Study area

CAKR is located along the Chukchi Sea, just north of Kotzebue, Alaska (Fig 1). The study area within the monument is located between 67–67.75°N latitude and 163–164.75°W longitude. The Monument is located on a coastal plain tundra ecosystem dominated by an open, low, mixed shrub-cottongrass (Eriophorum L.) tussock tundra [13] interspersed with well-drained hills supporting a variety of alpine lichens, forbs, and dwarf shrub species.

The climate in the study area is classified as cold-continental [14]. The mean annual temperature ranges from –7°C along the DMTS to –6°C in southern CAKR [15]. Summer temperatures in the region typically fluctuate between 2° and 18°C, while temperatures along the Chukchi Sea coast range from 4° to 13°C [3]. Winter temperatures typically range from −26° to −15°C [3]. The 1961–1990 mean January temperatures in the two closest villages of Kivalina and Noatak were –20° and –23°C, while the mean July temperatures were 11° and 14°C, respectively [15]. The modeled mean annual precipitation along the DMTS in CAKR ranges from 320 to 380 mm, with montane areas receiving up to 510 mm [16]. More than one-half the annual precipitation occurs as rain from July through September; August is the wettest month [3]. Snowfall has been recorded in every month of the year, but persistent snow cover generally lasts from mid-October to mid-May. The Chukchi Sea is covered in ice from mid-November or December through May or June [3].

Bedrock is predominantly calcareous and of Paleozoic age. Soils are poorly developed due to the cold climate, low precipitation, and the presence of continuous permafrost. Lowland soils of the monument are typically of the order Gelisol, suborder Turbel (with ice wedge terrain), and most frequently of the group Aquiturbels [13]. The DMTS haul road bisects the northern portion of the Monument. Immediately south of the haul road, the Tahinichok Mountains rise from the flat coastal plain to a maximum elevation of 502 m. South of this mountain range, Monument lands are predominantly wet rolling hills and coastal plain. The area north of the haul road is generally low-lying and gently sloping, draining into the Wulik River approximately 10 km north of the Monument boundary.

The CAKR coast typically receives high winds though wind patterns are influenced by local topography [3]. Mean monthly winds at Kotzebue, about 100 km south of the port site, average above 10 knots from September through April and blow from the east [17]. Mean wind speeds are comparable during the summer months (with an average of 10.5 knots) but are from the west. August and September are the windiest months, while the most extreme winds are associated with winter storms. Winds at the Port Site are typically from the northeast in winter months. In summer, Port Site winds are highly variable and strong winds from the south-southwest are common [3]. At the mine site, winds are predominantly from the northeast and southeast in winter months and highly variable in summer months [3]. Because of the NE-SW orientation of the road (Fig 1), prevailing easterly winds tend to drive dust to the north side of the road, which has made the north side a zone of higher contaminant accumulation than the south side in past studies [1, 2, 3].

Study design

Re-measurement of Zn, Pb and Cd concentrations in moss tissue was nested within a larger study examining the effects of contaminants on vegetation in CAKR. Hasselbach et al. [1] found that deposition of these metals decreased logarithmically as a function of distance from the DMTS road. To address the question of vegetation community change as a function of contaminant concentrations (in a separate component of this study), we established permanent plots using a stratified-random design based on the two dominant land cover classes and several distance-from-road classes. Our overarching goals were to choose a sampling design that would reflect the exponential pattern of heavy metal decrease with increasing distance from the road while controlling for natural differences in vegetation community types and elevation. The most common vegetation types along the DMTS were Upland Moist Dwarf Birch-Ericaceous Shrub and Upland Moist Dwarf Birch-Tussock Shrub [13]. Together, these closely related classes accounted for 66% of the area along the DMTS road corridor out to a distance of 4000 m. Both of these community types are fairly rich in the lichens and bryophytes sensitive to Zn, S, and alumino-silicate road dust rich in crustal elements. To avoid confounding the pollution signal with the natural variation in vegetation due to elevation, we avoided sampling in any alpine communities and restricted our sampling to between 60 and 250 m in elevation ( = 130 m).

Sampling distances from the road were chosen to obtain an even spread of values along the heavy metals deposition gradient models reported by Hasselbach et al. [1]. Twelve non-linear “transects” were created according to the distances in Table 1, with points at 10, 50, 100, 300, 1000, 2000, and 4000 m. To help model fine-scale spatial autocorrelation, one additional plot (termed an “autocorrelation plot”) was established in each transect at a random distance 10 to 20 m from the 1,000, 2,000, or 4,000 m plot in each transect. Each “transect” consisted of a group of points within a land cover class at specified distance classes. Points were chosen at random in ArcGIS 9.2 [18] in close geographic proximity to a line, but did not represent a strictly linear feature, per se. Random selections were determined using a random number generator assigned to up to 10 pixel groups of interest closest to the transect line at a particular distance. Linear transects were tested in the planning stages, but rejected as it was not possible to find both the correct land cover classes and distances along a straight line.

Two groups of reference plots were chosen at random from points in CAKR with sufficient representation of the two desired land cover classes, and at a distance of at least 40 km from the DMTS. Each reference area had either 2 or 3 replicates of each land cover class, with a total of 10 plots (Table 1).

The 2001 sampling methods are reported in Hasselbach et al. [1] and summarized in Table 2. Note that the statistical design strata are defined differently in the two studies: in addition to road-based distances, Hasselbach et al. [1] uses Port Site -based distances (strata 4, 5). Sampling points from both projects are shown in Fig 2. The 2001 and 2006 sampling efforts each occurred during the month of July.

Field methods

After sample plot locations were selected in GIS, precise coordinates (error ≤ 1 cm horizontal) were uploaded into Trimble GeoExplorer XH and GeoExplorer XT GPS’s [19]. Researchers located each point on the ground with an estimated real-time horizontal position error ranging from 10 to 200 cm and permanently marked and recorded each location in GPS. GPS plot data were later post-processed for sub-meter accuracy (10–50 cm in most cases). Locations for plots 10 and 50 m from the DMTS were adjusted for greater precision using a measuring tape from the roadbed’s toe slope contact point with the tundra.

Within 10 m of each plot (but at the same distance to the DMTS), approximately 2 L (fairly compacted and yielding 40–80 g dry weight) of the moss Hylocomium splendens was collected. Field duplicates were collected from 10% of the sample sites and during the moss tissue cleaning process, an additional 10% of samples were subdivided into laboratory duplicates for laboratory quality assurance.

Laboratory analysis

The 2001 moss sample analysis is summarized in Hasselbach et al. [1]. The most recent 3–5 years of the 2006 moss tissue were collected by snipping the moss at the corresponding node demarcating annual growth segments. This tissue was then ground, homogenized and analyzed at the same laboratory (University of Minnesota Research Analytical Laboratory, St. Paul, MN) using nearly identical protocols. Moss tissue was analyzed for total sulfur, total nitrogen and a suite of 15 metals. Detailed laboratory analysis methods are provided in S1 File.

Bayesian spatial regression and prediction

Spatial regression models, using Bayesian methods [12], were fit to Zn, Pb, and Cd concentration in H. splendens tissue from the two independent sampling efforts in 2001 [1] and 2006. After model fitting, posterior predictive distributions [11] were used to simulate predictions from each model on a common spatial grid, which allowed spatial comparison between the two years to assess changes for each element.

The Bayesian regression modeling followed Hasselbach et al. [1] (2005). The measured concentration of Zn, Pb and Cd were transformed to natural logarithm values. Let Y_i denote the transformed value for the ith location, then the regression model was (1) where d_i is log (distance from road), s_i is a binary indicator of north or south of the road, d_i: s_i is their interaction, β₀, β₁, β_2, and β₃ are regression parameters, U_i is a random effect for site (to account for duplicate samples per site) with variance θ₁, Z_i is a spatially autocorrelated random effect, and ϵ_i is uncorrelated (independent) error, which includes the replicate laboratory sampling, with variance θ₀. The spatial autocorrelation for Z_i among locations was modeled with the exponential autocovariance model, (2) where h is the distance (m) between any two locations. There are four covariance parameters: the nugget θ₀, the variance of the random effect for location θ₁, the partial sill θ₂, and the range θ₃. We used flat priors on β₀, β₁, β_2, and β₃ and the reference prior [20] on θ₀, θ₁, θ₂, and θ₃. The posterior distribution of all parameters was sampled using Markov chain Monte Carlo methods (MCMC, [21]). We used a “burn-in” of 4,000 iterations and then retained each 10th iteration of the next 2000, leaving a sample from the posterior distribution of 200 iterations for further inference. A separate model was fit for each element, Zn, Pb, or Cd, to the data from each year, 2001 and 2006.

A prediction grid of 2,374 points, consisting of five strata (Table 3) was generated by defining grid points that varied in density among strata. Denser prediction grids were better able to capture more rapid changes in Zn, Pb, and Cd nearer to the haul road. We placed as many points in each stratum as was computationally feasible (Fig 3). For each of the 200 MCMC iterations, the sampled parameters and observed data were used to predict values at the prediction grid locations yielding 200 prediction surfaces for each element and each year. All prediction values were back-transformed to the original scale for further summaries of the posterior predictive distributions, which we denote as for the kth simulated prediction at the jth prediction site for year equal to 2001 or 2006. The point-wise predictions, across all strata, were used for mapping. In addition, block averages were computed within each stratum, for each element, each year, and each of the 200 prediction surfaces. We computed the mean stratum concentration as

(3) for the kth prediction surface for stratum t, where was the number of prediction grid points in the tth stratum (Table 3). The mean estimate for each stratum, then was the average of over the k = 1,…, 200 simulations, the standard error was the standard deviation among the 200 values, and the 2.5^th and 97.5^th percentiles were computed from the ordered values of . We also computed the average change from 2001 to 2006 as the average of over the k = 1,…,200 simulations, and the percent change as the average of 100 *.

Again, the standard errors were computed from the standard deviation among the 200 values, and the 2.5th and 97.5th percentiles were computed from the ordered values from the 200 simulations of the average change and percent change. The probability of decrease was computed as , where I(a) is the indicator function (equal to 1 if the expression a is true, otherwise it is 0. Both the prediction grid and the model output maps were created using ArcGIS software [17]. To provide a measure of the uncertainty of each point prediction (on each year and for each element), the reciprocal of the coefficient of variation (CV), computed as the estimate/standard-error-of-estimate, was divided into four quantiles which were mapped with increasing size. Thus, not only was each point prediction color-coded by concentration class, but was sized according to the CV, where the largest dots had the lowest CV (the most certainty).

Regression of distance to road vs. concentrations

Hasselbach et al. (2005) graphed the unmodeled moss tissue concentrations of Zn, Pb and Cd in field samples against the distance from the DMTS haul road. We present the same data in Fig 4, updated with the 2006 samples. A reduction of concentration is apparent for the 2006 points relative to the 2001 points, with spatial regression coefficients provided in Table 4. All elements showed a significant decrease with increasing distance from road, as seen by the credible intervals which are far from zero. Changes from 2001 to 2006 are apparent, with either a lowering of the intercept values (Cd and Zn), or a steeper slope (Pb and Zn), indicating lower concentrations of Cd, Pb, and Zn in 2006. Also, there is generally a difference due to side-of-road, either by causing a shift in the intercept or through a significant interaction with slope.

Spatial patterns of heavy metals distribution model

Maxima and minima in raw data

The 2001 and 2006 ranges of Cd, Pb, and Zn in the tissue data set (unmodeled) are presented in Table 8 alongside earlier ranges from the Arctic Contaminants Research Program [6]. The 2006 maxima for these three elements (24, 301, and 993 mg/kg respectively) were generally about one-third of 2001 values. The 2006 minima were slightly higher than the 2001 minima for Cd and Zn, and slightly lower for Pb. Notably, the 2006 Pb maximum dropped below the Alaska Department of Environmental Conservation’s recommendation of 400 mg/kg in soil cleanup value for residential areas and well below the 800 mg/kg in industrial areas [22].

Differences between North and South sides of road

The differences in Zn, Pb and Cd concentrations between the north and south side of the road observed for the 2001 modeled data (Tables 5–7) and in Hasselbach et al. [1] were still evident in the 2006 data. In 2006, the modeled moss tissue concentrations on the north side of the road exceeded those in the south side in every stratum for every element except for one which was equal. Within-strata mean modeled moss tissue concentrations ranged from 0 to 100% greater on the north side in Strata 1–4 with an average of 53% for all elements in all strata (S1 Table). The mean percentage by which concentrations-within-strata on the north side of the road exceeded those on the south side ranged from 74% (range 55–100%) for Pb to 46% (range 35–59%) for Cd and 38% (range 0–81%) for Zn. The 2006 elemental concentration differences between north and south were generally highest in Strata 1 and 2.

The comparison of percent change over time by side of road told a different story. Percent decreases were higher on the south side in Stratum 1 for all elements. Because the 95% credibility interval spanned zero in average change for some element-strata-side of road combinations, the change in north and south sides over time and the percent exceedance of north-to-south sides contain moderate to high levels of uncertainty for these combinations (S1 Table). For those combinations in Strata 2–4 in which the 95% credibility intervals were entirely negative for average change, both Zn and Pb concentrations decreased more on the north side than the south between 2001 and 2006 (12–57% for Pb and 19–133% for Zn). For strata in which the 95% credible interval for average change did not span zero, the percentage decrease between 2001 and 2006 ranged from –42 to –51% (N) vs –18 to– 58% (S) for Zn, and –39 to –47 (N) vs. –30 to –52 (S) for Pb.

In Stratum 1, the mean north:south ratio of pooled moss tissue concentrations of Zn, Pb and Cd was 1.5 in 2001 and 1.8 in 2006 (S2 Table). This pattern of increase was reversed for strata 2 through 4. In these strata, the north:south ratio, while still above 1 (ranging from 1.1 to 1.6) decreased in 2006 compared to 2001. In sum, for Stratum 1 while elemental concentrations were much lower in 2006 than in 2001, the north side improved less than the south side. For strata 2–4, the north side improved more than the south side.

Changes in deposition patterns

There was a substantial reduction of dust-borne contaminant deposition between 2001 and 2006, and reductions were greatest in areas where heavy metals concentrations were previously highest–Stratum 1, the area surrounding the Port Site, and the north side of the road (Figs 8–10, S4, S11 and S18 Figs). The several miles surrounding the Port Site had previously been a high deposition zone because standard operations generally left the doors of the Concentrate Storage Buildings open prior to 2001, thus allowing the steady winds to disperse 55% Zn and Pb concentrates onto the surrounding tundra [3]. This is reflected in the mine’s 2003 sampling of NANA-owned lands adjacent to the Port Site, which revealed moss tissue concentrations at the Port Site of 1,720 mg/kg of Pb and 8,120 mg/kg of Zn [3]. Vegetation at the port site was reported in 2003 to be highly stressed and/or dead. Tundra soils in this area reached maxima for Cd, Pb and Zn of 3,600, and 11,500 and 15,000 mg/kg, respectively [3]. The north side of the DMTS haul road had previously accumulated high deposition because of the prevailing winter easterly winds that carry road dust preferentially to the north of the road during the 8 winter months. Concentration contours of all three contaminants still showed a moderate bulge near the Port Site in 2006, but much of the area in this zone had dropped into lower tissue concentration classes than in 2001, presumably due to dust control measures (see below).

Generally, the initially high decreases in modeled tissue concentrations between 2001 and 2006 lessened with distance from the road, and Cd and Zn both eventually showed moderate (though nonsignificant at the 95% level) probabilities of positive percent change with increasing distance. While the diminishing decreases in concentrations occurred in strata 2–4 for Zn and Pb, Cd showed no significant trends beyond Stratum 1 and moderate probabilities (56–85%) of increase in strata 3–5 on the south side of the road. The reasons for these increases in Cd—if real—are unknown; possibilities could include a greater mobility of Cd relative to Zn or Pb, changes in the Cd concentrations in the Zn and Pb concentrates over time, or other factors. In 2003, Cd represented 0.12% of the Pb concentrate and 0.33% of the Zn concentrate (Exponent 2007), thus production differences could potentially play a role.

Spatial variability of modeled concentrations was low in Stratum 1 with the exception of one area of increase near the northern boundary of CAKR, which attained percent change values up to +18% (with probabilities of increase from 51–97%). Convincing support for this hotspot comes from probabilities of increase: Cd showed a probability of increase of ≥95%, while Pb and Zn ranged from 49–83% and 51–74% respectively, with the highest probabilities immediately adjacent the road. It is conceivable that this increase is a relict of a 1992 Pb concentrate spill of 15 tons [23] and/or a 2004 spill of 1.4 tons of gravel contaminated with concentrate that occurred over a two mile stretch of the haul road in this vicinity. Both 2001 and 2006 sampling points fell within 300 m of these spill sites, thus the potential to detect them was higher than at other spill sites without sample points in close proximity. Spatially intensive sampling of former spill sites would be required to determine whether residues from these incidents have led to ongoing heavy metals dispersal.

Differences between North and South sides of road

Several factors favor greater dust dispersal and long-distance transport on the north side of the road than the south. First, the topography on the north side is generally flatter than the rolling landforms on the south. Second, the terrain generally drops in elevation to the north and rises fairly rapidly to the south. Third, the easterly winds that prevail during the 8 winter months tend to transport particles to the north for more of the time in a given year. Not surprisingly, the differences between elemental concentrations on the north vs. south side of the road noted in 2001 (Hasselbach et al. 2005) persisted in the 2006 data. For Stratum 1, the north to south differences in elemental concentration increased in 2006 from their 2001 levels for all elements both individually (S1 Table) and pooled (S2 Table). The opposite occurred in more distant strata though ratios remained above 1:1 (S2 Table). Theoretically, if the heavy metal deposition decreased toward zero, the N/S ratios at some future date would approach 1. The decrease in values of Strata 2 through 4 could signal exactly this sort of improvement trend. As the stratum with the highest Zn, Pb and Cd concentrations in the study area, Stratum 1 may receive too much ongoing localized contamination (and especially on the north side) to show any improvement in the north relative to the south. Dust control measures may have the potential to reduce the north:south differences in Stratum 1 over time: if dust-control measures continue a general abatement of releases, the north side should see more improvement, which could help equalize the north and south sides in Stratum 1.

Technical improvements

Beginning in 2001, the Red Dog Mine made a number of technical alterations to its operations in order to reduce the amount of fugitive dusts escaping to the environment [24]. Among these measures were: replacing the tarp-covered concentrate haul trucks with a new fleet of trucks with hydraulically-covered lids, implementing a seasonal truck rinse station for use in summer months, keeping the large doors of the concentrate storage building doors closed by constructing enclosed unloading facilities, constructing negative-pressure bag houses in the concentrate loading and unloading facilities, enclosing the conveyor system that moves concentrates from the concentrate storage buildings to vessels at the Port Site, applying dust palliatives to the roadbed, and segregating road traffic to reduce concentrate tracking by vehicles. The noted decreases in fugitive dust deposition cannot be attributed to decreases in production during the 2001–2006 period. In 2001, the mine shipped 1.2 million metric tons of lead and zinc concentrates (pers. comm. Teck, Inc; S2 File). The quantities rose from this starting value by approximately 130,000–175,000 tons during the following 5 years.

Investments in road dust control and other emissions mitigation at the Red Dog Mine Site led to analogous reductions in fugitive dust emissions there. In an evaluation of heavy metals emissions at the mine site (about 30 miles from the CAKR boundary), SENES [24, 25] found that factors such as road dust control and a deepening pit that better contained emissions were key in these reductions. Only one period of slight increase of fugitive dust emissions at the mine site was noted (from 2004–2005), which appeared to correspond to increased truck traffic to transport a growing body of waste rock from the mine pit to a storage area nearby [25].

Concentrations in Southern CAKR

There were no clear trends in deposition patterns of Zn, Pb or Cd in the reference area of southern CAKR (Stratum 5). The average and percent change values for Cd and Zn in Stratum 5 had only a 54% and 48% chance of positive change for Cd and Zn, respectively. Pb had an 85% probability of an 18% decrease, with a change from 1.9 mg/kg to 1.5 mg/kg. Uncertainty in these results stems partly from the large area included in the model, the variability of the sample concentrations, the relatively low sample size, and the low contaminant concentrations. Our main finding for Stratum 5 is that while no trend is evident, this area is still experiencing low levels of ongoing deposition of Zn and Pb. While concentrations in southern areas are still fairly low, concentrations for Pb and Zn remained above the arctic baseline values reported in Ford et al. [6]. The 2006 mean modeled concentrations of Pb and Zn were 1.5 and 49 mg/kg respectively, compared to Ford et al.’s Alaska background values of 0.6 and 37 mg/kg respectively. The 2006 value for Cd in this study (0.15 mg/kg) matched Ford et al.’s Alaska background value. The variability among raw samples was surprisingly high for the reference area. 2006 tissue samples ranged from 28 to 74 mg/kg for Zn, 0.8 to 5.7 mg/kg for Pb, and 0.13 to 0.30 mg/kg for Cd. The mean of the raw values of Cd (0.18 mg/kg) was slightly higher than the modeled value and Ford et al.’s background. There are no baseline measurements of the reference area prior to the mine’s opening in 1989, therefore natural sources of deposition such as weathering of mineralized outcrops cannot be ruled out. Elemental fingerprinting (e.g., via isotopes, speciation) could help to establish the source more definitively [26, 27].

There does not appear to be any obvious geographic basis to explain the variability in Zn, Pb or Cd concentrations or spatial patterns within the reference area (Stratum 5). Moreover, it is also conceivable that deposition of Zn, Pb and Cd in Stratum 5 decreased between 2001 and 2006 but that this change was too small to be reflected in the 3–5 year moss tissue sampled. Mosses are highly adapted to retain atmospheric deposition, especially at low levels of deposition beneath the higher thresholds at which they reach saturation or begin to leak. Brumelis and Brown [28] reported that heavy metals can be mobile between older and younger segments within H. splendens tissue and that metals can migrate from higher concentration source areas to lower concentration sink areas. Brumbaugh et al. [29] found significant differences in elemental accumulation between first year and 3–5 year tissue of H. splendens, with higher metals concentrations in the older tissue. Unlike the stressed feather mosses adjacent to the haul road, reference area mosses tended to retain green (and presumably absorptive) tissue in the range of 6–8 years before beginning to decompose and release retained elements. In environments with low inputs of heavy metals, high retention and within-tissue redistribution may make detection of short term temporal concentration changes more challenging. Small but consistent quantities of laboratory error can also lead to larger proportional error in tissue with low concentrations. Likewise, trend assessment in low deposition environments is inherently challenging because small changes in concentration represent a much higher proportional change than in areas of higher deposition.

In interpreting percent change maps for low deposition environments, it is important to consider that small changes of only a few mg/kg in actual concentrations can at times appear inflated in importance in percent change mapping because of low starting values. In areas of higher starting deposition, even lower percentage change values represent higher changes in actual concentrations. Maps of actual differences in concentration (Average Change) are presented to provide a cross-check on the importance of the changes presented (Fig 7, S3, S7, S10, S14 and S17 Figs).

Environmental fate of contaminants

Moss biomonitoring enables an integration of moss accumulation of contaminants during a narrow window. The current analysis pertains only to contaminants in and on moss within a 3–5 year sampling period. The fugitive dusts that either deposited earlier than this timeframe, or dusts that previously washed into the organic soil profile over the past 24 years are likely to have resulted in a considerable pool of residual Zn, Pb and Cd. Accumulated contaminants in soils can be further dispersed by wind and storm events. This redistribution of sediments and contaminants will be particularly pronounced in low-precipitation environments such as this, characterized by sparse vegetation and low surface moisture during much of the year [30, 31]. Smaller particles will be more mobile and carried further away from the source. Mining-associated contaminants have been reported to be deposited kilometers to hundreds of kilometers from source previously [32, 31]. For small particle sizes such as those of the ore concentrates, factors important in determining metal fate will include wind-driven transport near the soil surface, surface creep and saltation, as well as suspension [32]. This study and Hasselbach et al. (2005) [1] provide evidence for the distance over which contaminants can be transported, with elevated levels observed across the studied northern section of the CAKR, up to 4 km from the road, and low levels of contaminants found tens of kilometers from the road in the southern CAKR (discussed above).

The Pb and Zn-enriched dusts deposited on the tundra initially contain these metals in reduced sulfide forms (Exponent 2007), which tend to exhibit low bioaccessibility [33]. Over time chemical weathering and exposure to air will oxidize these contaminants to more bioavailable forms. A 2007 risk assessment [3] assumed 100% bioavailability of a large suite of contaminants and concluded that the risks to a wide range of wildlife, plants and humans via ingestion and direct contact were to: 1) small mammals from elevated boron; 2) Willow Ptarmigan (Lagopus lagopus) from elevated Cd near the mine site (outside of CAKR); and 3) the cover of lichens and bryophytes out to 2,000 m from the haul road. Sublethal effects that could influence population numbers and health were not considered, however, and it is possible that the below-ground contaminant reservoir could contribute to these types of effects for a range of ecological receptors that live in or eat soil and vegetation and then introduce them into the broader food web [34]. Because of the early history of fugitive dust dispersion, the soil-based pool of contaminants could persist for many decades even if no further release of fugitive dusts occurred. The active layer of the soil profile is generally less than 1 m deep above permafrost in CAKR [13]. This suggests that contaminants have the potential to be trapped in the organic soil horizons unless released into perched water as they become more bioavailable. Within this century, Arctic warming is projected to deepen the active layer and warm permafrost temperatures to the thaw point in northern CAKR [35]. This is likely to bring changes in the hydrological regime and organic carbon fluxes that have the potential to change the rate of transfer of legacy contamination to surface and ground waters.

An area over 100 km² in CAKR (approximately 2000 m on each side of the 32 km road easement) is likely to have some level of reduced lichen and bryophyte cover and extensive lichen mortality [3]. If the decreases in Zn, Pb and Cd noted in this study persist, there should be some level of lichen recovery in the more distant areas of the affected zone, though the timing and spatial patterns of recovery are difficult to predict. It is also unknown how the soil reservoir of Zn, Pb, Cd and sulfides may affect lichens with extensive underground tissue, e.g., the very common Cladonia spp. and Cetraria spp. Due to the slow growth rates of these nonvascular plants (6.2 mm/yr for Cladonia stygia (Fr.) Ruoss common in the study area [36], however, it is likely that their recovery will take many decades even at low future levels of contaminant input. A subsequent manuscript will use the simulations developed in the present study to make estimates of the terrestrial area above elemental concentration thresholds established for biological metrics such lichen species richness, lichen cover, lichen mortality, and the blackening of the moss Hylocomium splendens.