwise impacting health (Shupe et al. 1992). No Mg means exceeded the 100-ppm limit of Socha et al. (2003). Most guidelines ignore Mg as a potential toxin (Soltanpour and Raley 1993; Carson 2000; Higgins et al. 2008). Mean Na at some sources was double the 300-ppm limit of Socha et al. (2003), but according to studies on cattle, only two of our measurements (2110, 3275 ppm) were great enough to reduce feed intake and body weight (Embry et al. 1959; Weeth et al. 1968; Alves et al. 2017).
Fe and SO4 were sometimes concerning
Even our maximum Fe concentration (40 ppm) is unlikely to directly impact livestock production and health (Wright 2007; Olkowski 2009), but Fe can indirectly impact livestock by reducing water palatability and consumption. Mean Fe in some of our sources exceeded 8 ppm, and Fe at 8 ppm reduced water consumption by dairy cattle about 25% over 22 h (Genther and Beede 2013). Factors that reduce water consumption (e.g., fecal contamination, high TDS) often reduce weight gain (Willms et al. 2002; Patterson et al. 2004; but see Petersen et al. 2016).
Mean SO4 of several water sources exceeded the 500-ppm limit of Socha et al. (2003), and some source means exceeded the 1 000-ppm limit cited elsewhere (Carson 2000; Higgins et al. 2008). With feedlot steers, 1730 to 2360 ppm of SO4 reduced gain 0.05 to 0.1 kg d–1 (Loneragan et al. 2001; Patterson et al. 2004; Sexson et al. 2010). Assuming similar effects for range cattle, about 1–1.5% of our SO4 values could reduce weight gain. Prolonged exposure to our two greatest SO4 concentrations (4820, 9591 ppm) would be a concern: Steers drinking water containing 4650–5800 ppm of SO4 suffered severe illness and high mortality (Hamlen et al. 1993; Patterson et al. 2004).
Anticipating and preventing water quality problems
Dissolved solids that never reached concerning levels in our study may exceed safe levels elsewhere in the northern Great Plains. However, general patterns we observed in our area should reflect patterns occurring elsewhere, because the same physical processes regulate water quality everywhere. For example, reservoirs had greater concentrations in dry years, which indicates evaporation in dry years increases concentrations more than runoff and leaching in wet years. Additionally, the tendency of creeks toward greater concentrations in wet years reflects increased runoff and leaching in wet years. Evaporation, runoff, and leaching seem likely to have qualitatively similar effects across the region. Other patterns likely to hold across much of the region involve differences among creeks, reservoirs, wells, and springs and between spring and fall. F and Na are most likely to be problematic in wells, whereas Fe is most likely to be concerning in creeks and reservoirs. SO4 is similarly likely to attain high levels everywhere, except risks seem lower in reservoirs. Concentrations are unlikely to differ consistently between spring and fall.
Our research revealed water quality patterns in the northern Great Plains, and knowledge of these patterns provides a starting point for identifying water quality problems. However, water quality problems can arise unexpectedly. For example, SO4, Na, and TDS occasionally and unexpectedly reached concerning levels, and these occurrences cannot be predicted with our model, nor can they necessarily be detected with practical monitoring efforts. A solution to this problem is ensuring livestock can access sources of different types (e.g., wells and creeks; springs and reservoirs) so that they can use taste and smell to seek out water with acceptably low SO4, Na, and TDS. This strategy is also effective for Fe, which was frequently elevated in creeks and reservoirs and usually low in wells and springs. A similar strategy pertains to F, a cumulative toxin that reached greater concentrations in wells than other sources. Regularly rotating livestock through pastures with wells will help prevent chronic overexposure to F. When livestock are restricted to one or a few similar sources that are prone to intermittent water quality issues, monitoring becomes important. Inexpensive electrical conductivity meters roughly estimate TDS, and these TDS estimates accurately predicted SO4 (simple linear regression R2 = 0.78), our most concerning solid.
The authors thank Susan Reil, Brooke Shipp, and Susan Bellows for assistance with sampling.