3

Analysis of Inputs to Coeur d’Alene Lake

The committee was tasked to evaluate current water quality in Coeur d’Alene (CDA) Lake, the lower rivers, and the lateral lakes, with a focus on observed trends in nutrient loading and metals concentrations. This task necessitates looking at the water quality not only in the Lake itself (see Chapters 4, 5, and 6), but also in the major inputs to the Lake. Because CDA Lake is a surface water–dominated system, this chapter focuses on the inputs of the two major river systems: the CDA River and the St. Joe River. The trends observed in the river monitoring network data over the past 30 years provide clues about the effectiveness of the Superfund cleanup efforts that have already taken place in the watershed. They also provide a window to help prioritize and plan for future remediation efforts and to better understand the role of potential changes in land use and population in the watershed.

This chapter begins with three narratives that describe (1) how metals from historical mining activities have moved through the landscape, (2) how the Superfund activities begun in the late 1990s have affected sources of metals in the watershed, and (3) ongoing sources of phosphorus in the watershed. The chapter then analyzes data on water flows and metal and phosphorus concentrations and fluxes collected over the past 30 years in the major river systems to reveal the most important trends in the inputs of these constituents from the watershed to CDA Lake.

SOURCES OF METAL INPUT TO COEUR D’ALENE LAKE

Ore deposits are composed of primary minerals (or mixtures of minerals) from which metal(loid)s can be profitably extracted. Mineral extraction beneficiation¹ and processing of these deposits inherently create massive quantities of wastes, all of which are enriched to one degree or another with the potentially toxic trace elements characteristic of the ore deposit. The most important contaminants in the wastes from the CDA activities included silver (Ag), lead (Pb), zinc (Zn), cadmium (Cd), arsenic (As), and antimony (Sb) (NRC, 2005).

The release of large volumes of tailings until 1968 was one of the most important factors in the widespread metal contamination in the CDA watershed. Once wastes are released or escape the immediate physical area of primary activities, hydrologic and geochemical processes distributed them widely over the basin. Particularly important in this case is the relatively limited dilution of contaminated particulates that occurs in the CDA watershed before the mine wastes are ultimately deposited in a large, deep lake of great societal value. This section describes the processes that contribute to metal inputs to the Lake, including the sources and stockpiles of metals

___________________

¹ Beneficiation is the process by which the mineral-rich ore is concentrated, usually involving flotation in these types of mines.

established up until 1968 (when the effective tailings impoundments were completed), and the physical and biogeochemical processes that determine the fate of those metal(oid)s.

Character of Mine Wastes in Different Basins

Mining, milling, and smelting in the CDA basin created the primary, secondary and tertiary contamination typical of mineral extraction (Moore and Luoma, 1990). Historic mining operations typically scattered the immediate wastes of the operation (primary contamination) across the local landscape in an “ill-defined patchwork of waste rock, mill tailings, furnace slag and flue dust” (Moore and Luoma, 1990). In simplistic terms, waste rock is the material that is removed to uncover the ore deposits. Mill tailings are mostly fine-grained particulate wastes that result from the combination of physical and chemical processing used to separate and concentrate the metals targeted for extraction from the raw ore. A high percentage of the ore materials that are processed end up discarded as highly contaminated fine-grained tailings (Downs and Stocks, 1977). The concentrate from the milling process is refined by smelting, which produces flue dust and slag in which metal contaminants are even more concentrated than in tailings. Smelter complexes typically created intensely contaminated soils and groundwater, often in close proximity to larger population centers (Moore and Luoma, 1990). Fine particulates and gases are also emitted into the atmosphere by smelting for distribution over tens to hundreds of square kilometers.

Primary wastes, by definition, create localized contamination issues. Constraining wastes to the footprint of the operation can limit risks to downstream waterways, floodplains, and resources like lakes (Luoma and Rainbow, 2008). Unfortunately, during the western U.S. mining boom between 1850 and 1950, environmental considerations like constraining waste distribution were of little or no concern.

Secondary particulate contamination is created when the discarded tailings and other mine wastes are washed downstream, mixed with other sediments in transport, and re-deposited in lower areas. Bookstrom et al. (2013) found lead-rich sediments (defined as more than 1,000 μg Pb/g dw sediment) throughout the CDA basin, co-enriched with zinc, silver, copper, cadmium, iron, and manganese (and in some cases, arsenic, antimony, and mercury). Floods can disperse these tailings across the floodplain, with more intense floods moving larger sized particles. During the U.S Environmental Protection Agency’s (EPA’s) remedial investigations in 2001, 1,080 mining-related areas on the CDA floodplains were identified as sources of either primary or secondary wastes, varying in size from less than an acre to hundreds of acres (NRC, 2005). Tertiary contamination occurs as a result of biogeochemical reactions within the secondary deposits.

Upper Basin

The upper basin, and specifically the South Fork of the CDA River upstream from Elizabeth Park, is dominated by primary waste deposits, including waste rock and small mill deposits. These can be seen amidst the workings at the mouths of the underground mines (deposits visible in Figure 3-1). Most of the mining in the basin was done in steep terrain through a complex of 200 underground mines and a few very small pits. Thus, the deposits are large in number but limited in volume and area.

A common source of metal-contaminated water in the upper basin are adits, which are horizontal passages leading into a mine for the purposes of access or drainage. Where adits cut through aquifers in the region, adit waters characterized by low pH and high dissolved metal concentrations are created—a consequence of oxygenated aquifer water reacting with sulfide mineral ores.

The section of the upper basin along the South Fork of the CDA River between Elizabeth Park and Pinehurst is commonly known as “the Box” and includes the Bunker Hill smelter, the largest ore-processing facilities, and associated population centers (Figure 3-2). Due to the intense contamination typical of smelting activities, both primary and secondary wastes are deposited in this part of the upper basin. The river gradient here is more moderate and the floodplains are broader than upstream. The basin is bordered by steep valley walls, such that primary wastes from the mining facilities were deposited directly onto the floodplains or released into streams. Facilities and housing were built on top of the vast amounts of mine tailings deposited in this region (NRC, 2005).

The efficiency of metal removal from the ore evolved over time, which affected the nature of the wastes. Fine-grained wastes deposited at Cataldo early in the mining history contained 10,000–30,000 μg Pb/g dry weight

(1–3 percent lead; Bookstrom et al., 2013). The finer, more easily mobilized wastes from beneficiation created in later operations (NRC, 2005) contained 10,000–14,000 μg Pb/g dry weight at Cataldo. But whatever the era, deposits of tailings have consistently contained lead concentrations up to and more than 100 times natural levels.²Box et al. (2001) estimated that much of the floodplain in the Box was covered with metal-contaminated alluvium with an average thickness of approximately 1.3 m.

As of 2005, the upper basin was a continued source of metals to the river system and carried 20,000 to 70,000 metric tons (MT) of particulates downstream per year (Bookstrom et al., 2013). Of the approximately 800,000 MT of mined lead historically lost directly or indirectly to streams, Bookstrom et al. (2001) estimated that 24 percent (200,000 ± 100,000 MT) resided in the upper basin. These wastes have been the focus of remediation to date.

___________________

² Based largely upon sediments obtained from cores that extended to the pre-mining era, Bookstrom et al. (2013) calculated pre-mining Pb soil concentrations were 31 ± 19 μg/g dw. Estimates from several studies reviewed by the NRC (2005) and from CDA lake cores (Horowitz et al., 1993) are within one standard deviation of this estimate. Median concentrations of Pb in fine-grained floodplain sediments in the lower basin are 2,900 μg/g dw.

A few attempts were made at constraining wastes from the upper basin in the first 70 years of mineral extraction. For example, at least two in-stream tailings ponds were built between 1900 and 1933, but these both failed during large floods, leaving behind thick deposits of contaminated sediments over the floodplain the dams had engulfed (Bookstrom et al., 2013). A few unlined tailing-settling ponds were built off-stream between 1928 and 1968, including the Central Impoundment Area (CIA; Box 3-1) in the Box. But the large number of unregulated operations scattered over a wide area of the watershed limited the overall effectiveness of these efforts (Bookstrom et al., 2013; NRC, 2005). Other mining areas in the West of comparable size to the Silver Valley began employing effective settling basin technologies in the 1930s (Bookstrom et al., 2013). But it was not until 1968, more than 80 years after the first mining claim discharged tailings into the waterways, that all operating mills and mines in the Silver Valley had effectively impounded their tailings (Morra et al., 2015).

Lower Basin

The lower basin includes the main stem of the CDA River extending from Cataldo Landing to CDA Lake. This area is characterized by broad alluvial floodplains that range in width from 300 m at Cataldo, below the confluence of the North and South Forks of the CDA Rivers, to 5 km at the river mouth at CDA Lake (NRC, 2005). The lower basin is now a source of secondary contamination, created when the discarded tailings and other wastes from the upper basin were washed downstream, mixed with other sediments in transport, and re-deposited in the lower basin.

Floods have been instrumental in dispersing tailings from the upper basin across the floodplains of the lower basin, with more intense floods moving larger sized particles. Channel aggradation was caused by sedimentation of mine wastes and reduced forest cover during the early phases of mining. This increased the extent and severity of overbank flooding as the mineral extraction progressed (Bookstrom et al., 2013; NRC, 2005). Between 1893 and 2004, at least 40 discharge episodes peaked at 17,000 cfs (481 m³/s) or more, inundating the floodplain of the CDA River valley, such that “on average, much of the valley floor was flooded about every 2.5 years” (Bookstrom et al., 2013). Floods with peak discharge of about 70,000 cfs (1,982 m³/s) or more occurred in 1933, 1974, and 1996. By 1900, mill tailings deposited in the upper reaches of the CDA basin had reached CDA Lake and had affected as much as 25,000 acres (10,117 hectares) along the South Fork and main stem of the CDA River (Long, 1998). Thus, the accumulation of mine wastes that dominate the sediments of CDA Lake began 120 years ago and continues today.

The river channel in the lower 44 km of the lower basin has a meandering pattern and, for most of the year, has essentially a zero gradient (NRC, 2005). After floods and spring runoff subside, the channel of the CDA River in the lower part of the lower basin is backflooded in summer/early fall (as controlled by the Post Falls Dam, see Chapter 1). This has created delta-like floodplains in the lower basin that are metal-enriched and are now permanent features.

The CDA River in the lower basin is about 60 to 100 m wide and 5 to 15 m deep, with a bottom composed of sandy lead-rich sediments averaging about 3 m thick (Bookstrom et al., 2013). These riverbed sediments contain about 51 percent of the lead in all sediments on the floor of the CDA River valley (Bookstrom et al., 2001). The other 50 percent of the lead is scattered across the floodplain. Bookstrom et al. (2013) estimated that mill tailings covered about 60 km² of the 80-km² floor of the main stem of the CDA River valley. As river channels naturally meander and cut new banks, metal-rich deposits on the floodplains are remobilized into the river—yet another process that is accentuated by floods.

Hence, the lower basin comprises an immense stockpile of metal-enriched particulates poised for transport to CDA Lake. Indeed, in the 1999–2000 water year, the South Fork of the CDA River delivered only about 20 percent of the total lead load to CDA Lake, with the remaining 80 percent derived from erosion of the bed and along the course of the main stem of the CDA River below the confluence of the North Fork (Clark, 2003, Fig. 12).

Absence of Watershed Scale Dilution

An important issue with secondary wastes in the CDA basin is minimal dilution of particulate material with uncontaminated sediments. When metals in primary waste deposits enter streams and rivers as solutes and particulates, dilution is determined by the area of the watershed (Helgen and Moore, 1995). For example, metal

contamination in sediments can be detected 650 km downstream from the Clark Fork mining complex in Montana, but the concentrations of those metals progressively decline downstream from their source by as much as 100-fold as a result of inputs of unenriched particles and waters from numerous tributaries (Axtmann and Luoma, 1991; Helgen and Moore, 1995). The CDA basin differs from this situation in that the river reach between the mineral extraction operations (the Superfund Site) and the Lake is relatively short (~80 km) (Bookstrom et al., 2013). The North Fork of the CDA River is the only large unenriched tributary to enter the CDA River below the mining district (and ~60 km from CDA Lake).

The dilution effect of tributaries can be reversed if the bed and banks of the river below the tributary are contaminated by secondary and tertiary wastes (Hornberger et al., 2009; Axtmann and Luoma, 1991). On average, the North Fork of the CDA River has about twice the discharge of the South Fork and about five times the sediment load—both largely uncontaminated, which NRC (2005) estimated should result in dilution of South Fork metal concentrations by 25–35 percent. The expected dilution, however, is not evident in sediments or soils from the levees, lateral lakes, or the river bottom in the lower basin. Bookstrom et al. (2013) found that lead concentrations between 1988 and 1993 in river-side levee sediments of the lower basin (below the North Fork confluence) varied from 3,300 μg/g dw on riverbanks to 3,800 μg/g dw on levee back-slope uplands, similar to the average upstream. In lateral floodplains, median lead concentration increased with water depth, from 1,900 μg/g dw in lateral marshes, to 2,100 μg/g dw in littoral margins of lateral lakes, and 4,400 μg/g dw on the bottoms of lateral lakes. These sources of contamination in the lower basin will continue to contribute metals to CDA Lake until they are isolated from the river.

Floodplain Geochemistry and Dissolved Metal Inputs

Biogeochemical reactions affect the distribution, partitioning, mobilization and bioavailability of metals in river valleys contaminated by fluvial deposition of mine wastes (Nimick and Moore, 1991). In the CDA basin, the primary mined ore minerals included mostly metal and metalloid sulfides with some carbonates mixed with the sulfides (NRC, 2005). These were formed in anoxic environments deep within the earth’s crust. When these deposits are exposed to oxygen in air or water, the minerals are oxidized, releasing the metals along with the secondary products abundant in such ores (including reduced iron, manganese, sulfate, and hydrogen ions). The sulfuric acid produced by the oxidation reactions results in waters with pH < 2. High concentrations of iron, manganese, aluminum, and other metals and metalloids are common in the low-pH waters generated by mining sulfide deposits. Acidic groundwater and overland flow (runoff) then moves from un-remediated, mining contaminated floodplains to nearby rivers and streams. Inputs of dissolved metals from groundwater are an especially important source of metal enrichment of rivers and streams in low-flow conditions, while particulate-bound metals in overland flow dominate inputs to rivers during high-flow conditions (Moore and Luoma, 1990).

When the low pH waters emanating from mining activities are neutralized (e.g., by mixing with oxidized, higher pH waters containing alkalinity), the reduced iron and manganese precipitate as various oxide and hydrous oxide phases. Most of these form multi-phase complex particulates that also include clays and organic material (Jenne and Luoma, 1975). Colloidal or nanophase iron and manganese, which take longer to aggregate and settle, are also common in groundwaters and surface waters of the CDA region (Langman et al., 2020). Because metal cation sorption to clays and clay minerals such as iron, aluminum, and manganese oxides increases with increasing pH, dissolved metal cations present at low pH become particulate-bound as pH increases, with the strength of binding to the particulate phases determining the equilibrium distribution of metals between particulates and solution. Binding strength to iron oxides among metals of interest in the CDA basin follows the order Pb >> Zn > Cd. In anoxic environments, like subsurface wetland sediments and lateral lake bottom sediments, particulate metal sulfides are formed under neutralized conditions where sufficient sulfide exists (Balistrieri et al., 2002; see Chapter 6). Thus, the fate of metals in the CDA valley involves mineral dissolution, metal mobilization and migration, and mineral re-precipitation and metal adsorption that are controlled by both pH and redox conditions. Although pH and the availability of oxygen will determine which processes dominate spatially and temporally, the permeability of the soils or sediments, organic content, and microbial activity of the depositional environment are also important (Balistrieri et al., 2002).

The distribution of metals between particulate and dissolved phases differs depending on the metal and the location’s redox status. For example, Balistrieri et al. (2002) noted that riverbanks and natural levees of the CDA River are more oxidizing environments and have lower zinc than lead concentrations because zinc is preferentially leached from these sediments when they are inundated. During the major rain-on-snow flood of 1996, they observed that zinc/lead ratios of suspended sediment were less than 1, indicating that most of the suspended sediment was mobilized from oxidizing environments (i.e., surface soils). In contrast, permanently inundated (and reduced) bed sediments in the river channel are relatively more enriched in zinc because zinc is not mobilized from the sulfide-rich sediments. During the major 1997 spring flood, zinc/lead ratios of suspended sediment were greater than 1, indicating that fines winnowed out of riverbed sediment predominated over sediment from oxidizing environments.

Groundwater can play an important role in delivering mining contaminants to river systems, especially during summer and fall, when the rivers gain from the groundwater. The groundwater below contaminated floodplains of the CDA valley can be of low pH. Complex reactions occur when acidic groundwaters contact less acidic waters in the river or in the hyporheic zone (the subsurface zone where groundwater first mixes with river water). In particular, Paulson (2001) proposed that dissolved Fe(II) oxidizes to Fe(III) and precipitates within the oxygenated hyporheic zone of the stream channel, absorbing lead and reducing its transport to the river. Because of their lower binding intensity, significant concentrations of zinc and cadmium remain in solution in the drainage that then enters surface waters. This was supported by the observation that dissolved zinc was essentially conservative in the free-flowing reach of the South Fork of the CDA River between the Bunker Hill Superfund Site and Cataldo despite significant dilution and a slight increase in pH. The release of zinc from particulates in acidic groundwaters and its limited re-adsorption when acidic groundwaters are neutralized (see Chapter 6) is one justification for the high priority given to remediating acid mine drainage from the upper basin.

Barton (2002) calculated metal inputs to the South Fork of the CDA River during summer and fall (July, September, and October 1999) when the river was gaining groundwater. He found that dissolved zinc loads from tributaries were less than 10 percent of the dissolved zinc entering the river; hence, groundwater dominated zinc inputs. There was not a concomitant gain of dissolved lead from groundwater, probably because lead was retained in the hyporheic zone by the precipitating oxides as described above. Zinsser (2019) confirmed that groundwater inputs remain the main source of loading of zinc and cadmium to the CDA River during periods of low flow.

THE SUPERFUND REMEDY AND ITS EFFECTS ON METAL INPUTS

This section describes the remedy at the Superfund site because of the likely significant impacts of the remedy on both metals and nutrient loading to CDA Lake. In general, the chosen remedies have attempted to control sources of contamination in the watershed and/or reduce exposures of humans and wildlife to contaminated material in the river basin. This might suggest that metal fluxes to CDA Lake would have declined over the entire period of remedy implementation; indeed, as shown later in this chapter, lead inputs to CDA Lake have been declining for about the past ten years, but cadmium and zinc inputs to CDA Lake have been declining for at least 30 years. Nonetheless, the water quality impacts of remedy implementation have been highly variable, with periods of both increasing and decreasing fluxes of lead into the Lake from the watershed. This narrative discusses the timing of specific remedial activities, including an analysis of land use changes coincident with the implementation of the remedy. Remedial activities are then revisited later in this chapter’s quantitative analysis of metals and phosphorus loading to the Lake, to further examine what impacts the Superfund cleanup has had, and will continue to have, on the quality of water entering CDA Lake via the CDA River.

Operable Units and Remediation Activities

Operable Units 1 and 2

The Superfund remedy for the Bunker Hill site has been divided both in space and time into three distinct operable units (OUs). As described in Chapter 1, OU-1 and OU-2 are encompassed within the 3 × 7 mile Box, from the vicinity of Kellogg on the eastern end to Pinehurst on the western end (Figure 3-2). Remedial actions

began in 1986 when public areas in OU-1 and OU-2 were targeted for “fast-track” cleanup. Active remediation within OU-1 primarily focused on managing lead-contaminated soil in residential areas.

OU-1 remedies were designed to ensure that less than 5 percent of children have blood lead levels of 10 micrograms per deciliter (μg/dL) or greater; and that less than 1 percent of children have blood lead levels of 15 μg/dL or greater. Remediation was achieved through a number of actions, including removing soil to a depth of 12 inches (30 cm) that had greater than 1,000 milligrams of lead per kilogram soil (mg/kg), achieving a geometric mean yard soil lead concentration of less than 350 mg/kg for each residential community, and installing protective soil/vegetation barriers, culverts, and retaining walls to prevent runoff from re-contaminating remediated properties. Because removal of soil was conducted to a depth of 12 inches, contaminated material likely remains at greater depths.

The 1992 Record of Decision (ROD) for OU-2 also addressed non-populated areas, commercial areas, and other common-use areas including (1) the former industrial complex and Mine Operations Area in Kellogg; (2) the Smelterville Flats (the floodplain of the South Fork of the CDA River in the western half of OU-2); (3) hillsides, creeks, and gulches; (4) the CIA; and (5) the Bunker Hill Mine and associated acid mine drainage (EPA Region 10, 2015). Major surface remedial activities included demolition of the smelter and fertilizer plants (1995–1998); removal of contaminated material from hillsides and gulches and placement in constructed repositories (1995–2000); removal of contaminated material from the Smelterville Flats (1995–2005); and revegetation,

stabilization of the land area, and creek reconstruction in the gulches and on the hillsides after removal of contaminated materials.

Remedial actions related to surface contamination in OU-1 and OU-2 were largely complete by 2008, with institutional controls allowing site use and reuse. The Central Treatment Plant (CTP), originally constructed in 1974 to treat acid mine drainage from the Bunker Hill Mine, has been upgraded to expand capacity and improve metals removal from groundwater over time. Most recently (2018–2020), a slurry wall and extraction wells have been installed to capture contaminated groundwater that was reaching the river, and this water is now being treated at the CTP (see Box 3-1 for details).

Operable Unit 3

OU-3 extends over a much larger land area (Figure 3-2) and includes the upper and lower basins of the CDA River. In the upper basin portion of OU-3, remedial actions are occurring, while for the lower basin portion, investigations and feasibility studies are ongoing, with pilot actions forthcoming. Remediation in the lower basin is complicated by the fact that metal contamination is present in the lower river beds, river banks, and floodplains, and in the lateral lakes; flooding continues to redistribute contaminants across the region; and groundwater contamination is widespread.

The ecological risk assessment (CH2M Hill and URS Corp, 2001) identified nine chemicals of potential concern in surface water, including antimony, arsenic, cadmium, copper, lead, manganese, mercury, silver, and zinc. Lead and zinc raised the greatest level of concern to EPA and were felt to serve as models for two distinct transport mechanisms. That is, lead is typically present in particulate form, whereas zinc is more likely to be present in dissolved form, except during floods. Across the upper basin, significant deposits of tailings (millions of tons) had built up on floodplains above the dams constructed in the early 1900s. Poor maintenance and flooding not only transported these contaminated materials downstream to the lower basin, but significant amounts of tailings remained behind the dams. There was no attempt to characterize groundwater sources of contamination, which is of concern since much of the dissolved zinc load that enters the CDA River during low discharge derives from groundwater.

The ROD for OU-3, issued in 2002, focuses on a large area that extends from near the Idaho-Montana border west through the Idaho Panhandle into the State of Washington. It includes communities, floodplains, rivers, tributaries and lakes, as well as Pine Creek and the portion of the South Fork of the CDA River within the Bunker Hill Box. OU-3 also includes areas where mine wastes were used for road building, fill, or construction. The ROD established a 30-year interim remedial plan that included final cleanup for human health exposure to residential/community soils and interim cleanup for ecological protection based on benchmarks. It is important to note that the ROD does not meet protectiveness standards or applicable or relevant appropriate requirements, it did not include groundwater or CDA Lake, and it did not allow for practice of tribal or subsistence lifestyles.

In 2012 an interim ROD amendment (EPA, 2012) was created to clarify and modify water collection and treatment actions from the initial RODs for OU-2 and OU-3. Relevant to OU-3, the amendment included consolidation/isolation source control at upper basin mine and mill sites, construction and maintenance of repositories, remediation of road surfaces, protection of remedies from erosion and recontamination, and groundwater remediation in Ninemile and Canyon Creek watersheds. These remediation actions are complex, involving multiple actions over many years. For example, 19 mine and mill sites in Ninemile Creek have seen remedial activity that commenced in 1992 and continues today (see Box 3-2).

Future Remedial Activities

The remedial efforts associated with OU-3 have largely been implemented in the upper basin, and many remediation challenges remain, particularly along the South Fork of the CDA River, which provides 90 percent of the metals input to the main stem of the CDA River. Zinc has been the primary driver of cleanup activities in this area, and remediation activities to address zinc in the Ninemile and Canyon Creek watersheds have been extensive. Based on data collected at Harrison and analyzed by the committee, the combination of improved treatment at the CTP and remediation of major zinc sources in the South Fork of the CDA River has lowered zinc inputs to the Lake by about 45 percent, a reduction of 384 MT/yr. Note that the reduction simply due to the removal of zinc by the CTP has been approximately 116 MT/yr (see Box 3-1). The effect of remedial activities on lead, which is particle-associated, is more difficult to discern, but analysis of data from the U.S. Geological Survey (USGS) and by the committee (vide infra analysis) indicate that removal of materials during remediation may have either exposed materials with higher lead concentrations or led to other disturbances of the river and floodplain of the lower CDA River that temporarily increased total lead fluxes into the CDA River from 2005 to 2010, but that fluxes have been decreasing since. This information will be important to consider as remedial activities are undertaken in the lower basin.

Compared to the upper basin, more limited remedial activities have been conducted in the lower basin. In addition, the lower basin has different and more complex hydrology, including large areas of wetlands and floodplains as well as river bed materials that are highly enriched in metals that present additional remediation challenges. Slowing of river flow at mile 160 near Cataldo causes much of the sediment that enters the lower river to deposit in the lower basin. Although the levels of metals entering this stretch of river have been reduced by remedial activities in the upper basin (particularly in the Box), the now contaminated river sediments, poorly stabilized river banks, and beaches in this stretch represent exposure risks and a continuing source of metals to CDA Lake because contaminated sediments are continuously transported downstream. EPA data (Prestbo, 2021a) and the committee’s analysis of lead data later in this chapter indicate that the main stem of the CDA River downstream of Cataldo is a lead source, likely due to mobilization/transport of riverbed sediments, with < 15 percent of the load being derived from bank erosion. In addition to the river itself, the lower basin covers 18,000 acres (7,284 hectares), much of which has been contaminated during flooding events. Remediation challenges in this area include complex hydrology due to the fluctuation of lake levels, changing sediment supplies and streamflow conditions, limited access, and multiple stakeholders. The lower basin has large areas of waterfowl habitat that are contaminated above acute (80 percent of area) or chronic (95 percent of area) levels for lead (EPA, 2020).

Ongoing remedial efforts in the lower basin are shown in Box 3-3, with a focus on lead contamination and potential human and wildlife exposure risk. Efforts near the Cataldo Mission and recreation area from 1995 to 2002 focused on removal of contaminated tailings, erosion control, and capping. In 2003–2004, remediation near the Rose Lake boat launch occurred (Moreen, 2021b; EPA, 2010), which included bank stabilization to prevent erosion

and capping to limit exposure of recreational users. Construction of the Trail of the Coeur d’Alenes (2000–2004) provided both capping of contaminated material and recreational opportunities (see Figure 3-3 for a map showing a portion of the trail and the river reach from Cataldo to Rose Lake). Recreation sites continue to be a top priority in the lower basin, with a combination of cleanup, communication, and access restrictions being used to minimize exposure, but these remedies are not aimed at reducing inputs to the rivers or the Lake.

Due to the complexity of remediating contaminated wetlands and floodplains that serve as waterfowl habitat, EPA has begun projects to convert uncontaminated agricultural land to wetlands to provide more desirable, clean habitat for birds in the region (e.g., Schlepp project 2007–2014; 2015 Robinson Creek project; Moreen, 2021b). Although this does not remediate contaminated areas, it does reduce wildlife exposure to contaminants by diverting the birds to cleaner sites that have been excavated and capped and have low risk of recontamination during flooding. Future lower basin remedial activities will include source control, additional agriculture-to-wetland conversion activities, and actual wetland remediation, such as soil treatment to reduce lead bioavailability (Moreen, 2021b).

As in OU-1, OU-2, and the upper basin of OU-3, property remediation has been occurring in the lower basin. Priorities are to control sources in and out of the river, manage exposure risks, and prevent floodplain (re)contamination. Some remedies are identical to those elsewhere, such as excavation and replacement of soil in residential areas (850 properties since 2019; EPA, 2020) along with remedies to protect sites from recontamination.

River, Riverbank, and Floodplain Sources of Contaminants

As previously mentioned, a major challenge in the lower basin is remediation of the large amount of lead contamination in floodplains, river banks, and river bed sediments. Sediment cores collected in transects of the CDA River in 2012 (CH2M Hill, 2015, 2017; Maul Foster and Alongi, 2020) show typical concentrations of thousands of mg/kg of lead in the river sediment, although this does vary, with some stretches of the river showing much lower levels (tens of mg/kg). For a particular transect, levels are usually relatively consistent, but variation

does occur with depth. For the ~3 m cores collected in 2012, the largest concentrations were found in the deeper portions of the core, potentially indicating deposition of less contaminated sediment from either upstream or the floodplain during more recent times (although the flood plains/banks of the river are also contaminated). In other stretches, the most contaminated material is at the shallow depth. The 2019 cores were only ~0.66 m in depth and showed comparable lead and zinc levels that did not vary with depth. Over a 1-mile stretch, sediment concentrations could vary from tens of mg/kg to > 25,000 mg/kg for both lead and zinc.

These varying levels reflect the complicated bathymetry of the system, with scour and deposition affecting metals concentrations in different stretches as water flows and flooding transports materials. Laboratory erosion experiments generally showed coarser, non-cohesive surface sediment but more cohesive, fine material at depth (CH2M Hill, 2015, 2017; Maul Foster and Alongi, 2020). The depth of the erodible layer varied from only the top 5 cm of the core to 20 cm. Erosion pins that were installed as part of the 2019 sampling and monitoring have the potential to provide valuable information regarding sediment transport and deposition.

Remedies in the lower basin are not without risk of recontamination. For the contaminated river banks and beds, dredging would remove the metals from the system, but this plan also risks resuspension of contaminated material. Capping has a lower risk of release during implementation, but must be robust enough to ensure the contaminated sediments are not re-exposed in the future during flooding or changing system hydrodynamics.

A field experiment at Lane Marsh (see Figure 3-3B) used thin layer capping, a method designed to mimic the deposition of clean sediment in multiple applications to allow vegetation to recover. Results showed that a thickness of 6 inches (15 cm) reduced lead exposure, and vegetation recovered within two years. Active capping using biochar, which is also being tested in the lower basin, could increase the sorptive capacity of the capping material (Knox et al., 2014; Wang et al., 2018). As noted in Lane Marsh, transport of contaminated material to a capped site from an upstream or neighboring untreated area may re-contaminate a treated area, and this could be exacerbated during a flooding event in which contaminated materials are more widely dispersed. It is also necessary to account for the connection between the river and the flood plains in remediation strategies. If remediation were to focus primarily on highly contaminated river bed sediments, the relative contribution of the floodplains may increase as they continue to replenish the bed sediments, unless remediation also isolates those sediments from the river or reduces major flood frequencies.

The 2020 Optimization Report for OU-3 (EPA Region 10, 2020) states that “there is considerable uncertainty in evaluating discharge of metals from sediments both spatially and temporally” and it emphasized that a conceptual site model is critical to implementing remedial activities given the amount of lead present in the floodplains and riverbanks/riverbed that can be transported to the Lake. A conceptual site model that includes detailed hydrodynamics is being developed by EPA (Prestbo, 2021a; EPA, 2020). Any models or analysis of the lower CDA River must give full consideration to the evolving nature of the channel and floodplain system as it continues to respond to the changing rates of sediment input and metals input from both the South Fork and North Fork CDA watersheds.

Future Remedy Priorities

EPA’s goal is to adaptively manage remedial activities in the lower basin, with a focus on prevention of sediment (and thus lead) transport from the lower basin of the CDA River to the Lake (EPA, 2020). In 2017–2018, multiobjective decision analysis was used to prioritize and select remedies and projects for the lower basin (Prestbo, 2021b). Overall, wetland restoration and capping scored high in this process. Current projects in the planning phase include Gray’s Meadow restoration (Figure 3-3B) to reduce metals, improve water quality, and provide wetland habitat, with construction likely to begin in 2022. In Lane Marsh, research on thin layer capping continues. Pilot testing of biochar amendments to limit lead bioavailability in wetlands is also occurring. Near Harrison, dredging and capping of the riverbanks is the identified remedy, with characterization set to begin in 2022 and construction planned for 2026. The Dudley reach of the CDA River (Figure 3-3B) is highly contaminated and hydrologically challenging (deep channel), such that new technologies are likely to be needed to limit downstream transport of metals. Pilot implementation of dredging/capping scenarios is planned for 2024 or 2025.³ Waste consolidation areas for these efforts are currently being sited.

Assessment of Land Use Changes Coincident with Remedy Implementation

As mentioned in Chapter 2, the USGS has been measuring metal concentrations and fluxes within the upper and lower CDA basin as part of the Basin Environmental Monitoring Program to evaluate the long-term effects of the Bunker Hill Superfund site remediation. These data are further analyzed by the committee (later in this chapter)

___________________

³https://cumulis.epa.gov/supercpad/SiteProfiles/index.cfm?fuseaction=second.stayup&id=1000195

to better understand the potential effects of remedial actions not only on metals, but also on sediment and phosphorus. As a complement, the committee analyzed land cover information to determine if it could reveal changes indicative of Superfund remedy implementation. This analysis made use of the National Land Cover Database (NLCD),⁴ a national dataset with information on land use and land cover for various time periods for all areas of the United States. Each grid cell represents a particular land cover category and was derived from classification algorithms that processed Landsat satellite imagery. There are NLCD datasets available that represent 2001, 2004, 2006, 2008, 2011, 2013, and 2016 and 2019; these datasets all use the same land use descriptions and codes and can be compared directly. In addition, a 1992 NLCD dataset is available, but the land-use classification system was different than those beginning in 2001. Finally, there is a historical dataset, compiled and published by USGS, that provides land use and land cover from the 1970s and 1980s. This dataset is in a different format (polygons versus raster) and uses a categorization scheme different from either the 1992 NLCD data or the 2001–2019 data. To compare land use in OU-1 and OU-2 over time, it was necessary to reclassify the earlier datasets. Once the categories were realigned to the 2001–2019 classification system, the CDA watersheds that encompass OU-1 and OU-2 were isolated and the land use areas were determined using ARCGIS Pro. Comparison of the land cover in these areas is presented comprehensively in Table 3-1 and Figures 3-4A, B, and C for years 1992, 2001, and 2019.

The data in Table 3-1 highlight several key trends in land use in and around OU-1 and OU-2, although caution must be exercised in comparing recent data to pre-2001 data and especially to the historic data for 1970–1985 due to changes in methods and instrumentation. First, barren land represented a major portion of the land use prior to major remediation efforts that started in 1992, which is not surprising since barren land is a common characteristic of floodplains and forested lands severely impacted by mine wastes (Moore and Luoma, 1990). The acreage of barren lands has been reduced dramatically since the 1970–1985 time period, indicating the success of landscape revegetation that was part of the Superfund activities. Second, throughout the 1970–1985 period, forest land was the dominant land use in the area. This is in large part because the areas within these watersheds extend beyond the mining impacted areas in OU-1 and OU-2. However, there has been an increase in the amount of evergreen forest since 2001, which has coincided with a decrease in shrubland and herbaceous land cover. In other words, since 2001 acreage of evergreen forests has increased and acreage of shrubs has decreased, but the total of the two has remained relatively constant. Third, developed lands (the sum of all four developed categories) have been increasing since 1992 from 7.29 to 9.48 km². Finally, acreage of emergent wetlands appears to have increased over the past decade. These trends are consistent with expectations of improved landscape within an area of ongoing remediation.

SOURCES OF PHOSPHORUS TO COEUR D’ ALENE LAKE

Beyond metals, the other major input of concern from the CDA basin to CDA Lake is that of nutrients that might increase in lake productivity, which could enhance anoxia in the hypolimnion and result in the possible release of metals from lake bed sediments. Nutrient inputs to CDA Lake are the primary concerns of the 2009 Lake Management Plan (LMP; IDEQ and CDA Tribe, 2009) and have been the subject of investigation by both parties over the past decade. In particular, in 2020 IDEQ and the CDA Tribe produced the report, Coeur d’Alene Lake Management Program: Total Phosphorus Nutrient Inventory, 2004–2013 (IDEQ and Tribe, 2020), which includes an accounting of phosphorus sources and loads to the Lake, using monitoring data, estimates from modeling exercises, and previous reports to determine how the sources of phosphorus to CDA Lake had changed from 2004 to 2013. That report noted that nitrogen had not yet been subjected to the same analysis, although this was planned for the near future. The nitrogen to phosphorus ratios (such as TN/TP and SRP/DIN)⁵ for CDA Lake suggest that this system may be mainly phosphorus-limited, but that during the summer the system may trend toward co-limitation or even nitrogen limitation at times (discussed further in Chapter 5). The 2020 Phosphorus Inventory report was used extensively to inform the section below, which describes the major point and nonpoint sources of phosphorus to CDA Lake.

___________________

⁴https://www.usgs.gov/centers/eros/science/national-land-cover-database

⁵ TN = total nitrogen, TP = total phosphorus, SRP = soluble reactive phosphorus, DIN = dissolved inorganic nitrogen = NO₃ + NH_4.

TABLE 3-1 Distribution of Land Uses for the Sub-Watershed Areas Including and Surrounding OU-1 and OU-2

SOURCE: National Land Cover Database.

Phosphorus Point Sources

Point sources to the Lake include wastewater treatment plants (WWTPs) that discharge to the watershed, including the CTP that treats groundwater from the CIA. With regard to municipal wastewater, most of the more than 175,000 people in Kootenai County (see Chapter 1) are serviced by WWTPs of the cities of Coeur d’Alene and Hayden. The WWTPs for both of these communities have state-of-the-art advanced phosphorus removal, and their phosphorus discharge permits are extremely low (i.e., effluent total phosphorus < 100 μg/L relative to typical wastewater concentrations of 5,000–7,000 μg/L). But even more importantly, the effluents from these facilities are discharged to the Spokane River downstream of CDA Lake, so these plants (and the population they service) are not nutrient inputs to the Lake.

Table 3-2 shows the nine WWTPs that either directly discharge to the Lake (Plummer and Harrison WWTPs) or upstream of the Lake into the CDA River (Page, Mullan, and Smelterville WWTPs) and St. Joe River (St. Maries, Santa-Fernwood, Clarkia, and Emida WWTPs). Of these WWTPs, the Plummer facility employs advanced phosphorus removal and has average effluent total phosphorus concentrations of ≈ 80 μg/L. The other facilities are conventional primary/secondary systems and have average effluent total phosphorus concentrations of 1,500–2,000 μg/L. The overall phosphorus export from these facilities averaged 8.3 MT/yr for the 2009–2017 period (IDEQ and CDA Tribe, 2020). Compared to the total phosphorus inputs from the CDA and St. Joe Rivers (about 133 MT/yr; see later analysis), smaller tributaries to the Lake, and direct atmospheric deposition onto the Lake, these WWTP discharges only account for 6 percent of the annual total phosphorus budget.

The phosphorus in conventional municipal wastewater discharges generally has very high bioavailability (≈ 80 percent; Li and Brett, 2012), whereas the particulate phosphorus draining forested catchments that probably dominates CDA Lake inputs has been shown to have low bioavailability (< 20 percent) (Ellis and Stanford, 1988; Ekholm and Krogerus, 2003). When taking phosphorus bioavailability into account, the discharges from these WWTPs could have a more substantial impact on primary production in CDA Lake than 5 percent suggests (i.e., WWTPs account for 23 percent of the bioavailable phosphorus loading to CDA Lake). Advanced nutrient removal processes in WWTPs can achieve 95 percent phosphorus removal using biological treatment (i.e., effluent total

phosphorus concentrations ≈ 250 μg/L) and 99 percent removal using iron or aluminum based chemical treatment (i.e., effluent total phosphorus concentrations ≈ 50 μg/L) (Li and Brett, 2012, 2015). Given that all of the WWTPs that discharge to the Spokane River downstream of CDA Lake already implement advanced phosphorus removal, consideration of this technology for the WWTPs that discharge upstream of CDA Lake could reduce concerns about these phosphorus inputs into the future.

Phosphorus from the Central Impoundment Area

It has been estimated that the Bunker Hill CIA contributes 5–15 tons⁶ of TP/yr to the CDA River (Clark and Mebane, 2014; IDEQ and CDA Tribe, 2020). This facility stores gypsum waste from a former phosphoric acid fertilizer plant. Gypsum is a common soil amendment that is often used to reduce phosphorus runoff from over-fertilized soils

___________________

⁶ All the load values in the Phosphorus Inventory report are in US short tons.

(Watts and Torbert, 2009). Because gypsum is calcium sulfate (CaSO₄·2H₂O), which easily solubilizes (K_sp = 10^−4.58), dissolved calcium concentrations in these deposits would likely be very high, which should result in the formation of calcium phosphate minerals. These have been shown to have widely varying bioavailability depending on which specific mineral complexes form (Li and Brett, 2013). For example, the phosphorus in calcium phosphate (CaHPO₄) has been shown to be ≈ 100 percent bioavailable, whereas the phosphorus in apatite [Ca₅(PO₄)₃(OH,F,Cl)] had zero bioavailability (Li and Brett, 2013). Other mineral complexes, such as tricalcium phosphate [Ca₃(PO₄)₂] probably have intermediate bioavailability depending on the specific environmental conditions (Deubel and Merbach, 2005). The phosphorus released from this facility could have had a substantial impact on productivity in CDA Lake, but this depends on what types of calcium phosphate minerals are formed.

Superfund remediation activities have targeted treatment of the gypsum pond wastes in the CIA. Indeed, EPA has completed construction of a groundwater cut-off wall and groundwater collection system for the CIA that will route the retrieved groundwater through the CTP (see Box 3-1). Although the CTP was designed to capture

metals being released from the Bunker Hill complex, it is also capturing phosphorus, and the recent upgrades are expected to result in substantial reductions in total phosphorus loads, at least during times of low to moderate river discharge. Indeed, Moreen (2021a,b) presented data suggesting that the upgraded CTP was removing more than 98.4–98.8 percent of total phosphorus with an effluent discharge loading of less than 0.045 kg/day (or about 0.016 MT/yr). (For perspective, the analysis later in this chapter concludes that average fluxes of total phosphorus from the CDA River are currently about 67 MT/yr.)

Nonpoint Sources of Phosphorus

The major nonpoint sources of phosphorus in the CDA basin include stormwater-mediated erosion of soils and sediments, agriculture and forestry, nonmunicipal community wastewater treatment systems and septic systems, and atmospheric deposition, but data are not available to determine the relative importance of these various sources.

TABLE 3-2 Permitted Municipal Wastewater Treatment Plants in the CDA Basin

¹ From Woods and Beckwith (1997).

² From average over all years within the period of record to 2 significant figures, unless otherwise noted.

³ Plummer data are for 2014–2017, after plant upgrades and process testing were completed.

⁴ Assumed total phosphorus concentration, based on average from other sites to 1 significant figure.

⁵ Average monthly data not available. Estimate based on monthly maximum flow.

SOURCE: Table 12 from IDEQ and CDA Tribe (2020).

An important nonpoint source of phosphorus to CDA Lake is soil erosion that mobilizes particulate phosphorus during storm events. This phosphorus export during erosion is probably anthropogenically enhanced in some places in the watershed (e.g., due to forestry activities and unpaved roads), but the extent to which erosion is enhanced above natural background levels is not currently known. Although some of this phosphorus could be derived from legacy mine tailings that have accumulated in the lower basin, the analysis presented later in this chapter shows that mine-affected watersheds deliver a similar amount of total phosphorus to the Lake as the watershed that is largely free of historic mining activity (total phosphorus loadings for the CDA and St. Joe Rivers are 67 and 66 MT/yr, respectively). Thus, forests, which are the dominant land cover throughout the CDA basin, and hence forestry practices, are probably the largest determinant of phosphorus loss from the watersheds. Agriculture received scant mention in the CDA Lake Phosphorus Inventory report because it is a small percentage of the land use/land cover (0.55 percent from Figures 1-19 and 1-20). Although it is possible that agriculture (other than forestry) plays a small role in the overall export of phosphorus to CDA Lake, there are insufficient data to evaluate its role. The uncertainty about the role of agriculture as a source of nutrients should be addressed in future reports.

The population in the CDA basin that is not serviced by municipal WWTPs is either on septic systems or connected to small, non-municipal community wastewater systems, the latter of which often discharge to a lagoon, drain field, soil adsorption system, or some other type of land application system. The Phosphorus Inventory report catalogued 17 such small community systems in the vicinity (less than 500 ft) of the shores of CDA Lake and concluded that they discharged less than 0.1 tons/yr of TP to CDA Lake. These small community wastewater systems can be found on the CDA Story Map.⁷

Septic systems are generally considered a “low-tech” approach to wastewater treatment. However, the high affinity of phosphate for iron, aluminum, and calcium minerals in soils means that phosphorus export from these systems to surface waters can be lower than for conventional primary/secondary WWTPs, which typically only remove ≈ 50 percent of the phosphorus during treatment and usually discharge directly to rivers or lakes. The contribution of on-site septic systems in the immediate vicinity of CDA Lake to overall phosphorus loading is difficult to estimate with available information. The Phosphorus Inventory report concluded that in 2008 there were 3,969 septic systems less than 1 mile from the Lake in Kootenai County, accounting for 1.3–1.5 tons P/yr. This calculation assumed that a three-bedroom home (with three to four people) generates wastewater with a total phosphorus content of 9 mg/L and flow of 250 gal/day (946 L/day), which equates to a load of about 1.04 kg P/person-year. One review indicates that humans discharge an average of 0.61 kg P/person-year associated with their feces and urine (Rose et al., 2015). The Phosphorus Inventory report assumed 90 percent soil retention, which is supported by Robertson et al. (2019). The report acknowledged that there would be population growth in the lakeshore areas beyond 2008; indeed Table 1-4 in Chapter 1 lists 35,599 people in 2021 in the nearshore watersheds. Hence, there is considerable uncertainty about the phosphorus loading from septic sources calculated in the Phosphorus Inventory Report. Furthermore, IDEQ does not keep records on the proportion of the population within the CDA basin connected to WWTPs versus septic systems, although the CDA Story Map shows the region’s structures including nearshore properties, population by census blocks, and the boundaries of city limits and small community wastewater systems. A significant proportion of residents in some cities (e.g., St. Maries, South Fork Sewer District) live outside city limits but are served by sewer systems.

Direct atmospheric inputs of phosphorus to CDA Lake are estimated to be 6 MT/yr. This estimate is based only on loading to the surface of CDA Lake, because atmospheric loading to the watershed is accounted for in the tributary estimates. Rather than being based on direct measurements at CDA Lake, these estimates are based on studies done at Flathead Lake by Ellis et al. (2015) and at Fernan Lake by LaCroix (2015), the latter of which is within the watershed and close to CDA Lake, and a global compilation of atmospheric phosphorus deposition estimates (Tipping et al., 2014). The phosphorus deposition estimate used for CDA Lake (i.e., 30 ± 34 kg/km²·yr) is similar to the average estimate for all of North America (42 ± 39 kg/km²·yr) from Tipping et al. (2014).

___________________

⁷https://storymaps.arcgis.com/stories/41e74951a3224d6aad6523b10a9bff09

Superfund and Lake Management Plan Activities on the Landscape That Affect Phosphorus Loading

Over the past three decades there have been a variety of Superfund activities as well as activities done in response to the LMP (outside the purview of Superfund) that may have altered phosphorus loading from the watershed. For example, revegetation of areas denuded by mining waste is often accomplished by hydroseeding to accelerate the establishment of a vegetated land cover. The mulch used for hydroseeding likely has a high concentration of soluble phosphorus, and additional phosphorus in the form of fertilizers may be applied. EPA has estimated that Superfund-related revegetation efforts were associated with total phosphorus inputs of 14 MT in the lower basin (1997–2021), 35 MT in the Box (1997–2005), and 10.8 MT in the upper basin (1997–2021) (personal communication, Ed Moreen, EPA, late 2021). Expressed as a total mass, this is about 60 MT over the 25-year period (1997–2021). It is unlikely, however, that all of this phosphorus was mobilized and transported to the river because the fertilization process is designed to retain the phosphorus on the soil so it can stimulate plant growth. Fertilization for forestry purposes is also a possibility, but the extent and intensity of such activities is not currently known. The analysis of the total phosphorus flux for the CDA River near Harrison (shown in Figure 3-23, later in this chapter) indicates a total of about 1,880 MT for these years (assuming that the 2021 value is the same as the 2020 value). Even if all of the Superfund-applied phosphorus were delivered to the Lake (an extreme assumption) it would constitute only about 3 percent of the total flux from the CDA watershed to the Lake (60 MT/1,880 MT).

There are also numerous restoration activities that occur as part of the LMP and/or are facilitated by the Basin Environmental Improvement Project Commission that might affect phosphorus loading to CDA Lake, but there is no quantification of the phosphorus load reduction from these efforts. Examples include (1) purchase of more than 2,200 acres (890 hectares) by the CDA Tribe for fish and wildlife habitat restoration and enhancement of streams for in-stream habitat; (2) advertisement of low-interest loans for purchase of direct-seed (no till) equipment for farmers; (3) shoreline protection on the St. Joe River (~8.5 miles or 13.7 km) and the CDA River (~11 miles or 18 km) via National Resources Conservation Service and Soil and Water Conservation district programs; (4) stabilization of eroding riverbanks and creeks; and (5) construction of new WWTPs in the city of Plummer, in Heyburn State Park, and for newer developments in and near the city of Coeur d’Alene, as well as upgrades to existing WWTPs (Brunner, 2017). The Restoration Partnership⁸ also lists efforts undertaken by a variety of partners (including EPA) to return natural resources to a healthy condition in the CDA basin.

Phosphorus loading data from Zinsser (2020) and presented later in this chapter show extended periods of increased phosphorus fluxes from the CDA River followed by a period of decreased phosphorus fluxes during the 2010s. In addition, the committee’s analysis of trends in total phosphorus in rivers not significantly influenced by mining or remediation also show decreases in recent years, suggesting other drivers of total phosphorus change such as land use, wildfire, or climate. It is possible that some remediation activities could lead to total phosphorus release due to pH changes or that total phosphorus could be affected by broader changes such as revegetation associated with the much-improved air quality in the mined area.

Given the LMP’s emphasis on reducing phosphorus loadings, a focused effort needs to be carried out on an ongoing basis to record all activities in the CDA Lake watershed that are expected to influence phosphorus loads (positively or negatively). Some type of model (which could be very simple) should be applied so that estimates of the cumulative phosphorus management activities can be evaluated over time and then compared with observed changes in phosphorus loading (as measured at river monitoring stations). Discrepancies between these estimated and observed changes could help to build better models for predicting how management actions (or climate change) may influence phosphorus loadings in the future. Such predictive models are central to an adaptive management for controlling phosphorus loadings going forward.

___________________

⁸restorationpartnership.org

ANALYSIS OF INPUTS TO COEUR D’ALENE LAKE

The following quantitative analyses investigate inputs of water, lead, zinc, cadmium, sediment, and phosphorus into CDA Lake and throughout the CDA watershed over the past 30 years. Using river data collected by the USGS, the analyses paint a general picture of improving water quality conditions in the rivers and lowering of contaminant loads to CDA Lake.

Water

The water inputs to CDA Lake are dominated by the two major tributaries: the St. Joe River (drainage area 4,788 km²) and the CDA River (drainage area 3,797 km²). (Minor inputs to the Lake include direct precipitation on the Lake, which is about 1 percent of the river inputs, and groundwater inputs, which have not been estimated.) The total area of all tributaries to CDA Lake is 9,474 km² (Wise, 2021). This means that 51 percent of the total drainage area of the Lake is in the St. Joe River watershed and 40 percent is in the CDA watershed. The remaining 9 percent is in other small, unmonitored watersheds surrounding the Lake.

For purposes of this analysis of inflows to the Lake, long-term discharge records from the two major rivers were used to create a single time series of estimated inflows to the entire Lake. The two records used are the CDA River at Cataldo (USGS gage 12413500) and St. Joe River at Calder (USGS gage 12414500). The period of record used in this section is water years 1987–2020 (October 1, 1986–September 30, 2020). For the analysis described here, the discharge at each of these two streamgages is multiplied by the ratio of the total drainage area of that tributary to the drainage area at the streamgage location (for the St. Joe River at Calder that ratio is 1.8034, for the CDA River at Cataldo it is 1.2135), and the two tributary values are summed. Then, to extrapolate to the ungaged portions of the watershed, these values are multiplied by the ratio of the total watershed area of the Lake to the total watershed areas of these two tributaries (i.e., 1.1036).

Based on the discharge record for water years 1987–2020, the mean daily discharge into the Lake is 209 m³/s, and the total range of daily values is 15–3,842 m³/s. There is a strong seasonality to the record, with the highest discharges (based on median monthly values) in May and the lowest in September. The distribution of daily discharge values over the entire period is shown in Figure 3-5. This figure shows that the median discharge for May is 518 m³/s and the median discharge for September is 35 m³/s. This pattern comes about as a result of the timing of precipitation (higher values in the winter and spring), the timing of snowmelt, and the influence of high rates of evapotranspiration in the summer and early fall. The cause of the highest discharge events are typically rain-on-snow events. With the warming conditions now taking place in the region (see Chapters 1 and 10), it is

reasonable to expect that the potential for very high discharges from rain-on-snow events will change and that the peak annual discharges will come earlier in the year.

The dataset can be explored for trends over time. For the annual mean discharge, using a Mann-Kendall test for trend (Mann, 1945), the slope computed is +0.17%/yr but the p-value⁹ for the trend is 0.81 (meaning that evidence for the trend is very weak). The slope was computed using a Thiel-Sen slope estimate (Helsel et al., 2020) based on the logarithms of the discharge values. The annual mean values are shown in Figure 3-6A. Trends in low-flow, as characterized by the annual seven-day minimum discharge, were also not significant. They are shown in Figure 3-6B, which indicates a trend slope of −0.21 percent/yr with a p-value of 0.49. Finally, the records were examined for trends in the annual maximum daily discharge. This is shown in Figure 3-6C, which shows a trend slope of +0.13 percent/yr with a p-value of 0.93. There is no significant trend to these measures of discharge over the period 1986–2020. There are slight, but insignificant, indications of increases in average and high flows and decreases in low flows.

One other type of change that is commonly considered in discussions of streamflow trends, particularly in watersheds where there is a significant amount of seasonal snow pack, is a measure of changes in the timing of annual runoff. Here, each water year was evaluated, and the date on which half of the total water-year volume had entered the Lake was determined; these dates range from mid-March to late-May. The relationship between this half-volume date and year is shown in Figure 3-6D. Overall, there is no significant trend in the half-volume date (Mann-Kendall p-value is 0.38), and this nonsignificant trend has a positive slope (the date tends to become later each year). Climate change projections would suggest that this trend may become negative at some point in the future (a tendency toward earlier runoff), but at the present time this is not what the observations are indicating. There does appear to be a shift in the direction of earlier runoff starting in about 2008, but the signal is not conclusive.

Overall, none of the major features of the daily discharge pattern of the river inputs to the Lake has changed over the period 1987–2020 in a manner that can be considered statistically significant. It is plausible that climate change and land use change will result in changes in the future, but at this point such changes are not apparent.

Metals and Sediment

To assess the input of metals from the CDA watershed to the Lake, it is necessary to have a model that estimates contaminant transport at key locations in the river/lake system. The model could be purely statistical or some combination of mechanistic and statistical. What is crucial is that the model be able to take the existing set of concentration and discharge measurements made over the period of record and produce unbiased estimates of contaminant concentration and flux for all days in the period of record. Only by having an estimate of conditions on all days does it become possible to do mass balance calculations, define the spatial and temporal trends, understand the past drivers of the system, and build hypotheses about future concentrations and fluxes in the system.

The committee’s analysis uses the statistical method/model known as Weighted Regressions on Time, Discharge, and Season (WRTDS; Hirsch et al., 2010) to make inferences about the history of concentration and flux, on a daily time step, based on the types of records that are typically available in the rivers of the CDA Lake watershed (typically on the order of 250 observations for each of the key contaminants at a given monitoring location over nearly three decades). WRTDS has been used extensively for many river systems in the United States, including in this watershed (see Zinsser, 2020). More sophisticated methods could have been used if the dataset (sampling frequency) had been larger.

The committee’s analysis considered four monitoring locations, highlighted on the map in Figure 3-7. Figure 3-7 shows six sites, but two of them (the North Fork of the CDA River at Enaville and the St. Joe River at Ramsdell) have metal fluxes well below 1 percent of the values for the CDA River near Harrison. Hence, these two sites are only considered later in the chapter, in the results on phosphorus. The South Fork of the CDA River at Elizabeth Park lies just upstream of the Box, and the South Fork of the CDA River near Pinehurst lies just downstream of the Box. The CDA River near Harrison is the point at which the CDA River enters CDA Lake and it integrates the inflows from the South and North Forks of the CDA River. The flow at this location is often

___________________

⁹ The p-values reported for this trend test and the subsequent discharge trends discussed are all adjusted for the influence of long-term persistence of discharge at an annual time scale using the zyp package in R. The specific method for adjustment is the Zhang method of prewhitening (Zhang et al., 2000).

influenced by backwater from CDA Lake, which is regulated by the Post Falls Dam. The Spokane River near Post Falls is the regulated outflow from CDA Lake. As the dam only permits surface withdrawals, the material that is discharged would be expected to reflect the conditions in the epilimnion.

River water quality data appear highly chaotic. For example, the record of total lead values measured for the South Fork of the CDA River near Pinehurst is shown in Figure 3-8A (note the logarithmic scale with concentrations ranging over three orders of magnitude). Figure 3-8A suggests a downward concentration trend, but from the data alone, any sort of trend quantification would be highly uncertain. Another way to look at the same data is to consider how total lead varies with river discharge, shown in Figure 3-8B. This figure shows that total lead concentration is weakly related to discharge at low discharges (less than 10 m³/s), but at higher discharges, total lead concentrations increase substantially with discharge. In addition to these relationships, concentration varies with season, such that for any given discharge, the concentrations tend to be highest in the winter and lowest in the summer. The key to making meaningful interpretations of the water quality data is to build a statistical model that captures these relationships so that the apparently chaotic behavior can be quantified and understood.

The concept of the WRTDS model is to use statistical smoothing to partition the variations in concentration into components that are related to season of the year, watershed hydrologic condition (characterized by the daily mean discharge on the day of sample collection), long-term trend, and a random component (the unexplained portion of the variation). The WRTDS model fitted to these data is shown in Figure 3-8C, which is a contour plot showing the expected value of the total lead concentration for any combination of date and discharge throughout the period of record (regardless of whether that discharge was actually observed on that date). This WRTDS model can then be used to produce several types of outputs, including estimates of the concentration on any given date in the period of record. These daily values can be summarized by taking mean values of these estimates by month or by year. In addition, the model can be used to produce estimates of the expected number of days that concentration might have exceeded some threshold value in any given month or year.

Although the annual averages of concentration are somewhat informative about the long-term trends in lead in the river, these estimates are very strongly influenced by the particular pattern of discharge values that happened in each year. Years such as 2017 or 1996, which had very high discharge, tend to produce high values of mean concentration, and years of low discharge tend to produce lower mean concentrations. Estimates of trends ideally should not be strongly influenced by the specific pattern of wet and dry years. Rather, they should integrate across the seasonally specific frequency distribution of discharge values to produce estimates that are free of the impact of these interannual variations in discharge. The method for doing that is called flow-normalization (see Hirsch and DeCicco, 2015).

The flow-normalized annual mean concentrations are shown by the smooth curve shown in Figure 3-9A, along with the 90 percent confidence intervals around this estimate (based on a bootstrap calculation described by Hirsch et al., 2015). The confidence interval is very wide in the early years, because so much lead was moving in particulate form during high discharge events, and the relationship between discharge and concentration at these

higher discharges is subject to a high degree of variability. In later years, these high discharge events were much less important contributors to lead in the river; thus, the uncertainty becomes much smaller. The concentration trend at Pinehurst over the period of record, 1992–2020, is a decline in the mean of 54 μg/L (from 71 to 17 μg/L), or a 77 percent decrease.

In thinking about metrics of progress, it makes sense to consider not only the change in mean concentration but also the change in flux (often called “load”). The WRTDS model can also produce estimates of flow-normalized flux in a manner similar to the flow-normalized concentration; the random variable being considered on each day is the flux (the product of concentration and discharge) rather than just the concentration itself. The results for the South Fork of the CDA River at Pinehurst are shown in Figure 3-9B. The metric of change over the full record is a decline of 77 MT/yr (from 98 to 21 MT/yr), or a 79 percent decrease. Trend results for flow-normalized flux are highly relevant to this study because they are building blocks of mass-balance analyses for segments of the river system and the Lake, although the concentration results are useful when thinking about concentrations experienced by the biota (including humans) that interact with the river water on its way to the Lake.

In the sections below, trends in lead, sediment, cadmium, and zinc concentration and flux are assessed using the WRTDS approach for the four monitoring locations shown in Figure 3-7. The text and figures are designed to provide a spatiotemporal analysis of the trends in each of these metals and some interpretation of the results. The section on total lead, which includes Figures 3-8 and 3-9, is the longest because trends on total lead are evaluated to better understand the role of the Superfund remedy in reducing total lead inputs to CDA Lake. See Appendix A for an explanation of the methods.

Throughout the descriptions of trend results for concentrations and flux, “likelihood” terminology is used to describe the degree of statistical uncertainty of the results. The concept is described in Hirsch et al. (2015) and is based on a new perspective of reporting to replace the traditional “p-value” approach for evaluating water quality trends (see McBride et al., 2014; McBride, 2019). The change between the starting year and ending year of the period being analyzed are evaluated using a block bootstrap procedure (see Hirsch et al., 2015), and the result is expressed as the likelihood that the true trend has the same sign as the estimated trend. For example, in Table 3-3, the dark blue shading indicates that the likelihood that the change from the start of the trend period to the end is indeed negative, is greater than 95 percent, and this is called a “highly likely downward trend.” The full set of likelihood categories used is shown below:

Total Lead

Using the same approach described above for the South Fork of the CDA River near Pinehurst, Idaho, additional evaluations of total lead trends were conducted at the South Fork of the CDA River at Elizabeth Park, Idaho; the CDA River near Harrison, Idaho; and the Spokane River below the Lake outlet. The major results for all four of these sites are shown in Table 3-3 and Figure 3-10.

Throughout the record, the flow-normalized flux at the downstream end of the South Fork (Pinehurst) is always greater than at the upstream site (Elizabeth Park). Furthermore, although fluxes have declined at Elizabeth Park, the rate of decrease, either in mass terms or percentage terms, is less steep than at Pinehurst. This is what one would expect given the high intensity of the remediation effort within the Box. The results show that, if there is to be more decrease in total lead flux from the South Fork CDA watershed, it will need to come almost entirely as a result of future mitigation measures above Elizabeth Park versus any additional remediation done in the Box.

Another observation (data not shown) is that the lead transport past Pinehurst is strongly punctuated with periods of high flow events, which transport large amounts of lead. That analysis shows that the net erosion of high-lead sediments in the Box had been strongly suppressed, but in the area above the Box, the tendency for high-flow events to move large amounts of high-lead sediments has continued. This is another indicator that, if future investments are going to be made to control lead transport from the South Fork of the CDA watershed, they would need to focus on erosion in the area upstream of Elizabeth Park.

The flux of total lead out of the lower CDA River (into CDA Lake) is only minimally related to the inputs coming from the South Fork. Throughout the 1990s, the ratio of flow-normalized flux of total lead out of the lower CDA River to the flux into the lower CDA River was about 7:1. This means that most of what was entering the Lake during those years was not derived from the South Fork (or the North Fork, which delivered about one tenth the amount of lead

TABLE 3-3 Trends in Total Lead: Flow-Normalized Annual Mean Concentration (upper table) and Flow-Normalized Annual Flux (lower table)

NOTES: The period of record is different among the four sites, so the first water year is designated in the column labeled “First year.” Trends are computed over three time periods, and for each time period the total change, expressed in percent, is displayed. The color code for the boxes is based on the uncertainty about the trend direction. Dark blue indicates a highly likely downward trend (likelihood > 95 percent), light blue indicates a likely downward trend (likelihood between 70 and 95 percent), no shading indicates that the trend direction is highly uncertain (likelihood of a downward trend is between 30 and 70 percent), pink shading indicates a likely upward trend (likelihood of an upward trend between 70 and 95 percent), and red shading (which does not appear in this table but will appear in other similar tables) indicates a highly likely upward trend (likelihood of an upward trend > 95 percent). The last two columns indicate the estimated value of flow-normalized concentration or flux for the first and the last year of the record. SF = South Fork.

as the South Fork) but rather was derived from lead stored in the lower CDA River basin as a legacy of a century of mining activities within the South Fork watershed. In 2000, the situation changed, such that the flow-normalized flux from the lower CDA River basin to the Lake rapidly increased, and this increase lasted most of the decade, even though inputs to the reach were continuing to decline. By approximately 2009, outputs from the lower basin had more than doubled compared to 2000, but inputs to the lower basin had declined by about 45 percent. Thus, the ratio of output to input for the lower CDA River had grown to approximately 27:1 by 2009. Total lead fluxes out of the lower CDA River began decreasing again around 2009 and have been decreasing to the present time.

This general pattern of a decline in total lead flux from the lower CDA basin up until about 2000, then a flux increase between 2000 and 2009, followed by another decline to the present, is also seen for the other constituents investigated in this chapter (see subsequent discussions). The Committee’s hypothesis about these patterns is that the 2000–2009 period was a time when the lower river (including the floodplain, bed, and banks) was readjusting to major changes in inputs of lead and suspended sediment brought about by the Superfund remediation in the upper basin, such that the decreased input from upstream was approximately balanced by an increase in lead flux generated from within the lower CDA reach. This readjustment involved very significant erosion and downstream transport of legacy sediments, primarily derived from the bed of the lower CDA River, which are rich in lead¹⁰ and are thick (in some cases tens of meters). After about 2010, this period of readjustment ended and declines in total lead flux once again followed a roughly exponential decline. The results suggest that further reduction of South Fork sources will have minor impacts on delivery of total lead to the Lake.

Studies conducted by CH2M used a different method for calculating annual lead flux values (CH2M, 2015) and did not compute a flow-normalized flux at these locations. However, that report also identified the large discrepancy between lead entering the lower CDA River from upstream and the amount discharged to the Lake.

___________________

¹⁰Bookstrom et al. (2013) defined lead-enriched as > 1,000 mg/kg dw. Lead concentrations in a 1-mile stretch of river bed sediments can vary from 10 to greater than 25,000 mg Pb/kg.

Considering 1988–2012, they estimated total lead flux from upstream of about 34 MT/yr. Through very detailed studies of processes taking place in the reach, they also estimated bank erosion of about 32 MT/yr and bed erosion of about 250 MT/yr. They also identified an important sink (about 68 MT/yr) in floodplain deposition. Their estimate of the net flux out of the watershed and into the Lake was about 250 MT/yr. For the comparable time period, the WRTDS method estimates a flux to the Lake of 594 MT/yr. The large discrepancy between the two calculations is likely related to differences in approach. First, it does not appear that CH2M considered the statistical issue of retransformation bias associated with the use of regressions based on the logarithms of concentration (see Cohn et al., 1992; Helsel et al., 2020, pp. 256–257). The very high degree of natural variability of this dataset could account for underestimation of as much as 40 percent. Second, the CH2M calculations assumed that the rating curve (the relationship between concentration and discharge) was trend-free over the entire period. Third, their calculations are based on estimating suspended sediment fluxes and then multiplying these fluxes using an estimated mass of lead per unit mass of sediment.

Even though the committee’s findings are quantitatively different from the CH2M findings, the most important result is consistent: the vast majority of lead transported to the Lake in recent decades is derived not from contemporary inputs of lead from the mined area, but from erosion of legacy lead-rich sediments in the lower CDA watershed. The Committee’s analysis strives to look at the dynamics over time in this imbalance of inputs and outputs, whereas CH2M considers the entire period of 1988–2012 to be static such that year-to-year variations in inputs or outputs are only related to the year-to-year differences in streamflow conditions, rather than being related to evolving river system dynamics that arise in response to the complex history of sediment and lead inputs due to mining and remediation.

The results for the Spokane River below the CDA Lake outlet show very substantial declines in the total lead fluxes out of the Lake—a decrease of 86 percent over the period 1991 to 2020. They also show that Lake outputs in relation to Lake inputs have changed substantially, from outputs being about 7 percent of inputs in 1993 to outputs being less than 1 percent of inputs in 2020. This can be explained by the fact that over time the inputs to the Lake have become more and more dominated by the particulate phase and these particles deposit very readily as they move through the Lake (see later discussion of the dissolved lead).

Figure 3-11 shows the history of total lead flux through the watershed in combination with the history of major remedial activity. What is clear is that the major landscape modification activities in OU-2 correspond in time with the decreases in flux from the South Fork of the CDA River, but the large increase in flux from the lower CDA River follows those activities by five to ten years.

Figure 3-12 provides another way to look at the history of flow-normalized total lead fluxes by tracking the downstream pattern of lead transport in 1993 and 2020. What it clearly shows is that, even in 1993, the South Fork watershed provided an input that was small compared to the output from the lower CDA River and by 2020 the upstream input was an even smaller component of the output of the lower CDA River. It also shows how small the output of the Lake is in comparison to the inputs, indicating that something close to 99 percent of the total lead that enters the Lake is deposited in the bottom sediments of the Lake and never reaches the outlet.

The major take-away messages from Table 3-3 and Figures 3-11 and 3-12 is that the inputs of total lead from the South Fork of the CDA River have been greatly reduced over this nearly three-decade history. Further decreases in total lead from the South Fork basin will have limited consequences for the future of lead inputs to the Lake. Rather, understanding the changes in the lower basin from 2000 to 2010 is crucial to understanding how the delivery of lead from this reach will change going forward and how it will respond to remedial strategies. Box 3-4 provides another analysis, different from the WRTDS approach, of the total lead data that demonstrates the effectiveness of the Superfund remedy in reducing loads of lead to CDA Lake.

Sediment

The fluctuations of the flow-normalized flux of lead described above might be explained by differences in the parts of the legacy materials that have eroded over time. The stratigraphy and chemistry of the bed material of the CDA River is documented extensively in CH2M Hill (2017). The bulk lead concentrations of the sediments vary considerably in the vertical dimension. The oldest materials, which are considered to predate the mining era,

have a lead content of less than 100 mg/kg sediment. The deposits from the early mining period, when milling processes were very crude and large amounts of lead remained in the waste materials discharged downstream, had lead concentrations of ~10,000–20,000 mg/kg, with an observed maximum of 70,000 mg/kg. Sediments from the mid-20th century when ore processing was more effective had lead concentrations of ~5,000–10,000 mg/kg. Finally, near the surface are the modern deposits, representing the periods after the end of mining activity through the Superfund remediation, which have lead concentrations of ~1,000–3,000 mg/kg. The committee analyzed data on the sediments that moved during high discharges (above 250 m³/s or approximately the 95th percentile of the distribution of daily discharges) over the past two decades. Concentrations of total lead were in the range of 2,000–5,000 mg/kg, consistent with a mixture of the mid-20th century deposits and the modern deposits.

The two relevant data sets are the USGS data on suspended sediment concentration, total lead, and dissolved lead; and the Basin Environmental Monitoring Program, which made direct measurements of lead in several particle size fractions. One hypothesis for the complex patterns of flow-normalized lead flux seen at Harrison emerges from the examination of the suspended sediment records for the South Fork near Pinehurst (1989–2020), the North Fork at Enaville (1989–2020), and the CDA River near Harrison (1999–2020). Sediments were sampled only about six times per year in recent years, and a consistent sampling pattern for all three sites only covers the years 1999–2020. Figure 3-13 shows the flow-normalized flux estimates for sediment at these three sites from 1999 to 2020 and the 90 percent confidence intervals on these estimates.

Even with these low sampling frequencies, it is possible to say that the flow-normalized sediment flux for both the South Fork and the North Fork of the CDA River have declined steadily and substantially over the period 1990–2020. If one looks at the fluxes on a per unit area basis, for the South Fork the decline is from about 40 MT/km²/yr to about 12 MT/km²/yr. For the North Fork the decline is from about 29 MT/km²/yr to about 13 MT/km²/yr. It is reasonable to conclude that some of this substantial decline in sediment yields from the South Fork watershed is the progress of the Superfund remediation, since stabilization of mine tailings and other waste material was an important remedial objective. However, it is notable that sediment yields also decreased substantially in the North Fork watershed. This may be due to some remediation carried on outside

of the Superfund project, but it is also possible that an important driver of the decline for the North Fork and the South Fork sediment fluxes may be some other activity, such as improved forest management and harvest practices in these watersheds. Regardless of the causes of these trends in sediment delivery to the lower CDA River, from about 2005 to 2015 the flux of sediment out of the lower CDA River far exceeded the input from upstream, meaning a considerable part of the sediment transported out of the lower CDA River came from storage in the lower basin. The findings of CH2M Hill (2017) suggest that most of this sediment was derived from the lead-enriched sediments in the bed of the lower CDA River.

Figure 3-14 shows the flow-normalized mass balance for the lower CDA River. One possible explanation for the increase in total lead flux from the CDA River during the 2000s, followed by declines in the 2010s, is that the channel of the lower CDA River has been adjusting to the declining inputs from upstream (as hypothesized earlier). If the river was carrying a declining amount of sediment from upstream (brown decline in Figure 3-14), increased erosion of the lead-enriched channel bed in the lower basin could occur. Over time, the net effect of this erosion would be a decreasing slope through most of the lower basin’s channel length and a gradual decline in the net scour rate (as seen in the final several years in Figure 3-14). There are two implications of this hypothesis that should be considered in projecting future lead transport to the Lake. The first is that as a new equilibrium channel configuration is reached, there may be substantial declines in the rate of scour of lead from the lower CDA River and thus a lower rate of delivery of lead to the Lake. The second is that managing the future inputs to the Lake demands more scientific attention to the changes underway in the lower CDA watershed. Such analysis must consider the transport capacity of the river to be dynamic, responding to the decades-long decreases in sediment delivered from the upstream watershed.

Without an understanding of the changes over the past three decades it will be difficult to project how legacy lead transport will change in the future. In the long run, lead transport will decline over time if the riverbed becomes depleted of lead-rich sediments. However, a crucial fact to consider is the magnitude of the legacy lead that underlies the river. The detailed mapping and stratigraphy study by CH2M Hill (2017) estimates the mass of lead in the lower river bed as between ~5 million and 11 million MT. The analysis of lead transport presented in Table 3-3 indicates that the rates of transport of lead from the CDA River to the Lake as of 2020 are about 790 MT/yr. Using the lower of the two estimates of lead in the bed material, this suggests that, at current rates, the lead would be eroded from the river bed in about 6,000 years. This estimate illustrates that the amount of lead in the lower basin is extremely large in comparison to the rates of delivery to the Lake. The establishment of a new equilibrium condition for the channel, dictated by the dramatic declines in sediment input from both the South Fork and the North Fork, will probably drive the delivery of this legacy lead in the coming decades. If the current rates of delivery of lead to the Lake are considered to be unacceptably high, then the questions that will need to be answered going forward are as follows: How rapidly will the channel of the lower CDA River readjust to the much lower sediment inputs? What will be the pattern of lead flux declines in response to this readjustment?

If some engineering modifications are proposed, how will they change this trajectory of channel readjustment and lead delivery?

Two kinds of observations will be needed to project how lead delivery to the Lake will evolve into the future. The first are observations of dissolved and total lead as well as suspended sediment into and out of the lower CDA River, at least at the existing monitoring locations of Enaville (North Fork of the CDA River), Pinehurst (South Fork of the CDA River), and Harrison. The goal of this data collection should be to obtain about a dozen high-flow samples every year at each site. This needs to be done in conjunction with ongoing data analysis to determine the total fluxes at each location, change in lead and sediment storage, and average lead content of the sediment transported. This could be enhanced by using insitu turbidity sensors, as turbidity is a statistical surrogate for sediment and lead, but the current level of monitoring is still needed to ensure that the calibration of the statistical relationships remains relevant. Second, there also needs to be a system of regular observations of scour and deposition of bed materials along the 60-km length of the main stem of the CDA River. Modern boat-operated bathymetry technologies (including perhaps autonomous boats) can probably be employed as an efficient system for repeated riverbed change mapping. This would need to be coupled with a data analysis approach that documents the total amount of change and then relates it to the hydrologic conditions that drove the change.

Together, these data can then be used to build and calibrate a reach-scale model that can predict future lead transport to the Lake. In short, the lower CDA River is not simply a “conveyer belt” delivering lead from upstream to the Lake. Rather, it is a dynamic system that will respond to the changing delivery of lead and sediment in a manner that increasingly delivers legacy lead to the Lake or that increases the storage of new sediments that are less contaminated than the legacy materials and thus reduces delivery of lead in the coming decades. An adaptive management strategy will be needed to continuously assess the success of any future control measures (including the no-action option) that may be considered.

Dissolved Lead

Understanding dissolved lead is important because it is potentially more bioavailable in the river and the Lake. In the case of input from the South Fork, the lead flux is dominantly in the suspended fraction (about 5 percent is dissolved) and this has not changed substantially over the past 30 years. WRTDS can be used to estimate the expected number of days for each year when the concentration of dissolved lead for the CDA River near Harrison exceeds relevant aquatic life criteria values—for example, both the EPA acute criteria level of 65 μg/L and the chronic level of 2.5 μg/L (which assume a hardness of 100 mg/L CaCO₃—see Chapter 9). These calculations use the WRTDS model and the actual discharge on each day to compute a probability that the concentration would have exceeded the criteria value, and then these are summed over the water year to compute an expected number of days for each year (Figure 3-15). The number of days above the water quality criteria is small for the 65 μg/L threshold but very large for the 2.5 μg/L threshold. In both cases, it is difficult to say that there is a meaningful trend, although the past decade suggests a substantial decrease in the number of days the river may have exceeded these criteria. These are worst-case examples for CDA Lake, since the highest concentrations are found near the mouth of the CDA River and decline with distance as they mix with Lake water.

Finally, comparing results for total and dissolved lead over the history of the datasets, on average the flow-normalized dissolved lead flux for each year is about 3 percent of the flow-normalized total lead flux, declining slightly over time (Table 3-4). Over the past 25 years, the total lead flux to the Lake has been almost entirely in the suspended fraction, suggesting that future remediation strategies should focus on reducing particulate lead.

TABLE 3-4 Trends in Dissolved Lead; Flow-Normalized Annual Mean Concentration (upper table) and Flow-Normalized Annual Flux (lower table)

NOTES: The period of record is different among the three sites, so the first water year is designated in the column labeled “First year.” Trends are computed over three time periods, and for each time period the total change, expressed in percent, is displayed. The color code for the boxes is based on the uncertainty about the trend direction. Dark blue indicates a highly likely downward trend (likelihood > 95 percent), light blue indicates a likely downward trend (likelihood between 70 and 95 percent), no shading indicates that the trend direction is highly uncertain (likelihood of a downward trend is between 30 and 70 percent), pink shading indicates a likely upward trend (likelihood of an upward trend between 70 and 95 percent), and red shading (which does not appear in this table but will appear in other similar tables) indicates a highly likely upward trend (likelihood of an upward trend > 95 percent). The next two columns indicate the estimated value of flow-normalized concentration or flux for the first and the last year of the record. The last two columns compare the estimated values of dissolved lead to total lead for the years 2000 and 2020.

Total and Dissolved Cadmium

Cadmium was analyzed using the same approach used in the previous section on lead, but focusing on total cadmium with only brief mention of the dissolved phase. The reason is that there is a strong similarity between the results for total and dissolved cadmium at all sites, and the dissolved cadmium constitutes a very large fraction of the total (unlike the case with lead). The overall behavior of the total cadmium data (summarized in Table 3-5) is much more regular and much less variable than that of the total lead data. The results for dissolved cadmium are summarized in Table 3-6. Comparing the total and dissolved cadmium data, for flow-normalized flux over the 1997–2020 period, the dissolved cadmium makes up about 83 percent of the total cadmium, and this percentage is consistent over the whole record. For flow-normalized concentration, dissolved cadmium makes up about 96 percent of the total cadmium.

Throughout the record, the total cadmium concentrations (Figure 3-16A) tend to be highest at Pinehurst, and the values at Elizabeth Park are nearly as high. With the substantial dilution from the North Fork of the CDA River, the concentrations at the mouth of the CDA River are much lower (as of 2020, about one-fourth of those at Pinehurst). Finally, as a result of both dilution and loss processes (i.e., settling and biological uptake), the levels below the Lake are about 15 percent of the levels at Harrison, and about 4 percent of the levels upstream at Pinehurst. The total cadmium flux results (Figure 3-16B) show, by far, the highest annual flow-normalized values being at Harrison, with the Pinehurst values second highest and about a half the values at Harrison. The flux at Elizabeth Park is generally about half of the value that it is at Pinehurst. Finally, the flux at the outflow (Spokane River) is much smaller than it is at the inflow at Harrison, being only about one-third of the Harrison value. An interesting feature that stands out is that all of the records—except for the CDA River near Harrison—show something like a linear or exponential decline in flux over the period of record. At Harrison there is a plateau between about 2000 and 2010, which is similar to features seen with total lead as well. It is indicative of some type of readjustment happening within this reach around this period of time (see previous discussion on page 88).

TABLE 3-5 Trends in Total Cadmium: Flow-Normalized Annual Mean Concentration (upper table) and Flow-Normalized Annual Flux (lower table)

NOTES: Trends are computed over three time periods, and for each time period the total change, expressed in percent, is displayed. The color code for the boxes is based on the uncertainty about the trend direction. Dark blue shading indicates a highly likely downward trend (likelihood > 95 percent). The last two columns indicate the estimated value of flow-normalized concentration or flux for the first and the last year of the record. NF* indicates “not feasible” because the data set starts in 2001.

TABLE 3-6 Trends in Dissolved Cadmium: Flow-Normalized Annual Mean Concentration (upper table) and Flow-Normalized Annual Flux (lower table)

NOTE: Trends are computed over three time periods, and for each time period the total change, expressed in percent, is displayed. The color code for the boxes is based on the uncertainty about the trend direction. Dark blue shading indicates a highly likely downward trend (likelihood > 95 percent). The next two columns indicate the estimated value of flow-normalized concentration or flux for the first and the last year of the record. The last two columns compare the estimated values of dissolved cadmium to total cadmium for the years 2000 and 2020. NF* indicates “not feasible” because the data set starts in 2001.

Figure 3-17 shows snapshots of the flow-normalized flux for total cadmium through the system in 2001 and 2020. First, in contrast to the total lead history, as of 2020 the inputs of cadmium from the South Fork of the CDA River to the lower CDA River were 43 percent of the outputs of the lower CDA River to the Lake (1.3/3). For lead, the 2020 inputs of lead from the South Fork of the CDA River to the lower CDA River were only about 3 percent of the output of the lower CDA River to the Lake (21/790). This suggests that further reductions in cadmium coming from the South Fork are likely to be important to reducing cadmium inputs to the Lake. Second, the Lake is a sink for cadmium, with outputs that are 30 percent of the inputs, but less so than for lead, for which Lake outputs are only about 1 percent of Lake inputs. Both of these differences are primarily a result of the fact that cadmium is transported mostly in the dissolved phase, but lead is transported mostly in particulate form.

Figure 3-18 shows the yield (flux per unit area) of total cadmium at each of the monitoring sites. What is particularly noteworthy is that in the 1990s the yield at Pinehurst was substantially greater than the yield at Elizabeth Park. This means that the incremental area between the two monitoring sites was delivering more total cadmium per unit area than was coming from the watershed area above Elizabeth Park, which is not a surprise since this incremental area was known to be the part of the watershed with the greatest amount of contamination. Moving toward the present, the yields for these two sites have become virtually equal, suggesting that the areal average yield is now no different in the Box than in the watershed upstream of the Box. This is illustrative of the effectiveness of the Superfund remediation within the Box, with respect to cadmium.

In general, all indications from Tables 3-5 and 3-6 are that total cadmium fluxes and concentrations are continuing to trend downward, at least since the year 2010. Compared to the total lead record, the total cadmium record shows more continuous trends and the magnitude of the seasonal and year-to-year fluctuations are lower. The uncertainty of results is much lower than is the case for total lead, which is what one might expect given that so much of the total cadmium is in the dissolved form and its transport is not subject to the highly variable processes of deposition and resuspension that are so important for total lead. Reductions in cadmium fluxes from the South Fork of the CDA River have been very substantial (a decline of 66 percent at Pinehurst since 1992). Further reductions of cadmium inputs to the Lake will probably depend both on further decreases in flux from the South Fork as well as on controlling the losses of cadmium from storage in the lower basin.

An additional analysis done by the committee (data not shown) is that in the years before about 2005, the total cadmium concentrations were high at both low and high discharge, but after about 2005 the discharge versus concentration relationship became essentially monotonic, with concentrations declining with increasing discharge. This indicates that the primary source of total cadmium is now in the base flow, presumably coming from groundwater.

Looking at the dissolved cadmium concentrations of CDA River inflows to the Lake, in the 1990s the number of days per year that concentrations exceeded the EPA chronic criteria of 0.43 μg/L (assuming 50 mg/L CaCO₃ hardness—EPA, 2016) was nearly 365. By 2020, this had decreased to about 350 days per year. For the EPA acute criterion of 0.94 μg/L, between 1996 and 2000 the dissolved cadmium concentration exceeded the criterion about 325 days per year; this has decreased to about 50–100 days of exceedance per year for 2015–2020. These results are further indications of substantial improvement in cadmium discharges to the Lake. They should be thought of as worst-case numbers in the sense that these represent concentration values where the CDA River enters the Lake and do not account for the dilution and losses that may take place within the Lake.

Total and Dissolved Zinc

Total and dissolved zinc were considered at the same four sites where lead was analyzed, with results shown in Tables 3-7 and 3-8. Dissolved zinc is generally more than 80 percent of the total zinc, except for the CDA River near Harrison, and trends at all four sites are generally very similar (in percentage terms).

Figure 3-19 summarizes the flow-normalized total zinc concentration records at these four sites. Throughout the record, the concentrations tend to be highest at Pinehurst; the values at Elizabeth Park are nearly as high. With the substantial dilution from the North Fork of the CDA River, the concentrations at the mouth of the CDA River

TABLE 3-7 Trends in Total Zinc: Flow-Normalized Annual Mean Concentration (upper table) and Flow-Normalized Annual Flux (lower table)

NOTES: The period of record is different among the sites, so the first water year is designated in the “First year” column. Trends are computed over three time periods and for each time period the total change, expressed in percent, is displayed. The color code for the boxes is based on the uncertainty about the trend direction. Dark blue shading indicates a highly likely downward trend (likelihood > 95 percent). The last two columns indicate the estimated value of flow-normalized concentration or flux for the first and last year of the record.

TABLE 3-8 Trends in Dissolved Zinc: Flow-Normalized Annual Mean Concentration (upper table) and Flow-Normalized Annual Flux (lower table)

NOTES: The period of record is different among the sites, so the first water year is designated in the “First year” column. Trends are computed over three time periods, and for each time period the total change, expressed in percent, is displayed. The color code for the boxes is based on the uncertainty about the trend direction. Dark blue shading indicates a highly likely downward trend (likelihood > 95 percent). The next two columns indicate the estimated value of flow-normalized concentration or flux for the first and last year of the record. The last two columns compare the estimated values of dissolved zinc to total zinc for the years 2000 and 2020.

are much lower (as of 2020 they are about one-fourth the magnitude of those at Pinehurst). Finally, as a result of both dilution and loss processes (i.e., settling and biological uptake), the total zinc concentrations below the Lake are about 20 percent of the levels at Harrison and about 5 percent of the levels upstream at Pinehurst.

The flux results in Figure 3-19B show the highest annual flow-normalized total zinc values being at Harrison, with the Pinehurst values slightly less than half of the values at Harrison. This means that the lower basin is a significant source of total zinc. The flux at Elizabeth Park is generally about half of the value that it is at Pinehurst. Finally, the flux at the outflow is only about one-third of the Harrison value. A substantial portion of the observed decrease over time at Pinehurst is almost certainly the result of continuous improvements in treatment by the CTP. In 1993, 283 MT/yr of total zinc entered the river between Elizabeth Park and Pinehurst, and in 2020, this was reduced to 74 MT/yr. Based on the information in Box 3-1 and recent performance data, up to 75 percent of this amount could be from removal at the CTP (potentially less, depending on what CTP performance was in 1993). An interesting feature of all of the records except for the CDA River near Harrison is the exponential decline in total zinc flux over the period of record. At Harrison, there is a plateau between 2000 and 2010, similar to what was seen with total lead, indicative of some type of readjustment happening within this reach around this time.

Figure 3-20 shows the yield (flux per unit area) of total zinc at each of the monitoring sites. What is particularly noteworthy is that in the 1990s the yield at Pinehurst was well above the yield at Elizabeth Park. This means that the incremental area between the two monitoring sites was delivering more total zinc per unit area than the area above Elizabeth Park (which is not a surprise, since this was known to be the part of the watershed with the greatest amount of contamination). But, moving toward the present, the yields for these two sites become virtually equal, suggesting that the areal average yield is now no different in the Box than in the watershed upstream of the Box.

All indications are, at least since the year 2010, that total zinc fluxes and concentrations are continuing to trend downward. Overall, and compared to the total lead record, the total zinc record is much more limited in its variability. This is to be expected given that so much of the total zinc transport is in the dissolved form and not subject to the highly variable processes of deposition and resuspension that are so important in the case of total lead. Figure 3-21 shows the declines of 45 to 62 percent at all four sites from the early 1990s to 2020 and indicates that the South Fork watershed as well as sources within the lower basin continue to be nearly equal contributors to the fluxes to the Lake. There is a strong similarity to the shapes of the schematic figure for the two time periods, but the latter is about half as wide as the former, with the exception that the reach between Elizabeth Park and Pinehurst narrows the most (because it has been the focus of the most intensive Superfund cleanup activity).

The spatial and temporal patterns of total zinc flux trends are very similar to those for cadmium. The inputs to the lower CDA River in 2020 are 44 percent of the outputs of the lower CDA River (this was 43 percent for cadmium). The total zinc outputs of the Lake are 47 percent of the inputs (this was 30 percent for cadmium). Similar to the

findings for cadmium, it is clear that further reductions in zinc entering the Lake will depend both on reductions in the lower CDA River as well as on reductions in the losses of zinc from the contaminated sediments in the lower basin.

The number of days that dissolved zinc at the inflow to the Lake is above the acute and chronic life criteria of 117 μg/L (corrected to 100 mg/L CaCO₃) shows a decline from about 365 days per year in the late 1990s to about 200 to 250 days per year around 2020 (Figure 3-22). Although this is a substantial improvement, it still represents a large number of days above these criteria.

Phosphorus

Before proceeding into a discussion of the available data about total phosphorus (TP) in the rivers of the CDA watershed and total phosphorus at the outlet of the Lake, it is worth reviewing what has been written on this topic in recent years. In the report, Coeur d’Alene Lake Management Program: Total Phosphorus Nutrient Inventory, 2004–2013 (IDEQ and CDA Tribe, 2020), the following summary of key trend findings is provided.

In reading this summary, it is very important to note that the data used in preparing this report only covered the time frame through 2013.

The other publication that discusses river water quality trends is Trends in Concentrations, Loads, and Sources of Trace Metals and Nutrients in the Spokane River Watershed, Northern Idaho, Water Years 1990–2018 (Zinsser, 2020). The overall summary in that report regarding trends in total phosphorus in the years 2009–2018 is this: Of the nine monitoring sites considered for this period, five were characterized as showing “somewhat likely down” or “likely down” trends in flux, and the other four were characterized as “about as likely as not” to be downward trends. From a concentration perspective, six of the sites showed “somewhat like down” or “likely down” trends, and the remaining three sites were “about as likely as not” to be downward trends. None of the sites showed even moderately convincing evidence of increases over this period. This finding stands in sharp contrast with the statement from the previous report, but is understandable given that the IDEQ and CDA Tribe (2020) study considered a different period (1991–2013) than the Zinsser (2020) report (2009–2018). Zinsser (2020) did consider one site (South Fork of the CDA River near Pinehurst) that had data covering the same period as IDEQ (1991–2013) and it did show an approximate doubling of fluxes between 1990 and 2007, but then a decline from 2007 to 2018. In addition to differences in the period of record evaluated, the statistical method used by Zinsser (2020) was one that allowed for the characterization of non-monotonic trends, which was not a part of the IDEQ method. It should also be recognized that the periods of record covered a relatively narrow window of time within which to separate trends from persistent long-term stochastic variability (Cohn and Lins, 2005).

Against this backdrop of a potentially changing set of conditions, the synthesis presented here employs the same basic method used by Zinsser (2020) but using total phosphorus data through the end of 2020. This synthesis evaluates data at all six sites, which includes the four included in the analysis of metals and sediment but adds two additional sites in watersheds that did not have any Superfund cleanup activity. The results of the WRTDS results for flow-normalized concentration, flux, and yield of total phosphorus are shown in Figure 3-23. It should be noted that a similar exercise was not done for nitrogen because the total nitrogen data set was considered too short for this analysis.

Looking at the yield graph (Figure 3-23C), clearly the highest total phosphorus yield is for the South Fork of the CDA River near Pinehurst. This is likely a result of the fertilizer-production waste that is located in that watershed and perhaps the additions of fertilizer to the landscape for stabilization purposes, particularly during the 2000s. The yield is generally more than twice as high as the next upstream site, the South Fork of the CDA River at Elizabeth Park. The CDA River near Harrison ranks next and is always lower than Pinehurst and the pattern is similar but shifted to the right by about five years. This reflects the dilution effect of the North Fork of the CDA River and it also suggests that the lower CDA River reach may have stored phosphorus from the high input time of 2000–2008 and then released it a few years later.

The St. Joe River ranks next, being the watershed with more agricultural activity and population as compared to the North Fork basin, which has a yield that is about two-thirds of the St. Joe River yield. The total phosphorus yield for the outlet below the Lake is lower than any of the others, and this largely reflects the substantial losses

that take place in the Lake. Finally, the unmonitored part of the watershed is shown on the figure as a horizontal dotted line. It is based on the USGS 2012 SPARROW model (Wise, 2021), and there is no information available about possible trends in this unmonitored part of the watershed. It is difficult to argue that this nearshore area should show a lower yield than some of the relatively unmined watersheds (such as the North Fork and the St. Joe) because parts of it have substantial human populations and some parts have agriculture. Better characterization of this unmonitored area should be a high priority (see Chapter 8). However, since it is only about 16 percent of the land area of the total watershed, even a substantial increase in the estimated yield for this portion would have a relatively small effect on the total inputs to the lake. Using the estimates shown here, if the estimated flux of the unmonitored area were doubled, it would result in only a 7 percent increase in the total phosphorus flux to the Lake.

The temporal patterns of total phosphorus concentration, flux, and yield at Pinehurst and Harrison are complex. At Pinehurst, flow-normalized flux increased 114 percent from 1989–2006 and then decreased from 2006–2020 by 43 percent. For both low and high discharges, concentrations of total phosphorus rose dramatically between 1998–2010, with some concentrations higher than 200 μg/L at high discharge. The committee speculates that these very elevated total phosphorus concentrations were related to high rates of erosion of surficial sediments (soil and stream banks and bed) coincident with a period of intense landscape modification as part of the Superfund remedy. Once those processes settled down and vegetation became more established, the rates of erosion and transport of sediments high in total phosphorus would have decreased.

For the CDA River near Harrison, the pattern is similar. The increase in flow-normalized flux from 1991–2010 was +178 percent and then −27 percent from 2010–2020. This pattern might be explained by the same drivers as those above Pinehurst, but at Harrison it might have been extended in time by the re-working of recently deposited materials from the South Fork (which may have diminished between 2011–2020 as the landscape of the lower basin became more stable). These observations of patterns and hypotheses about causative factors were also addressed by Zinsser (2020). The difficulties in explaining the strong reversal of total phosphorus concentrations and fluxes at these two sites between 2005 and 2010 points to the need for a much more robust scientific effort to understand the historic and current drivers of total phosphorus movement in the watershed. Undoubtedly one of the key issues for maintaining high water quality in CDA Lake will be control of total phosphorus inputs, and without a strong scientific understanding of these very large changes over the past three decades, any conclusions about future strategies for phosphorus control would have to be viewed as highly uncertain.

Table 3-9 presents the overall results for total phosphorus concentration and flux for all six of the sites considered here, and it shows downward trends in total phosphorus flux at almost all locations in the most recent ten years.

Looking at the flow-normalized total phosphorus flux values for 2020 in Table 3-9, the lower CDA watershed value is 38.6 MT/yr (14.3 from the South Fork and 24.3 from the North Fork). The output of the lower CDA River is estimated to be 67 MT/yr, which means that as of 2020 the lower basin is a source of about 28.4 MT/yr. Similar to the metals, there is clearly a legacy of phosphorus in the lower basin that is being released over time, at least since about 2010. This source can be expected to continue into the future, even if inputs to the lower basin decrease.

The flow-normalized total phosphorus input to the Lake (excluding the unmonitored areas) is about 133 MT/yr in 2020 (in almost exactly equal amounts from the CDA River and the St. Joe River). The output from the Lake is estimated to be about 32 MT/yr. Hence, losses of phosphorus in the Lake are about 100 MT/yr, or about 25 percent of the phosphorus input to the Lake becomes output. The columns labeled “2010–2020” in Table 3-9 show trends over the most recent decade. For total phosphorus concentration, trends range from −18 to −29 percent, with all but one being considered “highly likely” to be truly downward trends (and the other being categorized as “likely” to be downward). (Box 3-5 argues why this ten-year trend is valid and not a result of the last two years of the record.) For total phosphorus flux, the range of trends runs from −20 to −37 percent, with likelihood categories similar to those for phosphorus concentration.

There is concern among various stakeholder groups in the CDA region that increasing phosphorus inputs to CDA Lake or phosphorus concentrations in the Lake could have serious consequences for the Lake’s future water quality, which is not unreasonable. The documents and presentations available to the committee, with the exception of Zinsser (2020), generally suggest that phosphorus inputs to the Lake have been rising in recent years. However, ongoing increases in either phosphorus flux or concentration in the inputs to the Lake are not indicated by the data from the major rivers for the past decade.

TABLE 3-9 Summary of Trends in Concentration and Flow of Total Phosphorus from the Subwatersheds o CDA Lake. Upper Table Is for Trends in Flow-Normalized Annual Mean Concentration, Lower Table Is for Trends in Flow-Normalized Annual Flux

NOTES: The period of record is different among the six sites, so the first water year is designated in the column labeled “First year.” Trends are computed over three time periods, and for each time period the total change, expressed in percent, is displayed. The entry “NF*” indicates that calculations for that trend period are not feasible because the record does not extend back to 2000. The color code for the boxes is based on the uncertainty about the trend direction. Dark blue indicates a highly likely downward trend (likelihood > 95 percent), light blue indicates a likely downward trend (likelihood between 70 and 95 percent), no shading indicates that the trend direction is highly uncertain (likelihood of a downward trend is between 30 and 70 percent), pink shading indicates a likely upward trend (likelihood of an upward trend between 70 and 95 percent), and red shading indicates a highly likely upward trend (likelihood of an upward trend > 95 percent). The last two columns indicate the estimated value of flow-normalized concentration or flux for the first and the last year of the record.

Going forward, it will be important to use trend analysis methods that allow for depiction of trends as having changing slopes over time and even be non-monotonic (the CDA River near Harrison provides an excellent example of that). The trend analysis will need to be revisited frequently. For example, the Chesapeake Bay trend analysis of river nutrient flux data is carried out annually at nine major sites and biannually for a set of more than 100 additional sites. These analyses are published roughly a year after the end of the final water year of the period being analyzed, and those reports provide results for the most recent ten years as well as results covering the whole record (typically around 35 years).

All of the sites considered (whether in the mined and Superfund affected areas or not) showed strong indications of a downward trend in total phosphorus concentration and flux over the 2010–2020 period, and the mean concentrations at the end of this period were highest in sites with the greatest intensity of mining and cleanup activity and lowest at the outlet of the Lake. It is also worth noting that a major improvement in the mitigation of phosphorus outflows (the CTP) began in late 2020 and is not yet reflected in the results at Pinehurst (or the sites downstream from Pinehurst). Based on current inputs to the plant, the total release without any treatment would be 1.5–2 MT/yr, and the amount actually being released is < 0.02 MT/yr (Box 3-1). Based on past and current treatment effectiveness, the CTP was a minor contributor to the phosphorus loads over the past decade, and any further improvements would not be expected to meaningfully alter the loads to the river or Lake. So, it is difficult to propose any empirical basis for concern about rising total phosphorus levels in the watershed or downstream of the Lake. The one possible exception to that is for the unmonitored tributaries close to the Lake, which constitute

16 percent of the total watershed land area. The Committee has not been able to obtain any empirical evidence of trends in total phosphorus from this area although it is logical to have some concern that they might be increasing in some areas due to growing population.

To make an argument about the threat to the Lake from future increases in phosphorus from the watershed, it will be necessary to first understand the reason for these observed decreases (in both mined and unmined watersheds) and then develop a scenario under which these may increase. Factors that may have influenced the total phosphorus decreases within the CDA watershed include the following: (1) declines in availability of phosphorus on the landscape following the end of major fertilizer applications (through hydroseeding) during the intense phase of Superfund landscape revegetation activities; (2) better controls of the waste from the fertilizer plant; (3) riverbank erosion control carried out under the LMP that may be limiting internal loading of phosphorus from river banks to river water; (4) improvement in the vigor of vegetation (perhaps due to air quality improvements since the end of the mining era), and/or (5) improved forest practices that may reduce erosion and hence reduce phosphorus loss. It is also possible that regional climate change may have some influence on total phosphorus fluxes but the mechanisms for that are unclear. Understanding these changes, especially outside of the Superfund area, should be an important research topic to be considered in pursuit of the LMP goals. Without a good conceptual model of the reasons for past decreases, it is difficult to have confidence about the impacts of future total phosphorus control strategies. The parties involved in the protection of the watershed and the Lake should focus on trying to understand the dynamics of phosphorus over the past 30 years.

Summary of Trends across Multiple Constituents and Locations

There are several commonalities and several differences between the trend patterns of the four constituents evaluated here (lead, cadmium, zinc, and phosphorus) across the monitoring sites. Table 3-10 considers all four of these constituents at four sites in the watershed arranged from upstream to downstream. It is a narrative, intended to present general patterns rather than focusing on precise numerical results.

It is also useful to look at the how the frequency distributions of concentration vary across the multiple sites and over the duration of the records and to compare the concentrations at sites downstream of the Superfund remediation with those not downstream from the remediation. The approach was to take the WRTDS model for four sites (St. Joe River, CDA River at Harrison, North Fork of the CDA River, and the South Fork of the CDA River) and estimate the concentrations one would expect on each day of the period of record. Box plots were constructed representing all the days in each of two different periods: water years 1993–1996 and 2017–2020. It was not possible to do the 1993–1996 period for any metals for the St. Joe River because data do not exist for that period). Box plots were constructed from these daily concentration estimates rather than being constructed from the raw data because the distribution of the raw data values is strongly influenced by the particular set of days that happened

TABLE 3-10 Summary of the Temporal Patterns of Change for Four Constituents at Four River Monitoring Sites

to be sampled. The goal was to make the comparisons across sites and time periods that are representative of the general behavior of the individual watersheds, showing the broad range of temporal variations that are driven by the variations in discharge and season. The result should also be responsive to trends that may exist between the two time periods being evaluated. The two sets of years that were examined both contain individual years of very high discharge (1996 and 2017). The result of this analysis for total lead is shown in Figure 3-24.

The St. Joe River data provide an upper bound to what might be considered modern regional background levels of total lead. There is likely to be some small amount of lead mining and/or milling that has taken place in this watershed, but it was never sufficiently intense to warrant this watershed being included in the Superfund remedy. The North Fork of the CDA River in the more recent period shows estimated daily lead concentrations that are slightly higher than those of the St. Joe, although they are still much less than those in the CDA River basin (Figure 3-24). Even though the North Fork watershed was not part of the Superfund remedy, it is clear that the concentrations declined between the two periods, presumably because of the decline in mining activity and/or remediation that has happened outside of the Superfund. It would not be surprising if at least some of the soil contamination in the North Fork and St. Joe watersheds originated from smelting activities in the basin of the South Fork. Nonferrous smelters resulted in widespread dispersion of secondary contamination through the first century of mining activity in the western United States. Moore and Luoma (1990) cited soil contamination with arsenic, lead, and cadmium that still affected vegetation and cropland in the late 1980s and covered 300 km² surrounding the Anaconda smelter in the Clark Fork mineral extraction complex in Montana. Declining trends in metals contamination would be expected with revegetation and recovery of those soils over time.

Moving to the South Fork of the CDA River in the earlier period, the median concentration is about 30 times higher than that of the North Fork of the CDA River (and about 350 times higher than that of the St. Joe River). The change at the South Fork of the CDA River between the two periods (expressed on the basis of median values) is about 80 percent, which one can attribute to the end of mining and the Superfund remedy. This is an impressive decrease, yet an additional 99 percent reduction from current levels would have to occur for lead levels at the South Fork of the CDA River to reach modern background levels. For the CDA River at Harrison, the reduction in the median between the two periods is about 30 percent; in order for levels here to reach the low levels seen in the St. Joe River, a further 99.6 percent reduction from current levels would be needed.

Figure 3-25 shows a similar analysis for total zinc. It shows that in the more recent period, the St. Joe River had concentrations that were about a factor of two lower than the North Fork. Similar to the lead analysis, there

is a substantial decrease (about 75 percent) in concentrations from the earlier period to the later one, indicating improvements in this much less contaminated watershed, even in the absence of the Superfund remedy. For both the South Fork of the CDA River and the CDA River at Harrison, declines in median total zinc concentrations between the two periods are around 55 percent. It is logical that the reductions should be similar because the declines come from improvements in the South Fork basin, and the CDA River at Harrison results simply reflect the dilution of South Fork concentrations by water from the North Fork of the CDA River. A further reduction of about 99 percent from current levels would be required for the CDA River inputs to the Lake to be similar to a regional background level, over and above the 55 percent reduction that has already taken place.

For cadmium, Figure 3-26 illustrates the distribution of concentrations during the two time periods for the North Fork of the CDA River at Enaville, the South Fork of the CDA River near Pinehurst, and the CDA River

near Harrison. It was not possible to create a WRTDS model of total cadmium for the St. Joe River because the vast majority of the data (55 out of 59 observations) were less than the reporting limit of 0.03 μg/L.

The declines in total cadmium at Pinehurst and Harrison between the two time periods were both about 60 percent, and the concentrations at the CDA River near Harrison are approximately what would be expected based on the South Fork values modified by dilution with the “clean” North Fork water. Comparing the recent period for the CDA River at Harrison to the background level represented by the North Fork of the CDA River, the concentrations at the CDA River at Harrison would need to decrease by an additional 98.5 percent to be equivalent to background.

What all of these figures for lead, zinc, and cadmium indicate is that the remediation has, in most cases, brought about a sizable reduction in concentrations over the past two and a half decades (the one exception being lead for the CDA River near Harrison). Nevertheless, the metals concentrations remain one to two orders of magnitude above regional background levels.

Comparison of Trends by Location

Another perspective is to view the trends in flow-normalized flux for all four constituents at a single site on one graph, with each of the records rescaled to a common range defining the maximum flow-normalized flux for that constituent to a value of 1. Figure 3-27 shows the results for the South Fork of the CDA River at Elizabeth Park. For this site, the three metals follow a virtually identical temporal pattern. All of them decline from the early 1990s, ending up at about 40 to 50 percent of their maximum value in 2020. The phosphorus record is shorter than the others, but it also declines throughout its record. For phosphorus, the rate of decline appears to be accelerating in the final years as compared to the other constituents. Figure 3-27 suggests that the Superfund remediation in the upper basin above Elizabeth Park has had a very favorable impact on the three metals, yet all of them continue to be far above regional background levels. The reason for the decline in phosphorus is not well understood.

Moving downstream, Figure 3-28 shows the results for the South Fork of the CDA River near Pinehurst, just downstream of the Box. At this site, all three metals have declined in a roughly exponential pattern, with cadmium and zinc having virtually identical time histories and declining slightly more than 60 percent since 1992. The decline in lead has been steeper than for cadmium or zinc, with a total decline of about 80 percent since 1992. There is a much longer record of phosphorus at this site than at the Elizabeth Park site. Phosphorus was trending upwards steadily from the late 1980s to about 2006 and has turned downward since then, falling by about 40 percent from its highest value in 2006. The reasons for this pattern are not clear but they do suggest that the early years of the Superfund remedy may have caused the mobilization of phosphorus from the gypsum-fertilizer wastes in the Box.

Phosphorus inputs to this watershed segment may have also increased because of vigorous revegetation efforts, including hydroseeding, which resulted in large additions of phosphorus to the Superfund landscape. Many of these activities were at or near completion by the mid-2000s; thus, this source of phosphorus has greatly diminished since that time. There is no reason to expect it to reverse and trend upward in the future.

Figure 3-29 shows the flow-normalized fluxes for the four constituents at the outflow of the CDA River near Harrison. The temporal patterns here are quite different than those observed at Pinehurst. For zinc and cadmium, the declines are both steep initially, they are nearly constant from about 2000 to 2010, and they return to fairly steep declines after 2010. The initial rate of decline is greater for cadmium than for zinc, but after 2000 they follow nearly identical patterns. The lead record is more complex, with a downward trend from 1990 to 2000 and then a sharp reversal from 2000 to about 2009. This must represent a release from sources within the lower CDA

watershed (because one sees no such increase in the inputs to the lower CDA watershed). Then, after 2009, the decline is somewhat erratic and fairly modest in slope (slightly more than 20 percent over the 2009–2020 period). Total phosphorus rises from 1990 to 2010 (similar to the inputs from the South Fork) and then turns down from 2010 to 2020 and roughly matches the timing and relative decline of lead. This suggests conditions stabilizing during the most recent decade.

Finally, Figure 3-30 shows the trends in flow-normalized fluxes for the Spokane River below the Lake outlet. In the early years, the relative rate of decline in lead was much greater than for zinc. But, the interpretation of the figure requires care because of the large differences in the record starting dates (1991 for lead and zinc, 2001 for cadmium, and 2003 for phosphorus). To compare their rates of decline in the common period of record (2003–2020), the results for those years are plotted again in Figure 3-31, which shows that the declines for zinc,

cadmium, and phosphorus are nearly identical over these 18 years. Using cadmium as an example, the rate of decline in the early years of this period (around 2004) was about 2 percent per year and in the last few years it has been about 3.5 percent per year. The recent decline in zinc has been about 2.7 percent per year, and for phosphorus the recent decline is only about 2.4 percent per year. However, for lead the decline started later than the others but has become relatively much steeper, with declines of about 6.4 percent per year.

CONCLUSIONS AND RECOMMENDATIONS

Inputs of lead, cadmium, and zinc to CDA Lake reflect the century-long legacy of mine waste deposition in the Lake’s watershed. The frequent floods that transport wastes downstream, the geochemical reactions within groundwater that mobilize metals, and the minimal dilution that occurs between the source of the primary contamination and the Lake all contribute to ongoing metal deposition in the Lake. Although the Superfund remediation has reduced metal inputs from the upper basin, the lower basin comprises an immense stockpile of metal-enriched particulates poised for transport to CDA Lake. Reducing metal inputs in the future will increasingly depend upon controlling the drivers of inputs from the lower basin as remediation progresses. Based on the committee’s analysis of numerous data sets to better understand changes in metal, sediment, and nutrient loading to CDA Lake over the past 30 years, the following detailed conclusions and recommendations are made.

1. Observed land use changes in the upper basin are consistent with a recovering landscape where remedial activities, revegetation, and improvements to air pollution are ongoing. The analysis of land use suggests that halting of mining activities and revegetation activities are reducing the extent of barren land and altering the balance of vegetated land cover, although the urban footprint has not changed. The reduction in barren land through remedial activities and the associated stabilization of soils have likely contributed to lowering metal loads from the upper basin to the river and Lake.
2. Rates of streamflow entering CDA Lake over the past three and a half decades have been remarkably free of trends. This is not only true for average annual flow rates but also true for annual low flows and annual high flows. Also, there does not appear to be a shift in the timing of runoff. These findings stand in sharp contrast with other parts of the western United States, which have generally seen substantial declines in streamflow and a shift of the center-of-mass of annual runoff to earlier in the year.
3. Cadmium, lead, and zinc concentrations and loads into the mainstem Coeur d’Alene River from the South Fork have declined over the past 30 years, and Superfund activities have likely contributed to this decline. For the South Fork of the CDA River at Elizabeth Park, fluxes of the three metals have declined since the early 1990s, with 2020 values being 40 to 50 percent of their maximum. Similarly, at the South Fork of the CDA River near Pinehurst, just downstream of the Box, fluxes of zinc and cadmium have declined slightly more than 60 percent since 1992 while the decline in lead flux has been about 80 percent. Stabilization of the landscape, capping, and sequestration activities have likely been effective at reducing fluxes of particle-associated lead. For zinc, remedial activities in the upper basin and particularly in the Box, including the continuous improvements at the Central Treatment Plant, have helped substantially lowered concentrations and fluxes. Continued remediation efforts and institutional controls are required to maintain the progress that is underway.
4. Reductions of total lead fluxes from the South Fork of the CDA River were offset by processes in the lower basin that released lead from roughly 2000–2010, such that present-day lead inputs to CDA Lake are still substantial. That is, lead concentration and flux at Harrison decreased from 1991 to 2000, increased from 2000 to 2010, and decreased from 2010 to 2020, for a net overall increase of 27 percent. Overall, lead flux to the Lake at Harrison was still 1.3 times higher in 2020 compared to the 1990s because of the increase in fluxes between 2000–2010. In 2020, lead fluxes into the lower basin at Pinehurst were only 2.6 percent of lead flux to CDA Lake at Harrison, demonstrating that there are large

4. reservoirs of metals in the river sediments and floodplains of the lower basin. Furthermore, it is clear that future decreases in lead fluxes into the Lake will be determined much more by evolving storage and release mechanisms in the lower basin than by further efforts at controlling pollution in the South Fork watershed. The committee’s analysis of total lead in high-flow discharges at Harrison shows that lead concentrations in these flows have decreased over time, suggesting that remediation is having a beneficial effect. Remediation of the lower basin will require careful planning so as not to remobilize metals and transport them to the Lake.
5. Sediment transport in the CDA watershed has been going through a period of adjustment, driven by substantial decreases in delivery of sediment from both the North and South Forks of the CDA River to the lower CDA River. The committee hypothesizes that this has brought about scouring of sediment in the lower basin, leading to an increase in legacy lead delivered to the Lake. The slope of the channel downstream profile should decline over time as the river adjusts to dramatically lower sediment inputs, such that this change in profile should eventually result in a decrease in the rate at which legacy lead- and phosphorus-enriched sediments are scoured from the lower CDA river bed and banks. To better understand the future trajectory of sediment (and hence lead and phosphorus) delivery to the Lake, there will need to be a substantial increase in sediment data collection at the upstream and downstream limits of the lower CDA River, coupled with repeated measurement of channel slope and geometry and predictive modeling of the readjustment of channel transport and morphology. This will also be vital to evaluating any proposed remedial measures dealing with the legacy sediment and lead stored in the lower basin.
6. There have been downward trends in cadmium and zinc concentrations and fluxes throughout the CDA basin, and fluxes of both metals to the Lake (measured at Harrison) were lower in 2020 than in 1992 (by 63 and 45 percent, respectively). At the CDA River at Harrison, the cadmium and zinc fluxes leveled off during 2000–2010 but are declining again in the most recent decade. Unlike total lead, as of 2020 the inputs of total cadmium and total zinc from the South Fork of the CDA River to the lower CDA River were 43 and 44 percent (respectively) of the outputs of the lower CDA River to the Lake. This suggests that further reductions in cadmium and zinc coming from the South Fork are likely to be important to reducing inputs of these metals to the Lake. Targeted studies and trend data show that the primary sources of cadmium and zinc are now base flow, presumably coming from the groundwater system. The Central Treatment Plant is now a minor source of zinc.
7. Over the past decade, total phosphorus fluxes and concentrations at monitoring sites in the CDA River, the St. Joe River, and the Spokane River below the lake outlet have all been declining (typically 20 to 30 percent reductions during the 2010–2020 decade). In the case of the CDA River, this is a reversal of the trend observed over the prior decade. Like lead, total phosphorus flux to the Lake in 2020 was higher (by 2.3 times) than in the early 1990s. Projecting future trends of phosphorus in CDA Lake will require a sustained effort at monitoring and regular data synthesis for phosphorus across the whole watershed (including the 16 percent of the watershed that is not in the CDA or St. Joe River watersheds), with monitoring efforts closely connected to research (including modeling) aimed at understanding the reasons for this current decline. Without a better understanding of phosphorus transport in the whole watershed, there is no basis for projecting future phosphorus transport or the potential for future increases in phosphorus in the Lake.
8. All of the wastewater treatment plants that discharge to the CDA basin should be considered for advanced phosphorus removal, similar to those that discharge to the Spokane River downstream of CDA Lake. Phosphorus emanating from wastewater treatment plants is likely to be much more bioavailable than nonpoint source phosphorus inputs from the basin. By improving understanding and taking prudent action, CDA Lake can prove to be an exception to the broader rule of worsening trophic status and increasing algal blooms seen in lakes elsewhere in United States.

REFERENCES

This page intentionally left blank.