This report was prepared in cooperation with the Colorado Water Conservation Board, Reclamation, Colorado River Water Conservation District, FWS, Bureau of Land Management, and Natural Resources Conservation Service (NRCS). The purpose of this report is to (1) summarize information regarding selenium in the lower Gunnison River Basin; and (2) describe and identify monitoring and research efforts that may support a more complete understanding of selenium sources, mobilization, transport, fate, and the effects of mitigation projects in the lower Gunnison River Basin. Multiple Federal, State, and local agencies are investigating salinity and selenium issues in the lower Gunnison River Basin, and the SMP maintains a working database of planned and active selenium-mitigation projects being led by Reclamation. This report could support and complement the goals of the SMP by identifying data gaps to better understand selenium in the lower Gunnison River Basin.

Selenium is an essential trace element and dietary micronutrient for humans and animals, but the intake of elevated levels of selenium can be toxic to humans and wildlife (Stadtman, 1974; Gissel-Nielsen and others, 1984; Lemly, 1996, 2002). In the semiarid Western United States, weathering of selenium-bearing rocks and derived sediment contributes dissolved selenium to rivers by natural and human-caused processes (Seiler and others, 1999, 2003; Presser, Hardy, and others, 2004; Presser, Piper, and others, 2004). In western Colorado, the Cretaceous Mancos Shale of marine origin is the primary selenium-bearing formation. Rainfall runoff, wind, infiltration, recharge, and groundwater and surface-water interaction processes naturally weather and erode the selenium-bearing rocks, releasing dissolved selenium to the aquatic environment. These same weathering and erosion processes also generate salinity (dissolved solids) because they release major ions (sulfate, nitrate, calcium, sodium, and chloride), dissolved organic carbon, and other trace elements to groundwater and surface water from rocks and soils in the surrounding basin. Weathering processes are accelerated by water use on the land surface in irrigated areas (Butler and others, 1996; Seiler and others, 2003), and irrigation of agricultural fields or in suburban areas in semiarid climates is considered a primary mechanism of mobilization and movement of selenium from geologic sources to streams (Butler and others, 1996; Seiler and others, 1999, 2003).

Selenium present in aquatic systems can cycle in and out of the water column because of complexation, adsorption, evaporation, and oxidation-reduction (redox) chemical reactions (McNeal and Balistrieri, 1989; Lemly, 2002). Dissolved and particulate selenium accumulate through the aquatic food chain, and the toxic effects can result in invertebrate mortality (deBruyn and Chapman, 2007; Conley and others, 2013; Henry and Wesner, 2018) and mortality, decreased reproduction, and deformities to fish and birds (Lemly, 1985; Ohlendorf and others, 1986; Ohlendorf, 1989; Presser, 1994). The U.S. Environmental Protection Agency (EPA) has established a primary drinking-water standard for selenium of 50 μg/L (EPA, 2003), and, in 1997, the State of Colorado adopted the criteria of 4.6 μg/L for a chronic aquatic-life standard for selenium (Colorado Department of Public Health and Environment [CDPHE], 2008). EPA guidance recommends the use of chronic aquatic-life criterion that include egg-ovary, whole body, muscle, and water-column concentrations (EPA, 2016).

In Colorado, numerous streams are included on the State’s list of impaired waters through provisions of Section 303(d) of the Clean Water Act (33 U.S.C. §1251 et seq.). These streams are on the so-called “Colorado Section 303(d) List,” because concentrations of dissolved selenium exceed the chronic aquatic-life standard of 4.6 μg/L (fig. 1). Many selenium-impaired streams in Colorado are associated with nonpoint source discharge from areas underlain by selenium-bearing Cretaceous shale geologic formations ­­­­such as the Mancos Shale in western Colorado (Butler and others, 1996; Tuttle and others, 2005) or the Pierre Shale and the Niobrara Formation in central and eastern Colorado (Gates and others, 2009; Ivahnenko and others, 2013; Bern and Stogner, 2017) (fig. 1). The Gunnison River Basin in western Colorado is a major tributary to the Colorado River. Many tributaries in the lower Gunnison River Basin are underlain by the Mancos Shale and several are on the Colorado Section 303(d) List for dissolved selenium concentrations that consistently exceed the aquatic-life standard (figs. 1, 2; table 1).

In 2009, the FWS issued a PBO that evaluated the effects of water-use projects on endangered species in the Gunnison River Basin. The PBO stated, “the ongoing operation of irrigation projects and other water uses in the basin will continue to contribute selenium to the Gunnison and Colorado Rivers at levels that adversely affect the endangered fishes and their designated critical habitat and are inhibiting the survival and recovery of the endangered fishes” (FWS, 2009, p. 15). To address these concerns, the PBO called for the creation of a program to implement selenium-reduction efforts. Since its formation, the SMP has supported many improvement projects, studies, and education throughout the lower Gunnison River Basin (Ward and others, 2021). Monitoring data indicated a significant downward trend in selenium concentrations since the 1980s (Reclamation, 2011; Henneberg, 2016, 2018, 2020; Henneberg and others, 2021), and in 2021, the Gunnison River was removed from the list of impaired waters. This milestone represents achievement of the first SMP goal to comply with the State of Colorado aquatic-life standard for dissolved selenium (Reclamation, 2011) and could allow for more focus on the second goal, recovery of endangered fish. This topic could be especially important because aquatic organisms are primarily exposed to selenium through the food web (Chapman and others, 2010), and water-quality standards based on dissolved selenium concentrations alone do not meet EPA recommendations for selenium criteria, which include both fish-tissue and water-quality components (EPA, 2016).

Located in western Colorado, the main stem of the Gunnison River originates in the West Elk Mountains near Crested Butte and flows southwest toward Gunnison before turning west and then northwest toward its confluence with the North Fork of the Gunnison River (fig. 2). The North Fork of the Gunnison River flows west from the headwaters in the northeast part of the lower Gunnison River Basin; the Uncompahgre River drains the southern and western parts of the lower Gunnison River Basin, originating in the San Juan Mountains near Ouray, Colo. The Uncompahgre River joins the Gunnison River near Delta, Colo., and the confluence of the Gunnison and Colorado Rivers is in Grand Junction, Colo. (fig. 2).

Physiography of the Gunnison River Basin is characterized by high mountains (elevation up to greater than 14,000 feet) on the eastern and southern edges of the basin, where winter snowpack accumulation supports annual spring snowmelt and runoff that supplies surface water to lower elevations (Sprenger and others, 2024). For the 1981–2010 period of record, average annual precipitation in the mountainous areas ranged from 21 to 71 inches per year with the greatest precipitation falling at the highest elevations (PRISM Climate Group, 2012; fig. 3). A semiarid climate and desert landscape prevail (fig. 4) in lower elevations and downstream parts of the basin, where average annual precipitation ranged from 7 to 20 inches per year for the 1981–2010 period of record (PRISM Climate Group, 2012; fig. 3).

The distribution of Paleozoic and Mesozoic sedimentary and volcanic rocks that overlie Precambrian intrusive and metamorphic rocks in the Gunnison River Basin (Lohman, 1965; Tweto, 1979) was established by the geologic history of sediment deposition, tectonics, and volcanism in the central Rocky Mountains during the past 1.7 billion years (Hettinger and others, 2003). The lower Gunnison River Basin is north of the Uncompahgre Plateau along the southeast edge of the structural Piceance basin, west of the Sawatch Uplift (within Delta, Gunnison, Mesa, Montrose, Ouray Counties; figs. 2, 5). The Mesozoic Morrison and Dakota Formations cap the Uncompahgre Plateau and dip gently to the north–northeast toward the structural Piceance basin (Hettinger and others, 2003). Coal and lignite deposits occur in the Dakota Formation on the west side of the lower Gunnison River Basin (Hettinger and others, 2003), and these coal deposits, in addition to Mancos Shale, could be potential sources of selenium to selenium-impaired streams on the western side of the lower Uncompahgre River (Butler and others, 1996; Adriano, 2001; Tuttle and others, 2005; Paschke and others, 2014). Laramide faulting and folding of the Sawatch Range, Elk and West Elk Mountains, and Uncompahgre Plateau contributed further to complex geologic structure in high-elevation parts of the Gunnison River Basin, and much of the Paleozoic section is missing in some areas because of an erosional unconformity on top of the Precambrian rocks (Lohman, 1965). The Mancos Shale (fig. 6) overlies the Dakota Sandstone. The Mancos Shale dips gently toward the north–northeast off the Uncompahgre Plateau and west–northwest off the Elk and West Elk Mountains into the structural Piceance basin. Tertiary volcanic rocks are intruded into the stratigraphic section in a series of calderas that formed the San Juan volcanic field (fig. 5).

The Mancos Shale occurs primarily in the lower-elevation and semiarid western half of the Gunnison River Basin (fig. 7) and is a major source of selenium (Mast and others, 2014; Tuttle and others, 2014a, b). The Mancos Shale is a marine shale 4,000–5,000 feet thick that was deposited in a Cretaceous inland sea (Lohman, 1965; Hettinger and others, 2003). The Mancos Shale consists of carbonaceous and organic fissile shale containing limited interbedded sandstones (Lohman, 1965). The Mancos Shale underlies the valleys at Grand Junction, Colo., and along the lower Gunnison and Uncompahgre Rivers. These large valleys in the lower Gunnison River Basin, “owe their origin to the ease with which the thick, soft Mancos Shale has been eroded” (Lohman, 1965, p. 66). The Mancos Shale is flat in the lower Gunnison River Basin, forming low adobe hills (fig. 4) dissected by rainfall runoff and arroyos.

Quaternary glacial deposits occur at high elevations in the mountainous areas of the basin, and Quaternary alluvial, colluvial, and aeolian deposits occur in the lower elevations along drainages (Tweto, 1979; fig. 7). In the lower Gunnison River Basin, there are Quaternary terraces above stream channels that can be coarse grained (Hettinger and others, 2003); however, Quaternary deposits and soils formed along streams in areas overlying bedrock of Mancos Shale are generally fine grained and derived from the weathering of Mancos Shale (Lohman, 1965; Tweto, 1979).

Land cover in the Gunnison River Basin is characterized by alpine and forested areas, pasturelands, cultivated areas, and mixed-use development (Dewitz and U.S. Geological Survey [USGS], 2021; fig. 8). Forested areas, perennial ice and snow, and barren land compose about 55 percent of the river basin and are primarily in the mountains and higher elevations where the snowpack and precipitation are greatest. At lower elevations with less precipitation, native shrublands, grasslands, and wetlands are present and compose about 38 percent of the Gunnison River Basin area. Pasturelands are primarily along the river valleys and represent about 3 percent of the basin area. Cultivated crops (irrigated agriculture) represent only about 1 percent of the land cover in the basin, but this irrigated agriculture is concentrated in the lower parts of the Gunnison and Uncompahgre River Basins in an area overlying Mancos Shale and Quaternary alluvial deposits (fig. 7). Developed areas represent about 1 percent of the basin area, and Gunnison, Montrose, and Delta are the largest municipalities.

Surface water is the primary source of water supply for municipal and agricultural uses in the lower Gunnison River Basin. Natural surface drainages and canals move water from headwater areas to downstream reservoirs for domestic and municipal uses and to agricultural fields to irrigate pastureland and crops. The Gunnison Tunnel (fig. 2) transfers water from the Gunnison River Basin to the Uncompahgre River Basin to support municipal water demands and irrigated agricultural fields in the lower Uncompahgre River Basin. Irrigation canals distribute surface water to irrigate lands in the valleys and lower basin, and irrigation water is a source of recharge to the shallow groundwater system (Mills and others, 2016, Thomas and others, 2019).

The Mancos Shale is the primary geologic source of selenium to groundwater and surface water in the lower Gunnison River Basin. Seawater, volcanic ash, and pyrite are the primary original sources of selenium to the deposits of Mancos Shale (Byers and others, 1938; Davidson and Powers, 1959; See and others, 1991; Kulp and Pratt, 2004; Tuttle and others, 2014a, b), and in the unweathered shale, selenium is primarily associated with organic shale and pyrite (Tuttle and others, 2014a, b). The weathered shale and derived materials contain secondary weathering products of iron oxides and soluble selenium-bearing salts (Tuttle and others, 2014a, b). The soluble salts that have accumulated in the weathered zone through thousands of years (Tuttle and others, 2014a, b) are the most soluble forms of selenium in naturally weathered Mancos Shale and its derived materials (Mast and others, 2014).Taken together, the weathered and unweathered Mancos Shale could be an ongoing substantial nonpoint source of selenium to the Colorado River, and the present-day distribution of selenium within the Mancos Shale and its derived materials could provide useful information for selenium management in the lower Gunnison River Basin.

Selenium is mobilized from Mancos Shale and derived materials by natural weathering processes and human activity. In undeveloped natural areas, particulates of Mancos Shale can be mobilized to drainages and streams by wind and by local storms and precipitation. Land disturbance in developed areas caused by road construction, agriculture practices, oil and gas development, and residential and urban development can liberate particulates of Mancos Shale and increase wind and water mobilization of selenium. In areas underlain by Mancos Shale and its derived materials, application and deep percolation of water dissolves soluble selenium-bearing salts and related minerals from the unsaturated weathered zone, which can release selenium, sulfur, nitrogen, major ions, and other trace elements to the water table (Mast and others, 2014; Thomas and others, 2019; Mast, 2021). Human activities that apply water to the landscape include domestic water use (lawn irrigation, water infrastructure, and septic tank leachate), irrigation water application (flood and sprinkler irrigation), and irrigation water delivery and return flows (leaking canals, laterals, ditches, and ponds) (Thomas and others, 2019). Dissolved selenium also can be released directly to surface water by runoff, sediment transport, or present-day streams that are in contact with Mancos Shale and its derived materials (Bern and others, 2023).

Selenium-bearing salts in the lower Gunnison River Basin formed from the oxidation of pyrite in shale during thousands of years (Tuttle and others, 2014a, b), and there is little evidence of primary pyrite remaining from the original Mancos Shale in near-surface weathering environments (Mast and others, 2014). Laboratory experiments and geochemical modeling corroborate findings that selenium is primarily released from soluble salts and gypsum dissolution (Mast and others, 2014). In previously irrigated soils, soluble forms of selenium can be depleted (flushed) and most selenium is associated with organic matter that was stable in near-surface weathering conditions. Selenium is also released from previously irrigated materials, but at a much slower rate (Mast and others, 2014). High concentrations of extractable selenium and nitrate were measured in previously unirrigated soils and bedrock. Nitrate salts in previously unirrigated materials are derived from weathered organic matter in the original shale rather than from agricultural sources (Mast and others, 2014), indicating that there are natural sources of nitrate in the lower Gunnison River Basin. Nitrate in groundwater acts as an oxidizing agent and inhibits the reduction of selenium. Oxidized forms of selenium are more soluble than reduced forms, so the presence of nitrate contributes to elevated concentrations of selenium in water (Mast and others, 2014; Mills and others, 2016; Thomas and others, 2019). Sources of nitrate from human activity, such as synthetic fertilizers, could also contribute to elevated selenium concentrations in the lower Gunnison River Basin (Wright, 1995, 1999).

Selenium can be transported from source areas to receiving streams and biota by surface water, groundwater, and sediment. The major rivers in the lower Gunnison River Basin originate in the high mountains, where winter snowpack accumulation supports annual spring snowmelt and runoff that supplies surface water to lower elevations. The annual hydrograph is characterized by fall and winter base-flow conditions when streamflow is supported by groundwater discharge followed by the annual spring snowmelt and runoff in May or June (Schaffrath, 2012).

Some of the surface water from headwater parts of the basin is diverted by tunnels and canals to lower parts of the basin where it is distributed to irrigated areas by canals and ditches. In irrigated areas, the annual surface-water hydrograph is similar to the natural hydrograph, having a characteristic fall and winter nonirrigation (no irrigation) season base-flow condition but having spring runoff sustained for a longer period by irrigation return flows (surface water and deep percolation) through the summer irrigation season (Mast and others, 2014). Selenium concentrations in surface water are inversely related to flow, and the greatest selenium concentrations typically occur during the nonirrigation season when flows are lower and primarily due to groundwater discharge to the streams (Mast and others, 2014). About one-half of the annual selenium load to Loutsenhizer Arroyo and Sunflower Drain is estimated to come from groundwater base flow (Mast and others, 2014; Thomas and others, 2019).

The shallow groundwater system is supported by recharge (deep percolation) from streams, irrigation canals, irrigated fields, seasonal spring snowmelt, and infrequent summer thundershowers (Butler and others, 1996). On a regional scale, the Mancos Shale is not water bearing and is considered a confining unit rather than an aquifer because of its great thickness and low permeability (Lohman, 1965). Before the development of irrigation, shallow groundwater in the lower Gunnison River Basin was likely insubstantial (Reclamation, 1982; Butler and others, 1996). On a local scale, a small number of wells obtain water from unconfined groundwater that comes from the weathered zone of the Mancos Shale and (or) the alluvium that fills former arroyos in the Mancos Shale (Lohman, 1965; Butler and others, 1996). Data from a 30-well monitoring network installed on the east side of the Uncompahgre River have led to an increased understanding of the shallow groundwater system in the lower Gunnison River Basin (Thomas, 2015; Thomas and Arnold, 2015). Wells completed in alluvial sediment were unconfined, allowing groundwater to flow through unconsolidated material. Wells completed in Mancos Shale, however, were confined and recharge to the confined aquifer is likely along fractures and bedding planes (Thomas and others, 2019). Lithologic and groundwater-level data obtained during well installation were used to map the hydrogeology of the shallow groundwater system, including extent and thickness, consolidated bedrock surface, and potentiometric surface (Arnold, 2017). Study of the well network showed that the isotopic composition of the shallow groundwater system was similar to that of local rivers, indicating that groundwater is recharged by irrigation water. Additionally, groundwater ages ranged from 6 to 20 years (Thomas and others, 2019).

As groundwater moves from recharge to discharge areas, it transports dissolved constituents downgradient to receiving streams, drains, canals, and wetlands. Previous studies of groundwater quality in the lower Gunnison River Basin indicated that high calcium, sodium, sulfate, selenium, and dissolved solid concentrations in groundwater were ubiquitous in areas underlain by Mancos Shale (Butler and others, 1996). Previous study results also showed an inverse relation between groundwater levels and dissolved solid concentrations–the highest groundwater levels and lowest dissolved-solid concentrations occurred during the irrigation season (Butler and others, 1996). Irrigation recharge dominates the hydrologic system during the summer irrigation season, raising groundwater levels (Mills and others, 2016; Thomas and others, 2019; Mast, 2021), increasing streamflow, and diluting selenium concentrations in groundwater (Butler and others, 1996) and surface water (Mast and others, 2014; Mast, 2021).

Selenium-bearing sediment can be transported from natural areas by wind and water erosion. Erosion of steep slopes of Mancos Shale greater than 20 degrees, such as in arroyos and adobe hills, plays an important role in the transport of selenium and gypsum (Elliott and others, 2008; Tuttle and others, 2014a, b). During erosion, salt and selenium are either removed in solid phases and transported to the valley floor as soil creep or entrained in runoff and wind. Schumm and Gregory (1986) provide a discussion of alluvial deposition of Mancos Shale material.

The fate of selenium, once mobilized and transported, is of interest in the lower Gunnison River Basin because selenium bioaccumulation may prevent the recovery of endangered fish species. Diet is the primary pathway of selenium exposure for aquatic organisms (Luoma and Fisher, 1997; Luoma and Rainbow, 2005; Luoma and Presser, 2009; Chapman and others, 2010; Presser and Luoma, 2010). In fish and birds, selenium accumulates in tissues and is maternally transferred to eggs (Luoma and Presser, 2009; Chapman and others, 2010). Reproductive impairment caused by maternal transfer results in embryotoxicity and teratogenicity, which are the primary results of selenium toxicity in fish (Chapman and others, 2010).

Selenium enters the food web and diet of consumers from water and particulates, and the factors linking water, particulates in water, and fish-tissue concentrations are an area of active research (Presser and Luoma, 2010; Presser, 2013; Jenni and others, 2017; Luoma and Presser, 2018; Presser and Naftz, 2020; Brandt and others, 2021). In the stream environment, selenium in water and sediment is chemically reduced in the presence of organic matter to bioavailable forms of selenium such as organic particulates, biofilms, and phytoplankton (Presser and Luoma, 2010). These reactions commonly occur at the sediment-water interface in biofilms such as periphyton on the streambed where dissolved selenium is bioconcentrated at the base of the food web through accumulation into phytoplankton, periphyton, or other organic matter (Presser and Luoma, 2010; Presser, 2013; Brandt and others, 2021). The bioavailable selenium is then bioaccumulated through macroinvertebrates to predator fish and birds (Presser and Luoma, 2010). The dissolved concentration of selenium in the water column is not necessarily a good indicator or predictor of selenium accumulation in or toxicity to fish, which highlights the need to evaluate fish recovery with selenium-toxicity studies of the food web in targeted locations (Presser and Luoma, 2010; Presser, 2013). In 2016, EPA updated the recommended national chronic aquatic life criterion for selenium, moving from a water-column basis to standards that include fish egg-ovary, whole body, muscle, and water-column concentrations (EPA, 2016).

Characterization of geologic sources includes geologic mapping, soil studies, and geochemical characterization of geologic materials and soils. Geologic mapping and additional targeted studies of geologic sources are identified as data gaps. In addition, developing geospatial representations of the geologic sources of selenium could inform surface-water loading models for the lower Gunnison River Basin and spatial analysis for evaluating selenium-mitigation projects.

In 2012, the USGS began to design and implement a 30-well regional groundwater-monitoring network on the east side of the Uncompahgre River in irrigated areas underlain by Mancos Shale (Thomas, 2015; Thomas and Arnold, 2015). The entire 30-well network (fig. 12) was sampled during the irrigation season in 2014 and during the nonirrigation season in 2015 to evaluate regional groundwater conditions (Thomas and others, 2019), and cores collected during the monitoring well installation were used for geochemical source characterization (Mills and others, 2016) and hydrogeologic mapping (Arnold, 2017). As of 2021, one well in the network had been destroyed and another was no longer accessible. In addition to the 30-well network, a total of 4 wells (2 paired wells) were installed in 2012, adjacent to a leaky canal near Olathe, Colo. (Linard and others, 2017). Multiple studies have contributed to the understanding of groundwater and its role in selenium mobilization and transport in the lower Gunnison River Basin (for example, Butler and others, 1996; Mast and others, 2014; Mills and others, 2016; Arnold, 2017; Thomas and others, 2019; Mast, 2021; refer to the “Conceptual Model of Selenium Occurrence in the Lower Gunnison River Basin” section of this report). However, the system is complex and could be better defined through additional study.

The USGS investigated interactions between surface water and groundwater in Sunflower Drain, an agricultural drainage near Delta (Mast, 2021). Results indicated that groundwater inflow in the study reach is likely focused rather than diffuse and could be related to the distribution of irrigated land and features such as tile drains. The project also involved three pilot studies for techniques to investigate groundwater discharge to surface water: (1) fiber-optic distributed temperature sensing to detect groundwater discharge zones, (2) radon-222 as a geochemical tracer of groundwater discharge, and (3) passive seismic technique to estimate depth to bedrock; these methods all showed promise for investigating groundwater discharge to surface-water systems (Mast, 2021).

The USGS and other entities routinely sample surface water in the lower Gunnison River Basin. The sampling includes a variety of measurements and analytes, including streamflow, dissolved selenium, and other inorganic chemical analyses. The SMP workgroup coordinates sampling with partners such as USGS, CDPHE’s Colorado Water Quality Control Division, and State and local entities to ensure uniformity and data comparability. The monitoring may differ by sampling agency and station depending on individual site objectives, some of which may address objectives outside the scope of the SMP. Refer to the SMP formulation document (Reclamation, 2011) for additional information on coordination with partners.

The SMP surface-water monitoring network consists of USGS sites located primarily on the main-stem of the Gunnison River, North Fork of the Gunnison, and Uncompahgre Rivers with additional sites on several larger tributaries (fig. 2). Sites in the surface-water network have been strategically placed to account for concentrations and loads of dissolved selenium and other constituents from six major contributing areas (CAs) to the lower Gunnison River Basin. Data collected through the monitoring program supplement a database that provides the basis for statistical analysis at various scales (for example, refer to Henneberg, 2016, 2018, 2020; Henneberg and others, 2021). Data are stored in the USGS National Water Information System database at https://doi.org/10.5066/F7P55KJN. The monitoring network includes both core and ancillary sites (table 1; figs. 2, 6). Core sites have USGS streamgages; ancillary sites do not. The list of ancillary sites may change annually as data gaps are filled, operations change, and cooperator research is completed.

Additional sampling of rivers and streams in the lower Gunnison River Basin is done by the CDPHE. The CDPHE sampling does not usually include streamflow measurements, which limits the utility of the data for load calculations; however, some of the data are collected at USGS streamgages and have associated streamflow measurements that can be used for load calculations. The data collected by CDPHE primarily aid in the determination of instream concentrations for State regulatory purposes, and the data are useful for determining dissolved selenium and other constituent loads from the six major CAs (refer to the following section, “Contributing Areas”).

Fluvial sediment is often thought of as a physical condition of the stream system, but it also can affect chemical or toxicological characteristics in the aquatic ecosystem. Owing to a combination of physical and chemical processes, along with aquatic physiochemical conditions, fluvial suspended sediment and streambed sediment are both sources of and vectors for inorganic constituents and nutrients (Horowitz, 2008a, b). Solid-state transport of trace elements, such as selenium, is common and can occur at concentrations of orders of magnitude greater than those within the dissolved phase of the water column (Horowitz, 2008a).

Sediment derived from the physical weathering of surficial rock and soil may retain selenium from within the crystalline lattice structure of minerals. Additional retention can occur along or near the surface of particles where selenium complexes are held by sorption or complexation reactions with unbalanced surface charges. These unbalanced surface charges can be caused by clay mineralogy or surface coating of either organic matter or iron and manganese oxides and oxyhydroxides (Förstner and Wittmann, 1981; Förstner, 1989; Horowitz, 1991; Foster and Charlesworth, 1996). Changes in reduction-oxidation potential (Eh) and pH can cause changes in the properties of organic matter and iron and manganese oxides that can result in either additional sorption or desorption of selenium to or from these surface coatings (Horowitz, 1991; Foster and Charlesworth, 1996).

Knowledge of selenium on fluvial sediment is limited in western Colorado and is often based on assessments of Federal Irrigation Project Areas (hereafter, Federal Project Areas [FPAs]) such as the Uncompahgre, Grand Valley, and Dolores FPAs (Butler and others, 1995, 1996). Concentrations of selenium on sediment in the Grand Valley and Uncompahgre FPAs are elevated in comparison to upstream selenium concentrations on bed sediment (Vaill, 2000; table 5). Analysis of bed sediment in the Grand Valley and Uncompahgre FPAs showed that, of the total selenium, less than 1 percent is soluble and between 2.5 and 16.9 percent is surface sorbed. Speciation of the selenium (soluble and sorbed) on the bed sediment showed that, at most sites, most of the selenium was not selenite (Butler and others, 1996). This result is unlike findings from Fujii and others (1988) whose samples from the San Joaquin Valley in California showed that most of the soluble and sorbed selenium was selenite. This contrast may indicate that geochemical processes or soil chemistry in the San Joaquin Valley differ from those in the lower Gunnison River Basin.

Because of the potential reactiveness of selenium sorbed to surface coatings, and the hypothesis that these forms of selenium are more readily available, the term “available selenium” can be used to describe them. The selenium that is retained from the parent material within the lattice structure is less reactive and is referred to as “total extractable selenium” and is quantified by chemical digestions of the minerals and laboratory analysis (for example, Fujii and others, 1988). The differing reactivity of these two states of selenium affects the relative solubility, bioavailability, and toxicity of selenium to organisms. However, differing laboratory analysis methods used to target the soluble, sorbed, and total forms of selenium can produce inconsistent results, and a direct toxicity assessment may be a better means to identify the environmental importance of these concentrations (Tessier and others, 1979; Chao, 1984; Horowitz, 1991, 2008b; Hall and others, 1996).

Comparisons of suspended-sediment flux between the USGS streamgage at Colorado River near Cisco, Utah (09180500; USGS, 2021b), and key upstream locations can be used to evaluate potential selenium from source areas. Findings from Horowitz and others (2001) demonstrate that, in water years 1996 and 1998, as much as 72 percent of the annual load of selenium at the site Colorado River near Cisco, Utah, was in the filtered phase and 28 percent of the total load was measured in the solid phase (sediment). In a wet year (water year 1997), the percentage in solid phase increased to 46 percent, and there was only a small increase in the dissolved selenium load (18 tons in 1996 and 1998, 19 tons in 1997). Specifically, the sediment load ranged from 4.0 to 4.2 million tons (1996 and 1998, respectively) to 9.8 million tons in water year 1997. During 2015–17, USGS measured selenium on suspended sediment at selected sites in the lower Gunnison River Basin in critical habitat for endangered fish species (table 6). The goal was to investigate the role of sediment as a selenium load vector in surface water, including tributaries, where the particulate transport of selenium was largely uncharacterized in the lower Gunnison River Basin. Results showed that selenium concentrations on suspended sediment did not differ greatly, but slightly increased upstream from the site Gunnison River near Grand Junction, and concentrations were slightly higher in tributaries (Sunflower Drain and Loutsenhizer Arroyo) than in the main-stem Gunnison and Uncompahgre Rivers (table 6) (data from the USGS National Water Information System database; USGS, 2021b).

Increases in the surface area of a particle provide increased capacity for sorbed trace elements such as selenium (Horowitz, 2008b). For most sediment particles, surface area per mass increases as particle size decreases. This is especially true for platy particles such as clays where the surface areas are 1,000 times greater than those for the same mass of sand (Black, 1957). Particle size can affect sorbed chemistry concentrations, wherein concentrations increase as particle size decreases (Förstner and Wittmann, 1981; Horowitz, 1991). The enrichment of selenium on fine sediment has not been shown to be substantial in the study area. However, targeting only one specific particle size of suspended sediment, typically based on the separation of silt and clay (less than 0.0625 microns) and sands (0.0625 microns to 2 millimeters), may not be appropriate (Horowitz, 2008b). Understanding the specific particle size is important because there are typically substantial differences in the concentrations for each particle size and the locations where the sediment is deposited.

The importance of characterizing the role of sediment in the lower Gunnison River Basin is further stressed by deficiencies in the conventional assessments of total selenium using whole water (unfiltered) samples. The issue with this type of assessment is related to the quantity of sediment within the water sample and the laboratory reporting limit of the analysis. When there is a small concentration of suspended sediment and the concentration of the sorbed constituent is low, a larger volume of the sediment within the whole water sample is needed to accurately quantify a concentration. A whole water (unfiltered) sample analysis does not represent the total selenium in the water column because, in most conditions, it does not include selenium sorbed to suspended sediment.

Collection of water samples and separate suspended sediment samples allows computation of total selenium concentrations, which can differ from dissolved concentrations because of the presence of particulate selenium (for example, Presser and Luoma, 2006). In the samples collected in the lower Gunnison River Basin as part of the 2015–17 suspended sediment analysis (table 6), selenium associated with the suspended sediment represented between 7 and 17 percent of the total selenium concentrations, on average (table 7). Tributary locations (Sunflower Drain and Loutsenhizer Arroyo) had a greater concentration of selenium associated with suspended sediment, and higher total selenium concentrations than main-stem locations.

The causes for the decrease in the populations and ranges of the endangered fishes Colorado pikeminnow, razorback sucker, Gila cypha (humpback chub), and Gila elegans (bonytail) in the lower Gunnison River Basin have been a long-standing debate. Although some believe the introduction of nonnative fish and changes to the flow and water temperature regime to be the dominant causes of the declines (Valdez and Muth, 2005; FWS, 2011a), selenium accumulation has also been identified as a threat to endangered fishes in the basin (Hamilton, 1999; FWS, 2011b). Hamilton (1999) suggested that there is evidence that declining populations of these native species predate main-stem reservoirs in the upper and lower Colorado River and may be caused by elevated selenium concentrations. For example, Williams (1935) reported selenium concentrations as high as 220 µg/L at the mouth of the Uncompahgre River and 80 µg/L in the Gunnison River near the confluence with the Colorado River. Hamilton (1999) also maintained the potential importance of addressing selenium in recovery efforts, which is supported by additional studies of toxicity to larval razorback suckers in Utah (Hamilton, 1995; Hamilton and others, 1996).

The determination of appropriate dissolved selenium standards has been another area of discussion, and arguments have been made for a chronic exposure standard below 4.6 µg/L to be protective of aquatic life (Presser and Piper, 1998; Hamilton and Lemly, 1999; Lemly, 1999) or standards based on concentrations in fish tissue rather than concentrations in water (Hamilton, 2002, 2003; EPA, 2016). It is well established that selenium bioaccumulates through the food web (Luoma and Fisher, 1997; Luoma and Rainbow, 2005), and about 95 percent of the selenium in tissues is attributable to dietary exposure; only 5 percent is attributable to dissolved exposure (Fowler and Benayoun, 1976; Luoma and others, 1992; Roditi and Fisher, 1999; Wang and Fisher, 1999; Wang, 2002; Schlekat and others, 2004; Lee and others, 2006). Introduction of selenium into the food web occurs through the organic uptake of selenium by bacteria, algae, fungi, and plants, and the substitution of selenium instead of sulfur into proteins (Presser and Luoma, 2010).

The EPA-recommended chronic aquatic-life criterion for selenium includes two media (fish tissue and water) separated into four elements: fish egg-ovary (15.1 mg/kg dry weight), fish whole body (8.5 mg/kg dry weight), muscle (11.3 mg/kg dry weight), and water column (1.5 µg/L lentic and 3.1 µg/L lotic). In most circumstances, the fish-tissue elements take precedence (EPA, 2016, 2021a). The State of Colorado chronic aquatic-life standard is based on selenium concentration in water (CDPHE, 2008). Techniques such as muscle-plug sampling (Waddell and May, 1995; Osmundson and others, 2000; May and Walther, 2012, 2013) or determining tissue concentrations in surrogate species such as Catostomus commersonii (white sucker) and Lepomis cyanellus (green sunfish) (Osmundson and Skorupa, 2011) are used to assess selenium concentrations in threatened and endangered species. Other surrogate species that have been tested include Rhinichthys osculus (speckled dace), Gila robusta (roundtail chub), and Cyprinus carpio (common carp). Additional complications of measuring selenium in fish tissue may arise from seasonal changes in selenium concentration within female fish (pre- and post-spawning) when selenium is passed from the adult into the expressed eggs through selenium partitioning (Osmundson and Skorupa, 2011).

A comprehensive assessment of selenium and mercury concentrations in fish tissue across the entire upper Colorado River Basin found that selenium concentrations in endangered native species exceeded thresholds thought to be protective of health in 63 percent of Colorado pikeminnow and 35 percent of razorback sucker individuals (Day and others, 2020). The study compiled data from multiple sources spanning 1962 to 2011. Comparison of subbasins indicated that selenium concentrations in fish tissue were highest in the Gunnison, lower Green, and Colorado headwaters basins, and concentrations throughout the upper Colorado River Basin were elevated compared to other rivers in the Western United States. The two species with the highest mercury concentrations were Colorado pikeminnow and roundtail chub, of which 70 percent and 32 percent, respectively, of individuals had concentrations exceeding the mercury benchmark (Day and others, 2020). Studies have shown that, when selenium and mercury occur together, they may interact antagonistically and reduce each other’s toxicity and bioaccumulation, depending on the molar ratio (Ganther and others, 1972; Peterson and others, 2009; Dang and Wang, 2011; Sørmo and others, 2011). A molar ratio of selenium to total mercury at or above 1 could mean that the toxic effects of mercury are reduced (Ganther and others, 1972; Dang and Wang, 2011). In all tributaries assessed for the upper Colorado River Basin study, molar ratios of selenium to mercury were above 1 (Day and others, 2020).

A risk-management model methodology is presented in Presser and Luoma (2010) and could be used to determine important pathways, exposures, biogeochemistry, and hydrology for specific ecosystems. The model could be tailored to specific species after characterization of environmental partitioning, transformation, bioaccumulation, and trophic transfer for the selected ecosystem (Presser and Luoma, 2009; Presser, 2013; Jenni and others, 2017; Luoma and Presser, 2018; Presser and Naftz, 2020; Brandt and others, 2021). The variability of selenium bioaccumulation is ecosystem specific; local phenomena such as residence time, selenium speciation, carbon and nutrient availability, and food chain length all contribute to the potential for selenium to bioaccumulate in the food web at any given location (Presser and Luoma, 2010). This ecosystem methodology could be applied to developing site-specific selenium criteria that are protective of sensitive species (Presser and Luoma, 2009; Jenni and others, 2017; Luoma and Presser, 2018; Presser and Naftz, 2020; EPA, 2021a).

Brandt and others (2021) investigated the temporal effects of selenium partitioning and trophic transfer of selenium in the lower Gunnison River Basin with an approach similar to that outlined in Presser and Luoma (2010). Selenium concentrations in particulate fractions consumed by aquatic life were decoupled from total dissolved selenium concentrations, meaning that accumulation of selenium into the particulate phases consumed by aquatic life reflected higher dissolved selenium concentrations observed during previous periods of selenium mobilization (that is, concentrations in aquatic life lagged behind dissolved selenium concentrations in the water column). The ecosystem-scale selenium accumulation model parameterized to the food webs of the lower Gunnison River Basin predicted selenium concentrations in razorback suckers and Colorado pikeminnow in excess of the EPA National Selenium Chronic Standard for Aquatic Life (8.5 micrograms per gram dry weight whole body fish tissue; EPA, 2016) during peak flow and irrigation season, which are when razorback suckers and Colorado pikeminnow are reproductively active. These results indicate that contemporary total dissolved selenium concentrations, although decreasing for several decades (Reclamation, 2011; Henneberg, 2016, 2018, 2020; Henneberg and others, 2021), may not protect razorback suckers and Colorado pikeminnow (Brandt and others, 2021).

Best management practices are aimed at reducing the potential for selenium mobilization and transport by groundwater and surface-water systems. Through the SMP, Reclamation, in collaboration with other Federal, State, and local agencies, pursues BMPs to reduce groundwater recharge (deep percolation) and surface-water runoff, and thus selenium loading to streams (Ward and others, 2021).

Canal lining and piping is the primary strategy being implemented throughout the lower Gunnison River Basin for selenium and salinity mitigation (URS, 2013; Ward and others, 2021). Leakage from these water conveyance features is known to recharge the groundwater system, create downgradient wetland areas, and contribute substantial salinity and selenium loading to groundwater and surface-water systems (Butler and others, 1996; Butler and Leib, 2002; Mills and others, 2016; Linard and others, 2017; Thomas and others, 2019; Mast, 2021). On-farm improvements also are being pursued in the lower Gunnison River Basin and include a host of mitigation efforts such as improvements in irrigation systems (hardware), irrigation management and oversight by the farmer, soil health, and irrigation efficiency and uniformity (URS, 2013; Ward and others, 2021).

Ponds, water features, and wetlands are associated with golf courses, residential and commercial properties, and agriculture, and leakage from these water features can mobilize and transport selenium (Thomas and others, 2019). “Water Wise Pond Construction for Residential and Commercial Properties” (Isom and others, 2007) is a guide for BMPs for water conservation and pond management that was developed as a cooperative effort of the Gunnison Basin and Grand Valley Selenium Task Forces. Pond seepage (water that leaks from unlined ponds) contributes to groundwater that transports salts and selenium to streams (Mayo, 2008). A variety of BMPs for water features are site specific with considerations such as site selection, soils, types of ponds, lining options, maintenance, presence of a water table, and cost effectiveness. Best management practices that limit or eliminate seepage from ponds and water features could reduce the potential for selenium mobilization and transport.

Best management practices in areas of urban development, land-surface disturbance, and natural environments could also be beneficial in reducing selenium loading, although effects can be difficult to quantify, and additional monitoring could be beneficial to potentially determine efficiency and magnitude of selenium reduction. Urban development on previously irrigated or nonirrigated lands has the potential to increase or decrease selenium loading. Conversion of land from agricultural to residential subdivisions could result in substantially less irrigation-water application and thus reduced deep-percolation rates (Leib, 2008; Mayo, 2008; Richards and Moore, 2015). Golf course irrigation is another source of deep percolation to groundwater, and BMPs related to golf course irrigation and ponds could be important to consider in the lower Gunnison River Basin. Leaks from septic systems not connected to municipal sewers also could become an increasingly important concern (Mast, 2021). Land disturbance from human activities has the potential to expose and mobilize salinity- and selenium-bearing material, which then can be more easily transported during episodic phenomena such as windstorms, snowmelt, and intense thunderstorms (Lusby and Knipe, 1971; Bureau of Land Management, 1978; Elliott and others, 2008; Belnap and others, 2014; Bern and others, 2015; Duniway and others, 2019).

In addition to BMPs that control or limit the mobilization and transport of selenium, recent studies have investigated the potential for treatment practices, such as bioreactors, to remove selenium from water (Walker and Golder Associates Inc., 2010), and reactive media (Sottilare and Thomas, 2018). Sottilare and Thomas (2018) investigated the potential for in situ removal of selenium in groundwater with two media—apatite and zero-valent iron—and three methods of installation—permeable reactive barriers, deep aquifer remediation tools, and injectable permeable reactive barriers. Six candidate sites within the groundwater monitoring network on the east side of the Uncompahgre River (Thomas, 2015; Thomas and Arnold, 2015) were selected for geochemical modeling to evaluate selenium removal with zero-valent iron. Results indicated that zero-valent iron could be effective at removing dissolved selenium from groundwater in the study area (Sottilare and Thomas, 2018).