Switching predominance of organic versus inorganic carbon exports from an intermediate-size subarctic watershed
Abstract
Hydrologic exports of dissolved inorganic and organic carbon (DIC and DOC) reflect permafrost conditions in arctic and subarctic river basins. DIC yields, in particular, increase with decreased permafrost extent. We investigated the influence of permafrost extent on DIC and DOC yield in a tributary of the Yukon River, where the upper watershed has continuous permafrost and the lower watershed has discontinuous permafrost. Our results indicate that DIC versus DOC predominance switches with interannual changes in water availability and flow routing in intermediate-size watersheds having mixed permafrost coverage. Large water yield and small concentrations from mountainous headwaters and small water yield and high concentrations from lowlands produced similar upstream and downstream carbon yields. However, DOC export exceeded DIC export during high flow 2011, whereas DIC predominated during low flow 2010. The majority of exported carbon was derived from near-surface organic sources when landscapes were wet or frozen and from mineralized subsurface sources when infiltration increased.
Key Points
- Predominance of annual DOC versus DIC export switches with hydrologic conditions
- Switching highlights the importance of shallow versus deep flow path partitioning
- Flow dynamics and vegetation sources interact to determine watershed C exports
1 Introduction
Rivers are important components of the global carbon cycle [Cole et al., 2007] that function as integrators of terrestrial processes [Jenerette and Lal, 2005], biogeochemical reactors [Battin et al., 2009; Aufdenkampe et al., 2011], and transporters of carbon from terrestrial landscapes to oceans [Stets and Striegl, 2012]. Northern high latitudes store vast amounts of organic carbon in permafrost [Tarnocai et al., 2009]. Though the eventual fate of permafrost carbon is uncertain, dissolved inorganic carbon (DIC) yields from river basins are known to increase with decreased permafrost extent across the circumboreal [Tank et al., 2012], and permafrost thaw has been associated with increased mineralization of dissolved organic carbon (DOC) and inflow of DIC-rich groundwater to rivers in the Yukon River basin [Striegl et al., 2005; Walvoord and Striegl, 2007; Walvoord et al., 2012]. Streamflow is primarily generated from surface runoff in continuous permafrost landscapes, and its associated DOC chemistry suggests soil surface and active layer origin [Striegl et al., 2007; Spencer et al., 2008; Aiken et al., 2014]. As permafrost thaws, some organic carbon is mineralized and transported via new subsurface hydrologic pathways that intensify groundwater inputs to rivers and streams, altering the DOC:DIC mix [Walvoord and Striegl, 2007; Walvoord et al., 2012]. Such climate-related changes make it critical to decipher the source and fate of landscape contributions to riverine carbon flux at multiple scales in northern environments. DOC and DIC exports are fairly well known for the Yukon basin and its largest tributaries [Striegl et al., 2007; Tank et al., 2012; Guo et al., 2013]. Catchment studies in the basin have tended to focus only on DOC [Maclean et al., 1999; Carey, 2003; Petrone et al., 2006; Koch et al., 2013], which is more prevalent in headwaters than DIC. However, DOC and DIC exports for intermediate-sized watersheds are generally unknown. Forty tributaries having watershed areas >1000 km2 discharge directly to the Yukon River; 25 of those are in the range of 2000 to 10,000 km2 [Frederick et al., 2011]. This study addresses the gap in understanding of seasonal and annual dissolved carbon exports from those intermediate-size watersheds.
During 2010 and 2011, we combined water discharge and chemistry measurements and the multivariate U.S. Geological Survey (USGS) load estimator model (LOADEST) [Runkel et al., 2004] to derive seasonal and annual loads and yields of dissolved carbon species for the Beaver Creek watershed in interior Alaska (Figure 1). Our immediate objective was to identify and quantify subsurface hydrologic and surface landscape influences on Beaver Creek carbon chemistry and export. Our broader objective was to improve general understanding of the effects of mixed permafrost distribution and permafrost thaw on carbon exports from intermediate-size northern high-latitude watersheds. Will permafrost thaw be detectable in carbon exports from these watersheds?
Our investigation was motivated by the hypothesis that the DIC:DOC ratio of Beaver Creek annual exports will increase as the stream courses from mountainous headwaters, having continuous permafrost, through low-lying Yukon Flats, having discontinuous permafrost. This hypothesis recognizes the importance of subsurface hydrology on aqueous carbon export and suggests that increased infiltration in areas having discontinuous and/or thawing permafrost will result in increased subsurface mineralization or sorption of DOC, increased weathering of particulate carbonates, and increased DIC export [Striegl et al., 2005; Walvoord and Striegl, 2007; Tank et al., 2012; Walvoord et al., 2012].
2 Methods
2.1 Site Descriptions
Beaver Creek is a remote tributary of the Yukon River in Alaska, having a watershed area of ~7000 km2 (Figure 1). We sampled water chemistry at two previously established streamgaging stations on Beaver Creek to investigate water availability, flow path, and land cover effects on downstream carbon exports (Table 1). Beaver Creek above Victoria Creek (USGS Station #15452100; “Upstream” station) drains 3315 km2 and receives water from the White Mountains and nearby headwater catchments. The upper Beaver Creek watershed is underlain by continuous permafrost. Below Victoria Creek, Beaver Creek enters Yukon Flats National Wildlife Refuge, an extensive area of lowland forest, wetlands, and lakes. Beaver Creek near Michel Lake (USGS Station #15452300; “Downstream” station) is within the refuge, approximately 180 river km downstream. This station has a watershed area of 6164 km2, with the “Intervening Area” between the Upstream and Downstream streamgages having discontinuous permafrost. Stage was measured at 15 min intervals at both locations, and water discharge (Q) was determined from stage-discharge relationships. Station descriptions and water data are accessible at http://waterdata.usgs.gov/NWIS.
| Beaver Creek | ||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Station | USGS Station # | Latitude, N Longitude, W (NAD 27) | Watershed Area (km2) | Watershed Contribution (%) | Elevation Range (m) | Elevation at Gage (m) | Year | Discharge (km3) | Water Yield (mm) | DOC Load (Gg C) | DIC Load (Gg C) | DOC Yield (g C m−2) | DIC Yield (g C m−2) | DOC FWMC (mol C m−3) | DIC FWMC (mol C m−3) | DIC:DOC FWMC |
| Victoria (Upstream) | 15452100 | 65°48′20″N, 146°38′55″W | 3315 | 54 | 229–1595 | 229 | 2010 | 0.45 | 132 | 3.90 | 6.20 | 1.2 | 1.9 | 0.72 | 1.14 | 1.6 (0.1) |
| 2011 | 1.01 | 317 | 14.60 | 9.20 | 4.4 | 2.8 | 1.20 | 0.76 | 0.6 (0.0) | |||||||
| Ave | 0.73 | 225 | 9.25 | 7.70 | 2.8 | 2.4 | 0.96 | 0.95 | ||||||||
| Intervening Area | NA | NA | 2849 | 46 | 113–1100 | NA | 2010 | 0.15 | 52 | 3.70 | 4.00 | 1.3 | 1.4 | 2.08 | 2.25 | 1.1 (0.5) |
| 2011 | 0.36 | 126 | 13.20 | 6.50 | 4.6 | 2.3 | 3.06 | 1.51 | 0.5 (0.2) | |||||||
| Ave | 0.25 | 89 | 8.45 | 5.25 | 3.0 | 1.8 | 2.77 | 1.72 | ||||||||
| Michel (Downstream) | 15452300 | 66°13′18″N, 146°45′44″W | 6164 | 100 | 113–1595 | 113 | 2010 | 0.61 | 95 | 7.60 | 10.20 | 1.2 | 1.7 | 1.05 | 1.40 | 1.5 (0.3) |
| 2011 | 1.41 | 229 | 27.80 | 15.70 | 4.5 | 2.5 | 1.64 | 0.93 | 0.6 (0.1) | |||||||
| Ave | 1.01 | 162 | 17.70 | 12.95 | 2.9 | 2.1 | 1.34 | 1.17 | ||||||||
- a DIC:DOC FWMC includes standard errors in parentheses.
2.2 Sample Collection and Analysis
Beaver Creek water samples were collected at each station approximately monthly during the open water season and 5 times during ice cover. Water samples were collected near the centroid of flow. Water for DIC analysis was collected without headspace using a polypropylene syringe, filtered through a 0.45 µm Whatman glass microfiber syringe filter, and injected into sealed 30 mL serum bottles having N2 headspace. The serum bottles were acidified, and the headspace was analyzed for total inorganic carbon content in our Boulder, Colorado, laboratory [Striegl et al., 2001].
Ten additional liters of sample were filtered through a Geotech 0.45 µm capsule filter into a clean rinsed container and kept cold and dark for the return flight to Fairbanks. The chilled sample was sent overnight to Boulder, Colorado, for DOC analysis within 48 h of arrival by the platinum catalyzed persulfate wet oxidation method on an OI Analytical 700TM Total Organic Carbon Analyzer [Aiken, 1992].
Daily river constituent loads (mass C d−1) were calculated using continuous discharge data and water chemistry measurements at the Upstream and Downstream stations and the USGS Load Estimator (LOADEST) program [Runkel et al., 2004], following established modeling conventions. Three seasons are defined as spring = 1 May to 30 June, summer-autumn = 1 July to 31 October, and winter = 1 November to 30 April consistent with related published studies [Striegl et al., 2005, 2007; Dornblaser and Striegl, 2007, 2009]. “Water Year” (WY) is defined as 1 October to 30 September. LOADEST requires at least 12 direct measurements of flow and chemistry over a wide range of flow conditions in order to calculate loads. LOADEST centers the Q and chemical concentration data to eliminate collinearity and automatically selects one of nine predefined regression models to fit the data, based on the Akaike Information Criterion. Annual DIC and DOC loads for the Beaver Creek stations were calculated for WY 2010 and 2011 using 13 discrete water-quality measurements (Figure 2; supporting information Table S1). We additionally calculated the apparent water discharge and C loads and yields for the Intervening Area between the Upstream and Downstream gages on Beaver Creek.
Yields were calculated by dividing total Q or constituent load for a flow period by watershed area. Water yield (runoff) is presented as millimeters, and constituent yield is presented as mass C m−2 yr−1. Flow-weighted mean concentrations (FWMCs) are annual loads (moles) divided by annual discharge (m3).
3 Results
3.1 Water Discharge
Annual water discharge was roughly twice as great in 2011 as it was in 2010 at the Upstream and Downstream stations (Table 1 and Figure 2). The doubling of Q was consistent with data from the U.S. Department of Agriculture (USDA) Snow Telemetry (SNOTEL) station in the mountainous headwaters of Beaver Creek, which recorded 68.8 cm of precipitation in 2011 and 34.8 cm in 2010 (National Water and Climate Center, http://www.wcc.nrcs.usda.gov/snow/). For comparison, the average precipitation during 2007–2009 was 56.1 ± 2.7 cm, highlighting that 2011 was wetter than average and 2010 was drier than average. Annual precipitation at the SNOTEL station in Fort Yukon, Alaska, located about 50 km northeast of the Beaver Creek watershed, more than doubled from 2010 to 2011 (12.2 cm to 26.7 cm), with 2010 being the driest year in the period of record (1994–2013) and 2011 being the second wettest. Thus, the pattern in precipitation and wet versus dry years was not limited to the headwaters of Beaver Creek but encompassed the entire Beaver Creek watershed and Yukon Flats.
Beaver Creek hydrographs show a peak in Q just after ice out in early May, with peak Q at Downstream station trailing Upstream station by approximately 2 days. Peak Q occurred 1 week later in 2011 than it did in 2010 at both stations. In 2011, a large extended snowmelt peak in Q is evident, reflecting that snowpack at the headwaters SNOTEL station was more than twice as thick in late winter 2011 as it was in late winter 2010. Again, the same pattern in snow depth was observed at the Fort Yukon SNOTEL station.
3.2 Carbon Loads, Yields, and FWMCs
Table 1 lists LOADEST-derived DIC and DOC loads and yields for the Upstream and Downstream streamgages, and for the Intervening Area. LOADEST model fit for both DIC and DOC had R2 values from 0.91 to 1.00 for all LOADEST model runs, the ratio of estimated to observed loads ranged from 0.96 to 1.09, and the Nash-Sutcliffe Efficiency Index (E) ranged from 0.84 to 0.99, all indicating good fit using automatically selected predefined regression models (supporting information Table S2).
For 2010 and 2011, dissolved C loads increased from Upstream to Downstream stations, with DIC load increasing 50% and DOC load nearly doubling (Table 1). There were large differences in the relative amounts of DIC and DOC transported between years (Table 1). In low flow 2010, DIC comprised approximately 60% of the total dissolved C load at both Upstream and Downstream stations, with DOC comprising about 40%. Predominance switched in high flow 2011, with DOC contributing more than 60% of the total C load. This switching was also evident for the Intervening Area. Although watershed normalized DIC and DOC yields increased substantially between 2010 and 2011, the yields were similar at the Upstream and Downstream stations and the Intervening Area for each year.
FWMCs of DOC and DIC increased between Beaver Creek Upstream and Downstream stations in 2010 and 2011 (Table 1 and Figure 3). However, the proportion of DIC versus DOC export reverses between years, with DIC dominating export in 2010 and DOC dominating in 2011. Regardless of the reversal, FWMC DOC and DIC increased by approximately 1.4 and 1.2-fold, respectively, from Upstream to Downstream station in both years. When placed in a basin perspective, Beaver Creek watershed has larger DOC yield and smaller DIC yield than the entire Yukon basin, and the FWMC DOC at Beaver Creek Downstream site is nearly twice that of Yukon River at Pilot Station near the river mouth (Figure 1) [Striegl et al., 2007] (http:waterdata.usgs.gov/NWIS). But, the Yukon River has nearly twice the FWMC DIC [Striegl et al., 2007]. FWMCs of both DOC and DIC are substantially greater for the Intervening Area than for Downstream station.
Seasonal loads of DIC and DOC are listed in Table 2. Approximately half the annual DIC load was exported during summer-autumn at both Upstream and Downstream stations, for both years, with spring exports contributing another third. In contrast, approximately 70% of annual DOC export occurred in spring, with summer-autumn contributing much of the remaining load. As with DIC, the proportions of DOC seasonal loads did not appreciably change between years or stations, except that greater proportions of winter DIC and DOC loads were suggested in 2010 than in 2011.
| Station | Year | Season | Water Yield (mm) | Water Yield (% of Annual Yield) | DIC Load (% of Annual DIC) | DOC Load (% of Annual DOC) |
|---|---|---|---|---|---|---|
| Victoria (Upstream) | 2010 | Spring | 64 | 48.5% | 32.3% | 70.2% |
| Summer-Autumn | 57 | 43.2% | 52.4% | 24.9% | ||
| Winter | 11 | 8.3% | 15.3% | 4.8% | ||
| Victoria (Upstream) | 2011 | Spring | 155 | 48.9% | 35.2% | 72.6% |
| Summer-Autumn | 153 | 48.3% | 55.5% | 26.9% | ||
| Winter | 9 | 2.8% | 9.3% | 0.4% | ||
| Michel (Downstream) | 2010 | Spring | 45 | 47.4% | 36.0% | 69.6% |
| Summer-Autumn | 42 | 44.2% | 50.2% | 27.0% | ||
| Winter | 8 | 8.4% | 13.8% | 3.4% | ||
| Michel (Downstream) | 2011 | Spring | 126 | 55.0% | 39.9% | 78.2% |
| Summer-Autumn | 96 | 41.9% | 52.8% | 21.3% | ||
| Winter | 7 | 3.1% | 7.2% | 0.5% |
4 Discussion
As with the largest tributaries of the Yukon River basin [Striegl et al., 2007], Q was greatest during spring flush at Upstream and Downstream Beaver Creek gaging stations (Figure 2), as were concomitant DOC loads. Nearly three fourths of the annual DOC load was exported in spring (Table 2). This majority of DOC export during spring flush reflects routing of runoff across organically rich and frozen or mostly frozen soil surfaces, when the subsurface active layer is also frozen. DIC export, on the other hand, commonly reflects deeper subsurface flow paths [Walvoord and Striegl, 2007; Tank et al., 2012]. Thus, half the DIC export occurred in summer-autumn, when Q decreased and the active layer deepened, permitting proportionally more subsurface flow. This pattern held relatively constant across both stations and both years (Table 2). Both DOC and DIC annual loads increased substantially from 2010 to 2011 (Table 1), and runoff, which more than doubled between years, was the primary driver of annual C loads. While the proportions of DIC and DOC to the total C load did not change between Upstream and Downstream stations (Table 1), the proportions did change with annual runoff. DIC was approximately 60% of the total C load in 2010, whereas DOC comprised more than 60% of the total C load in 2011. This change reflects the doubling of Q between years, mostly during spring runoff (Figure 2), and therefore the relative importance of surface versus subsurface flow paths on C export. Late season snowpack depth and total annual precipitation throughout the region were twice as great in 2011 as 2010, driving interannual differences in Q at the Beaver Creek Upstream and Downstream stations and influencing the sources of aquatic C transported from the watershed. Interestingly, the change in the predominance of DIC versus DOC export is not observed at gaging stations on the Yukon River or on the two largest tributaries of the Yukon River, the Tanana and Porcupine Rivers, where DIC export always exceeds DOC export despite a wide range in Q [Striegl et al., 2007, 2012]. However, the ratio of DIC:DOC export was 7.2 in the Tanana River watershed, which has discontinuous permafrost, and was 1.5 in the Porcupine River watershed, which has continuous permafrost [Striegl et al., 2007]. Terrestrial weathering, some within-stream particulate inorganic carbon weathering, and respiration maintain high DIC concentrations in the Yukon River [Striegl et al., 2007]. At the catchment scale, DOC likely predominates C export [Maclean et al., 1999; Carey, 2003; Petrone et al., 2006; Koch et al., 2013], and at the large Yukon River basin scale, DIC predominates. However, at the intermediate size of the Beaver Creek watershed, predominance of dissolved C species switched, depending on water availability and flow path. This is important to consider when ascertaining the influence of permafrost thaw and increased infiltration on downstream carbon exports.
Watershed normalized yields of dissolved carbon species did not appreciably change between Upstream and Downstream stations in either year (Table 1), which might suggest that the entire watershed behaves similarly in terms of carbon export. However, ~72% of Q originated from mountains and upland forest upstream of the Upstream station during the study (Table 1, http://waterdata.usgs.gov/NWIS). Additionally, Yukon Flats are an extensive area of lakes and wetlands having high DOC concentrations [Halm and Guldager, 2013], and wetland abundance is commonly cited as a primary driver of DOC flux [Frey and Smith, 2005; Raymond and Oh, 2007; Lauerwald et al., 2012]. Percent lake area increases from 0.3% of the watershed above Upstream station to 3.4% of the Intervening Area between stations, and respective wetland area doubles from 5.1% to 10.2% (USDA Agriculture Watershed Boundary Data Set, http://datagateway.nrcs.usda.gov and 2006 National Land Cover Database, http://www.mrlc.gov/nlcd2006.php). Consequently, an increase in FWMC DOC could be expected as Beaver Creek courses through Yukon Flats. Table 1 and Figure 3 indicate a 1.4-fold increase in FWMC DOC and a 1.2-fold increase in FWMC DIC between Upstream and Downstream stations. If we just consider the 2849 km2 of the Intervening Area between those locations (Table 1), specific runoff from that region was only about 40% of runoff from the headwaters above the Upstream gage for the 2 year study. Hence, FWMC DOC from that region of Yukon Flats between the gages was actually much greater than the measured FWMC DOC at the Downstream gage. Assuming conservative behavior, FWMC DOC from the Yukon Flats region between the stations was 2.9 times greater than above the Upstream gage in 2010 and 2.6 times greater in 2011. The assumption of conservative behavior appears reasonable and is supported by the results of Wickland et al. [2012], who measured minimal biodegradation of DOC in 28 day incubations of samples from both stations during the open water season, as compared to the estimated 2 day water transit time.
Using the same logic, if wetlands and other lowlands are a primary source of the DOC leaving the Flats, the specific DOC loading from those low-elevation source areas must actually be greater than suggested by our above assumptions. This is because 62% of the Intervening Area has a higher elevation than the Upstream gage and likely has a DOC FWMC that is more similar to the Upstream watershed than the lowlands. Applying the Upstream DOC FWMC to that 62% and again assuming conservative DOC behavior, FWMC DOC from the low-elevation portions of the watershed between the gages would have been about 3.6 times greater than at the Upstream station in 2010 and 3.1 times greater than in 2011.
FWMC DIC from the Intervening Area also increased and was 2.0 times greater than at the Upstream gage in both 2010 and 2011 (Table 1 and Figure 3). Contrary to our hypothesis, DIC:DOC decreased both years as Beaver Creek coursed through Yukon Flats. Although FWMCs of both DIC and DOC increased, FWMC DOC increased more than FWMC DIC. We attribute this to the substantial addition of hydraulically connected lakes and wetlands in Yukon Flats as a carbon source. This additional source outweighed the potential effects of increased subsurface flow on DIC:DOC exports from unsaturated lowlands having discontinuous permafrost. This reinforces the importance of source area vegetation and associated organic C production and respiration in contributing to both DOC and DIC load.
5 Conclusions
Large DOC yield and small DIC yield from Beaver Creek watershed, relative to the Yukon basin as a whole, suggest that regions of flats, like Yukon Flats, represent sources of C that are more organic C rich and inorganic C poor than the collective average input of all contributing watersheds in the basin. The hydrology of Yukon Flats is also sensitive to permafrost thaw [Walvoord et al., 2012; Jepsen et al., 2013], and increased permafrost thaw and infiltration are expected to result in increased organic C mineralization [Striegl et al., 2005] and mineral weathering [Tank et al., 2012]. Yukon Flats and other flats regions warrant closer observation of carbon and nutrient biogeochemical dynamics as permafrost thaw progresses. Still, Beaver Creek watershed only represents < 1% of the Yukon basin, and large variability among contributing watersheds is expected. Small permafrost-dominated watersheds respond dramatically and unevenly to environmental forcing, and their interannual hydrologic variability is so large that it will likely require many years of observations to decipher permafrost thaw effects on C and other material exports [Carey et al., 2013]. Large rivers have less hydrologic variability and integrate the inputs of many small watersheds and hence are indicators of regional change [Striegl et al., 2005; Holmes et al., 2012; Tank et al., 2012].
Like the Yukon River and its largest tributaries, most lateral export of DOC in Beaver Creek occurred during spring flush, whereas most DIC export occurred in summer-autumn. Doubling of discharge from 2010 to 2011 was manifested by switching of the predominance of DOC and DIC in the total annual C load, verifying the importance of shallow versus deeper flow paths under different discharge regimes. Additionally, increased FWMCs of DOC and DIC along the Beaver Creek continuum highlight the contributions from prominent landscapes, especially the Yukon Flats region in the lower watershed. We attributed increased dissolved carbon yields to increased vegetation and respiration sources and the interannual shift in the predominance of dissolved C species to changed surface versus subsurface routing of flow, illustrating the interdependence of ecological, physical, and biogeochemical controls on watershed C exports. Rates of infiltration, permafrost thaw, plant production, and respiration will likely increase with climate warming. All these factors depend on water availability. This reinforces the need for long-term hydrological and biogeochemical observations of regions vulnerable to permafrost thaw and for improved coupling of terrestrial and aquatic models that accurately account for and project changes in ecosystem carbon dynamics at all scales [Walvoord et al., 2012; Kicklighter et al., 2013].
Acknowledgments
We thank Heather Best and the USGS Alaska Water Science Center for field and logistical support; Kenna Butler for DOC analyses; and Michelle Walvoord, Kimberly Wickland, John Crawford, Wolfgang Knorr, and four anonymous reviewers for their helpful comments that improved the manuscript. This research was supported by the USGS National Research Program. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.
The Editor thanks two anonymous reviewers for their assistance in evaluating this paper.
