The Butte mine and the ore smelter at Anaconda, Mont. (fig. 1), once constituted the largest copper-producing complex in the United States, operating from about 1870 to 1982 (table 1; Gammons and others, 2006). Increasing energy costs and decreasing copper prices since the early 1980s reduced the mining industry in Montana, however, leaving the present complex marked by an environmental and urban renaissance in the cities of Butte and Anaconda (Clark Fork Watershed Education Program, 2022).

Mining at the headwaters of the Clark Fork began in 1864 when gold was discovered in streambed deposits, and small-scale placer mining (separating heavily eroded minerals from sand or gravel streambed deposits) began in and near Butte, Mont. (Hoffman, 2001; Clark Fork Watershed Education Program, 2022). Prospectors at this time also found silver in abundance while searching for the lode source of the placer gold (Hoffman, 2001), and the first silver strike, discovered in 1865, marked the beginning of Butte’s successful silver mining episode, which peaked around 1887 (Chadwick, 1982). Many of the silver lodes mined at the time included copper veins (mostly in the form of sulfide minerals), but copper was in low demand because it was not as profitable as gold and silver. By 1893, the Northern Pacific Railroad arrived in Butte, which helped to lower the cost of supplies and provided a cheaper avenue to ship high-grade ore in time for the dawn of the electric age (around 1882) and the need for copper wire (Jenkins and Lorengo, 2002).

While working for a silver mining company in Butte, Marcus Daly discovered the largest copper sulfide deposit in America in 1880 (table 1) and began mining the extensive copper ore (Morris, 1997), which consisted of arsenic, copper, lead, and zinc (Lambing, 1991). By 1884, four large smelters were operating around Butte, and the world’s largest metallurgical plant was under construction in Anaconda, 30 miles west; by 1886, about 10 major ore-processing mill and smelter operations were along Silver Bow Creek (fig. 1; EPA, 2005). In 1902, 17 to 20 percent of all copper production in the United States came from Butte (Jenkins and Lorengo, 2002). To supply hydroelectricity to his sawmills in nearby Bonner, Mont., William A. Clark, one of the three “Copper Kings,” constructed the earth-fill hydroelectric Milltown Dam in 1907 about 7 miles east of Missoula (fig. 1) at the confluence of the Blackfoot River with Clark Fork (Clark Fork Watershed Education Program, 2022). However, when the dam was just months old, a record flood on the Clark Fork washed extensive (about 5.0 million cubic meters, m³) mining sediment contaminated with arsenic, lead, zinc, copper, and other elements downstream, where it settled at the base of the dam; the reservoir accumulated tailings-laden sediment for more than 80 years (Lambing, 1991). The dam was removed in 2008, and the Clark Fork channel was reconstructed in 2010 (EPA, 2021).

During the 1890s, mines around the world, including those in Montana, were consolidated under a single owner, the Anaconda Copper Mining Company (AMC) (Toole, 1950). In the 1920s, the primary milling and smelting facilities near the Clark Fork headwaters were moved to Anaconda (table 1), where most of the copper ore from the Butte area was processed (EPA, 2005, 2010). The small town of Butte became one of the most prosperous cities in the country, called the “Richest Hill on Earth,” and the Anaconda mine was the largest copper-producing mine in the world, producing more than $300 billion worth of metal in its lifetime. By 1917, more than 150 mines were in and near Butte (EPA, 2005). From 1910 through the 1920s, AMC acquired most of the mines in the Butte area. However, after severe financial setbacks during the Great Depression and after World War II and falling copper prices that continued to drop through the 1950s, the Anaconda mine closed in 1947 after producing 94,900 tons of copper (Malone, 2006). Its location later became part of the Berkeley Pit north of Butte in 1955, when AMC mining operations began to transition from underground to open-pit mining of a lower grade, porphyry-copper-style deposit.

Underground mining continued until 1976; however, by the late 1960s, the primary focus was on open-pit mining taking place in the Berkeley Pit (fig. 1, table 1) (Gammons and others, 2006). The Atlantic Richfield Company (ARCO) acquired AMC in 1977 (EPA, 2004); AMC remained a subsidiary and continued to function as AMC (Gammons and others, 2006), and the East Berkeley Pit operated until July 1, 1983, when AMC suspended their entire operations (Gammons and others, 2006). The pumps used to dewater the underground mines and Berkeley Pit operated until April 23, 1982, and ARCO announced they were suspending their Butte operations (except for the East Berkeley Pit) on July 1, 1982 (Duaime and McGrath, 2019). When the underground pumps in the Kelley Mine shut down, the underground mines and the Berkeley Pit began to fill with acidic water (Duaime and McGrath, 2019).

After 100 years of mining activity, Silver Bow Creek was a toxic stream with little to no aquatic life. Extensive deposition of mining wastes in the channels and floodplains had adverse effects on water quality and severely affected local aquatic ecosystems by affecting primary and secondary production, nutrient cycling, energy flow, and decomposition (Younger and others, 2004; Knott and others, 2009; Batty and others, 2010). Substantial mine waste deposits collected on the Clark Fork floodplain, all vegetation was gone, and in some places (such as Deer Lodge, Mont.), streamside tailings deposits (locally known as “slickens”) created the appearance of a barren, “slick,” wasteland (Rader and others, 1997).

Early efforts to retain the mine tailings in ponds were often small in scope or unsuccessful. Andrews (1987) estimated that 100 million tons of tailings containing large quantities of heavy metals and acid-producing pyrite were dumped or eroded directly into Silver Bow Creek and in the upper Clark Fork between 1880 and 1982. Mine tailings, commonly 90 percent of the ore separated by milling and flotation, contained concentrations of arsenic, cadmium, copper, lead, and zinc that were 10 to 100 times expected background values (Andrews, 1987). As they traveled downstream, tailings mixed with streambed sediments and were deposited as alluvium that became a secondary source of contamination to the Clark Fork (Moore and Luoma, 1990). The finer-grained deposits were continually resuspended and transported downstream by bed scour, lateral channel cutting, and overland erosion. Soluble salts within the deposits were flushed directly into the streams by surface runoff or leached through the floodplain soils to the alluvial aquifer (Lambing, 1991).

Initial efforts to decrease the quantity of tailings discharged directly into the river and to mitigate risks of heavy metal exposure to aquatic life and humans included the construction of three settling ponds on Silver Bow Creek (designated near Warm Springs) between 1911 and 1959 to capture tailings eroded from the upper part of the Clark Fork drainage basin (EPA, 2004). During the Warm Springs Pond Remedial Investigation, the EPA estimated that more than 14 million m³ of sediments were contained in the three settling ponds (EPA, 1990), preventing substantial quantities of mining and milling wastes from moving downstream into Clark Fork. Beginning in 1975, lime was added at the inlet of the ponds to induce precipitation of metals and thereby increase settling efficiency (Lambing, 1991). These ponds decreased the amount of contaminated sediments reaching the upper Clark Fork from Silver Bow Creek, but part of the trace element-rich sediments continued to move through the river system (Phillips, 1985). Large-scale mining operations in Butte ceased in 1982, but widespread public concern over the effects of mine tailings throughout the upper Clark Fork Basin persisted. In response, the EPA designated three areas affected by mine tailings from Butte to Milltown Dam as National Priority List. These areas included the Silver Bow Creek/Butte Area site, the Anaconda Smelter Site, and the Milltown Reservoir Sediments site, which were all established in 1982. The entire length of the Clark Fork from the confluence of Warm Springs and Silver Bow Creeks to Milltown Reservoir was incorporated within the Milltown Reservoir site and was designated the “Clark Fork River Operable Unit,” which includes more than 140 river miles.

Federal Superfund remediation activities began in 1983 and have included substantial remediation near Butte and the removal of the Milltown Dam near Missoula in 2008. Many dumps were capped with clean topsoil and revegetated to hold the waste in place, preventing it from eroding back into Silver Bow Creek, seeping into groundwater, and (or) blowing away in high winds (Clark Fork Watershed Education Program, 2022). Further remediation activities included in-place stabilization and neutralization of floodplain tailings, tailings removal, construction of dikes at the inlet to the Warm Springs ponds, construction of berms (low earthen dikes about 30 centimeters, cm, high) along streambanks and around floodplain tailings deposits from Warm Springs Creek to Deer Lodge, and planting vegetation in amended soils to provide stability against erosion (EPA, 2020). Small-scale demonstration projects, which focused on in-situ treatment of soils in the riparian zone and bank stabilization, also have been conducted (Pioneer Technical Services, 2002). These efforts have improved the efficiency of the Warms Springs treatment ponds, ongoing cleanup of Silver Bow Creek, and other activities completed in the Butte area.

Remediation in Clark Fork was implemented to mitigate some of the more acute sources of contamination (Pioneer Technical Services, 2002; EPA, 2004), and, to date (2023), all activities have been spatially restricted to the upper 45 km of the river. In consultation with the EPA and the Montana Department of Environmental Quality, the ARCO prepared major parts of the final Clark Fork Operational Unit Remedial Investigation and Feasibility Study, completed several in-situ demonstration projects and streambank stabilization projects, and conducted a Time Critical Removal Action at Eastside Road in Deer Lodge (EPA, 2004). To reduce the tailings deposited into the river, large berms were constructed between 1989 and 1990, and, in December 2004, the EPA issued a record of decision that included removal of Milltown Dam (EPA, 2004). The dam was breached on March 28, 2008, to facilitate removal and excavation of the contaminated sediment.

The data presented in this report were collected as part of an ongoing USGS long-term monitoring program that includes assessments of metals in surface stream water, fine-grained bed sediment, and aquatic macroinvertebrates and that was designed to help identify metal trends relative to naturally varying hydrologic conditions and implementation of remediation efforts (USGS, 2023). Since 1985, monitoring data have shown elevated concentrations of cadmium, copper, lead, and zinc within the upper reach of Clark Fork and decreasing concentrations with increasing distance downstream (Hornberger and others, 1997, 2009). Constituent-based trends also corresponded to bioassessment monitoring studies, where species diversity was lower in parts of the upper Clark Fork (McGuire, 2007).

To monitor water-quality changes related to remediation activities, the USGS began collecting data in the upper Clark Fork Basin in 1985 at six sites above Milltown Reservoir (fig. 1) (Lambing, 1991). The data were collected in cooperation with the State of Montana, the Montana Power Company, and the EPA. The site upstream from Missoula was added to the network in 1986, and the site near Galen was added in 1988. Another 14 sites were added between Silver Bow Creek at Opportunity and in the mainstem downstream from Missoula in 1993 (fig. 1; not all sites are shown) (Lambing and others, 1995). The long-term data collected from mainstem sites and major tributaries enabled identification of the primary source areas contributing sediment and metals to the Clark Fork and tracking of changes over time as remedial treatments were implemented. Concentrations and mean annual loads of suspended sediment and trace elements at network sites are published in annual reports (for example, Clark and others, 2020, 2021) and associated data are in the USGS National Water Information System (NWIS) database (USGS, 2023).

To evaluate the effectiveness of mine-waste remediation at the Clark Fork Superfund complex, Hornberger and others (2009) evaluated concentrations of copper and cadmium in fine-grained bed sediment and tissue of the Hydropsyche spp. (caddisfly; Pictet, 1834), from 1986 to 2006 collected along 200 km of Clark Fork between Silver Bow Creek and the former Milltown Dam. Concentrations of copper and cadmium in fine-grained bed sediment and caddisflies were lower in the upper reach at sites closer to remediation. A significant positive relation between metal bioaccumulation and stream discharge in the middle reach indicated a flow-induced redistribution of contaminants throughout the river. Correlations between stations in sediment concentration trends showed longitudinal decreases from upstream to downstream.

More recently, Sando and others (2014) analyzed water-quality data and characterized flow-adjusted trends in mining-related contaminants for 22 sampling sites distributed across the Clark Fork River Operable Unit for water years 1996–2010. Results of this study indicated moderate to large decreases in flow-adjusted concentrations and loads of copper and other metallic elements and suspended sediment in Silver Bow Creek upstream from Warm Springs Creek. However, in the Clark Fork reach from Galen to Deer Lodge, mobilization of copper and suspended sediment from floodplain tailings and streambanks was a large source of contamination. The Clark Fork reaches downstream from Deer Lodge were smaller sources of metallic elements. Also, during 1996–2010, small temporal changes were observed in arsenic loads and flow-adjusted concentrations in the Silver Bow Creek and Clark Fork reaches downstream from the Clark Fork at Opportunity.

The upper Clark Fork Basin is in west-central Montana within the northern Rocky Mountains physiographic province (not shown), which is characterized by rugged mountains and intermontane valleys (Fenneman, 1946). The Clark Fork valley is about 145 km long and ranges in elevation from about 4,800 feet near Galen, Mont., to about 975 meters near Missoula, Mont. The north-trending valley is flanked on the east by mountains along the Continental Divide and on the west by the Flint Creek Range.

Sites included in the current study (fig. 1, table 2) encompass the upper Clark Fork Basin upstream from the Clark Fork above Missoula and are on the mainstem channels of Silver Bow Creek and Clark Fork. Originating near Warm Springs, Mont., the 780-km Clark Fork of the Columbia River Basin originates at the confluence of Silver Bow, Willow, and Warm Springs Creeks and drains an extensive region of the northern Rocky Mountains in western Montana and northern Idaho. Across Montana, the river flows northwest and empties into Lake Pend Oreille in the Idaho Panhandle. It is the largest river by volume in Montana (Clark Fork Watershed Education Program, 2022). The Clark Fork above Missoula drains a watershed of 15,400 km². From Butte to near Missoula (148 river miles), six major tributaries enter the Clark Fork: Blacktail Creek, Warm Springs Creek, Little Blackfoot River, Flint Creek, Rock Creek, and Blackfoot River. Smaller perennial and intermittent tributaries drain the surrounding mountains and terraces on both sides of the Clark Fork valley.

In its upper 32 km, the Clark Fork is known as Silver Bow Creek, which originates in the mountains north of Butte from the confluence of Basin and Blacktail Creeks. The upper reaches of Silver Bow Creek in and near Butte contain numerous mine shafts, pits, mills, smelters, and tailings piles and ponds (fig. 3). Downstream from Butte, Silver Bow Creek flows west about 16 km and north about 16 km to its confluence with Warm Springs Creek, marking the start of Clark Fork. Between Butte and the confluence with Warm Springs Creek, large areas of the intervening basin were affected by production and dispersion of waste materials (rock, water, and smelter emissions) primarily from milling and smelting activities of AMC. About 8 km upstream from the confluence of Silver Bow Creek and Warm Springs Creek, Silver Bow Creek enters the Warm Springs ponds constructed during 1908–59 (CDM Smith, 2005) to retain and treat contaminated sediment. Upstream from the Warm Springs ponds, Silver Bow Creek at Opportunity represents the outflow of the Silver Bow Creek Basin above substantial retention and diversion structures. Passing east of Anaconda, Clark Fork flows north for about 43 river miles past the towns of Galen, Deer Lodge, and Garrison. (fig. 1).

In the reach above Garrison, the Clark Fork is a highly meandering river, but the river downstream becomes narrow and confined by highway and railroad embankments. A broad valley extends from the Warm Springs ponds to Garrison and is bordered by high terraces reaching several hundred meters above the river (Lambing, 1998). Near Garrison is the confluence with the Little Blackfoot River, and downstream 8 km east of Missoula is the confluence with the Blackfoot River.

Between Deer Lodge and Garrison, the extent of floodplain tailings (slickens) along Clark Fork is like that in the valley upstream from Deer Lodge (Smith and others, 1998). Although slicken areas are rare in this area, cutbank soils are enriched with metals, and metal concentrations in bed sediments are highly variable (Axtmann and Luoma, 1991).

The taxonomic richness of sensitive invertebrate species is higher here than in the upstream reach. For example, species of caddisfly (order Trichoptera) and stonefly (order Plecoptera) that are absent in the upper reach of the river first appear below the confluence of the Little Blackfoot, indicating a change or improvement in habitat quality, decreased trace element concentrations, and recruitment of organisms from a relatively unaffected source (McGuire, 2007).

Between Garrison and Drummond where the Clark Fork valley narrows, floodplain tailings (slickens) are less extensive than in the Deer Lodge valley (not shown), and channel meandering decreases (Lambing, 1998; Smith and others, 1998). Downstream from Drummond, the Clark Fork valley is narrow (less than 1.6 km wide), and meandering decreases further in association with the narrow valley and presence of highway and railroad embankments (Lambing, 1998).

Just beyond the confluence with the Little Blackfoot River near Garrison, the Clark Fork valley turns abruptly northwest across western Montana for about 77 river miles to the former location of Milltown Reservoir near Bonner. This river segment is more channelized, although it remains single-threaded and sinuous (Smith and others, 1998). The downstream reach of the Clark Fork from the Rock Creek confluence to below Missoula shows the lowest trace element concentrations in fine-grained bed sediments and biota because of dilution from the Rock Creek, Blackfoot River, and Bitterroot River drainages (Axtmann and others, 1997), although concentrations in this area are higher than in these tributaries (Axtmann and Luoma, 1991). From Turah Bridge, the Clark Fork flows through the area of the former Milltown Reservoir.

The primary surface-water uses in the 15,539-km² upper Clark Fork Basin include agricultural irrigation, stock watering, small-scale industry (Cannon and Johnson, 2004), cold-water trout fishing (Morey and others, 2002), and recreational activities.

Following a previous synthesis of monitoring data by Hornberger and others (1997) for 1985 to 1995, the current study is a synthesis of monitoring data collected from 1996 to 2016 to describe the spatial distributions and temporal changes in copper, arsenic, cadmium, and zinc concentrations in surface water, fine-grained bed sediment, and the tissue of a widely distributed benthic insect (macroinvertebrate) genus, Hydropsyche, a netspinning caddisfly. This study also characterizes the co-occurrences of these spatial and temporal concentration changes as they relate to remediation activities and evaluates changes in tissue concentrations as a response to metal concentrations in water and bed sediment. As with Hornberger and others (1997), data were selected herein (USGS, 2023) to provide independent indicators of metal concentrations in the river and changes in metal enrichment over time. However, the current study differs from the previous work by qualitatively comparing the spatial and temporal patterns in the data rather than by calculating trends, which was done in Sando and Vecchia (2016).

Sampling sites in Clark Fork were included in this study to identify patterns of contaminant transport and accumulation at key locations. The key findings describe the variability in different indicators of metal exposure. However, unlike Hornberger and others (1997), which compared pre- and post-remediation conditions, all data included herein were collected after initiation of remediation in the upper reach of the basin. Metal concentrations in bed sediment and insect tissue were consistently low at tributary sites; accordingly, except for the Blackfoot River, tributary sites were not included in our analyses. Data collected from the Blackfoot River near Bonner (USGS streamgage 12340000; USGS, 2023) were used to represent baseline conditions for comparisons with mining-effected sites in the Clark Fork mainstem.

The goal of the current study is to describe the spatial and temporal variations of arsenic, copper, cadmium, and zinc in surface-water samples (dissolved and total recoverable fractions), fine-grained bed sediment samples, and aquatic macroinvertebrate (insect) tissue concentrations from 1996 to 2016. Discussion of metal bioaccumulation is restricted to concentrations measured in caddisflies, a net-spinning, omnivorous, filter-feeding insect that is metal tolerant and ubiquitous in western streams (Morse and others, 2019).

Like many mining-affected sites, certain trace elements characterize the contamination in Clark Fork sediments, including copper, cadmium, arsenic, and zinc. An EPA ecological risk assessment of the lower Clark Fork (EPA, 1999) concluded that copper was most likely to cause adverse ecological effects, whereas bioaccumulated cadmium and zinc were not highly correlated with changes in the community composition (Luoma and others, 2010). Arsenic concentrations can be elevated in the Clark Fork through remediation activities that alter river pH resulting in increased solubility (Hornberger and others, 2009), but ecological effects of arsenic contamination are more localized than for copper because of the more restricted occurrence of arsenic (Luoma and others, 2010). Therefore, although water samples were analyzed for concentrations of arsenic, cadmium, copper, iron, lead, manganese, zinc, and other constituents, the data presented here include only copper, arsenic, zinc, and cadmium, and focus is placed more on copper in the results and discussion. Cadmium concentrations were near or below the detection limits in all samples, so cadmium results are summarized here but not included in all analyses or in the final discussion. Also, concentrations of all metals were highest in Silver Bow Creek, upstream from the Warm Springs ponds, and greatly reduced immediately below the outfall. So, the data collected from the Silver Bow Creek monitoring sites (OP and PO; table 2; fig. 1) were analyzed separately from data collected in the Clark Fork mainstem. Data collected from the Blackfoot River near Bonner (USGS streamgage 12340000; USGS, 2023) were used to represent baseline conditions for comparisons with mining-affected sites in the Clark Fork mainstem.

The highest concentrations of metals within the Clark Fork superfund complex have historically occurred above the Warm Springs ponds at Silver Bow Creek at Opportunity (fig. 1, OP; USGS streamgage 12340500). EPA freshwater criteria for metals are expressed in terms of the dissolved metal in the water column, so comparisons of results to the aquatic life criteria were made only with water sample concentrations measured in the dissolved fractions. During the 20-year period between 1996 and 2016, the grand annual mean (mean of April to August yearly means) copper concentration in the dissolved fraction of water samples collected from OP was 31.1 plus or minus (±) 15.4 µg/L (one standard deviation; table 3, fig. 5), which is above the acute and chronic aquatic life criteria of 24.3 µg/L. April to August mean dissolved copper concentrations at this site were highest in 1997 (59.6 µg/L; data are not available for 1996), then declined to 9.25 µg/L by 2016 (table 3; fig. 5B). Dissolved copper concentrations were much lower at the Silver Bow Creek at Warm Springs (site PO; USGS streamgage 12323750) about 7 river miles downstream. For the PO site, the grand annual mean (mean of April to August yearly means) dissolved copper concentration between 1996 and 2016 (4.65±1.42 µg/L) was below acute and chronic aquatic life criteria and lower than the grand annual mean concentration in the Clark Fork mainstem water samples (7.42 ± 2.30 µg/L, all sites combined: USGS streamgages 12323800, 12324200, 12324680, 12331800, 12334550, and 12340500) over the same period.

The grand annual mean (mean of April to August yearly means) dissolved zinc concentration also was much higher in OP (173 ± 139 µg/L) than at PO (4.89 ± 2.42 µg/L) or in the Clark Fork mainstem sites (24.6 ± 18.1 µg/L) and exceeded the aquatic life criteria of 197 µg/L (acute and chronic) for zinc (table 3). As with copper, the maximum April to August yearly mean dissolved zinc concentration was observed in 1997 (470 µg/L; no data in 1996; fig. 5D) at OP, and April to August yearly mean concentrations fell at this site and remained between 58.8 and 43.0 µg/L in more recent years, 2013–16. In contrast, April to August yearly mean dissolved arsenic concentrations were highest in surface-water samples from PO (20.8 ± 3.13 µg/L) compared with OP (10.0 ± 2.45 µg/L) or the Clark Fork mainstem sites (10.4 ± 1.10 µg/L; table 3). The 20-year peak in April to August yearly mean dissolved arsenic occurred at PO in 2001 (26.6 µg/L; fig. 5A), whereas the arsenic maximum April to August yearly mean at OP was 14.4 µg/L in 2004 (fig. 5A). April to August yearly mean dissolved cadmium concentrations were below 1.0 µg/L in most years at both OP and PO (fig. 5C). April to August yearly mean dissolved cadmium concentrations remained below the chronic criterion of 1.26 µg/L in all years at PO and in all measured years at OP except for 1997 and 1998, when the April to August yearly mean concentrations were 1.65 µg/L and 1.41 µg/L, respectively.

Among the different compartments sampled, fine-grained bed sediment samples contained the highest metals and arsenic concentrations (table 3, fig. 6). Similar to surface-water samples, the 20-year mean (mean of all years, sampled once per year) copper and zinc concentrations in fine-grained bed sediment were much higher in samples from OP (3,039±2,278 micrograms per gram [µg/g] copper and 5,542 ± 3,569 µg/g zinc) relative to the Clark Fork mainstem sites (mean of all sites; 899±313 µg/g copper and 1,427±555 µg/g zinc); and, as with water, the highest mean (mean of all years, sampled once per year) sediment arsenic concentration occurred in samples from PO (112 µg/g versus 81.4 µg/g at OP and 63.3 µg/g in the Clark Fork mainstem sites). Copper concentrations in fine-grained bed sediment exceeded the PEL of 197 mg/kg at OP in all study years (fig. 6B) but dipped below the copper PEL at PO in 2001 and 2003. The concentration of copper in fine-grained bed sediment peaked at OP in 2001 (9,023 µg/g) and at PO in 2010 (466 µg/g). Overall, annual (sampled once per year) copper concentrations at OP remained near or above 2,500 µg/g from 1996 to 2009, when concentrations began to sharply decline (446 µg/g in 2016). A different pattern was found at PO, where copper concentrations were lower, ranging from 196 µg/g (2003) to 466 µg/g (2010).

Zinc concentrations in fine-grained bed sediment (sampled once per year) far exceeded the PEL (315 mg/kg) in all years at both Silver Bow Creek sites (table 3, fig. 6D). The highest zinc concentration in fine-grained bed sediment was recorded in 2000 at OP (13,357 µg/g) and in 2004 at PO (1,097 µg/g). Peaks in sediment zinc concentrations above 10,000 µg/g also occurred at OP in 2000, 2001, and 2004, but dropped to below 5,000 µg/g after 2008. At PO, annual zinc concentrations remained near or below 1,000 µg/g during the study period.

Arsenic data were not collected in Silver Bow Creek until 2004, but concentrations exceeded the arsenic PEL (17 mg/kg) for sediment at both OP and PO in all years except 2016, when annual arsenic concentrations in fine-grained bed sediment (sampled once per year) fell to 16.9 µg/g at OP (concentrations at PO remained above 66 µg/g in all study years). The higher arsenic concentrations below the settling ponds were from liming treatments, which increased pH and mobilization of arsenic. Annual arsenic concentrations were highest in fine-grained bed sediment at OP early in the study period, peaking at 163–153 µg/g between 2004 and 2008, and at PO, annual concentrations fluctuated from 177 µg/g in 2004 to 96 µg/g in 2016. Annual cadmium concentrations in fine-grained bed sediment were higher at OP than at PO until about 2009, when concentrations at both sites were similar (fig. 6C) and exceeded the PEL of 3.53 µg/g-in all study years at both sites. Annual cadmium concentrations were lower than the other constituents in all sediment samples (fig. 6C), but elevated annual concentrations were noted at OP in 1996 (42.0 µg/g), 2000 (52.9 µg/g), 2001 (41.4 µg/g), 2003 (41.9 µg/g), and 2004 (43.8 µg/g).

Annual copper and zinc tissue concentrations (sampled once per year) were higher than dissolved and total-recoverable fractions (April to August yearly mean) in surface water but were an order of magnitude lower than fine-grained bed sediment (fig. 7). As with fine-grained bed sediment, annual concentrations in insect tissue were higher during the study period (2000–16) at OP (350 µg/g copper and 835 µg/g zinc: table 3, fig. 7) than at the Clark Fork mainstem sites (101 µg/g copper and 275 µg/g zinc) or at PO (32.7 µg/g copper, 175 µg/g zinc). Insects were not collected at OP until 2000, but between 2000 and 2016, peak annual copper concentrations occurred early, in 2001 (864 µg/g) and 2002 (987 µg/g), then declined to a range of 100–150 µg/g after 2011. The peak in annual copper concentration in insect tissue in 2001 corresponded with peak copper concentrations in surface water and fine-grained bed sediment. Tissue copper concentrations in samples from PO fluctuated during 2000–16 from 20.3 µg/g (2001) to 46.1 µg/g (2011). Similar patterns were found in annual zinc concentrations: highest at OP and peaking between 2000 and 2003 (with an additional peak above 1,000 µg/g in 2011 that did not correspond to a peak in the copper concentration).

As with surface water and fine-grained bed sediment, annual arsenic concentrations in insect tissue were higher at PO (14.7 µg/g) than at OP (12.2 µg/g) or at the Clark Fork mainstem sites (9.30 µg/g). Arsenic was not analyzed in tissue samples from any site before 2002. In subsequent years, annual concentrations in insect tissue were highest at PO in 2004 (27.3 µg/g) and at OP in 2009 (25.9 µg/g). Annual concentrations fell below 10 µg/g between 2012 and 2016 at OP but remained between 9.0 µg/g and 17.1 µg/g during this period at PO (fig. 7A). Annual cadmium concentrations in insect tissue from 1996 to 2016 averaged 4.97 µg/g and 1.12 µg/g at OP and PO, respectively, and averaged less than 2.00 µg/g among the Clark Fork mainstem sites. The maximum annual cadmium concentration in insect tissue measured during the study was 10.6 µg/g at OP in 2001.

Pearson product-moment correlation analysis was performed to identify potential relations between annual concentrations of arsenic, copper, cadmium, and zinc in insect tissue and bed sediment samples, and the April to August yearly mean values in dissolved and total recoverable fractions of water samples. Results (table 4) showed that annual concentrations of copper, cadmium, and zinc in insect tissue samples were poorly correlated with those in sediment and water, but annual or annual mean arsenic concentrations correlated significantly (Pearson correlation coefficient, r=0.60, probability, P<0.05). Significant, positive correlations also were determined between the April to August yearly mean concentrations of copper, cadmium, and zinc in the dissolved fraction of water samples and the annual values in fine-grained bed sediment. April to August yearly mean total recoverable concentrations of arsenic, copper, and zinc in surface water samples also correlated significantly with annual values in fine-grained bed sediment. Although the presence of a statistically significant relation in concentrations of metals between compartments provides weight of evidence for corresponding trends, the lack of a statistical relation with concentrations in tissue samples indicates the complexities of biogeochemical interactions.

Annual mean streamflow at Clark Fork above Missoula (USGS streamgage 12340500) between 1996 and 2016 was 2,184 cubic feet per second (ft³/s). This was lower than during the previous 20 years (1975–95; 2,813 ft³/s) by about 22 percent and lower than during 1930 to 1975 (4,868 ft³/s annual mean streamflow) by about 45 percent. (fig. 4). During the current study period, 9 years (45 percent) showed annual mean streamflow above the 50th percentile of all data summarized annually since 1930 (fig. 4). Annual mean streamflow exceeded the 50th percentile (15 percent) between 1975 and 1995. Within the current study period, the annual mean streamflow was highest (above the 75th percentile) at the Clark Fork above Missoula in 1996, 1997, and 2011, and lowest (below the 25th percentile) in 2000, 2001, 2004, and 2016. The 2011 flood on the Clark Fork was notable in terms of its magnitude and duration. The snowmelt runoff that year exceeded the historical peak streamflow for over 2 months, extending from mid-May through mid-July, and resulted in extensive bank erosion, sediment movement, and several avulsions, including one in the newly remediated channel segment at Milltown (Karin Boyd, P.G., Applied Geomorphology, Inc., written commun., March 2018). If the differences in annual mean streamflow between the current 20-year study period, 2006-2016, and previous years represented a decline in annual mean flow, the 20 years considered here can be regarded as a low-flow period, relative to conditions since 1930. This implies that suspended sediment and trace-element loads also were smaller from 1996 to 2016, which is an important consideration for monitoring changes in contaminant concentration resulting from cleanup and remediation. However, large variations in annual suspended sediment and trace-element loads do not necessarily mean that a change has occurred in the supply of a constituent, only that more or less runoff (from increased or decreased precipitation, for example) was available to transport materials.

Despite the data corrections to reduce the influence of flow on concentration patterns, several high-flow years corresponded with peaks in contaminant concentrations based on their annual mean values calculated April to August. In particular, the 2011 flood had a substantial effect on arsenic and copper concentrations in all compartments, and this 20-year peak in flow corresponded with the 20-year peak in arsenic and copper concentrations measured in insect tissue samples. In the other compartments, copper concentrations were higher during the 1996–97 high-flow years, but also in 2008 (total recoverable and fine-grained bed sediment samples) and 2009–10 (fine-grained bed sediment). Arsenic concentrations also were higher in fine-grained bed sediment samples in 2009 and 2014. Streamflow was near the 75th percentile in those years (fig. 4), but the lower arsenic concentrations in 2009 and 2014 water samples, relative to fine-grained bed sediment and tissue, suggests differences in contaminant supply and transport between the river bottom and the water column. Cadmium and zinc concentrations decreased in water and fine-grained bed sediment as streamflow increased from 2002 to 2013; however, concentration peaks corresponded with the 2011 flood in the dissolved fraction of water samples (slight peak in cadmium and a large peak in zinc), insect tissue (zinc only), and fine-grained bed sediment (zinc only).

This report presents the spatiotemporal patterns of mining-related contaminants, copper, arsenic, cadmium, and zinc, in surface water, fine-grained bed sediment, and macroinvertebrate (aquatic insect) tissue in the upper Clark Fork from near Butte to Missoula, Mont. Although copper is most likely to cause adverse ecological effects, arsenic concentrations have been elevated in the Clark Fork through localized remediation activities that alter river pH (Hornberger and others, 2009), so the following discussion focuses primarily on copper and includes arsenic where relevant. Overall, the patterns observed in this study were consistent with the trends analysis performed by Sando and Vecchia (2016) and other studies of trace-element concentration trends in the Clark Fork. These studies measured the highest metal and arsenic concentrations in Silver Bow Creek, where remediation has had the greatest effect, and reported downstream declines in bed sediment concentrations (Axtmann and Luoma, 1991; Hornberger and others, 2009) and in macroinvertebrate tissue (Farag and others, 1998). Fine-grained bed sediment concentrations and flow-adjusted concentrations of total recoverable copper and arsenic in water samples primarily decreased downstream, but varied more in tissue samples among sites, indicating no clear pattern of increasing or decreasing between sites, which further indicated a more complex interplay of factors affected biological uptake of trace elements than distance from the source. The reach of the Clark Fork from GG to DL continues to be a large source of metal contamination, and this strongly contributes to downstream transport of those constituents.

In this study, trace element concentrations, especially copper, often exceeded the chronic aquatic life criteria and consistently exceeded the sediment PEL for copper, particularly in the upper and middle river segments (figs. 8 and 10). Water-column concentrations were adjusted to reduce variations attributable to flow, but the effect of flow and downstream delivery of contaminants was still apparent in the data during high-flow years, particularly with copper and arsenic concentrations measured in caddisfly tissue samples, based on when peaks were observed in both concentration (fig. 8) and streamflow (fig. 4). The 20 years considered here were the wettest period since remediation started in 1983 (fig. 4), and this increase in precipitation may have affected patterns in contaminant concentrations, overall, by redistributing elements and altering downstream physicochemical processes, particularly during periods of higher flow and transport, such as the floods of 1996–97 and 2011.

Correlation analysis (table 4) indicated a uniqueness to the concentrations of bioavailable arsenic, copper, cadmium, and zinc, in that tissue concentrations correlated only with arsenic in the dissolved fraction of water samples and that tissue copper concentrations were independent of concentrations in water and sediment. Correlations were strongest for arsenic, copper, and zinc concentrations measured in fine-grained bed sediment and in the total recoverable fraction of water samples. This strong correlation is reasonable considering that copper, zinc, and cadmium (also lead) bind tenaciously to organic matter in soil, sediments, and suspended particulates and that copper can remain bound to insoluble complexes (Gale and others, 2004). Zinc and cadmium have a greater tendency than copper to dissociate from such complexes and can form soluble, ionic species (Kabata-Pendias and Pendias, 2001).

Hare (1992) reported that aquatic organisms do not always respond to changes in dissolved metal exposure, and Cain and others (2011) demonstrated that consumption of periphyton is an important route of metal exposure to benthic invertebrate grazers. However, the relation between exposure to dissolved metals and their bioaccumulation is relevant given that dissolved metals influence dietary sources. Monitoring of environmental contamination often emphasizes dissolved metal concentrations with the assumption that increased dissolved concentrations indicate increased uptake and bioaccumulation. However, simple relations between environmental contaminant concentrations and animal contaminant concentrations are rarely observed in nature because of the complex interaction of geochemical and biological factors (Luoma, 1989).

The ecological effects of pollutants often result from complex interactions of toxins with natural environmental processes. Flow condition, for example, is important for bioavailability, bioaccumulation, and toxicity to aquatic organisms. As contaminants are transported downstream, they move through a mosaic of complex redox and pH environments that have profound effects on the solids and solute phases and are transformed via microbial and inorganic reactions (Hochella and others, 2005), which can further increase the mobility of (potentially) harmful compounds. A combination of biological monitoring and measurements of water and sediment quality may provide a good indication of conditions and potential risks to aquatic ecosystems. Understanding how physical processes, such as streamflow and sediment transport and deposition, affect biological exposures might improve the basis for predicting biological risks from contaminants. In mining-affected rivers like the Clark Fork, metal concentrations in tissue residues of aquatic organisms may track the downstream contamination trend observed in fine-grained bed sediments. Therefore, bioaccumulation seems to be linked to some of the same processes that control the distribution of sediment-bound metals, even if sediments are not necessarily a direct pathway of exposure.

Long-term monitoring provides data that may be used to assess changes in stream ecosystems, to provide early warning of potential problems, to evaluate the efficacy of remedies, and to predict future adverse environmental effects on humans and the environment. Evaluating contaminated systems that require remediation and restoration provides critical information for environmental management and long-term stewardship. Monitoring data collected before and after remediation supports documentation of the effects of cleanup actions on metal concentrations. From an ecological point of view, effective remediation of mine waste integrates aspects of environmental health, scope and scale considerations, ecological services, biodiversity, and long-term consequences. Clements and others (2021) used a before- and-after, control-impact study design to quantify responses of benthic macroinvertebrate assemblages to stream remediation. They observed substantial reductions in metal concentrations and corresponding improvements of benthic assemblages after remediation at mining-affected watersheds across the western United States, including the Clark Fork. Recovery rates across these regions were consistent, and streams typically recovered within 10 to 15 years after remediation began, although episodic events changed recovery trajectories at some sites. Overall, despite variation in defining complete restoration in these watersheds, this study used multiple lines of evidence to provide quantifiable measures of the timing and completeness of recovery relative to reference conditions.

Adequate and appropriate monitoring procedures are the most direct measure of the success of remedial strategies and the resulting rates of recovery for populations, communities, or ecosystems. In short, the success of any restoration effort can be documented through a well-designed monitoring program that collects physical, chemical, and biological information to provide a comparison with conditions prior to cleanup activities. Without collecting and analyzing comprehensive monitoring data, land managers cannot objectively evaluate the success of a remedial or restoration action or determine whether remediation and restoration goals have been met (Finger and others, 2007).

In the upper Clark Fork, the goal of remediation from historical mining activities is to restore the ecosystem health to the extent possible. Successful recovery in the Clark Fork may be facilitated by continuation of an adaptive management strategy that includes collecting a comprehensive, multivariate dataset to evaluate whether established goals are being met and for subsequent adjustments and management, as needed.