Article pubs.acs.org/est
Climate-Change-Driven Deterioration of Water Quality in a Mineralized Watershed Andrew S. Todd,*,† Andrew H. Manning,† Philip L. Verplanck,† Caitlin Crouch,‡ Diane M. McKnight,‡ and Ryan Dunham§ †
U.S. Geological Survey, Denver, Colorado, United States University of Colorado, Institute of Arctic and Alpine Research, Boulder, Colorado, United States § U.S. Environmental Protection Agency, Denver, Colorado, United States ‡
S Supporting Information *
ABSTRACT: A unique 30-year streamwater chemistry data set from a mineralized alpine watershed with naturally acidic, metal-rich water displays dissolved concentrations of Zn and other metals of ecological concern increasing by 100−400% (400−2000 μg/L) during low-flow months, when metal concentrations are highest. SO4 and other major ions show similar increases. A lack of natural or anthropogenic land disturbances in the watershed during the study period suggests that climate change is the underlying cause. Local mean annual and mean summer air temperatures have increased at a rate of 0.2−1.2 °C/decade since the 1980s. Other climatic and hydrologic indices, including stream discharge during low-flow months, do not display statistically significant trends. Consideration of potential specific causal mechanisms driven by rising temperatures suggests that melting of permafrost and falling water tables (from decreased recharge) are probable explanations for the increasing concentrations. The prospect of future widespread increases in dissolved solutes from mineralized watersheds is concerning given likely negative impacts on downstream ecosystems and water resources, and complications created for the establishment of attainable remediation objectives at mine sites.
■
the subsurface flow of oxygen and water, as well as the surface and subsurface transport of weathering products. As such, significant change in climate conditions within mineralized areas has the potential to affect ARD production. In the mountains of the western U.S., mean annual air temperatures have generally increased by approximately 0.5− 1.0 °C/decade over the past three decades.13−15 Climate change is altering the snowmelt-dominated hydrologic cycle, as evidenced by observed declines in mountain snowpacks, shifts in the timing of snowmelt, and declining baseflows.16−20 Additionally, the gradual loss of ice features (glaciers, rock glaciers, and permafrost) in alpine areas in response to climate warming has been well documented.21−24 Each of these climate-related changes potentially could alter weathering processes in mineralized mountain watersheds, as well as the chemical, biological, and physical characteristics of the streams that drain them. We present a unique 30-year streamwater chemistry data set from the Upper Snake River, which drains a naturally ARDproducing, mineralized, alpine watershed in Summit County, Colorado, USA. The objective of this study is to evaluate trends
INTRODUCTION Acidic, metal-rich water draining from rocks high in sulfide minerals (acid rock drainage, or ARD) in the western United States is one of the greatest water quality challenges facing the region.1 The release of ARD from abandoned mines and associated mine wastes, specifically referred to as acid-mine drainage (AMD), represents a significant portion of the problem.2 However, natural ARD also exists in watersheds containing hydrothermally altered, pyritized bedrock where little or no mining has occurred.3,4 This natural ARD (hereafter simply ARD) also has a significant environmental impact, and commonly plays an essential role in the establishment of environmental baselines and cleanup standards at nearby mine sites.5,6 Acid rock drainage occurs when sulfide minerals present within rock are exposed to oxygen and water, resulting in the mobilization of H+ ions, SO4, and metals (e.g., Zn, Cu, Cd, Mn) through microbially mediated weathering reactions.7,8 Streams receiving ARD typically have a low pH and high metal concentrations that can limit both downstream aquatic ecosystem health and downstream human use of these water resources.9,10 Many components of ARD production and transport are influenced either directly or indirectly by the local climate. For example, sulfide oxidation reaction rates are known to be temperature-dependent.11,12 Similarly, local hydrology, which is directly linked to climate, largely governs © 2012 American Chemical Society
Received: Revised: Accepted: Published: 9324
May 18, 2012 July 23, 2012 August 2, 2012 August 17, 2012 dx.doi.org/10.1021/es3020056 | Environ. Sci. Technol. 2012, 46, 9324−9332
Environmental Science & Technology
Article
Figure 1. Map showing the Snake River watershed in Summit County, Colorado, including key cultural features (e.g., Pennsylvania Mine) as stars, snowpack telemetry (SNOTEL) sites as squares, and long-term monitoring locations (SW-044, SW-047, and SW-082) as circles.
conclude that unmined areas on the eastern side of the watershed produce the majority of ARD.4,28−30 No significant land disturbances have occurred in the watershed during the past 30 years. The USR watershed is underlain by fractured Precambrian metamorphic rocks consisting of gneisses, schists, and amphibolites that have been intruded by small Proterozoic granitic bodies and Tertiary-age, felsic-porphyry stocks.31 An iron fen, or ferricrete deposit (iron-hydroxide-cemented alluvium and colluvium), exists near the stream along most of its course. Disseminated and veinlet-hosted pyrite occurs in large zones of hydrothermal alteration related to the Tertiary intrusions.32 Veins and veinlets may contain other sulfide minerals including sphalerite, galena, chalcopyrite, and enargite.31 Natural weathering of this pyrite and associated sulfides results in USR water having low pH (typically 3.5−4.5) and elevated concentrations of SO4 and metals (Fe, Al, Zn, Cd, Cu, Mn; Figure 2; Table S1 in the Supporting Information (SI)).28,30,33 Surface elevations range from 3200 to 4000 masl. Mean annual precipitation at nearby snowpack telemetry (SNOTEL) sites in this elevation range is about 80 cm, with 60−80% falling as snow. No stream gage exists on the Upper Snake, and the closest gage is approximately 10 km downstream on the Snake River (gage 090475000 operated by the U.S. Geological Survey (USGS); Figure 1). Manual streamflow measurements made from April through July near the mouth of the USR indicate that flows at this site are well correlated (r2 = 0.80) with flows
in trace metal and major ion concentrations and address the fundamental question of whether climate change may be causing substantial changes in water quality in such watersheds. Alpine watersheds are sensitive indicators of environmental disturbances, and recent studies of alpine lakes provide examples of water chemistry changes apparently caused by climate warming.25,26 However, these studies were not located in watersheds with extensive mineralized bedrock, sampled waters were too dilute to be considered ARD, and trends in metal concentrations were not reported. We are aware of no other published long-term water chemistry data sets in mineralized watersheds, particularly those considered representative of natural background conditions with limited historical or recent anthropogenic activity.
■
SITE DESCRIPTION The Snake River watershed (∼155 km2) is located in central Colorado’s Rocky Mountains, immediately west of the Continental Divide (Figure 1). It hosts the Montezuma Mining District which was an area of extensive historical mining from the 1860s to the 1940s, and is now affected by both ARD and AMD from abandoned mines.27 The Upper Snake River (USR) watershed (above Deer Creek, ∼12 km2) is an essentially undeveloped alpine headwater portion of the larger Snake River watershed generally unaffected by AMD. The watershed contains some relatively small abandoned mines, but most are located high in the watershed and drain little or no water. Recent studies that investigated the source of ARD in the USR 9325
dx.doi.org/10.1021/es3020056 | Environ. Sci. Technol. 2012, 46, 9324−9332
Environmental Science & Technology
Article
71% of total annual runoff occurring in May, June, and July (Figure 2A). Surface-water chemistry data have been collected from the USR since the 1980s by various entities, including state and federal regulatory agencies, academic researchers, and the USGS (Table 1). Regulatory agencies have collected data because the watershed is considered a background sample location representative of premining conditions, and because it contributes significant metal loading to the greater Snake River system. Chemical conditions thus provide an important frame of reference in the establishment of remediation objectives for other parts of the Snake River affected by AMD. Clow examined data from SNOTEL sites and stream gages throughout the mountains of Colorado for the period 1978− 2007 and found clear trends in temperature, precipitation, and streamflow timing.15 Statewide, mean air temperatures increased at a median rate of 0.7 °C/decade, April 1 snowwater equivalent (SWE) declined at a median rate of 4.1 cm/ decade, and snowmelt and associated peak runoff shifted toward earlier in the year by 2−3 weeks. The central Front Range region, which includes the Snake River, displayed trends similar to statewide trends. However, the decline in April 1 SWE was less pronounced (1.2−2.7 cm/decade), as was the winter-ward shift of snowmelt and peak runoff (1.5−2.5 weeks).
■
METHODS
Water Chemistry Data. Surface-water chemistry data were compiled from existing studies for the USR (Table 1). Compiled data were collected at locations not more than 300 m above the Deer Creek confluence (hereafter site SW-044; Figure 1). Data not contained in journal articles, published reports, or student theses are presented in the SI (Tables S2, S3). All samples were field filtered, and samples for dissolved metals were acidified in the field. The compiled chemical data were screened for apparently erroneous determinations. A dissolved constituent determination was considered erroneous if its value was greater than twice the maximum value measured for the other samples. A pH measurement was considered erroneous if it was greater than 1.0 pH unit outside of the range of pH values measured for the other samples. The entire sample was removed from the data set if any constituent met the above criteria (15°. Mean annual and mean summer air temperature were not computed for years missing >20 days and summers missing >10 days of data, respectively. Discharge data for the Snake River gage are available for the entire period of interest, and upstream diversions and impoundments are minimal. The USGS ceased recording winter-time flows at this gage in 2005, so November−March flows from a gage ∼700 m downstream managed by the Colorado Department of Water Resources (SNAKEYCO) were used instead for the period 2006−2011 (http://www.dwr.state.co.us). The following climate indices were computed from the SNOTEL site data: mean annual air temperature (MAAT); mean summer air temperature (June−September; MSAT); total annual precipitation; maximum annual SWE and its percentage of total precipitation; SM50, defined as the day of the year on which half of the snowpack had melted (based on maximum annual SWE); and SM100, defined as the first day following SM50 on which no snowpack existed. MSAT was examined because alpine ground temperatures are expected to be most influenced by air temperatures in the summer when the ground is not insulated by snow. The following streamflow indices were computed: total annual discharge; mean low-flow discharge (August−April); mean high-flow discharge (May− July); mean September discharge; and Q20, Q50, and Q80, defined as the day of the year on which 20%, 50%, and 80% of 9327
dx.doi.org/10.1021/es3020056 | Environ. Sci. Technol. 2012, 46, 9324−9332
Environmental Science & Technology
Article
that are similar to or fall between neighboring months (Figure 2C,D). Over the past 30 years, dissolved Zn concentrations at SW044 have increased in all months of the year (Figures 3, S1). Upward trends are statistically significant with p < 0.03 for all months except December and January, which have the fewest data points postdating the 1980s (≤ 2). Concentrations have increased by as much as 350% (1000 μg/L) during low-flow months (August−April), which have the highest metal concentrations. For these months, Kendall line slopes are 149 (December) to 329 (September) μg/L/decade. During highflow months (May−July), concentrations have increased by as much as 200% (500 μg/L) and Kendall line slopes are 71 (June) to 173 (July) uμg/L/decade. In low-flow months, Zn concentrations appear to have increased more rapidly since the early 2000s than prior to this time. Identification of the primary source of Zn is beyond the scope of this work, but weathering of sphalerite (ZnS) from mineralized veins or weathering of vein or disseminated pyrite which may contain trace to minor amounts of Zn are two potential sources. Dissolved concentrations of other constituents show similar increasing trends at SW-044 (Figure 4). Concentrations are shown for September because this is the low-flow month with both the best temporal data coverage (particularly for recent years) and data from the most sources. Concentrations are shown for the high-flow month of June in Figure S2 for comparison. Mn is another relatively conservative metal of regulatory concern in the Snake River, and Ca and SO4 are two primary mineral weathering products. September dissolved Mn, Ca, and SO4 concentrations have increased 400% (2000 μg/L), 100% (7 mg/L), and 150% (100 mg/L), respectively. Upward trends are statistically significant with p < 0.01. Like Zn, concentrations of Mn, Ca, and SO4 have been increasing in most other low-flow months as well. Upward trends for these constituents also are apparent at sites lower in the Snake River watershed (SW-047, SW-117, SW-082; Figure 1). However, data records at these sites are typically more recent, smaller, and more complex due to more diverse upstream influences on water quantity and quality. Trends in Climate and Hydrology. The RKT indicates statistically significant upward trends in MAAT and MSAT in the USR vicinity (Figure 5; Table S4), with p-values ≤0.05 for both indices. Warming rates at individual sites range from 0.2 to 1.2 °C/decade, and median slopes for MAAT and MSAT are 0.34 and 0.33 °C/decade, respectively. No other indices display statistically significant trends (Table S4). This was unexpected given that Clow (2010) reported significant downward trends for SM50, Q20, Q50, Q80, and April 1 SWE for the central Front Range region.15 The discrepancy is likely due to the fact that three out of four years in the period 2008−2011, included in this study but not in Clow (2010), had above-average total precipitation and maximum annual SWE.15 Although not statistically significant, slopes computed in this study for the same five indices (assuming maximum annual SWE is nearly equivalent to April 1 SWE) are all negative. It is therefore possible that these indices are indeed declining gradually in the USR vicinity, but a longer data record and/or more sampling sites would be required clarify these trends. Mean September discharge trends downward, but this trend is not significant (p = 0.56; Figure 5C).
Figure 4. Trends in September dissolved Mn (A), Ca (B), and SO4 (C) concentrations in the Upper Snake River (SW-044). Different colors represent different data sources: yellow, ref 34; green, ref 62; orange, CDPHE (Table S2); red, EPA (Table S2); blue, USGS (Table S2); and purple, ref 30. p-values were computed using the Mann− Kendall Test, and lines shown are Kendall−Theil lines (see text).
the total discharge for the water year had passed by the gage (generally corresponding to the beginning, middle, and end of peak flow, respectively). Streamflow indices were analyzed for temporal trends using the MKT. Because climate data were available from three different sites, climate indices were analyzed using the Regional Kendall Test (RKT), which simultaneously considers data from multiple locations to test for a trend within a given region.37 As with the water chemistry data, trend magnitudes were estimated by computing Kendall lines.
■
RESULTS Trends in Water Chemistry. Given the large seasonal transients commonly observed in the chemistry of snowmelt dominated mountain streams38 (Figure 2), water chemistry from SW-044 was grouped by month for trend evaluation (Figures 3 and 4). For this discussion, we have chosen a suite of constituents (Ca, Mn, SO4, and Zn) which are relatively insensitive to variations in stream chemistry, most importantly variations in pH, and their concentrations are well above the detection limits of the analytical methods used.8 We focus on dissolved Zn concentrations because Zn is one of the primary metals of concern within the Snake River watershed, occurring at concentrations well in excess of Colorado’s table value criteria for the aquatic life classified use.39 Every month is not shown in Figure 3 due to space considerations, but missing months (shown in Figure S1) generally have concentrations 9328
dx.doi.org/10.1021/es3020056 | Environ. Sci. Technol. 2012, 46, 9324−9332
Environmental Science & Technology
Article
However, few studies have directly investigated linkages between ARD and climate. Plumlee (1999) presents data suggesting that similar types of mineral deposits produce different ARD levels when located in different climate zones.41 Furniss et al. (1999) find that radiocarbon dates of ferricrete deposits, which are formed by ARD, are correlated closely with warm-wet periods in the Holocene.42 Nordstrom (2009) considers the potential impacts of current climate changes on ARD, but focuses on evidence that surface water acidity and metal concentrations during the initiation of high-flow events (first-flush) will likely become yet higher due to longer and more intense dry spells.43 The possibility of increasing concentrations during intervening low-flow periods is noted, but decreased dilution under reduced stream flows is the only mechanism invoked. Multiple studies have reported increasing SO4 concentrations in mountain lakes and streams over the past three decades, and have attributed these to climate warming.25,26,44−48 Concentration increases of 50−150% (1−10 mg/L) are typically observed. Sampled waters were not collected in highly mineralized watersheds, and SO4 concentrations are well below ARD levels (generally
