Copper. Arsenic and cadmium concentrations were determined in this study. Cadmium and excess copper are extremely toxic to plants. Cadmium is an accumulative poison to all higher animals (3). Cadmium contamination of the human food chain is a global public health-concern (5). Arsenic is less toxic to both plants and animals; however, it is the major contaminant of the copper ore from the Butte mines. 

The study objectives reported in this paper were to:

o Map the copper, arsenic, and cadmium contamination of both the soils and an important forage grass throughout the ranch.

            o Contrast the metal concentrations measured on the ranch to a check plot in another area of the Deer Lodge Valley, which has been subjected to similar aerial deposition of pollutants but not contaminated by sediments carried by the Clark Fork River.

            o Contrast the metal contamination of the ranch and check plot to the concentrations typical in control areas not subjected to significant anthropogenic contamination.

            o Determine the metal concentrations of riparian slickens on the upper Clark Fork River flood plain at a limited number of exemplary sites along the run of the river between Butte and Missoula.

            o Determine the soil microbial activity in metal-contaminated sediments.

METHODS

Study Site Selection,

Sampling Grid Establishment for the Grant-Kohrs Ranch,

Check Plot, and Control Site

   he perimeter of the grid included some land external to the ranch boundaries. he center of each hectare was marked in the field.

    Each hectare was assigned to one of four topographic/plant community ones based on elevation relative to the Clark Fork River and vegetative canopy structure. The flora and plant community structure of the ranch and adjacent lands have been described elsewhere (12). The hectares comprising ach of the four zones are contiguous so the zones form discrete units fig. 3). The zones do contain hectares for which a part of the land surface could be more accurately assigned to a different zone if the topographic esolution 0£ the study was finer than 1 ha. For example, portions of hectares assigned to the riparian zone are more typical of the meadow zone.

   The riparian zone is immediately adjacent to the river at the lowest elevation. It is subject to frequent flooding and deposition of river-transported sediments originating from upstream sources. The plant canopy structure primarily is a mix of grasses and shrubs. This zone is a grass/shrub flood plain. The meadow zone comprises the second elevation step relative to the river. It has an elevated water table and is subject to occasional flooding but not the heavy deposition of river sediments that occurs in the riparian zone is managed for hay production and as pasture. A grass canopy is predominant throughout the zone.

   The bench zone is elevated approximately 20 feet above the meadow zone and forms eastern boundary of the ranch property. It is well above the long-term flood plain of the river and accordingly contains most of the principle buildings as well as the Visitor Center parking lot. In general, this zone is well drained and supports grass-dominated vegetation typical of xeric sites. This zone contains some wet areas less than 1 ha that support mesic vegetation.

    A check plot was established on the fluvial contour of Tin Cup Joe Creek, west-side tributary of the Clark Fork River. This site was on the Montana State Prison Ranch ill, 5 miles southwest of the Grant-Kohrs Ranch and 4 miles above the confluence of this tributary and the river (fig. 1). The alluvial oils of this check plot and the Grant-Kohrs Ranch site were formed of parent material from a similar Quaternary glacial outwash plain.

    During the period of operation of the Anaconda Smelter both ranches were on the 2 ug/g isopol for aerially deposited surface soil cadmium (1). Therefore, both ranches have been exposed to similar aerial deposition of metals. Any site differences in metal contamination are assumed to have originated from sediments deposited at Grant-Kohrs by the Clark Fork River.

    A control site, assumed to be free of significant anthropogenic metal contamination from water-borne sediments or aerial deposition, was established along the Blackfoot River 60 miles northwest of the Grant-Kohrs Ranch. This site is distant and upwind of the smelter and is 10 miles above the confluence of the Blackfoot and the Clark Fork Rivers (fig.1).

Soil and Vegetation Sampling Procedures

    Each of the ninety-four I-ha divisions of the ranch was subdivided into four square O.25-ha subplots. Random soil and vegetation samples were then collected from the proximate center of each of the four subplots and composited to form single samples for each of the ninety-four I-ha subdivisions.

   In each subplot of the hectare a 1.9-cm diameter soil core was extracted with a coring tube from the 0- to 25-cm increment of the soil profile. The four subplot core samples from each I-ha division were combined in a single plastic soil container. The stainless steel coring tube was dry brushed and cleaned with water, then five dummy cores were cut and discarded in the next hectare before collecting the four subplot samples from that hectare. Soils were air dried in closed cabinets before being capped and returned to the laboratory.

    In order to quantify any possible contamination of the soil samples from either the coring device or carryover of soil particles from one hectare to another, two types of field blanks were taken. Rinse blanks were taken by rinsing the coring tube with distilled water after completing the between- hectare cleaning. Approximately 100 mL of a distilled water rinse were collected in a Nalgene bottle. Scour blanks were taken by inserting the coring tube into wet technical grade laboratory silica sand after the between- hectare cleaning. The wet sand core was then transferred to a plastic soil container and treated like any soil sample. Field blanks were taken at a combined frequency of 10% of the soil samples.

    An aspect dominance survey confirmed that redtop bentgrass (Agrostis alba L.) was the most ubiquitous plant species on the ranch (12). It also is of major importance as a forage grass. Similar to the soils, a subsample of redtop bentgrass was collected from each of the four O.25-ha subplots and composited to form a single grass sample for each of the ninety-four 1-ha divisions. The grass was cut above any apparent rain splash line (more than 2.5 cm), placed in closed Kraft~ paper bags, and allowed to air dry prior to being sent to the laboratory. The grass subsamples were collected as close to the subplot center as plant distribution allowed in order to maintain the correspondence of the maps of metal distribution in vegetation and soils.

    The redtop bentgrass was collected for metal analysis at the end of July in order to provide an estimate of metal content of forage just prior to the first cutting of hay.

Study Site Selection and Sampling Transect Establishment for the Run of the River Sites

   The additional flood plain slickens sites were selected based on ease of access, topography, and sediment texture and color characteristics similar to the unvegetated slickens observed on the Grant-Kohrs Ranch. The four sites chosen (Rocker, Racetrack, Garrison, and Drummond) spanned approximately 75 miles parallel to Silver Bow Creek and the Clark Fork River (fig. 1). The Rocker and Racetrack sites were upstream of Grant-Kohrs while Garrison and Drummond were downstream.

    A 40-m transect was established at each slicken site. Soil samples (0-25 cm) were taken every 5 m along the transect line. The coring tube was lined with removable plastic cylinders, which also served as sample containers until the cores were transferred to bags for drying.

SOIL SAMPLE PREPARATION 

    The previously air-dried soil samples were dried at 35 °c in a forced draft oven for 48 hours. The soils were then transferred to heavy-gauge plastic bags, pulverized, and worked through a stainless steel Tyler- equivalent 9 mesh sieve which passes particles of 2 mm or less in diameter.

    Sample digestion procedures for potentially volatile metal determinations were adapted from Van Meter (19). A sample portion of 0.25 g weighed to four decimal places was transferred to a 2.5- by 25-cm acid washed Pyrex test tube. Ten millimeters of Instra-analyzed~ concentrated nitric acid was added and the tube sealed with an oxygen-methane torch. The sealed tubes were placed in capped sections of steel pipe and put into a 150 °C oven for 3 hours to digest the soil. After cooling, the Pyrex tubes were depressurized by opening a small vent hole with the torch then removed from the steel pipe. The tops of tubes were then removed with a glass tubing cutter. Undigested silicates and other resistant minerals were removed by filtration through glass wool. The sample solutions were then evaporated to dryness with gentle heat to remove the concentrated nitric acid. The dried samples were redissolved with 0.3 M Instra-analyzed@ nitric acid to a constant volume in 10 mL volumetric flasks. This solution was used for atomic absorption spectrophotometric analysis.  

VEGETATION SAMPLE PREPARATION 

    The previously air-dried plant samples were dried (24-48 hours) in Kraft paper bags to a constant weight in a forced draft oven at 35 °C. The vegetation was then ground in a Wiley Mill~ to pass a 20 mesh screen. Digestion procedures for 0.25 g aliquots of the vegetation samples were similar to those described for soils. The final sample solution was analyzed by atomic absorption spectrophotometry.

ATOMIC ABSORPTION SPECTROPHOTOMETRY

    Copper, arsenic, and cadmium determinations were made with a Varian AA-275BD atomic absorption spectrophotometer. The unit has a built-in deuterium background correction. Unless otherwise specified, all results are expressed as micrograms per gram on a dry weight basis.

    Copper was analyzed by standard flame aspiration technique with a typical instrument detection limit of 0.03 ug/g. Arsenic was analyzed by hydride vapor generation (Varian Model 65 Vapor Generator) with a detection limit of 0.0003 ug/g. Cadmium determination had an instrument detection limit of 0.03 to 0.05 ug/g by flame aspiration and 0.0007 ug/g by carbon rod analysis.

    Carbon rod analysis, although achieving lower detections, is less precise and more labor intensive than the other AA techniques so all samples were first screened by flame aspiration for copper and cadmium. For the Deer Lodge Valley samples all the soil determinations exceeded the flame or hydride detection limits, as did all the vegetation determinations except 19% of the plant cadmium values.  

QUALITY ASSURANCE FOR CHEMICAL ANALYSIS  

    Five certified federal reference standards representing a variety of sample matrices were regularly analyzed to assess the accuracy of the procedures. These were National Bureau of Standards Reference Materials Orchard Leaves (/11571), River Sediment (/11645) and Estuarine Sediment (/11646) plus EPA Municipal Digested Sludge and Trace Elements in Water reference standards. The average absolute percent bias (the deviation of the measured value from the certified mean) was calculated to estimate accuracy of the procedures.

    Replicate analyses were performed on the same samples and the average relative standard deviations calculated to quantify the precision of the methods.

    Percent recovery of standard additions was calculated for all three elements including both flame aspiration and carbon rod techniques for cadmium.

    Quality assurance determinations including field blanks for soil samples accounted for 24% of the analytic effort.

 CHAIN OF CUSTODY DOCUMENTATION

    Documentation of sample transfers, dates, and personnel involved in collection and processing were maintained for purposes of quality assurance and in the event that the metal data have relevancy to litigation or other legal process activities.

Data Analysis

    Normal probability plots of the chemical data subjected to a number of transformations indicated that the assumption of normality was reasonably met by logl0 transformations.

    Variances for between-group comparisons were generally not homogeneous. All of the standard analysis of variance calculations indicated significant differences between groups. Because of the unequal variances, these significant inferences were confirmed by both the Welch and the Brown-Forsythe one- way analysis of variance methods, which do not assume homogeneity.

    Dunnett's procedure (16) was used to contrast the apparently higher metal concentrations of soil and-grass samples for the four ranch zones to the check plot values. The standard error terms derived from the overall ANOVA's were used in this method. While moderate departures from homogeneity do not have serious consequences for the overall test of significance, single degree of freedom comparisons may be far from accurate (11).

    Mean values were reported as the back transformation of the logarithmic values for mean unless indicated otherwise.

Soil Enzyme Activity Methods

    A labeled hydrogen oxidation bioassay for soil microbial activity was performed on soil collected from a representative unvegetated slicken in the riparian zone. Certain metabolic processes of soil microbes consume hydrogen. The microbial conversion of the tritium isotope of hydrogen gas to water can be used to determine the level of soil microbe enzyme activity (13). Heavy metals at high concentrations are known to suppress enzyme activity.

A Teflon-coated spade was used to take triplicate 1-km soil samples from the top 10 cm at each of three locations:

1. An unvegetated slicken in the riparian zone of the Grant-Kohrs Ranch

2. A vegetated area near the perimeter of the unvegetated slicken which provided the first triplicate sample

3. The check plot on Tin Cup Joe Creek

   The soil samples were then air dried and sent to Dr. Robert D. Rogers, EG & G Idaho, Inc., Idaho National Laboratory, who performed the soil enzyme activity bioassay.

RESULTS AND DISCUSSION

Distribution of Heavy Metals on the Grant-Kohrs Ranch

    Major unvegetated slickens were identifiable from aerial photos, and were mapped in figure 2, along with the major topographic features of the Grant- Kohrs Ranch. Major slickens accounted for 0.6% of the ranch surface area.
   
    The average concentrations of total copper, arsenic, and cadmium in the 0- to 25-cm soil profile and the forage grass Agro8ti8 alba L. are summarized in table 3 for each of the four zones of the Grant-Kohrs Ranch along with the Tin Cup Joe Creek check plot and the Blackfoot control plot. Copper concentrations were greater than arsenic concentrations while cadmium occurred at the lowest levels. Soil concentrations of all three metals were one to two orders of magnitude greater than the grass concentrations.

    The concentrations of the three metals in the soils of the four zones of the Grant-Kohrs Ranch and the check plot were greatly elevated relative to the metal content of the soil from the control plot. The usual concentration of copper in uncontaminated soils ranges from 20 to 50 ug/g (1). Some workers have suggested wider ranges of 2 to 100 ug/g for soil copper, with a mean value of 20 ug/g (1). The natural arsenic content in virgin soils is reported as ranging from 0.1 to 40 ug/g, with an average of 5 to 6 ug/g (8). The cadmium content of typical soils is less than I ug/g (10,11). Vinogradov (20) reported a worldwide soil average of 0.5 ug/g. We suggest that the median value from the published literature is approximately 0.3 ug/g although even lower median values have been reported for many regions (4,18). The lower metal concentrations of the Blackfoot control plot soil are in good agreement while all soils collected in the Deer Lodge Valley contain elevated concentrations of these metals.

    The entire Deer Lodge Valley was subjected to aerial deposition of heavy metals during the 1884-1980 period of metal smelting at Anaconda (6.7.17). Munshower (7) had established that the Grant-Kohrs Ranch and the Tin Cup-Joe Creek check-plot, which is located on the Montana State Prison Ranch #1, are both on the approximate 2 ug/g isopol for surface soil cadmium resulting principally from aerial deposition. The metal levels of soils and grass from the four zones of the Grant-Kohrs Ranch were compared to soil and grass samples collected from the Tin Cup Joe Creek check plot in order to evaluate additional contamination originating from metal-laden sediments carried by the Clark Fork River and deposited on the flood plain of the ranch.

    In contrast to the check plot (51 ug Cu/g). mean soil copper at the Grant-Kohrs Ranch was significantly higher for the riparian zone (1630 ug Cu/g. p less than 0.01) and the meadow zone (184 ug Cu/g. p less than 0.05). Grant-Kohrs Ranch soil arsenic means were significantly higher for the riparian zone (176 ug As/g. p less than 0.01), the meadow zone (49 ug AS/g. p less than 0.05), and the Cottonwood Creek zone in contrast to the check plot (20 ug As/g). The Grant-Kohrs Ranch riparian zone mean soil cadmium concentration (5.0 ug Cd/g. p less than 0.05) was significantly higher than the average cadmium content (1.7 ug Cd/g) of the soil on the check plot.

    Copper content of vegetation in uncontaminated areas typically averages from 4 to 15 ug/g. While some plant species are reported as naturally accumulating higher levels (up to 50 ug/g), the grasses average 5 ug/g in clean areas (2,9). Grass arsenic means for uncontaminated areas are reported between 0.l and 0.7 ug/g, although other plant species from untreated soils may contain a wider range (0.01 to 5 ~g/g) of arsenic (8). The cadmium content of foliage from uncontaminated areas is less well defined. Much of the data reported prior to the 1970's as representative of vegetation cadmium levels (approximately 1 ug/g) for clean areas is unreliable (10,14). For example, use of atomic absorption spectrophotometry without proper background corrections gave erroneously high values for many biological materials. We suggest that the typical concentration of cadmium in foliage from uncontaminated areas is 0.05 ug/g or less. In contrast the copper, arsenic, and cadmium content of grass (Agrostis alba L.) collected in the Deer Lodge Valley was higher than the concentrations suggested as typical of grasses from uncontaminated areas.

    As with the soils, a statistical comparison was made of the mean metal content of the grass samples from the four zones of the Grant-Kohrs Ranch with the check plot means of metal concentration in grass. Only copper in grass from the riparian zone (10.4 ug Cu/g. p less than 0.05) was significantly elevated relative to the check plot (7.2 ug Cu/g).
    
    The measured arsenic concentrations of the grass samples (4 to 7 ug As/g) and some of the soil samples (maximum 1103 ug As/g) (table 2) collected at the perimeter of denuded areas in the Grant-Kohrs Ranch riparian zone in October 1982 were considerably higher than the means for the July 1983 samples. While the July 1983 samples were randomly selected, the October 1982 samples were taken from suspected areas of high metal contamination chosen because of the apparent impoverishment of the plant community. Compositing of the four grass or soil subsamples from each hectare before chemical analyses also contributed to lower July 1983 values by mechanically averaging the samples. Species accumulation and tolerance differences would influence the foliar metal levels. The fall 1982 grass was Deschampsia cespitosa (L.) Beauv. Finally, the metal content of the grasses can be expected to increase with progression of the growing season and curing of the grass. Munshower (7) demonstrated this seasonal increase for grass cadmium levels which reached a maximum in September and October; however, his data cannot be directly extended to the present conditions in the Deer Lodge Valley because an. undefined portion of the seasonal increase resulted from aerial deposition of stack emissions from the Anaconda Smelter operating at that time.

    Schematic maps of copper, arsenic, and cadmium content (ug/g, dry weight) of the soils and grass portray the values measured as representative of each of the ninety-four l-ha divisions of the Grant-Kohrs Ranch (figs. 4a to 4f).

    The highest metal concentrations are in the riparian flood plain zone comprising the western two-fifths of the ranch. Within the riparian zone the highest metal levels are generally found adjacent to the river or in association with partially closed old channels or sloughs. These appear to be areas of heavy deposition of sediments carried by the river during previous flood events. Although hectares with a high proportion of the soil surface denuded of vegetation tend to exhibit high metal concentrations, other hectares with fuller plant cover also can have high metal levels. 

    In the meadow zone the hectares with the highest metal contamination levels are generally adjacent to the Kohrs-Manning Irrigation Ditch. Overflow and sediment deposition during flood periods or maintenance removal of sediments from the irrigation ditch are likely responsible for the elevated metals found along the ditch.

   The high bench zone of the Grant-Kohrs Ranch is not subject to flooding from the Clark Fork River. This comprises the eastern boundary of the ranch. It contains most of the principal buildings and sheds as well as the Visitor Center and parking lot. The metal levels appear to be relatively uniform throughout this zone. Grass cadmium concentrations from the ranch house north are equal to, or less than, the 0.03 ug Cd/g detection limit.

    The copper, arsenic, and cadmium concentrations in the soils of the Grant-Kohrs Ranch were very highly correlated (Table 4). Soil copper concentrations could be used to estimate soil arsenic (r2 = 0.835) or soil cadmium (r2 = 0.880) concentrations, reducing analytic cost in certain types of potential future studies. Analysis for copper alone would allow the use of less costly soil digestion procedures and eliminate the need for hydride vapor generation for arsenic and carbon rod analysis for low-level cadmium. 

    The correlations between the metals in grass were significant (p less than 0.05), but the coefficients of determination were so low (As -Cu r2 = 0.188; Cd- Cu r2 = 0.250; As- Cd r2 = 0.064) that any regression equation would be an imprecise predictor. Soil with grass metal concentrations were also significantly (p less than 0.01) correlated for all three metals, but again the coefficients of determination were so low (Cu r2 = 0.305; As r2 = 0.077; Cd r2 = 0.101) as to preclude using soil metal concentration to estimate precisely the metal content of the grasses. Variable physicochemical characteristics of the soil, such as pH, organic matter, clay particles, and metal ion competition, combine with variable plant physiological factors to reduce the strength of soil metal and plant metal correlations as well as correlations between metals in the plant under field conditions. If the metals induce toxic effects on the vegetation, the roots are likely to be the site of principal toxic effects. Injury to the root system would reduce shoot uptake of the metals. Hectares with high soil metal concentrations will tend to have vegetation with high metal levels but precise estimates of plant concentrations will require chemical analysis of the vegetation of interest.

Heavy Metals Along the Length of the River Flood Plain

Alluvial sediments contaminated with high concentrations of copper, arsenic and cadmium have been deposited along at least a 75-mile length of the Clark Fork River riparian zone (Table 5). These unvegetated flood plain sites were located both downstream and upstream of the Grant-Kohrs Ranch. Soil concentrations of all three metals were at least two orders of magnitude above expected levels for noncontaminated soils. 

   The Rocker site is between the upstream Butte mining area and the downstream Warm Springs Settling Ponds and the Anaconda Smelter; this site is actually on Silver Bow Creek, which drains the Butte mining area and forms a major upper tributary of the Clark Fork River. The Drummond site, which was the farthest downstream, had the highest mean copper and cadmium levels and the second highest arsenic concentrations. The data for these four riverside sites, in conjunction with the soil metal concentrations measured on the Grant-Kohrs Ranch riparian zone, clearly show that portions of the flood plain along the entire upper Clark Fork drainage are heavily contaminated with toxic metals and that metal concentrations do not necessarily decline with distance downstream of the two suspected major source areas. River sediment sorting, which occurs by particle size and deposition processes probably, determines the distribution of toxic metal sediments in the drainage.

   Woessner and others (21) have documented the presence of sediments contaminated by toxic metals in deposits behind the Milltown Dam on the Clark Fork River at Missoula. They have shown that metals entrained behind the dam have entered the ground water table and contaminated drinking water wells in the adjacent community. Thus, the documented length of contaminates in the river include the 120 miles between Butte and Missoula or approximately 40 miles upstream and 80 miles downstream of the Grant-Kohrs Ranch.

Soil Microbe Enzyme Activity Levels

    The bioassays indicated that soil microbial enzyme activity was drastically depressed in all three replicate soil samples collected from the intensive study slickens on the Grant-Kohrs Ranch. The activity level of the three soil samples from the adjacent vegetated area was within the normal range. One sample from the check plot was 15% below normal so that the mean measured activity was reduced 5% (Table 6).

Quality Assurance of Metal Analyses and Soil Sampling

    Rinse blanks and sand scour blanks used to quantify potential soil sample contamination resulting from the soil sampling procedure were below the instrumental detection limits for 24 out of 27 (89%) of the sampling blank determinations (Table 7). The maximum detected copper content of a sampling blank was 14% of the observed minimum soil sample copper concentration and 0.6% of the average soil copper level. The maximum detected arsenic content of a sampling blank was 4% of the minimum soil sample arsenic concentration and 0.2% of the average soil arsenic content. The maximum detected cadmium content of a sampling blank was 0.1% of the minimum soil sample cadmium content and 0.025% of the average soil cadmium content. These data indicate that cross-contamination of soil samples was an insignificant source of experimental error.

   The average percent recovery of standard additions ranged from 81.6% for bon rod analysis of cadmium to 110% for hydride generation for arsenic (Table 8). The precision of replicated analysis of the same samples ranged 2.6% for flame aspiration of copper to 12.2% for carbon rod analysis of cadmium. Precision characteristically declines at the lower detection limits obtained by using the carbon rod. Both recovery and precision are well within desirable technical objectives. 

   The average absolute percent bias determinations for the five certified standards are summarized in table 8. With the exception of the arsenic determinations for the EPA Municipal Digested Sludge reference standard, bias ranged from 2.6% to 25%. The high bias for the sediment standards was to be expected as determinations of soil and soil sediment type materials are more variable than grass and water standards because of the complexity of the sample matrices. The calculated 72.9% bias for the municipal digested sludge standard was resultant from our determinations being low relative to the certified estimate of the mean. In spite of the 72.9% calculated bias our laboratory's determinations were still well within the certified 95% confidence band for this analyte (certified x = 17 ug As/g; certified 95% confidence band 0- 88.9 ug As/g).

    The Orchard Leaves and River Sediment reference standards were most similar to the grass and soil sample unknowns analyzed for this study. Overall, the bias determinations were acceptable to meet the objectives of the study.

 LITERATURE CITED

1. Bowen, H.J.M. 1966. Trace elements in biochemistry. Academic Press.

2. Demayo. A.; Taylor. M.C.; and Taylor. K.W. 1982. Effects of copper on humans, laboratory and farm animals, terrestrial plants and aquatic life. CRC Critical Reviews in Environmental Control 12(3): 183-257.

3. Doyle. J.J. 1977. Effects of low levels of dietary cadmium in animals. a review. J. Environ. Qual. 6(2): 111-116.

4. Flynn, A.; Granzmann, A.W.; Arneson, P.D.; and Oldenmeyer, J.L. 1975 Seasonal rhythms of cadmium and lead in moose. International. Conference on Heavy Metals in the Environment, Toronto, Ontario, October 27-31. Abstract p. C-230-232.

5. Friberg, L.; Piscator, M.; Nordberg, G.F.; and Kjellstrom, T. 1974. Cadmium in the environment. CRC Press, Cleveland, OH. 166 p.

6. Hartman, L. 1976. Fungal flora of the soil as conditioned by varying concentrations of heavy metals. Ph.D. dissertation. University of Montana. 147 p.

7. Munshower, F.E. 1972. Cadmium compartmentation and cycling in a grassland ecosystem in the Deer Lodge Valley, Montana. Ph.D. dissertation, University of Montana. 106 p.

8. .NAS. 1977a. Arsenic. Medical and biological effects of environmental pollutants. National Academy of Sciences, Washington, DC. 332 p.

9. NAS. 1977b. Copper. Medical and biological effects of environmental pollutants. National Academy of Sciences, Washington, DC. 115 p.

10. Nordberg, G.F. 1974. Health hazards of environmental cadmium pollution. Ambio 111(2): 55-66.

11. Page, A.L.; and Bingham, F.T. 1973. Cadmium residues in the environment. Residue Reviews 48: 1-44.

12. Rice, P.M.; and Ray, G.J. 1984. Floral and faunal survey and toxic metal contamination study of the Grant-Kohrs Ranch National Historic Site. Technical Report. National Park Service Contract No. CX- 1200-2-BO44. Gordon Environmental Laboratory, Botany Department. University of Montana. 92 p.

13. Rogers, R.D.; and McFarlane, J.C. 1982. A possible toxic-effects indicator. Hydrogen oxidation in soils as J. Environ. Qual. 11: 364-368.

14. Schuller, P.L.; and Egan, H. 1976. Cadmium, lead, mercury, and methyl mercury compounds. A review of methods for trace analysis and sampling with special reference to food. Food & Agriculture Organization of the United Nations FAO/WHO, Rome. 97 p.

15. Sokal. R.R.; and Rohlf. F.J. 1981. Biometry. The principles and practice of statistics in biological research. 2nd ed. W.H. Freeman & Go. San Francisco. 859 p.

16. Steele, R.G.D.; and Torrie, J.H. 1960. Principles and procedures of statistics with special references to the biological sciences. McGraw-Hill, New York. 481 p.

17. Taskey, R.D. 1970. Soil contamination at Anaconda, Montana: History and influence on plant growth. M.S. Thesis, University of Montana. 145 p.

18. Tjell, J.C.; and Houmand, M.F. 1978. Metal concentrations in Danish arable soils. Acta Agric. Scand. 28: 81-89.

19. Van Meter, W. 1974. Heavy metal concentration in fish tissue of the upper Clark Fork River. Montana University Joint Water Resources Research Center, Bozeman, MT. 37 p.

20. Vinogradov, A.P. 1959. The geochemistry of rare and dispersed elements in soils. Consultant Bureau, New York.  

21. Woessner, W.W., Moore, J.N., and Johns, C. water supply study, Militown, Montana. Geology, University of Montana, Missoula. 1984. Arsenic source and Interim Report, Dept. of 43 p.