Environmental Earth Sciences

, Volume 70, Issue 3, pp 1057–1082

A GIS and statistical approach to identify variables that control water quality in hydrothermally altered and mineralized watersheds, Silverton, Colorado, USA

Authors

    • U.S. Geological Survey
  • Raymond H. Johnson
    • U.S. Geological Survey
  • Barnaby W. Rockwell
    • U.S. Geological Survey
  • Jonathan Saul Caine
    • U.S. Geological Survey
  • Kathleen S. Smith
    • U.S. Geological Survey

DOI: 10.1007/s12665-013-2229-y

Abstract

Hydrothermally altered bedrock in the Silverton mining area, southwest Colorado, USA, contains sulfide minerals that weather to produce acidic and metal-rich leachate that is toxic to aquatic life. This study utilized a geographic information system (GIS) and statistical approach to identify watershed-scale geologic variables in the Silverton area that influence water quality. GIS analysis of mineral maps produced using remote sensing datasets including Landsat Thematic Mapper, advanced spaceborne thermal emission and reflection radiometer, and a hybrid airborne visible infrared imaging spectrometer and field-based product enabled areas of alteration to be quantified. Correlations between water quality signatures determined at watershed outlets, and alteration types intersecting both total watershed areas and GIS-buffered areas along streams were tested using linear regression analysis. Despite remote sensing datasets having varying watershed area coverage due to vegetation cover and differing mineral mapping capabilities, each dataset was useful for delineating acid-generating bedrock. Areas of quartz–sericite–pyrite mapped by AVIRIS have the highest correlations with acidic surface water and elevated iron and aluminum concentrations. Alkalinity was only correlated with area of acid neutralizing, propylitically altered bedrock containing calcite and chlorite mapped by AVIRIS. Total watershed area of acid-generating bedrock is more significantly correlated with acidic and metal-rich surface water when compared with acid-generating bedrock intersected by GIS-buffered areas along streams. This methodology could be useful in assessing the possible effects that alteration type area has in either generating or neutralizing acidity in unmined watersheds and in areas where new mining is planned.

Keywords

Watershed Hydrothermal alteration Water quality GIS Statistical analysis

Introduction

Abandoned mine lands (AML) investigations conducted by the U.S. Geological Survey between 1996 and 2001 in the Animas River watershed (ARW) near Silverton, Colorado, USA, resulted in one of the most comprehensive geodatasets for a watershed affected by hard-rock metallic mining (Church et al. 2007). Information on ARW geology, geophysics, geochemistry, hydrology, biology, remote sensing, base geographic, and digital terrain information is archived in a relational database and in geographic information system (GIS) coverage formats (Yager and Bove 2007; Sole and Rich 2007; Smith et al. 2007; Kimball et al. 2007; Bove et al. 2007a, b; Dalton et al. 2007). Several AML datasets were analyzed for this study. Federal land managers and the local ARW Stakeholders group are actively involved in remediation of abandoned mines in the ARW, a region that was the site of base-metal (Cu–Pb–Zn) and precious-metal (Ag–Au) hard rock mining from the late 1870s to 1991. An extensive legacy of mining has left behind hundreds of mines and prospects. Weathering of mine waste and hydrothermally altered bedrock that hosts the metallic mineral deposits causes oxidation of sulfide minerals forming acidic and metal-rich leachate that adversely impacts water quality and is toxic to aquatic life (Church et al. 2007; DeGraff 2007).

A principal goal of this study was to analyze AML datasets to advance our understanding of the physical factors that control water quality at the watershed scale (1.5–20 km2). Knowledge of the most important variables that produce acid rock drainage (ARD) is especially useful to land managers and mining companies to aid in identifying areas posing the greatest environmental risks where new mining will occur and that may present reclamation challenges once mining ceases. In contrast, when watersheds have physical variables that can reduce acidity, such as rocks with acid neutralizing capacity (ANC), this information is also useful in mine site planning and remediation efforts, as it can be used to help mitigate acid mine drainage and sequester metals (Yager and Caine 2003; Yager et al. 2005, 2008a, b).

Geographic and geologic setting

The San Juan Mountains exhibit a wide diversity of ecological characteristics due to their mid-latitude location, high elevation, extreme topographic relief, and consequent ecosystem variability. The San Juan Mountains encompass subalpine parks, grasslands, and wetlands that contain multiple ecosystems including alpine, spruce fir; mixed conifer, ponderosa pine, oak, Douglas fir, and aspen forests; parks and meadowlands, mountain shrub, and pinon-juniper woodlands; and shrub-steppe communities (Chapman et al. 2006). Relatively high regional precipitation, occurring primarily as snow in the fall and winter, results in the critical water recharge area to the semi-arid Colorado River basin.

The geology of the western San Juan Mountains study area is exceptional in that many diverse rock types representing geologic eras from the Precambrian to Cenozoic are preserved. The general stratigraphy of the study area consists of Precambrian crystalline basement rocks overlain by Paleozoic to Tertiary sedimentary rocks and a voluminous Oligocene volcanic cover (Yager and Bove 2007). The volcano-tectonic framework of the study area consists of early intermediate composition volcanism overprinted by caldera-forming eruptions. An arcuate primary drainage network delineates structural margins of the San Juan and the nested Silverton calderas (Fig. 1). Structures that developed coincident with caldera collapse served as flow paths for hydrothermal fluids during several episodes of hydrothermal alteration and base- and precious-metal mineralization (Casadevall and Ohmoto 1977; Bove et al. 2001; Yager and Bove 2007). The northeast-trending Eureka graben served as an important flow path for mineralization and formed in response to caldera resurgence of the San Juan and Silverton caldera complex. Where not locally overprinted by intense hydrothermal alteration, a regional propylitic alteration assemblage containing calcite, chlorite, and epidote pervades the entire study area, locally supplying ANC and alkalinity to surface water (Yager et al. 2005, 2009; Yager 2008). The study area was extensively glaciated during the Pleistocene epoch and has high topographic relief ranging in elevation from 2,800 m to over 4,200 m providing for excellent bedrock exposures.
https://static-content.springer.com/image/art%3A10.1007%2Fs12665-013-2229-y/MediaObjects/12665_2013_2229_Fig1_HTML.gif
Fig. 1

Location map of the Animas River watershed study area showing geographic features discussed in text. Shaded areas are RMMD Red Mountain mining district, EMD Eureka mining district, and SSMA South Silverton mining area. Note that E. of Carbon Lakes and Red Tributaries are informal names used to describe features discussed in the text. Red line Silverton caldera structural margin, purple dashed line San Juan caldera southern topographic margin

Previous work

Geochemical investigations in the study area

Issues addressed as part of past AML research included identifying the relative contributions of ARD and associated metal loads to surface water from mined watersheds and naturally altered and unmined areas. Research described in Bove et al. (2007b) focused on determining water quality geochemical characteristics of surface water and groundwater that is sourced in rocks having well-characterized geologic units and hydrothermal alteration types. Water sampled as part of the Bove et al. (2007b) study was classified as being either mining- or non-mining affected. Statistics were used to indicate the range in concentration for multiple metals, sulfate, and pH. The study characterized the chemistry of waters sourced in highly and weakly altered rock types, but no attempt was made to determine how watershed water chemistry might be influenced as a function of differing proportions and types of watershed variables. Mast et al. (2000b) used a geochemical mass-balance approach to determine that un-mined, hydrothermally altered watersheds can contribute as much as 56–90 % of dissolved metal loads to surface waters, but the possible influences of multiple geologic watershed variables on water quality were not assessed.

GIS and statistical approaches used for other study areas

Previous studies in other areas have utilized watershed-based GIS and statistical approaches to evaluate the influences of watershed physical characteristics on water quality. Wang and Yin (1997) used GIS to assess the effects of variable land use and land cover type on surface water quality. Investigations in subalpine to alpine areas have addressed the influence that terrain slope, vegetation, surficial deposits, and bedrock lithology have on water quality (Clow and Sucker 2000). Sliva and Williams (2001) used analysis of entire watershed area and a linear GIS stream buffer to assess influences of land use types on water quality.

A watershed-based and statistical approach was used in the central Colorado Rocky Mountains by Schmidt et al. (2009, 2010), which includes GIS analysis of generalized alteration types derived from analysis of advanced spaceborne thermal emission and reflection radiometer (ASTER) data, water quality determined for samples collected at watershed outlets, and macro benthic invertebrate data as an indicator of ecosystem health. Their work identifies inverse correlations between area of hydrothermal alteration and macro benthic taxa type and abundance. An approach similar to that of Schmidt et al. (2009) was used by Bruns (2005), who developed a GIS and statistical analysis methodology of watersheds to understand stream ecosystem response to potential land use impacts caused by coal mining in northeastern Pennsylvania and southern New York.

Methodology and datasets

Methods used in this study combined a GIS and a statistically based approach that builds on preliminary work (Yager et al. 2002; Yager and Caine 2003; Yager et al. 2008a, b). In this study, GIS was used to analyze digital terrain surfaces to create derivative watersheds and hydrologic stream networks that, in turn, could be used to analyze AML geology and alteration datasets in a geochemical and geologic context. The watersheds are used as a geographic frame of reference for determining area of geologic units and alteration types.

Work completed as part of this study differs from previous research in that data acquired using GIS analysis of geologic and detailed alteration maps completed during AML investigations (Bove et al. 2007a; Yager and Bove 2007) were used to test statistical correlations between water quality signatures and hydrothermal alteration mapped using three different remote sensing platforms. The remote sensing datasets utilized in this study included: (1) Landsat Thematic Mapper (TM) (Rockwell 2010), (2) ASTER (Rockwell 2010), and (3) Airborne Visible Infrared Imaging Spectrometer (AVIRIS) (Bove et al. 2007a).

Geomorphometric and GIS methods

A surface hydrologic modeling tool, the GIS Weasel (Viger et al. 1998; Viger and Leavesley 2007; Viger 2008), was used to analyze a 10 m resolution USGS digital elevation model (DEM) and extract watershed data including slope, aspect, flow direction, flow accumulation, and drainage network grids. Standard DEM preprocessing commands available in ArcGrid™ were utilized that are accessed by the GIS Weasel software application. The DEM preprocessing steps (fill, flow direction, and flow accumulation) are required for subsequent geomorphometric analysis because these functions control how hydrologic flow is routed on a digital surface having grid-cell xy coordinate locations and elevation information. Fill is calculated using a technique described by Jenson and Domingue (1988) that locates each cell whose elevation is lower than that of eight adjacent neighboring cells and raises the elevation of these cells to that of its nearest neighbors. Once erroneous topographic depressions are corrected using Fill, this permits the downhill flow across a DEM to be determined. Flow direction is calculated from the filled DEM, using the D8 algorithm (Jenson and Domingue 1988). The flowdirection surface is then used to calculate a flow accumulation surface, which delineates those higher elevation cells that contribute flow to a specific grid cell (Jenson and Domingue 1988; O’Callaghan and Mark 1984). A hydrologic stream network grid was established using the flow accumulation (flowacc) grid and the following ArcGrid™ conditional (con) function:
$$ {\text{stream}}{\text{net }} = {\text{ con}}\;\left( {{\text{flowacc}}\,{ > }\, 1,000,\, 1} \right) $$

A flow accumulation threshold of 1,000 was used in the Streamnet function. All 10-m resolution cells in the flow accumulation grid with more than 1,000 cells flowing into them are assigned 1, and all other cells are assigned no data. The grid-cell contributing area of 1,000 used in this study resulted in a realistic, basin-wide hydrologic network that most closely resembled the drainage network observed in the field.

The GIS Weasel software utilizes the Watershed function of ArcGrid™ to delineate watersheds that represent upstream contributing areas derived from a source grid (Viger and Leavesley 2007). Each delineated watershed has a unique numeric identifier that can be associated with attributes of other geodatasets and provides a basis for spatial and statistical comparisons (Fig. 2). Once watersheds were delineated, GIS Weasel parameterization routines were used to calculate areas of alteration and bedrock lithologies in each watershed derived from digital geologic and alteration maps (Yager and Bove 2007; Bove et al. 2007a). In addition, an AVIRIS-derived vegetation map (Dalton et al. 2007) was analyzed using GIS to determine intersections between vegetated and barren areas corresponding with alteration types for each watershed. The zonal statistics function available in ArcGIS™ was used to calculate the area of each watershed alteration type derived from the ASTER and TM alteration grids. Tables of watershed alteration types were imported into Mintitab™ software for statistical analysis.
https://static-content.springer.com/image/art%3A10.1007%2Fs12665-013-2229-y/MediaObjects/12665_2013_2229_Fig2_HTML.gif
Fig. 2

Watersheds and associated id’s delineated using the GIS Weasel software interface and the ArcGridTM watershed function. Watersheds in green (near neutral pH), coral (5.0–6 pH), red (<5 pH). GIS id’s are only shown for areas sampled at the watershed outlet or discussed in text. The id’s and corresponding names are listed as follows: 2 Dry Gulch, 3 South of Dry Gulch, 4 Prospect Gulch, 6 Tiger Gulch, 7 Georgia Gulch, 9 Fairview Gulch, 10 Cascade Gulch, 11 Porcupine Gulch, 13 Ohio Gulch, 14 Illinois Gulch, 16 Topeka Gulch, 17 Niagra Gulch, 19 Hancock Gulch, 25 Minnesota Gulch, 26 North Fork Cement Creek, 29 South Fork Cement Creek at Gladstone, 36 Ross basin, 38 Boulder Gulch, 39 Blair Gulch, 40 Swansea Gulch, 41 Idaho Gulch, 42 Kendall Gulch, 44 Northeast of Belcher mine, 46 Zuni Gulch, 47 Battle Ship Slide, 52 Red Tributary, 53 Middle Fork Mineral Creek (east watershed), 54 Middle Fork Mineral Creek (central watershed), 55 Middle Fork Mineral Creek (western watershed), 56 Paradise basin, 62 South of Brown’s Gulch, 63 South of Imogene Mine, 64 West of Imogene Mine, 65 Southeast of Chattanooga, 69 Big Horn Gulch, 70 Carbon Lake, 71 Arrastra Gulch, 73 Hematite Gulch, 75 Eureka Gulch, 76 Niagra Gulch (upper Animas River basin), 79 Picayune Gulch, 82 Placer Gulch, 83 California Gulch, 87 Cunningham Creek, 92 South of Bonner mine, 93 Brown’s Gulch

A 30-m GIS buffer was applied to the hydrologic network grid to determine areas adjacent to streams that intersect alteration types. GIS buffer-derived data were subsequently compiled for statistical comparison with results from whole watershed statistical analysis. GIS-buffered areas adjacent to a stream network could represent locations within a watershed that are more susceptible to hydrologic weathering due to sediment being transported along perennial and ephemeral stream channels where the erosive power of surface water runoff may be highest. The GIS stream buffer was only applied to the AVIRIS-based alteration map in which alteration types were extrapolated to include vegetated areas, resulting in nearly complete watershed mineral mapping coverage. The only unmapped parts of the AVIRIS-based alteration map are those areas covered by surficial deposits.

Remote sensing data

Landsat Thematic Mapper

TM measures six spectral bands between 0.4 and 2.4 μm. Bands 1, 2, and 3 overlap the visible, blue, green, and red parts of the electromagnetic spectrum between 0.4 and 0.7 μm. Bands 4 (0.7–0.9 μm), 5 (1.5–1.8 μm), and 7 (2.0–2.4 μm) measure in the near- to shortwave infrared (SWIR) parts of the spectrum. Absorption spectra detected by TM are too broad to delineate specific minerals, but the spectra do permit mineral groups such as clays and iron minerals important in hydrothermally altered areas to be mapped. The spatial resolution for TM data is 30 m. In the Animas River study area, only 2–3 % areas above tree-line that are devoid of vegetation are adequately exposed for TM mineral group mapping. TM band ratios are used to reduce the effect of uneven illumination and enhance spectral contrast (Rockwell 2010). A compound ratio of (5/7–4/3) was used to delineate the clay–sulfate–sericite minerals. The spectral influence of chlorophyll-bearing green vegetation (highlighted by the 4/3 ratio) was removed from the 5/7 ratio results. A compound band ratio of (3/1 × (3 + 5/4)) was used to highlight ferric iron minerals having strong charge transfer absorptions in band 1 and crystal field absorptions in band 4 (Fig. 3). TM data are available for the entire United States and are, therefore, pertinent to national-scale geoenvironmetal assessments. Alteration types mapped by TM data and area calculations determined by GIS are in Table 1.
https://static-content.springer.com/image/art%3A10.1007%2Fs12665-013-2229-y/MediaObjects/12665_2013_2229_Fig3_HTML.gif
Fig. 3

Distribution of mineral groups identified using ratio-based analysis of TM data. Yellow ferric iron + clay–sulfate–mica, Green clay–sulfate–mica only, Red ferric iron only

Table 1

Alteration type area (m2) determined by GIS analysis of Landsat–TM and ASTER alteration mapping for Cement Creek, Mineral Creek, and upper Animas River basin watersheds

Basin and Wathersheds

TM–clay–sulfate–sericite

TM–Fe

ASTER argillic +ferric Fe

ASTER quartz–sericite–pyrite (phyllic)

∑(all ASTER quartz–sericite–pyrite)

Cement Creek basin

Dry Gulch

485,100

538,200

169,425

135,225

135,225

South of Dry Gulch

0

0

0

0

0

Prospect Gulch

361,800

633,600

166,725

165,825

165,825

Tiger Gulch

0

4,500

1,350

1,575

1,575

Georgia Gulch

26,100

70,200

1,575

0

225

Fairview Gulch

0

900

0

0

0

Cascade Gulch

0

6,300

900

675

675

Porcupine Gulch

45,000

52,200

19,575

18,900

18,900

Ohio Gulch

302,400

327,600

92,250

78,975

79,875

Illinois Gulch

0

0

2,250

1,575

1,575

Topeka Gulch

99,900

139,500

42,750

37,800

38,025

Niagara Gulch

81,900

134,100

49,500

75,825

75,825

Hancock Gulch

0

1,800

225

1,350

1,350

Minnesota Gulch

0

40,500

6,750

1,125

2,475

South Fork Cement Creek (above M. Fork)

141,300

158,400

32,850

69,750

70,650

North Fork Cement Creek

29,700

98,100

5,850

17,325

17,325

Ross basin

0

16,200

0

0

0

Mineral Creek basin

Northeast of Belcher Mine

1,748,700

0

450

0

675

Zuni

1,351,800

36,900

48,375

39,600

39,825

Battleship slide

813,600

27,000

20,250

11,250

13,500

Red Tributary

2,673,900

156,600

48,600

13,050

14,625

South of Bonner Mine

612,900

900

1,125

0

0

Middle Fork Mineral Creek-East Basin

476,100

0

0

0

450

Middle Fork Mineral Creek-Central Basin

492,300

7,200

0

450

450

Middle Fork Mineral Creek-West Basin

1,165,500

34,200

6,075

2,025

2,250

Paradise Basin

2,814,300

85,500

13,275

2,025

5,175

South of Browns Gulch

2,406,600

228,600

52,425

91,800

91,800

Browns

2,124,900

75,600

18,675

24,975

24,975

South of Imogene Mine

833,400

1,800

2,475

675

675

West of Imogene Mine

409,500

0

1,575

900

900

Southeast of Chattanooga

811,800

0

1,350

1,125

1,125

Big Horn Gulch

3,865,500

15,300

10,575

0

0

East of Carbon Lake

666,000

0

2,025

1,350

1,350

Upper Animas River basin

Boulder Gulch

0

0

3,150

900

1,350

Arrastra Creek

900

2,700

4,725

0

1,575

Hematite Gulch

0

900

0

0

0

California Gulch

306,900

186,300

9,225

0

1,575

Blair Gulch

0

0

0

0

225

Niagra Gulch (upper Animas)

4,500

17,100

675

0

0

Swansea Gulch

0

900

0

0

0

Eureka Gulch

98,100

254,700

9,900

0

7,200

Picayune Gulch

0

8,100

450

0

0

Placer Gulch

106,200

230,400

8,775

225

1,125

ASTER

The ASTER sensor measures reflected solar radiation in three bands between 0.52 and 0.86 μm and six bands between 1.6 and 2.43 μm; and emitted radiation in five bands between 8.125 and 11.65 μm (ERSDAC 2005). Absorption features in the shortwave infrared spectral region between 1.0 and 2.5 μm are shown to be particularly useful for mapping silicates and carbonates (Hunt and Salisbury 1970, 1971; Rockwell 2009). The spatial resolution for the SWIR part of the spectrum is 30 m. Vegetation cover commonly obscures 88 % of watershed areas in the Animas River study area permitting 12 % of the area to be mapped by ASTER (Fig. 4). Minerals and mineral mixtures were identified from the ASTER data using spectroscopic analysis techniques described by Rockwell (2009). The identified minerals were then combined into alteration and lithologic types using a Boolean modeling technique based on standard definitions of the alteration mineralogy (Rockwell 2010). Modeled alteration types include separate argillic, four pyrite-bearing, and six propylitic and carbonate-bearing alteration assemblages. ASTER is available for the entire US and could be a valuable tool for mapping hydrothermal alteration as part of geoenvironmental assessments requiring coverage of large areas. Alteration types mapped by ASTER and area calculations determined by GIS are in Table 1.
https://static-content.springer.com/image/art%3A10.1007%2Fs12665-013-2229-y/MediaObjects/12665_2013_2229_Fig4_HTML.gif
Fig. 4

Distribution of alteration mapped by ASTER that was used in GIS alteration area determinations

Hybrid AVIRIS and field-based alteration map

In contrast to TM and ASTER, AVIRIS data are available only in selected areas of the US for detailed study of relatively small watershed areas or specific basins. A hybrid alteration map that integrates AVIRIS spectral interpretations of Dalton et al. (2007) and field-based mapping (Bove et al. 2007a) provides nearly 100 % bedrock coverage irrespective of vegetation cover (Fig. 5). AVIRIS has 224 narrow spectral bands between 0.4 and 2.5 μm that are in the visible, near infrared and mid-infrared parts of the spectrum. The narrow spectral bands enable individual minerals to be delineated due to characteristic absorption spectra of mineral species. The AVIRIS spectra used for the hybrid alteration mapping have a spatial resolution of 17 m2 and represent the highest resolution remote sensing dataset used in our analysis. AVIRIS mineral mapping was supplemented by field-based mapping and mineralogy studies augmented by X-ray diffraction analyses, which permitted the entire area of each watershed not covered by surficial deposits to be assigned to one of 21 alteration types (Bove et al. 2007a). This hybrid product contrasts with TM- and ASTER-derived alteration maps in which vegetated areas are not shown. Hydrothermally altered areas mapped with AVIRIS, however, show excellent correlation with those altered areas mapped by TM and ASTER (Rockwell 2010). Alteration from hybrid AVIRIS-derived map and area calculations determined by GIS are in Table 2.
https://static-content.springer.com/image/art%3A10.1007%2Fs12665-013-2229-y/MediaObjects/12665_2013_2229_Fig5_HTML.gif
Fig. 5

Hybrid AVIRIS- and field-based map of alteration. Prevalent areas in green are propylitically altered rocks. Brighter hues represent more intense alteration that has overprinted the propylitic assemblage. A detailed geologic map (Yager et al. Yager and Bove 2007) is available for download at http://​pubs.​usgs.​gov/​pp/​1651/​downloads/​; an alteration map is described in Bove et al. (Bove et al. 2007a, 2007) and available online at http://​pubs.​usgs.​gov/​sim/​2976/​

Table 2

Alteration type area (m2) determined by GIS analysis of hybrid AVIRIS mineral mapping for Cement Creek, Mineral Creek, and upper Animas River basin watersheds

Basin and Watersheds

Total area

Prop

Prop area (%)

Argillic

qsp

qap

∑(NAP)

NAP area (%)

Cement Creek basin

Dry Gulch

1,980,400

658,600

33

114,000

241,300

296,600

930,675

47

South of Dry Gulch

131,600

128,100

97

0

0

0

191

0

Prospect Gulch

3,295,000

1,449,600

44

80,200

983,800

286,100

1,394,301

42

Tiger Gulch

700,100

623,000

89

0

0

0

171

0

Georgia Gulch

1,304,300

1,058,400

81

0

9,000

7,400

17,538

1

Fairview Gulch

819,300

798,100

97

0

0

0

2,371

0

Cascade Gulch

1,471,900

931,700

63

0

0

0

86,930

6

Porcupine Gulch

793,500

477,500

60

19,700

92,500

19,100

117,929

15

Ohio Gulch

1,244,900

483,900

39

71,900

280,200

104,300

427,589

34

Illinois Gulch

784,500

725,900

93

0

0

0

91,677

12

Topeka Gulch

1,291,300

894,300

69

112,000

221,400

52,800

274,334

21

Niagara Gulch

2,185,300

720,900

33

19,200

422,800

184,700

688,804

32

Hancock Gulch

1,797,300

902,900

50

0

0

0

929

0

Minnesota Gulch

1,678,200

1,505,600

90

0

0

0

3,963

0

South Fork Cement Creek (above M. Fork)

4,157,900

2,389,500

57

0

0

0

87,713

2

North Fork Cement Creek

1,198,700

730,800

61

0

0

0

112

0

Ross basin

1,605,500

1,149,600

72

0

0

0

72,828

5

Mineral Creek basin

Northeast of Belcher Mine

1,752,000

1,154,100

66

0

0

0

1,034,500

0

Zuni Gulch

1,378,300

46,900

3

700

18,100

122,500

288,100

21

Battleship Slide

827,300

0

0

0

521,600

0

797,900

96

Red Tributary

3,045,600

398,800

13

0

371,300

0

1,773,400

58

South of Bonner Mine

616,000

0

0

0

148,600

0

556,700

90

Middle Fork Mineral Creek-East Basin

478,400

410,700

86

0

0

0

59,200

12

Middle Fork Mineral Creek-Central Basin

493,800

358,200

72

0

0

0

119,000

24

Middle Fork Mineral Creek-West Basin

1,170,300

761,000

65

0

0

0

2,100

0

Paradise Basin

2,995,400

1,220,000

41

0

0

0

1,651,400

55

South of Browns Gulch

2,721,800

357,800

13

164,900

156,400

69,100

619,800

23

Browns Gulch

2,267,300

1,232,500

54

63,400

38,800

1,500

115,000

5

South of Imogene Mine

834,400

484,400

58

0

0

0

0

0

West of Imogene Mine

2,267,300

1,232,500

93

63,400

38,800

1,500

115,000

5

Southeast of Chattanooga

834,400

484,400

95

0

0

0

0

0

Big Horn Gulch

402,900

376,500

98

0

13,800

0

13,800

3

East of Carbon Lake

816,600

773,500

95

0

19,700

0

19,700

2

Mineral Creek basin

Boulder Gulch

6,770,200

5,592,700

83

0

0

0

0

0

Arrastra Creek

10,387,000

7,888,200

76

0

0

0

107,500

1

Hematite Gulch

2,380,700

2,352,600

99

0

0

0

0

0

California Gulch

4,968,700

3,251,800

65

6,300

297,600

0

486,200

10

Blair Gulch

1,563,000

914,500

58

0

0

0

0

0

Niagra Gulch (upper Animas)

2,518,900

1,624,800

64

0

0

0

45,000

2

Swansea Gulch

2,130,500

1,053,100

49

0

0

0

65,900

3

Eureka Gulch

19,357,000

15,235,600

79

0

0

0

278,500

1

Picayune Gulch

4,181,000

3,413,400

82

0

0

0

1,000

0

Placer Gulch

480,470

3,326,900

69

28,800

142,200

0

622,100

13

Statistical methods

Linear regression analysis

Linear regression was used to test correlations between watershed variables (alteration type, terrain, and vegetation) and water quality signatures. The Kendall’s tau coefficient was used to measure the statistical dependence between two measured quantities—for example, the area of a specific alteration type and pH or metal concentration. Kendall’s tau is a non-parametric and statistically robust method, meaning that it does not make assumptions about a continuous probability distribution, e.g., that the data are normally distributed. The method also uses a correlation coefficient that measures the correspondence between two rankings to assess the probability that a resultant correlation is not due to chance (Kendall 1975; Helsel and Hirsch 2002). The significance or robustness of linear dependence between two measured variables is determined using the Kendall–Theil Robust Line, which determines whether a linear regression slope coefficient for the explanatory variable is significantly different from zero (Theil 1950; Helsel and Hirsch 2002). This statistical approach tests whether variables are negatively or positively correlated, or whether there is no statistical relationship between variables. Correlation measures the co-variation between independent variables and dependent water quality parameters; however, the reasons for correlations are determined from knowledge regarding the geology and geologic processes affecting the area.

Factor analysis

Factor analysis was completed on the remote sensing alteration datasets derived from TM, ASTER, and AVIRIS data, and associated watershed water quality parameters. Factor analysis was used to help identify distinct patterns of occurrence and relationships among multiple variables that could synergistically affect watershed surface water quality. The principal component analysis extraction method was applied to the datasets, which seeks a linear combination of variables such that the maximum variance is extracted from the variables. The maximum variance is removed and analysis proceeds iteratively until the desired number of factors is extracted. This method seeks the fewest factors that can explain the variance (correlation) of a set of variables. Factor loadings (1−n) are the correlation coefficients between variables (rows) and factors (columns). Rotated factor loading scores were used that enable groups of interrelated data to be determined. The squared factor loading is the percent of variance in that indicator variable explained by the factor and is equivalent to Pearson’s r (Helsel and Hirsch 2002).

Water quality data

Watershed and GIS-buffered area characteristics compiled using GIS were statistically evaluated to determine correlations between watershed physical properties and associated water quality signatures. Water quality analyses used for statistical comparisons were determined for samples collected at watershed outlets during synoptic sampling events (Table 3). A watershed outlet is defined as the lowest elevation location that can be sampled prior to surface water entering a main stem stream. The water quality determined at a watershed outlet represents a geochemical integration of all upstream water–rock–soil–sediment–atmosphere interactions and processes (Fig. 6).
Table 3

Water quality parameters for samples collected at watershed outlets in the Upper Animas River, Cement Creek, and Mineral Creek basins during base-flow conditions

https://static-content.springer.com/image/art%3A10.1007%2Fs12665-013-2229-y/MediaObjects/12665_2013_2229_Tab3_HTML.gif

Values in italics are ½ detection limit

Red OFR-00-53, blue USGS Professional Paper 1651 database (2007), green Unpublished Report to Water Quality Control Commission, Animas River Stakeholders Group, 2001 Placer gulch sample UA33 is downstream of Evening Star Mine (mid-basin). SpC specific conductance

https://static-content.springer.com/image/art%3A10.1007%2Fs12665-013-2229-y/MediaObjects/12665_2013_2229_Fig6_HTML.gif
Fig. 6

Example of water quality samples collected at watershed outlets; parameters pH, Al, and Fe concentration shown for comparison. Watersheds have greatly differing proportions of alteration types with Georgia Gulch having much less acidic conditions due to the presence of abundant propylitic rocks (green), and lower Al and Fe concentrations compared to Prospect Gulch. Detailed descriptions of alteration types in Bove et al. (2007a)

The primary source of water quality data used in this analysis is from a relational database described in Sole and Rich (2007) and was populated from data collected as part of AML investigations (Mast et al. 2000a, 2007; Leib et al. 2007; Kimball et al. 2007; Wright et al. 2007). Samples collected during base-flow conditions during late summer to fall, primarily between 1996 and 2002, were used for analysis. Water quality samples collected during base flow conditions have element concentrations that are relatively unaffected by snowmelt dilution and are thought to closely represent groundwater discharge to streams.

Results

GIS-derived statistics for primary rock units

GIS was used to analyze AML geodatasets to determine area of bedrock units, alteration types, and vegetation coverage for each watershed. The study area consists of 63 % igneous bedrock terrain. Paleozoic, Mesozoic, and Tertiary sedimentary rock units represent 6 % of the study area. Precambrian crystalline bedrock exposed primarily along the southern caldera margins represent 4 % of outcrops. Extensive, quaternary sedimentary deposits cover 26 % of the study area and range in thickness from a few meters to more than 30 m (Yager et al. 2000; Johnson and Yager 2006; Johnson 2007).

The most widespread igneous rock type is the Silverton Volcanics intermediate composition lavas and volcaniclastic units that form 52 % of the igneous rocks and that are host to 90 % of metallic mineral deposits (Bejnar 1957). Silverton Volcanics lava filled the area of subsidence caused by San Juan caldera collapse (Yager and Bove 2002, 2007). Intermediate composition volcaniclastic deposits and minor lava flows of the San Juan Formation, exposed primarily outside the Silverton and San Juan caldera margins, represent 28 % of the igneous rocks and are the second-most abundant igneous unit. Granitoid intrusive rocks of the Sultan Mountain Stock are preserved along the western and southern caldera structural margins and comprise 8 % of the igneous rock units. The rhyolite to dacite composition Eureka Member of the Sapinero Mesa Tuff, derived from the San Juan caldera, represents 6 % of the igneous rock units and is exposed along the southern caldera margins and along the Eureka graben.

GIS-derived statistics for alteration types

During AML investigations, AVIRIS- and field-based mineral mapping permitted 21 acid-generating alteration types to be identified (Bove et al. 2007a). GIS-derived area determinations from the AVIRIS data for the principal, acid-generating alteration types determined in the Bove et al. (2007a) study are listed in order of decreasing area and are as follows: (1) propylitic acid sulfate (prop-as: 22 %); (2) quartz–sericite–pyrite (qsp: 17 %); (3) weak sericite–pyrite–hydrothermal (wsp-hyd: 17 %); (4) quartz–alunite–pyrophyllite (qap: 6 %); (5) weak sericite–pyrite–propylitic (wsp-prop: 6 %); (6) Quaternary quartz–alunite–sericite (Qqas: 6 %); (7) mixed quartz–alunite and quartz–alunite–pyrophyllite (qap-s: 5 %); (8) weak sericite–pyrite (wsp: 4 %), and (9) argillic (arg: 4 %) (Table 1, ESM only). For a detailed discussion of the alteration types listed above the reader is referred to Bove et al. (2007a).

Areas of net acid production and effect on pH

Watershed NAP area calculations completed using the hybrid AVIRIS- and field-based alteration maps enabled comparisons between the sum of all NAP alteration types and pH. These data are useful for determining where acidic conditions are likely to occur. Acidic watersheds were not surprisingly found to have relatively high proportions of NAP alteration. The NAP area determined for 17 acidic watersheds with pH < 5.5 ranges from 2.4 to 96.0 % with a mean NAP area of 46 %. Acidic watersheds having small areas of NAP alteration are commonly impacted by mining features such as draining adits. For example, GIS analysis of North Fork Cement Creek and East of Carbon Lakes watersheds with highly acidic water quality lacked significant NAP alteration and thus by inference could be identified as being mining impacted (Table 2). In contrast, the NAP area for 29 watersheds with pH > 5.5 ranges from 0 to 24 % with a mean NAP area of only 4.1 %. Note that alteration assemblages containing mixtures of NAP and ANC, e.g., qsp-prop are included in the NAP area determinations. This is because the propylitic component may be insufficient to neutralize acidity attributed to weathering of the qsp component having NAP > ANC.

The Mineral Creek, Cement Creek, and upper Animas River basins have varying proportions of NAP rocks that are associated with multiple types of hydrothermal systems. The Mineral Creek basin in the western part of the study area is associated with (1) a low-grade Cu–Mo porphyry system that formed between South and Middle Forks of Mineral Creek and (2) with acid sulfate alteration centered on Ohio Peak–Anvil Mountain (OPAM) to the east (Bove et al. 2007a). Alteration associated with the Cu–Mo porphyry and acid sulfate systems is extensive with some watersheds in these areas being entirely and intensely altered. The Upper Cement Creek basin has significant areas of acid sulfate alteration within the Red Mountain mining district and is also impacted by alteration in the OPAM area that forms the drainage divide between Mineral and Cement Creek basins. Smaller areas of NAP alteration occur along vein structures of the Eureka graben. Alteration in the upper Animas River basin east of Silverton is mainly associated with caldera-related vein structures that formed in the South Silverton mining area and along the Eureka graben in California and Placer Gulch’s (Varnes 1963; Casadevall and Ohmoto 1977).

Areas and watershed elevation of ANC rocks and effect on pH

Propylitic alteration containing abundant secondary ANC minerals (calcite, chlorite, epidote) is pervasive, overprinting most of the igneous sequence (Burbank 1960; Bove et al. 2007a). Propylitic alteration was mapped in 67 % of the study area. Regionally, extensive propylitic alteration is preserved where it is not overprinted by more intense hydrothermal alteration having NAP and affecting 9 % of the study area (Fig. 7; Table 4).
https://static-content.springer.com/image/art%3A10.1007%2Fs12665-013-2229-y/MediaObjects/12665_2013_2229_Fig7_HTML.jpg
Fig. 7

Intensely altered propylitic rocks (red and yellow) in the Red Mountain mining district, overprinting the regional, propylitically altered igneous rocks (grayish-green), peaks at top of image. View to west

Table 4

Summary statistics determined by GIS for generalized propylitic rocks and net acid producing assemblages for (A) whole study area, and major basins; (B) watersheds that comprise each basin

 

Propylitic area (%)

Net acid producing rocks (%)

pH

A

 Generalized alteration groups (whole watershed)

67

9

5.82

 Generalized alteration types for basins

  Upper Animas River basin

77

3

7.12

  Cement Creek basin

62

20

3.56

  Mineral Creek basin

51

30

6.29

B

 Generalized alteration groups summary statistics for watersheds analyzed in each basin

  Upper Animas River

72

3

 

  Cement Creek basin

66

13

 

  Mineral Creek basin

53

25

 

pH was determined on September 8, 2009 at hydrologic gage stations; whole study area pH represented by water quality sample collected near gage A72 below Silverton, Colorado

GIS calculations were used to determine the percent area and mean elevation of propylitic rocks in each watershed. If propylitic rocks are exposed in a watershed at a high elevation relative to the mean watershed elevation, weathering of these rocks could produce surface water with initial alkalinity that might mitigate acidity caused by weathering of NAP alteration or mining areas exposed at lower elevations. The percent area of propylitic rocks was also calculated for watersheds having greater and less than pH 5.5 at watershed outlets (Table 5). A pH of 5.5, similar to that of rainwater, was used as a gage for acidic versus less acidic water quality. Watersheds with less acidic to neutral water quality (pH > 5.5) have an average of 74 % propylitic area (Fig. 8). Only three of the less acidic to neutral watersheds (Big Horn, Placer, and Porcupine watersheds) have a mean elevation of propylitic rocks that is lower than the mean watershed elevation. In contrast, acidic watersheds average 44 % propylitic area and 64 % of those watersheds have a mean propylitic rock elevation that occurs below the watershed mean elevation (Fig. 8).
Table 5

Mean watershed elevation, mean elevation of propylitic rocks, total area of propylitic rocks, and pH for each watershed

Watershed

Mean watershed elevation (m)

Mean elevation of propylitic rocks (m)

Pr propylitic rock area (%)

pH

Ohio Gulch

3,407

3,318

39

3.04

North Fork Cement

3,669

3,690

61

3.10

Prospect Gulch

3,551

3,503

44

3.22

Tiger Gulch

3,489

3,483

89

3.23

South of Browns Gulch

3,474

3,395

13

3.25

Red Tributary

3,524

3,576

13

3.32

South of Bonner

3,321

No propylitic

0

3.45

Niagra, Cement Creek Basin

3,441

3,300

33

3.51

East of Carbon Lake

3,517

3,516

95

3.54

Battleship Slide

3,395

No propylitic

0

3.79

Dry Gulch

3,562

3,451

33

3.89

Zuni Gulch

3,461

3,641

3

3.9

Minnesota

3,497

3,501

90

4.19

Browns Gulch

3,527

3,592

54

4.3

Topeka Gulch

3,401

3,370

69

4.88

California Gulch

3,758

3,736

65

5.32

South of Dry

3,453

3,459

97

5.74

Georgia

3,557

3,566

81

5.86

Placer

3,764

3,756

69

5.93

Ross

3,806

3,824

72

6.17

South Fork Cement

3,605

3,660

57

6.34

Porcupine

3,459

3,407

60

6.38

S.E. of Chattanooga

3,579

3,592

95

6.51

Middle Fork Central

3,504

3,585

72

6.56

Middle Fork East

3,437

3,471

86

6.78

South of Imogene

3,629

3,639

58

6.86

Paradise basin

3,639

3,714

41

6.86

Picayune Gulch

3,700

3,727

82

6.86

Swansea

3,545

3,625

49

6.90

Big Horn

3,612

3,609

98

6.92

N.E. of Belcher

3,512

3,515

66

7.10

Eureka

3,664

3,697

79

7.10

Niagra, Animas River Basin

3,826

3,830

64

7.27

Hancock

3,567

3,743

50

7.32

W. of Imogene

3,467

3,488

93

7.44

Boulder

3,668

3,685

83

7.45

Fairview

3,575

3,583

97

7.50

Blair

3,551

3,636

58

7.52

Illinois

3,555

3,583

97

7.50

Arrastra

3,595

3,697

76

7.92

Cascade

3,613

3,623

63

7.96

Middle Fork West

3,663

3,683

65

8.02

Hematite

3,678

3,686

99

8.24

Bold indicates elevation of propylitic rocks is higher than the mean watershed elevation, italics are watersheds with nearly equal mean elevation and propylitic mean elevation, non-italics and non-bold indicates mean elevation of propylitic rocks is lower than mean elevation

https://static-content.springer.com/image/art%3A10.1007%2Fs12665-013-2229-y/MediaObjects/12665_2013_2229_Fig8_HTML.gif
Fig. 8

Propylitic alteration area (% of total watershed area) versus pH. Acidic watersheds average 44 % propylitic rock; less acidic to near neutral watersheds average 74 % propylitic rock. Mining impacted watersheds are clearly delineated by watersheds having abundant propylitic rock, yet having low pH

Mining impacts can overwhelm the ANC properties of propylitically altered rocks. For example, Mineral and Cement Creek basins have acidic surface water quality with Cement Creek being more acidic, despite Cement Creek basin having 11 % greater area of propylitic rocks (Table 4). Cement Creek is heavily impacted by mining compared with the less heavily mined Mineral Creek basin, and this could explain the lower pH of Cement Creek. In contrast to the more highly altered Mineral and Cement Creek basins, the upper Animas River basin on the western margin of the study area has low acidity, an average of 77 % propylitic rocks and only 3 % NAP rocks. The upper Animas River basin water sampled near the basin outlet upstream from Cement Creek has a near neutral pH (Table 4).

Vegetated and non-vegetated areas

Vegetated and non-vegetated areas were delineated using GIS analysis of AVIRIS-derived maps (Dalton et al. 2007). Alteration and vegetation layers were overlain to determine non-vegetated areas that intersect hydrothermally altered rock. Non-vegetated areas enhance the potential for erosion because plants are unavailable for interception of precipitation. Lack of interception by plants enhances splash detachment, an important physical weathering mechanism (Kinnell 2005). Intensely altered areas lacking vegetation tend to have more highly acidic water quality compared with vegetated areas. Acidic watersheds have an average of nearly 13 % non-vegetated areas that intersect NAP alteration. In contrast, less acidic to neutral watersheds (pH > 5.5) have an average of 2 % non-vegetated areas that coincide with NAP alteration.

Morphometric parameters determined by GIS

The principal morphometric parameters that were statistically analyzed in relation to water quality signatures were basin area, stream length, and drainage density (Table 2, ESM only). These parameters are very similar for Mineral and Cement Creek basins. Upper Animas River basin watersheds are 26 % larger, have average stream lengths that are 32–37 % longer, and drainage densities that are nearly 8 % greater compared to Mineral and Cement Creek basins.

Slope and aspect were also considered to be potentially important morphometric parameters. Topographic relief in the study area is high and average watershed slopes are consistently steep with a gradient commonly >25°. This resulted in the slope showing no significant correlation with water quality. Additional data are needed to fully evaluate the role that aspect may play in influencing water quality. This is because aspect strongly controls the timing of snow melt runoff and thus influences when infiltration of surface water into soils and bedrock and subsequent weathering reactions occur. Southerly facing slopes, for example, experience earlier melting compared to northerly slopes, and snow lasts longer throughout a season on north facing slopes. To permit interpretations regarding potential effect on water quality signatures, data acquisition is required at different times during a hydrologic cycle for watersheds with similar proportions of alteration and with varying aspect.

Linear regression analysis of alteration data

TM data

Linear regression coefficients were determined for alteration assemblages mapped by analysis of TM data and water quality parameters for the combined, mined and unmined watersheds, and separately for watersheds having minimal mining impacts (Table 6). Despite only 2–3 % watershed areas being mapped by TM due to vegetation cover, statistically significant correlations were determined between TM-derived alteration types and water quality signatures. Negative correlations are observed between area of clay–sulfate–sericite1 alteration (TM–CSM), pH, and alkalinity. Aluminum is positively correlated with area of TM–CSM. Area of iron alteration (TM–Fe) is negatively correlated with pH and alkalinity.
Table 6

Linear regression coefficients for area of alteration types determined using TM, ASTER, hybrid AVIRIS and field mapping data, and water quality parameters for combined mined and unmined watersheds

Alteration type

pH

Fe

Al

Cu

Zn

SO4

Alkalinity

Specific conductance

TM

 Clay–sulfate–sericite alteration

−0.32

0.27

0.41

0.19

0.11

0.29

−0.56

 

(0.03)

(0.05)

(0.005)

(0.19)

(0.45)

(0.05)

(0.02)

 

N = 24

N = 24

N = 24

N = 25

N = 24

N = 24

N = 12

 

 Iron alteration

−0.38

0.38

0.43

0.38

0.35

0.29

−0.51

 

(0.0001)

(0.001)

0.0004

(0.002)

(0.45)

(0.004)

(0.002)

 

N = 34

N = 33

N = 33

N = 32

N = 24

N = 33

N = 20

 

ASTER

 ∑of all argillic alteration

−0.37

0.41

0.47

0.39

0.16

0.07

−0.32

 

(0.0009)

(0.0002)

(0.00003)

(0.0004)

(0.14)

(0.51)

(0.03)

 

N = 38

N = 37

N = 39

N = 37

N = 38

N = 38

N = 23

 

 Argillic plus ferric iron alteration

−0.38

0.39

0.44

0.41

0.15

0.14

−0.46

 

(0.001)

(0.0009)

0.0002

(0.0005)

(0.19)

(0.23)

(0.006)

 

N = 35

N = 35

N = 35

N = 35

N = 35

N = 35

N = 19

 

 Quartz–sericite–pyrite (phyllic)

−0.42

0.39

0.45

0.32

0.18

0.16

0.18

 

(0.002)

(0.004)

(0.0009)

(0.01)

(0.2)

(0.27)

(0.44)

 

N = 27

N = 26

N = 28

N = 27

N = 27

N = 26

N = 12

 

 ∑of all Quartz–sericite–pyrite alteration

−0.41

0.47

0.55

0.42

0.12

0.15

0.16

 

(0.0006)

(0.00007)

(0.000005)

(0.00004)

(0.3)

(0.22)

(0.34)

 

N = 34

N = 32

N = 33

N = 34

N = 34

N = 33

N = 19

 

 Propylitic or carbonate rocks

0.24

0.17

0.08

0.28

0.01

0

0.06

 

(0.07)

(0.2)

(0.57)

(0.03)

(0.9)

(1)

(0.71)

 

N = 28

N = 27

N = 28

N = 28

N = 28

N = 27

N = 21

 

AVIRIS

 Argillic alteration

0.25

0.28

0.46

0.31

0.23

0.31

0.04

0.28

(0.24)

(0.28)

(0.03)

(0.16)

(0.29)

(0.16)

(1)

0.20

N = 13

N = 13

N = 13

N = 13

N = 13

N = 13

N = 7

N = 13

 Quartz–sericite–pyrite alteration

−0.52

0.52

0.62

0.34

0.18

0.23

0.43

0.29

(0.003)

(0.003)

0.0003

(0.05)

(0.32)

(0.1)

(0.17)

0.09

N = 18

N = 18

N = 18

N = 18

N = 18

N = 18

N = 8

N = 18

 Quartz–alunite–pyropylitic alteration

−0.49

0.15

0.36

0.12

0.21

0.06

0.33

0.06

(0.003)

(0.54)

(0.11)

(0.63)

(0.37)

(0.83)

(0.45)

0.84

N = 12

N = 12

N = 12

N = 12

N = 12

N = 12

N = 6

N = 12

 ∑of NAP alteration

−0.41

0.29

0.39

0.19

0.14

0.28

0.28

0.30

(0.002)

(0.03)

(0.004)

(0.16)

(0.28)

(0.04)

(0.09)

0.02

N = 28

N = 28

N = 28

N = 28

N = 28

N = 28

N = 19

N = 28

 ∑of propylitic alteration:∑NAP alteration (ratio)

0.46

0.29

0.38

0.18

0.02

0.16

0.44

0.19

(0.0009)

(0.03)

(0.009)

(0.20)

(0.88)

(0.26)

(0.09)

0.17

N = 26

N = 26

N = 26

N = 25

N = 26

N = 25

N = 18

N = 26

Values in bold indicate statistical trends

P values are in parenthesis, N is number of watersheds used in analyses

Unmined watersheds were analyzed separately to determine if correlations exist when mining impacts are excluded. Mining impacts, especially draining mine adits that are observed at several mining areas in Colorado, can adversely impact water quality and complicate interpretation of the geochemical background signal attributed to natural weathering processes of hydrothermally altered areas (Yager et al. 2010). In general, an increase in correlation values was observed when mined watersheds were excluded from regression analysis (Table 7). TM–CSM showed significant trends with pH, Fe (which were not observed in the combined mined and unmined dataset), Al, and alkalinity. Correlation values between TM–Fe, pH, Fe, Al, and Cu also increased. Alkalinity had a lower, negative correlation value compared with the combined mined and unmined dataset.
Table 7

Linear regression coefficients for area of alteration types determined using TM, ASTER and hybrid AVIRIS and field mapping data and water quality parameters for unmined watersheds

Remote sensing platform and alteration type

pH

Fe

Al

Cu

Zn

SO4

Alkalinity

Specific conductance

TM

 Clay–sulfate–sericite alteration

−0.47

0.47

0.46

0.33

0.05

0.22

−0.62

 

(0.02)

(0.02)

(0.02)

(0.09)

(0.8)

(0.27)

(0.07)

N = 15

N = 15

N = 15

N = 15

N = 15

N = 15

N = 7

 Iron alteration (unmined)

−0.57

0.51

0.52

0.45

0.25

0.22

−0.45

 

(0.0002)

(0.0009)

(0.0009)

(0.004)

(0.11)

(0.16)

(0.03)

N = 22

N = 21

N = 21

N = 20

N = 21

N = 21

N = 14

ASTER

 ∑of all argillic alteration

−0.43

0.47

0.58

0.45

0.25

−0.02

−0.26

 

(0.002)

(0.0006)

(0.00003)

(0.0008)

(0.06)

(0.89)

(0.15)

N = 26

N = 25

N = 26

N = 25

N = 25

N = 26

N = 17

 Argillic plus ferric iron alteration

−0.47

0.47

0.56

0.49

0.19

0.04

−0.39

 

(0.002)

(0.0002)

0.0002

(0.001)

(0.19)

(0.79)

(0.05)

N = 23

N = 23

N = 23

N = 23

N = 23

N = 23

N = 14

 Quartz–sericite–pyrite (phyllic)

−0.50

0.54

0.60

0.47

0.29

0.11

−0.08

 

(0.004)

(0.002)

(0.0005)

(0.006)

(0.09)

(0.52)

(0.83)

N = 18

N = 18

N = 18

N = 18

N = 18

N = 18

N = 8

 ∑of all Quartz–sericite–pyrite alteration

−0.33

0.49

0.59

0.46

0.07

0.02

−0.04

 

(0.03)

(0.001)

(0.000007)

(0.002)

(0.3)

(0.89)

(0.87)

N = 23

N = 21

N = 22

N = 23

N = 23

N = 23

N = 14

 Propylitic or carbonate rocks

0.23

−0.05

−0.04

−0.34

0.07

0.04

0.08

 

(0.18)

(0.75)

(0.83)

(0.04)

(0.69)

(0.8)

(0.69)

N = 19

N = 18

N = 19

N = 19

N = 19

N = 19

N = 15

AVIRIS

 Quartz–sericite–pyrite alteration

−0.45

0.48

0.66

0.35

0.03

0.24

−0.6

0.33

(0.05)

(0.03)

0.003

(0.13)

(0.94)

(0.31)

(0.22)

0.14

N = 12

N = 12

N = 12

N = 12

N = 12

N = 12

N = 5

N = 12

 Quartz–alunite–pyropylitic alteration

−0.52

0.14

0.52

0.14

−0.14

0.04

−0.66

0.14

(0.13)

(0.76)

(0.13)

(0.76)

(0.76)

(1)

(0.30)

0.76

N = 7

N = 7

N = 7

N = 7

N = 7

N = 7

N = 4

N = 7

 ∑of all NAP alteration

−0.66

0.49

0.54

0.42

0.23

0.5

−0.63

0.45

(0.0002)

(0.006)

0.0004

(0.03)

(0.21)

(0.008)

(0.008)

0.01

N = 17

N = 17

N = 16

N = 16

N = 17

N = 16

N = 11

N = 17

 ∑of propylitic alteration:∑ of all NAP alteration (ratio)

0.63

−0.42

−0.53

−0.34

0.13

−0.34

0.64

−0.28

(0.0001)

(0.03)

(0.010)

(0.10)

(0.52)

(0.10)

(0.01)

0.16

N = 15

N = 15

N = 14

N = 14

N = 15

N = 14

N = 10

N = 15

Values in bold indicate statistical trends

P values are in parenthesis, N is number of watersheds used in analyses

ASTER data

Alteration types mapped by ASTER and area calculations determined by GIS and used in statistical analysis are in Table 6. Two alteration assemblages, argillic plus ferric iron (Aster–Arg–Fe) and quartz–sericite–pyrite–phyllic (Aster-qsp), were observed to have correlations with water quality parameters. All argillic and qsp alteration types were also separately summed and statistically analyzed to evaluate the cumulative effect that the ∑all NAP assemblages have on water quality.

Aster–Arg–Fe is negatively correlated with pH and alkalinity. The major metal cations Fe, Al, and the trace metal Cu are positively correlated with area of Aster–Arg–Fe alteration, with aluminum having the the highest correlation. Analysis of the ∑all argillic alteration assemblages showed a similar trend compared to the Aster–Arg–Fe data, but had a slightly lower negative correlation with alkalinity. Area of Aster-qsp alteration is negatively correlated with pH and positively correlated with Fe, Al, and Cu. Correlation values are higher for the ∑all qsp-bearing assemblages and the elements Fe, Al, and Cu, when compared to correlations obtained with the Aster-qsp alteration assemblage. The “propylitic or carbonate rock” assemblage is not significantly correlated with any water quality parameter.

Correlation values were generally higher for area of Aster–Arg–Fe and water quality parameters when mined watersheds were excluded from analysis (Table 7). This is especially true for pH, Fe, Al, and Cu; however, the negative correlation value for alkalinity observed in the combined mined and unmined dataset decreased for the unmined dataset. Correlations also increased for area of Aster-qsp and water quality parameters pH, Fe, and Al when mined watersheds were excluded. Correlation values for the ∑all qsp alteration assemblages in ummined watersheds yielded a lower correlation value with pH and a positive correlation with Fe, Al, and Cu that are similar compared to the dataset that includes both mined and unmined watersheds. The propylitic or carbonate rock assemblage showed no statistically significant trends with water quality parameters when mined catchments were excluded.

Hybrid AVIRIS and field mapping data

Significant correlation values are observed for three acid-generating alteration types (arg, qsp, and qap) mapped by hybrid AVIRIS- and field-based data (Table 6). In addition, correlations are observed between the ∑all NAP, the ratio ∑propylitic:∑all NAP, and water quality. Aluminum is positively correlated with area of argillic alteration. No additional water quality parameters are correlated with argillic alteration. Quartz–sericite–pyrite has a relatively high negative correlation with pH, and positive correlation with Fe and Al. A possible positive trend is observed between qsp and Cu. Quartz–alunite–pyrophyllite is negatively correlated with pH, but lacks additional correlations with other water quality parameters. The ∑all NAP alteration is negatively correlated with pH, positively correlated with Al, and yields weaker positive trends with Fe, specific conductance, and SO4. A positive trend is also observed for the ratio ∑propylitc:∑all NAP and pH.

Selected morphometric parameters are correlative with the area of propylitically altered bedrock. A positive regression value (R = 0.63, n = 40) is observed between area of propylitic rocks and total watershed area. Positive trends are also observed between area of propylitic alteration, stream length (R = 0.31, n = 40), and non-vegetated area (R = 0.66, n = 40).

Exclusion of mined watersheds resulted in very similar trends between area of qsp alteration, pH, Fe, and Al when compared with the dataset containing mined and unmined watersheds (Table 7). The largest increase in regression values in the unmined dataset was observed for both the ∑all NAP alteration types and the ratio of propylitic to NAP alteration and water quality. The ∑all NAP areas have a negative correlation with pH and alkalinity. Positive correlations were also observed for the ∑all NAP areas, and Fe, Al, Cu, specific conductance, and sulfate. The ratio ∑propylitc:∑all NAP resulted in positive correlations with pH and alkalinity. Fe and Al are negatively correlated with area of propylitically altered bedrock. Exclusion of mined watersheds does not increase the strength of morphometric correlations observed with area of propylitic alteration.

Area of vegetation

Correlations between non-vegetated and whole watershed areas of qsp alteration and water quality parameters were similar, both having negative correlations with pH (R = −0.41, n = 19) and positive correlations with Fe (R = 0.36, n = 19) and Al (R = 0.56, n = 19). Area of non-vegetated qap is negatively correlated with pH (R = −0.51, n = 12) and has a similar negative correlation value that is also observed for whole watershed area of qap alteration.

Hybrid AVIRIS and field mapping data (30-m GIS stream-buffered area analysis)

Linear regression analysis was performed on the intersected area of alteration types along a 30-m GIS stream buffer and water quality signatures. Area statistics for GIS-buffered areas of qsp alteration is in Table 3 (ESM only). In general, the stream buffered area analysis yielded weaker correlations compared with the total watershed area. Buffered area of qsp alteration is negatively correlated with pH (Table 8). Copper is also more highly correlated with GIS-buffered areas of qsp compared with whole watershed area analysis. Alkalinity is negatively correlated with area of qsp, although the small sample size (n = 6) requires additional data to confirm this correlation. The sum of NAP alteration is negatively correlated with pH and is positively correlated with Al, with the regression value being higher for Al compared with whole watershed analysis. Correlations for Fe, Al, and alkalinity increased when mined watersheds were excluded from regression analysis (Table 9). GIS-buffered propylitic areas along streams are not correlative with either pH or alkalinity.
Table 8

Linear regression coefficients for 30 m GIS stream buffered area intersecting alteration type and water quality parameters for mined and unmined watersheds

Buffer alteration type

pH

Fe

Al

Cu

Zn

Specific conductance

SO4

Alkalinity

Quartz–sericite–pyrite alteration buffer

−0.40

0.25

0.36

0.48

−0.12

0.03

0.03

−0.73

(0.05)

(0.23)

(0.08)

(0.01)

(0.58)

0.91

(0.91)

(0.06)

N = 14

N = 14

N = 14

N = 14

N = 14

N = 14

N = 14

N = 6

∑of non propylitic alteration buffer

−0.42

0.32

0.49

0.21

0.09

0.18

0.09

−0.17

(0.007)

(0.04)

0.002

(0.17)

(0.58)

0.25

(0.59)

(0.50)

N = 21

N = 21

N = 21

N = 21

N = 21

N = 21

N = 21

N = 12

Propylitic alteration buffer

0.17

−0.07

−0.01

−0.02

0.09

−0.05

−0.11

0.14

(0.12)

(0.50)

(0.90)

(0.83)

(0.42)

0.65

(0.33)

(0.30)

N = 39

N = 39

N = 39

N = 39

N = 39

N = 39

N = 39

N = 28

GIS stream buffer was applied to hybrid, AVIRIS and field mapping data

Values in bold indicate statistical trends

P values are in parenthesis, N is number of watersheds used in analyses

Table 9

Linear regression coefficients for 30 m GIS stream buffered area intersecting alteration type and water quality parameters for unmined watersheds

Buffer alteration type

pH

Fe

Al

Cu

Zn

Specific conductance

SO4

Alkalinity

∑of non propylitic alteration buffer

−0.48

0.45

0.67

0.38

0.12

0.36

0.24

−0.73

(0.03)

(0.05)

0.003

(0.09)

(0.63)

0.11

(0.30)

(0.06)

N = 12

N = 12

N = 12

N = 12

N = 12

N = 12

N = 12

N = 6

GIS stream buffer was applied to hybrid, AVIRIS and field mapping data

Values in bold indicate statistical trends

P values are in parenthesis, N is number of watersheds used in analyses

Factor analysis

TM data

A four factor model was used to explain 84 % of the variance for TM data with factors one and two accounting for 60 % of the variance (Table 4, ESM only). Factor 1 loadings (29 % of the variance) indicate the highest correlation for areas of iron and clay–sulfate–sericite alteration. There is also a corresponding, weak inverse correlation with alkalinity that is likely caused by the increased weathering of hydrothermally altered areas and possible lack of propylitic rocks that have ANC. Factor 2 represents 26 % of the variance and has a weak negative correlation with the TM_Mod_Fe_2 classification defined by Rockwell (2010). Aluminum, Fe, SO4, and specific conductance all have inverse correlations with pH. Factor 2 loading scores are consistent with a correlation between increased area of iron alteration, corresponding low pH and an increase in major metal cations (Al and Fe), SO4, and specific conductance.

ASTER data

A six factor model was used to explain 92 % of the variance for the ASTER dataset, with factors one and two accounting for 56 % of the variance (Table 5, ESM only). Factor 1, with 33 % of variance indicates a strong, positive correlation for areas of clay, ferric iron, and qsp alteration. The high correlations for Al and Fe are consistent with weathering of hydrothermally altered rocks in watersheds having abundant aluminum-bearing clay minerals in addition to oxidation of pyrite and other sulfide phases. Weak correlations are observed for pH and specific conductance that would be expected in these more highly altered watersheds. Factor 2 represents 24 % of the variance and shows correlations between the multiple subtypes of propylitcally altered rocks mapped by ASTER. There is no associated correlation, however, between propylitic rocks and pH. This finding could be a result of the limits of ASTER data in completely mapping a watershed because of vegetation cover that ultimately reduces the detectable area of propylitically altered bedrock having ANC. Factor 3 represents 12 % of the variance with the metals (Fe, and Al), SO4, and specific conductance being inversely correlated with pH. Factor 4 represents 10 % of the variance and shows correlations between pH, Cu, and Zn. Correlative Cu and Zn is consistent with weathering of a base metal sulfide mineral assemblage containing copper and zinc sulfides that are common in polymetallic veins in the study area.

The importance of the propylitic assemblage in mitigating ARD can be inferred in factor 5 (9 % of variance). Low pH and low alkalinity are correlated with watersheds having a relatively small area of propylitic alteration. Thus, minimal ANC is supplied by propylitically altered rocks to mitigate NAP. Factor 6, representing 5 % of the variance has a weak positive correlation with pH and a corresponding positive correlation with the mixed, regional propylitic to weak qsp and minor sericite assemblage. Metal concentrations Al and Fe are also negatively correlated. This suggests that ANC and possible decreased metal abundances attributed to weathering of propylitically altered rocks as part of this mixed assemblage overwhelms any NAP produced by the weak qsp and minor sericite-bearing rocks.

Hybrid AVIRIS and field data

A six factor model was used to explain 80 % of the variance in the hybrid AVIRIS- and field-based dataset (Table 6, ESM only). Factor 1 represents 24 % of the variance and shows an inverse correlation between pH and the NAP alteration types argillic, qsp, qap, and ∑qsp + qap. Al, Fe, and Cu concentrations, sulfate and specific conductance are inversely correlated with pH. Factor 2 explains 24 % of the variance and identifies positive correlations among the morphometric parameters of stream length, total area, and propylitically altered rocks. In addition, both vegetated, non-vegetated areas, and area of quaternary deposits are all positively correlated. Weaker positive correlations are observed for wsp and ∑of all NAP. These patterns are consistent with large watersheds having an abundance of propylitically altered rocks and longer total stream lengths. Factor 3 explains 11 % of the variance. Variables that are correlated include the NAP alteration types qsp, ∑qsp + qap, wsp-hp and Zn. It is uncommon that Zn is correlated with alteration variables mapped by the TM and ASTER datasets. Thus, hybrid AVIRIS-field maps may be a more useful tool in identifying possible impacts to water quality from weathering of mineralized areas containing base metals. Factor 4 represents 8 % of the variance and shows correlations between sulfate and specific conductance. Factor 5, representing 8 % of the variance shows correlations between the ratio ∑propylitic:∑non-propylitic, pH, and alkalinity. No correlation is observed for only the area of propylitically altered rocks indicating that the ratio of ANC (propylitic) to NAP (non-propylitic) is the controlling factor in determining alkalinity and pH, and not just the area of ANC. Factor 6 consists of 6 % of the variance and indicates a positive correlation between drainage density and wsp-hyd alteration. Wsp-hyd alteration may be more prone to weathering that could cause an increase in drainage density compared with other alteration types that do not show a similar trend.

Discussion

GIS was valuable in quantifying areas of alteration derived from TM, ASTER, and hybrid AVIRIS-field mapping datasets. While it is useful to characterize the primary igneous units that had formed prior to alteration, weathering of hydrothermally altered bedrock and associated mineral deposits is the primary cause of resultant water quality. This is because there are few igneous units in the study area that have not been altered to at least some degree, modifying the primary igneous minerals to a secondary alteration mineral assemblage. GIS analysis identified the study area as being comprised of 67 % propylitic ANC rocks that are overprinted by 6 % NAP rocks. The low NAP:ANC ratio is informative because it highlights that only a small proportion of NAP alteration is necessary to adversely impact water quality. Mining impacts that are usually coincident with areas of NAP alteration add to the acid and metal loads to surface water and groundwater. Where the NAP:ANC ratio is high, water quality impacts are severe that have toxic effects on aquatic life. Two basins with a high average NAP:ANC ratio are Mineral Creek and Cement Creek. Severe to acute Al, Cu, Zn, and Cd toxicity is observed for trout and sensitive taxa in the Mineral and Cement Creek basins (Besser et al. 2007). In contrast to areas that are pervasively altered to NAP assemblages, watersheds having mostly vein-related mineralization are characterized by alteration confined to narrow zones adjacent to veins that do not pervade adjacent propylitic rocks. The south Silverton mining area contains several watersheds with prevalent vein-related mineralization. GIS analysis delineates small areas of NAP alteration in the south Silverton mining area and large areas of propylitic alteration consistent with water quality having neutral pH, and low metal concentrations with the exception of Zn in some watersheds that can locally sustain trout (Besser et al. 2007).

The average elevation of propylitically altered rocks relative to the mean watershed elevation determined by GIS was found to be an important factor affecting water quality. Less acidic to neutral watersheds tend to have propylitically altered rocks that form at an average elevation that is higher than the mean watershed elevation. In contrast, the average elevation of propylitically altered rocks in acidic watersheds is usually lower than the mean watershed elevation. When propylitically altered rocks occur at high elevations in mountain watersheds where precipitation is commonly greater and snow melt is more persistent throughout the spring and summer, this provides a consistent source of alkalinity to help mitigate ARD.

A watershed grouping schema was developed for mined and unmined watersheds using a limited numbers of variables including alteration type area determined using GIS analysis of AVIRIS data, surface water pH at watershed outlets, historical records, and field observations. Land managers could benefit from such a schema that would aid in quickly assessing watersheds in terms of possible geoenvironmental impacts. This analysis used pH data because it is a quickly acquired field parameter that can also be a proxy for other water quality impacts such as high metal and sulfate concentrations. Five watershed groupings were determined based on similarities in pH, area affected by NAP alteration, and mining impacts (Fig. 9). Group 1 watersheds may represent the best case scenario for possible reclamation success and mitigating mining impacts due to the large watershed size and low NAP:ANC ratio. In contrast, a high NAP:ANC ratio for group 5 watersheds may pose the greatest geoenvironmental challenges due to acidic surface water conditions that existed prior to mining. Mining impacts for group 3 watersheds are clearly delineated due to the low NAP:ANC ratio, yet having acidic pH. Group 2 and 4 watersheds have minimal mining impacts and have acidic surface water quality because of varying proportions of NAP alteration. This type of grouping could aid in prioritizing reclamation resources for geoenvironmental cleanup and also aid mining companies in identifying watershed NAP and ANC characteristics. Where information about specific mineral deposit types is also known, a prediction of the types of water quality impacts could be more completely discerned.
https://static-content.springer.com/image/art%3A10.1007%2Fs12665-013-2229-y/MediaObjects/12665_2013_2229_Fig9_HTML.gif
Fig. 9

Watersheds grouped based on percent NAP alteration area, mining impacts, and pH. Group 1 (green) have 10–20 % NAP area, are mining impacted, and pH ranging from 5.07 to 5.21; Group 2 (yellow) have 30–40 % NAP, have minimal mining impacts, and pH ranging from 3.58 to 4.88; Group 3 (coral) have <10 % NAP, are mining impacted, and pH ranging from 2.88 to 4.19; Group 4 (light blue) have 50–100 % NAP, have minimal mining impacts, and pH ranging from 3.45 to 3.90; Group 5 (red) have 50–85 % NAP, are mining impacted, and pH ranging from 3.04 to 3.89. Watershed id’s correspond with those labeled in Fig. 2

Statistical analysis in this study was effective in determining the association between watershed characteristics and water quality signatures. Previous preliminary studies also established correlations between water quality and areas of alteration mapped using the hybrid AVIRIS- and field-based data (Yager et al. 2008a, b). It was not known, however, whether less detailed alteration maps derived from TM and ASTER would also be effective in determining correlations with watershed properties. Analysis in this study revealed correlations with water quality parameters despite relatively small percentages of watershed areas being mapped by TM and ASTER alteration datasets because of vegetation cover. This has important implications for national geoenvironmental and mineral assessments, especially in areas of the western United States that are more suitable to spectral mapping by remote sensing methods because the western regions are often not as highly vegetated as eastern US terrain.

Linear regression analysis identified correlations between mineral groups mapped by TM (TM–CSM, TM–Fe), major metal cations (Fe, Al) in surface water and low pH. These results were achieved despite a relatively small watershed area being mapped by TM. TM data are available for the entire US and suggest that when more detailed mineral maps are not available, TM could be useful in identifying potential water quality impacts from hydrothermally altered areas. ASTER was able to map a larger watershed area compared with TM and linear regression analysis revealed correlations between Aster–Arg–Fe, Aster-qsp, major metal cations (Fe, Al) in surface water and low pH. ASTER data are also available for the entire US and when more specific mineralogic data such as the presence of propylitic rocks containing acid neutralizing assemblages calcite and chlorite are required, ASTER can be used to augment mapping by TM. In general, correlations between NAP assemblages and water quality signatures are improved when mining impacted watersheds were excluded from analysis, suggesting that the GIS and statistical methods used in this study could be effective in unmined and hydrothermally altered areas to identify possible naturally occurring sources of acidity and metals to surface water.

AVIRIS mineral mapping have the most complete watershed area coverage and the highest correlations in regression analysis. The area of qsp was determined to have the highest correlations with major metal cations (Fe, Al) and low pH. Quartz–alunite–pyrophyllite and argillic alteration types mapped by AVIRIS are each correlated with low pH. Argillic rocks are also correlated with aluminum. Correlations were also observed between increasing areas of propylitic rocks mapped by AVIRIS, watershed area, increasing pH and alkalinity. Similar correlations using propylitic bedrock mapping by ASTER were lacking that is likely due to the much smaller watershed area mapped by ASTER. This suggests that to understand the acid neutralizing capacity of a watershed, more complete watershed area mapping is required.

Statistical correlations are generally higher for whole watershed area analysis when compared to linear correlations that involve analysis of altered areas intersected by a 30-m GIS stream buffer. In-stream processes were hypothesized as possibly being more important in controlling water quality signatures because of the effectiveness that the hydrologic stream network has in conveying sediment and associated contaminants to watershed outlets. In general, altered areas intersected by a GIS buffer were less highly correlated compared with alteration areas calculated for whole watersheds. This finding suggests that the whole watershed analysis is a better geochemical integrator of water quality signatures derived from weathering of hydrothermally altered areas. Most watersheds near Silverton, Colorado have steep slopes; therefore, physical and chemical weathering work to effectively transport sediment and simultaneously leach contaminants to streams throughout the entire area. A GIS stream buffer, if applied to a lower gradient region, may result in higher correlations with water quality signatures compared to steep terrain because downslope slumping and transport of sediment may be less important in rapidly conveying sediment to streams allowing for a greater water–sediment contact time.

Factor analysis allowed trace metal patterns to be identified that are consistent with weathering of sulfide-bearing hydrothermally altered areas. Only correlations between Cu and selected alteration assemblages were identified in linear regression analysis. Thus, correlations with other trace metals besides Cu in factor analysis is significant because it identified associated geoenvironmetal indicators that were predicted based on the mineral deposit types that are prevalent in the study area. An example of base metal correlations is observed for the ASTER data set in which Cu and Zn are inversely correlated with pH, reflecting increased trace metal abundances with acidic conditions. Patterns were also identified in factor analysis of AVIRIS data involving propylitically altered areas. In these analyses, the ratio ANC:NAP is correlated with pH and alkalinity, highlighting the importance of the relative proportions of acid neutralizing- and net acid producing-rocks in controlling acidity.

Conclusions

This analysis methodology emphasizes the importance of watersheds as key geographic units for interpreting water quality signatures in a geologic and morphometric context. Sampling strategies devised when a project is initiated could maximize the potential benefits of watershed scale analysis. Such a strategy would involve sampling watersheds of interest at the outlet during base flow conditions when concentrations of contaminants are not diluted by snowmelt or high precipitation. Analyses of watershed outlet samples would provide integrated geochemical information for all up-stream contributing areas, and constituents measured would be a result of the water–soil–atmosphere interactions along the entire upstream flow path. Standard field parameters would be collected including pH, specific conductance, dissolved oxygen, alkalinity, in addition to metal cations, anions, and sulfate. Mapping of geology, alteration, vegetation, terrain, and delineation of the hydrologic network would permit each watershed or basin to be well characterized and subsequently analyzed with GIS tools. Resultant data for watershed physical characteristics compiled from GIS would provide a basis for statistical analysis. This methodology would likely be valuable, especially in unmined areas, because the information acquired would begin to establish a geochemical baseline that could be compared to data collected once mining began or ceased. This approach would also enable watersheds to be ranked as to potential geoenvironmental impacts prior to and after mining.

Future analysis requires that basin-scale (10 s to 100 km2) areas be analyzed using a GIS and statistical methodology similar to that used in this study. In order to upscale to larger areas, conservative geochemical constituents that are collected at basin outlets will need to be evaluated. Sulfate, pH, or major metal cations (Al and Fe) may provide the best evidence for large-scale impacts to water quality from hydrothermally altered and mineralized areas. Geochemical data determined for stream sediments collected as part of the National Uranium Resource Evaluation (NURE) or during USGS Bureau of Land Management mineral assessment studies may be useful datasets for up-scaling. The stream sediments, while not providing a direct link to water quality effects, may provide an indication of possible bioavailable elements that could potentially impact water quality and aquatic life.

Footnotes
1

It is emphasized here that the TM-CSM is not mapping marble in the study area, but marble is permissive in this TM grouping defined by Rockwell (2010). The principal mineral groups mapped in hydrothermally altered areas include clays, sulfate (alunite), and sericite.

 

Acknowledgments

The authors wish to thank the USGS, Animas abandoned mine lands (AML) team members who all provided useful discussions and data during preparation of this report. Special thanks to Stanley E. Church (USGS, ret.) who co-led the successful AML program. The Animas River Stakeholders Group, led by William Simon was inspirational in addressing watershed and mining-related issues in Silverton between 1996 and present. Roland J. Viger (USGS) provided useful guidance in using the GIS Weasel software. Denise Dundon provided GIS support. USGS reviewers Mary Ellen Benson, Eric Anderson and two anonymous external reviewers provided useful critique of this work. This study was supported by the U.S. Geological Survey Mineral Resources Program, the Mineral Environmental Assessment Methodologies Project and the Central Mineral and Environmental Resources Science Center. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the US Government.

Supplementary material

12665_2013_2229_MOESM1_ESM.docx (17 kb)
Supplementary material 1 (DOCX 17 kb)
12665_2013_2229_MOESM2_ESM.docx (27 kb)
Supplementary material 2 (DOCX 28 kb)
12665_2013_2229_MOESM3_ESM.docx (17 kb)
Supplementary material 3 (DOCX 17 kb)
12665_2013_2229_MOESM4_ESM.docx (16 kb)
Supplementary material 4 (DOCX 16.1 kb)
12665_2013_2229_MOESM5_ESM.docx (15 kb)
Supplementary material 5 (DOCX 15 kb)
12665_2013_2229_MOESM6_ESM.docx (18 kb)
Supplementary material 6 (DOCX 19 kb)

Copyright information

© The Author(s) 2013

Open AccessThis article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.