Dam tailwaters compound the effects of reservoirs on the longitudinal transport of organic carbon in an arid river

Reservoirs on rivers can disrupt organic carbon (OC) transport and transformation, but less is known how river reaches directly below dams contribute to OC processing. We compared how reservoirs and their associated tailwaters affected OC quantity and quality by calculating particulate OC (POC) and dissolved OC (DOC) fluxes, and measuring composition and bioavailability of DOC. We sampled the Yampa River near Maybell, Colorado, USA, and the Green River above and below Fontenelle and Flaming Gorge reservoirs as well as their respective tailwaters from early snowmelt to base flow hydrological conditions. In unregulated reaches (Yampa River, Green River above Fontenelle reservoir), DOC and POC concentrations increased with snowmelt discharge. POC and DOC concentrations also increased with stream discharge below Fontenelle reservoir, but there was no relationship between DOC and stream flow below Flaming Gorge reservoir. The annual load of POC was 3-fold lower below Fontenelle Reservoir and nearly 7-fold lower below Flaming Gorge reservoir, compared to their respective upstream sampling sites. DOC exported to downstream reaches from both reservoirs was less bioavailable, as measured with bioassays, than DOC upriver of the reservoirs. Lastly, tailwater reaches below the reservoirs generated OC, exporting potentially 1.6–2.2 g C m d of OC to downstream ecosystems. Therefore, the effect of impounding rivers on C fluxes is greater than the impact of the reservoirs alone given the additive effect of tailwater reaches below dams, which may produce and export comparable amounts of likely autochthonous carbon to downstream reaches.


Introduction
Unregulated streams and rivers compose a continuous ecosystem where a gradient of physical processes drive biological processes from headwaters to the river deltas Flow regulation by damming has converted most rivers into a series of lotic and lentic reaches (Ward and Stanford, 1983;Benke, 1990), affecting OC cycling and transport (Ward and Stanford, 1983;Miller, 2012;Stackpoole et al., 2014). Reservoirs on rivers may trap particulate OC (POC) (Friedl and Wuest, 2002;Downing et al., 2008;Tranvik et al., 2009), and transform and produce dissolved OC (DOC) 10 (Mash et al., 2004;Knoll et al., 2013). Increased water residence time (Vörösmarty et al., 1997;Sabo et al., 2010) allows for OC to be respired, incorporated into microbial production, or buried while production of autochthonous or microbial DOC increases (Mash et al., 2004;Knoll et al., 2013). Reservoirs may increase (Parks and Baker, 1997), decrease (Miller, 2012;Knoll et al., 2013), or not alter DOC concentrations to 15 downstream ecosystems (Parks and Baker, 1997;Nadon et al., 2014). Prior work has shown DOC fluxes increased longitudinally in the upper basin of the Colorado River, but then decreased with the presence of large reservoirs in the lower basin (Miller, 2012;Stackpoole et al., 2014). Conversely DOC fluxes increased in the lower Missouri River despite the presence of large reservoirs (Stackpoole et al., 2014). Similarly, DOC com-20 position did not change from upstream to downstream of reservoirs in boreal-forested rivers in northern Ontario where catchment characteristics had a stronger influence compared to the presence of impoundments (Nadon et al., 2014). These basin wide, large-scale studies have given insight into longitudinal OC fluxes in light of flow regulation by dams, but have not necessarily captured OC dynamics in the river reaches POC may consist of sloughed algae from production within the river reach (Webster et al., 1979;Perry and Perry, 1991). Algae in tailwaters may increase DOC concentration via exudation (Baines and Pace, 1991;Bertilsson and Jones, 2003). For example, autochthonous DOC flux was correlated with gross primary production in the tailwater of Glen Canyon Dam on the Colorado River and may equate to 7-91 % of gross primary production in the tailwater (Ulseth, 2012). Tailwater ecosystems produce algal OC, which is exported (Perry and Perry, 1991;Lieberman and Burke, 1993), transformed, buried, or consumed. These tailwater reaches can be found directly downstream of most, if not all, dams (Ward and Stanford, 1983); yet, coupled reservoir and tailwater OC fluxes are unclear within the context of riverine OC budgets. 15 We studied a series of reaches on the Green and Yampa Rivers located in the upper basin of the Colorado River, USA to quantify the role of hydrological regulation on OC quantity and quality. Within this objective, we asked the following questions: (1) How do OC concentration, fluxes, and DOC composition and bioavailability vary temporally in hydrologically regulated reaches compared to free-flowing rivers? (2) How 20 does the bioavailability and composition of DOC vary longitudinally in a river altered by reservoir-tailwater ecosystems? We addressed these questions by quantifying POC and DOC concentration and DOC composition and bioavailability in regulated and unregulated river reaches in the upper Colorado River basin. We sampled from the onset of snowmelt, where we expected transport processes to dominate, and during base Introduction

Study site
The Colorado River basin is heavily regulated (Nilsson et al., 2005) with 7 large impoundments (reservoirs with > 0.5 km 3 storage capacity) within its watershed, making it an opportune river basin to study DOC dynamics in regulated rivers. We selected sam- 5 pling sites to capture OC processes of non-regulated reaches, above and below reservoirs, and also to capture tailwater OC dynamics. In total, we selected seven sites on the Green and Yampa Rivers located in the upper basin of the Colorado River ( Fig. 1). Two sites served as unregulated reaches: the Green River above Fontenelle reservoir near La Barge, Wyoming and the Yampa River near Maybell, Colorado. To capture 10 potential longitudinal changes in OC transport and DOC composition and quality, we continued our sampling downstream starting above Fontenelle reservoir. Fontenelle reservoir had a mean water capacity of 0.26 km 3 and a mean water residence time of 0.13 yr for 2011 (Table A1). We sampled below Fontenelle dam at two locations: directly below the dam ( Fig. 1) and another site 39.6 km downriver to measure the OC 15 dynamics in the tailwater (referred to as Fontenelle tailwater). Further downstream, we sampled above Flaming Gorge reservoir. During 2011, Flaming Gorge reservoir had a mean volume of 4.08 km 3 and a mean residence time of 1.61 yr (Table A1). We sampled at two locations below Flaming Gorge dam, immediately below the dam and 25.7 km further downriver to capture tailwater effects on OC dynamics (referred to as 20 Flaming Gorge tailwater). Our sampling sites were located at US Geological Survey gaging stations, except for tailwater sites. For tailwater sites we assumed no change in discharge over these short distances from the dams. We sampled from the onset of snowmelt to base flow in one year to compare processes during runoff and base flow in these snowmelt driven rivers. We sampled at Introduction high-sustained flows into late July (Fig. 2). All sampling sites were accessible by car, and collection of samples took place over a two-day period for each round of sampling.

Sample collection
We collected samples to quantify DOC concentration, composition, and bioavailability as well as particulates for POC. We used acid-washed polyethylene Cubitainers to col-5 lect water from each sampling site. We immediately placed the collected water on ice, and filtered within 8 h of collection. We used a pre-rinsed Supor capsule filter (0.2 µm capsule filter; Pall SUPOR AcroPak 200) to filter approximately 4 L of water from each site for the bioassay experiments described below. From the remaining water, we collected samples for absorbance spectroscopy measurements and DOC concentration.

10
The water samples were filtered into acid-washed, pre-combusted 40 mL amber glass vials. Four replicates were immediately acidified with 400 µL of 2N HCl for later DOC analysis, and four other replicates were used for spectral measurements. All DOC samples were kept cold and dark until analyses. We analyzed DOC samples on a Shimadzu TOC 5000A in Laramie, Wyoming. Individual samples were run a minimum of three 15 times to estimate analytical precision. The coefficient of variation for replicate runs of the same sample was < 2 %.

DOC composition
We used spectrophotometric absorbance to evaluate DOC chemical composition. We used a scanning spectrophotometer to measure the absorbance of DOC from 200 20 to 600 nm on a Beckman DU spectrophotometer. We scanned the DOC samples using a 5 cm quartz cuvette and used deionized water as the blank. To characterize the structure of DOC, we used the absorbance measured at 254 nm, normalized by the DOC concentration to calculate specific ultraviolet absorption, otherwise referred to as SUVA 254 (L mg C −1 m −1 ). SUVA 254 indicates the aromaticity of the C compounds and Introduction ues indicating more aromaticity of the DOC compounds (Chin et al., 1994;Weishaar et al., 2003). Additionally, we calculated the spectral slope ratio (S R ). S R is the ratio of the spectral slopes (S 275-295 : S 350-400 ) at wavelength regions of 275-295 and 350-400 (Helms et al., 2008). S R is inversely correlated with DOC molecular weight and has been shown to shift in response to DOC photo alteration (Helms et al., 2008). By 5 using SUVA 254 and S R we expected higher SUVA 254 values and lower S R values for more aromatic, higher molecular weight DOC, and lower SUVA 254 , higher S R for lower molecular weight DOC.

Bioassay experiments to estimate bioavailability of DOC
Bioassay experiments, where we measured the decline in DOC over time, represent 10 the potential bioavailability of DOC to the microbial assemblage (del Giorgio and Davis, 2003). For each bioassay experiment, we added 1 L of 0.2 µm filtered river water to an acid-washed, pre-combusted glass jar. Then, we inoculated the filtered river water with 10 mL of 0.7 µm filtered water (pre-combusted glass fiber filter; Whatman GF/F). We inoculated deionized water with 10 mL of 0.7 µm filtered water as a control. We ran the 15 bioassays in triplicate, with one control per site and date. We incubated all bioassay experiments in the dark throughout the experiment. We collected DOC samples from the bioassay jars every few days for 28 days and then preserved and analyzed as described above. The decline in DOC concentration over time was fit to a 1st order exponential decay model (del Giorgio and Pace, 2008) such that, where ln DOC total is the natural log transformed total DOC concentration (mg L −1 ), ln DOC initial is the natural log transformed initial concentration of DOC (mg L −1 ), k is the decay rate (d −1 ) and t is incubation time (d). We used the lm function (

Particulate organic carbon
We filtered 0.2-5 L of river water, depending on the site and amount of sample visibly retained on the filter, through pre-combusted glass fiber filters to estimate POC (precombusted glass fiber filter; Whatman GF/F). Triplicate samples were dried at 60 • C, weighed for dry mass, and combusted at 500 • C. Following combustion, we re-wetted 5 the filters to account for potential clay dehydration, re-dried, and weighed again. The ash free dry mass (AFDM, mg L −1 ) of the particulate samples was calculated as the difference in the dry mass (mg) and combusted mass (mg) divided by the volume filtered (L). To estimate POC, we assumed that 45 % of the AFDM was OC (Whittaker and Likens, 1973).

Statistical analyses
To address longitudinal changes on OC transport, we compared the mean POC and DOC concentrations, SUVA 254 , S R , and bioavailability as k (d −1 ) at each site along the Green River with the closest upstream site. We used a paired t test to evaluate if the mean OC concentrations, spectral data, and bioavailability statistically differed between 15 sites in relation to upstream or downstream of Fontenelle or Flaming Gorge reservoirs and their respective tailwaters (Dalgaard, 2008). We used R (R Development Core Team, 2012) to conduct all statistical analyses.

Fluxes of DOC and POC
We calculated the daily fluxes and annual loads of DOC and POC for each sampling 20 site. Daily fluxes of DOC and POC were calculated as: where Flux d was the daily DOC or POC flux (g d  (Table 2), and therefore we were confident in extrapolating daily Flux d from predicted [OC] d and Q d for the calendar year of 2011. These daily fluxes were then summed to estimate the 2011 annual loads for DOC and POC for each sampling site.

20
Longitudinal OC concentrations, and the subsequent OC load for 2011 fluctuated in the presence of reservoir-tailwater ecosystems along the Green River (Table 3). POC concentrations were lower below both Fontenelle and Flaming Gorge dams compared to upstream of the reservoirs (Fig. 2). POC concentrations averaged 1.2 mg L −1 above Fontenelle reservoir compared to 0.4 mg L −1 below the dam (paired t test, p = 0.002). Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Lower concentrations of POC directly below Fontenelle dam translated to an annual POC load that was nearly 3-fold lower compared to upstream of the reservoir (Table 3). POC concentrations (Fig. 2) and subsequent OC loads (Table 3) above Fontenelle reservoir also had the most temporal variability, similar to the Yampa River, compared to the other sampling sites along the Green River. POC concentrations above Flaming

5
Gorge reservoir averaged 1.4 mg L −1 compared to directly below Flaming Gorge dam, which averaged 0.2 mg L −1 (paired t test, p < 0.0001). The annual POC load for 2011 was reduced nearly 7-fold directly below Flaming Gorge dam compared to upstream of the reservoir (Table 3).
Although POC loads were lower directly below both dams compared to above, 10 POC concentrations increased within their tailwaters ( Fig. 2). POC concentrations rebounded 2-fold at the Fontenelle tailwater site relative to directly below the dam (paired t test, p < 0.0001) and averaged 0.9 mg L −1 . Higher POC concentrations resulted in an annual POC load that was nearly 3-fold greater at the Fontenelle tailwater site compared to directly below the dam (Table 3). POC concentrations averaged 0.6 mg L −1 at 15 the Flaming Gorge tailwater site, which equated to a 3-fold increase in concentration (paired t tests, p < 0.0001) and an annual POC load for 2011 nearly 4-fold greater relative to directly below the dam (Table 3). Variation in DOC concentrations and the annual load of DOC along the Green River was less pronounced than that for POC. Mean DOC concentrations did not vary above 20 and below either reservoir during 2011 (paired t test, p > 0.1 for both reservoirs). However, snowmelt DOC concentrations above both reservoirs were greater than DOC concentrations directly below their respective dams, and vice versa during base flow conditions. Similar to the Yampa River, DOC concentrations peaked prior to peak discharge above Fontenelle reservoir, but peaked with discharge below the reservoir (Fig. 2 the annual DOC load within the 39.6 km reach (Table 3). DOC concentration averaged 3.7 mg L −1 at the Flaming Gorge tailwater sampling site compared to 3.6 mg L −1 directly below Flaming Gorge dam (paired t test; p < 0.0001). The increase in DOC concentration between the two sampling sites equated to an annual load of 244 Mg of DOC within the 25.7 km Flaming Gorge tailwater reach (Table 3).

DOC bioavailability and composition
DOC bioavailability, as measured by the decay rate k (d −1 ) of DOC, was lower directly downstream of both reservoirs than the bioavailability of the DOC upstream (Fig. 4, Table B1). Mean bioavailability above Fontenelle reservoir was 0.0036 d −1 compared to 0.0024 d −1 directly below the dam (paired t test, p = 0.0005). Average DOC bioavail- 15 ability was 2-fold greater above (0.0030 d −1 ) Flaming Gorge reservoir compared to directly below the dam (0.0014 d −1 , paired t test, p = 0.0002). Some seasonal variation in bioavailability was measured where k (d −1 ) was higher during onset of snowmelt, but decreased and remained relatively constant through the remaining snowmelt and base flow conditions (Table B1).

20
Bioavailability was greater at the tailwater sites relative to sampling sites directly below their respective dams (Fig. 4 directly downstream of the dam. S R increased, and SUVA 254 decreased from above Flaming Gorge reservoir to directly below its dam (Fig. 3). DOC composition directly below Flaming Gorge dam reflected less aromatic and smaller molecular weight DOC compared to more aromatic and larger molecular weight DOC above the reservoir (t test, SUVA 254 : p < 0.0001, S R : p < 0.0001). In comparison, SUVA 254 and S R re-5 mained relatively unchanged above and below Fontenelle reservoir, although composition shifted at each site during the transition from snowmelt to base flow (Fig. 3). DOC composition, based on SUVA 254 and S R , varied little through the tailwater reaches. SUVA 254 decreased from directly below Fontenelle reservoir to the Fontenelle tailwater site, indicating less aromatic DOC (mean difference = 0.04 L mg C −1 m −1 , 10 paired t test, p < 0.0001), but there was no statistically significant difference in S R between these sites (paired t tests, p = 0.1). Flaming Gorge tailwater reach had no effect on SUVA 254 or S R (SUVA 254 : paired t test, p = 0.27, S R : paired t tests, p = 0.08).

Discussion
Impoundments on rivers may disrupt longitudinal OC transport (Ward and Stanford, 15 1983;Miller, 2012;Stackpoole et al., 2014); however, the combination of reservoirtailwater ecosystems on OC dynamics is less understood than the impact of reservoirs alone. By measuring OC concentrations and DOC composition and bioavailability, we found that longitudinal OC dynamics fluctuated in the presence of reservoir-tailwater ecosystems. POC concentrations and DOC bioavailability were reduced below both 20 reservoirs compared to upstream reaches and OC was produced within the tailwaters. These combined effects of reservoirs and corresponding tailwater river reaches likely increased the impact on OC cycling compared to the presence of impoundments alone by magnifying the transformation of both POC and DOC.

Temporal OC dynamics
Hydrological seasonality drove variation in POC and DOC concentrations in the upper Green and Yampa Rivers (Fig. 2). The hydrological flushing hypothesis posits that terrestrial carbon within the watershed accrues during low flows and is flushed into streams and rivers during the initial infiltration of melt water during the onset of 5 snowmelt (Hornberger et al., 1994;Boyer et al., 1997). Therefore, our findings of peak OC concentrations preceding peak discharge were not surprising above Fontenelle reservoir and at the Yampa River sampling site. This pattern indicates the terrestrial supply of DOC is exhausted, resulting in hysteresis between DOC concentration and stream discharge (Hornberger et al., 1994;Finlay et al., 2006;Ågren et al., 2008). Peak concentrations of OC coincided with peak discharge below Fontenelle reservoir, which was likely driven by a combination of factors including dam operations and longer residence time of water in the reservoir relative to the river. Riverine DOC sources, and therefore bioavailability and composition, can be seasonally dependent. The initial flushing of terrestrial OC from a watershed during early 15 snowmelt can be more bioavailable than base flow DOC because shallow sub-surface runoff from the catchment can export stored terrestrial OC into aquatic ecosystems (Michaelson et al., 1998;Pacific et al., 2010;Pellerin et al., 2011). In contrast, DOC composition in semi-arid and arid rivers reflected autochthonous DOC during base flow conditions as opposed to high flows, due to the contribution of algal and mi-20 crobial exudates from increased primary production (Westerhoff and Anning, 2000). High-sustained flows during spring and summer 2011 in the Yampa and Green Rivers (Fig. 2) likely decreased the onset and magnitude of primary production (Uehlinger, 2000), which could account for our findings of stable, as opposed to increasing, DOC bioavailability after peak snowmelt. Furthermore, increased SUVA 254 and S R indicated 25 that base flow DOC likely comprised more aromatic, but smaller molecular weight carbon molecules than snowmelt DOC, likely due to microbial or photo-transformation of DOC (Helms et al., 2008;Kraus et al., 2011;Miller, 2012) (Weishaar et al., 2003;Goodman et al., 2011). Transformation of DOC, rather than production of labile DOC from algal-exudation supports our DOC bioavailability findings.

Longitudinal DOC dynamics
Not only total water storage, but also the type of reservoir may alter DOC dynamics and 5 longitudinal transport. The annual DOC loads (Mg yr −1 ) above Fontenelle reservoir and directly below the dam were similar (Table 3). In comparison, the annual DOC load increased from upstream to downstream of Flaming Gorge reservoir, regardless of the hydrograph. These annual estimated DOC loads for the Green River were similar to the upper Colorado River with comparable drainage areas (  (Table A1), but most of the storage was within Flaming Gorge reservoir. The difference in the relationship of DOC fluxes with watershed area between the upper Colorado River and Green River suggest reservoir type (i.e. many small vs. a few large reservoirs) and not just total water storage capacity of the basin may drive a decrease 20 in OC loads in this semi-arid watershed.
Although we do not have the appropriate data to adequately budget OC for either of the reservoirs, residence time likely drove, at least in part, the longitudinal DOC concentration and flux patterns we observed in relation to the reservoirs. Increased residence time due to impounding a river reduces water velocity, which allows POC to settle and 25 allows more time for the production and transformation of DOC (Mash et al., 2004;Kraus et al., 2011;Knoll et al., 2013). A similar shift in DOC concentration and timing of peak discharge occurred above and below natural lakes in snowmelt-dominated 6094 Introduction  ., 2011). The timing of reservoir filling and dam operations resulted in an arid reservoir (Westerhoff and Anning, 2000) and two temperate reservoirs (Knoll et al., 2013) to fluctuate between net source and net sink of DOC to downstream reaches. Also, seasonal shifts in reservoir primary production drove a reservoir in California to shift between a DOC source 5 and sink (Kraus et al., 2011). A combination of residence time and autochthonous production within reservoirs may lead to production (Parks and Baker, 1997;Kraus et al., 2011) or loss of DOC (Kraus et al., 2011;Miller et al., 2012;Knoll et al., 2013) to downstream ecosystems, likely driven by magnitude of hydrological variation such as high vs. low flow years (Knoll et al., 2013). 10 DOC composition differed from upstream to downstream of both Fontenelle and Flaming Gorge reservoirs. All SUVA 254 < 3 L mg C −1 m −1 , which indicates that DOC across our sampling sites was of low aromatic content (Weishaar et al., 2003), similar to values found in the Colorado River (Miller, 2012). Despite this low range of values, DOC composition below Flaming Gorge dam was less aromatic (as indicated by SUVA 254 ) 15 and reflected lower molecular weight OC (as indicated by S R ) compared to DOC composition upstream (Fig. 3). These small, but statistically significant, changes could be due to photodegradation (Brooks et al., 2007;Kraus et al., 2012;Cory et al., 2014) coupled with autochthonous production of DOC (Chin et al., 1994;Nguyen et al., 2002). A similar decreasing SUVA 254 pattern from upstream to downstream of reservoirs was 20 reported below Lake Powell and Lake Mead in the lower Colorado River basin (Miller, 2012). In addition, DOC bioavailability was reduced below Flaming Gorge dam compared to upstream of the reservoir (Fig. 4). This pattern along with our absorbance data indicates that DOC exported from Flaming Gorge reservoir was likely a combination of transformed and microbially produced DOC. In comparison, DOC composition did not 25 vary above and below Fontenelle reservoir based on SUVA 254 and S R metrics. But, similar to Flaming Gorge reservoir, bioavailability of DOC was significantly lower below Fontenelle dam compared to above the reservoir (Fig. 4). Reduced bioavailability below the dam indicates that even with no observed spectral changes (i.e., SUVA 254 and Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | S R ) above and below Fontenelle reservoir, DOC processing or transformation occurred within this reservoir ecosystem, but to a lesser extent than the larger Flaming Gorge reservoir. POC concentrations were 10-fold lower downstream of Fontenelle and Flaming Gorge dams when compared to upstream of both reservoirs (Fig. 2), and subsequently 5 resulted in a reduction of POC flux and annual load. This pattern is well established for large impoundments on rivers; impoundments allow for the settling of POC (Friedl and Wüest 2002;Downing et al., 2008;Tranvik et al., 2009). We did not measure within reservoir OC fate, but the fate of buried POC in reservoirs was likely a combination of preservation (Downing et al., 2008), mineralization to CO 2 (St. Louis et al., 2000;Knoll 10 et al., 2013), and transformation to DOC (Meyer et al., 1998).

Tailwater organic carbon transport
Despite the low POC concentration emanating from Fontenelle and Flaming Gorge dams, POC concentrations increased to 40-74 % of above reservoir concentrations within the short tailwater reaches. There were no perennial tributaries in either tailwa-15 ter; therefore the POC fluxes from the tailwater reaches were likely of autochthonous origin. Primary production likely drove this flux of POC from both tailwater reaches (Table 3). Tailwaters have high primary production (Webster et al., 1979;Davis et al., 2011) where algae and particulates are sloughed during discharge releases from the dam (Perry and Perry, 1991). The annual POC load from Fontenelle reservoir tailwater 20 was similar to the annual POC load into the reservoir, indicating that an equivalent amount of the POC load reduced from above to below the reservoir was generated within the 39.6 km tailwater. Similarly, the Flaming Gorge tailwater generated about half the amount of POC that entered the reservoir (Table 3). We estimated POC area-specific fluxes from the tailwaters by dividing the difference in POC annual load 25 (g yr −1 ) by reach area (m 2 ) and 365 (d −1 ). The POC daily flux from Fontenelle tailwater was 1.9 g C m −2 d −1 and from Flaming Gorge tailwater 1.3 g C m −2 d −1 . Although we did not measure primary production in Flaming Gorge tailwater, primary production in 6096 Introduction  (Hall et al., 2015). Also, these POC area-specific flux estimates were within the upper 50th percentile of gross primary production measurements from 72 streams showing that primary production can support this high OC flux (Bernot et al., 2010). The POC flux likely consisted of current primary production and organic matter from primary production accrued throughout 5 the year. Low discharge releases from Fontenelle and Flaming Gorge dams during the winter months combined, with the increased flows released from the dam during the onset of our sampling (Fig. 2), likely flushed the organic matter that had accrued within the tailwaters throughout the year (Parks and Baker, 1997;Brooks et al., 2007). Fontenelle and Flaming Gorge tailwaters were likely a source of autochthonously 10 produced DOC. The daily estimated flux of DOC from the tailwaters were 4 to 6-fold lower than POC fluxes, 0.3 g C m −2 d −1 for both Fontenelle and Flaming Gorge tailwaters. Autochthonous DOC fluxes were similar from the tailwater segment directly below Lake Powell on the Colorado River (0.3-2.1 g C m −2 d −1 ) and these fluxes were positively correlated with gross primary production (Ulseth, 2012). In addition, au-15 tochthonous DOC fluxes in the Grand Canyon reach of the Colorado River ranged from 0.09-0.39 g C m −2 d −1 (Ulseth, 2012). Although estimated DOC fluxes were lower than POC fluxes, they were within the lower 50th percentile of primary production rates across 72 streams in North America (Bernot et al., 2010) indicating that authochthonous derived DOC flux, similar to POC flux, was potentially a substantial pro-20 portion of primary production (Hotchkiss and Hall, 2014) from these tailwater ecosystems. The DOC flux from tailwater algae had a minimal effect on total DOC composition, given our absorbance data. However, bioavailability was higher at the Fontenelle tailwater compared to directly below the dam, suggesting freshly produced, labile DOC. Increased total DOC bioavailability from the tailwater likely produced an export of labile 25 DOC, potentially subsidizing the microbial food web in the downstream reaches.
The net effect of dams on the reduction of OC transport was essentially low (6-14 %), despite large changes in POC concentration and perhaps composition, as well as DOC composition. This finding affects how impoundments are viewed from an OC Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | cycling perspective. The effect of impounding rivers on OC fluxes is potentially underestimated because total concentration based fluxes do not represent transformation processes in river-reservoir-tailwater ecosystems. Given that impoundments have increased the capacity of rivers to transform or store DOC and POC (Kraus et al., 2011;Miller, 2012), regulation of rivers in the Western United States has changed OC cycling 5 in these ecosystems by altering the timing, magnitude, and composition of OC to downstream ecosystems (Miller, 2012;Stackpoole et al., 2014). The tailwater ecosystems contributed to the effect of reservoirs on OC transport in rivers by increasing the export of likely autochthonous OC downriver. Therefore, reservoirs regulate OC transport by reducing POC and altering the composition and bioavailability of DOC. The effect of impounding rivers on C cycling is larger than the reservoirs alone because of the additive impacts of tailwater reaches, which produce and then export a comparable amount of autochthonous OC than what is likely stored behind dams. To assess the effects in terms of regional carbon budgets, we need to consider not only reservoirs in regards to their capacity to transform terrestrial OC (Knoll et al., 2013), but also the additive 15 effects of their tailwater ecosystems. Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Nadon, M. J., Metcalfe, R. A., Williams, C. J., Somers, K. M., and Xenopoulos, M. A.: Assessing the effects of dams and waterpower facilities on riverine dissolved organic matter composition, Hydrobiologia, 744, 145-164, doi:10.1007/s10750-014-2069-0, 2014 Fragmentation and flow regulation of the world's large river systems, Science, 308, 405-408, 2005. 5 Nguyen, M. L., Baker, L. A., and Westerhoff, P.: DOC and DBP precursors in western US watersheds and reservoirs, J. Am. Water Works Ass., 94, 98-112, 2002. Pacific, V. J., Jensco, K. G., and McGlynn, B. L.: Variable flushing mechanisms and landscape structure control stream DOC export during snowmelt in a set of nested catchments, Biogeochemistry,99,[193][194][195][196][197][198][199][200][201][202][203][204][205][206][207][208][209][210][211]doi:10. Table B1. Mean decay rates (k, d −1 ) and mean percentage (%) loss of DOC over 28 day bioassay experiment. Mean percentage loss was calculated by dividing the difference in DOC concentration (mg L −1 ) from day 0 to day 28 by the initial DOC concentration at day 0 and then multiplying by 100. 95 % confidence intervals were calculated for k d −1 and % loss of DOC from 3 replicate bioassay experiments for each site and each date.  . Annual DOC load (Mg yr −1 ) plotted against drainage area (km 2 ) from the Green, Yampa, and upper reaches of the Colorado River (< 50 × 10 3 km 2 , data from Stackpoole et al., 2014;Miller, 2012). Linear regression line was fitted through data from the upper Colorado River where DOC load (Mg yr −1 ) = 1675.7 + 430.1 × km 2 , P = 0.0001, r 2 = 0.98)