- Email: [email protected]
Using stable isotopes of dissolved oxygen for the determination of gas exchange in the Grand River, Ontario, Canada
w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 7 8 1 e7 9 0
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Using stable isotopes of dissolved oxygen for the determination of gas exchange in the Grand River, Ontario, Canada Terra S. Jamieson a,*, Sherry L. Schiff a, William D. Taylor b a b
Department of Earth Sciences, University of Waterloo, 200 University Avenue West, Waterloo, ON, Canada N2L 3G1 Department of Biology, University of Waterloo, 200 University Avenue West, Waterloo, ON, Canada N2L 3G1
article info
abstract
Article history:
Gas exchange can be a key component of the dissolved oxygen (DO) mass balance in
Received 17 August 2012
aquatic ecosystems. Quantification of gas transfer rates is essential for the estimation of
Received in revised form
DO production and consumption rates, and determination of assimilation capacities of
21 October 2012
systems receiving organic inputs. Currently, the accurate determination of gas transfer
Accepted 1 November 2012
rate is a topic of debate in DO modeling, and there are a wide variety of approaches that
Available online 21 November 2012
have been proposed in the literature. The current study investigates the use of repeated measures of stable isotopes of O2 and DO and a dynamic dual mass-balance model to
Keywords:
quantify gas transfer coefficients (k) in the Grand River, Ontario, Canada. Measurements
Dissolved oxygen
were conducted over a longitudinal gradient that reflected watershed changes from agri-
Gas exchange
cultural to urban. Values of k in the Grand River ranged from 3.6 to 8.6 day1, over
Impacted river
discharges ranging from 5.6 to 22.4 m3 s1, with one high-flow event of 73.1 m3 s1. The k
Stable isotopes
values were relatively constant over the range of discharge conditions studied. The range in discharge observed in this study is generally representative of non-storm and summer low-flow events; a greater range in k might be observed under a wider range of hydrologic conditions. Overall, k values obtained with the dual model for the Grand River were found to be lower than predicted by the traditional approaches evaluated, highlighting the importance of determining site-specific values of k. The dual mass balance approach provides a more constrained estimate of k than using DO only, and is applicable to large rivers where other approaches would be difficult to use. The addition of an isotopic mass balance provides for a corroboration of the input parameter estimates between the two balances. Constraining the range of potential input values allows for a direct estimate of k in large, productive systems where other k-estimation approaches may be uncertain or logistically infeasible. ª 2012 Elsevier Ltd. All rights reserved.
1.
Introduction
Oxygen is considered to be a key parameter for the evaluation of water quality in aquatic ecosystems. In turbulent streams,
gas exchange can be an important component of the dissolved oxygen (DO) balance, mediating the effects of DO production and consumption due to chemical and biological processes by driving concentrations toward atmospheric equilibrium. Two
* Corresponding author. Present address: Agriculture and Agri-Food Canada, 1791 Barrington Street Suite 405, P.O. Box 248, Halifax, NS, Canada B3J 2N7. Tel.: þ1 902 407 7663. E-mail addresses: [email protected] (T.S. Jamieson), [email protected] (S.L. Schiff), [email protected] (W.D. Taylor). 0043-1354/$ e see front matter ª 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.watres.2012.11.001
782
w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 7 8 1 e7 9 0
theories are widely used to describe gas transfer in natural waters: the two-film model, mainly applied to standing waters, and the surface-renewal model, commonly used in flowing waters (Chapra, 1997). The surface-renewal theory, assumes that in an agitated liquid, as in the Grand River, turbulence extends to the surface and therefore no laminar boundary layer exists (Danckwerts, 1951). That is, it assumes that the surface is continually being replaced with fresh liquid from a mixed vessel. The gas transfer velocity for O2 entering or leaving the water column is then dependant on the liquid surface renewal rate, as well as the diffusion rate of the gas in water (Chapra, 1997). For surface renewal models, the mass flux of O2 in rivers is generally assumed to be a product of the deviation of DO from saturation in the water column, multiplied by a mass-transfer coefficient: GE ¼ kðDOsat DOÞ
(1)
Where GE is gas exchange, the mass flux of O2 in DO (concentration time1), k is the gas transfer coefficient, and DOsat is the concentration of DO at atmospheric equilibrium. For rivers, the k value is typically treated as a first order rate constant expressed on a time1 basis, where the water column is assumed to be well-mixed with a homogeneous DO concentration. Determining k is essential for modeling DO in aquatic environments; i.e., estimating rates of production and consumption and determining the assimilation capacity of systems receiving organic inputs. Currently, the accurate determination of k is a topic of debate, and there are a wide variety of approaches for determining gas exchange rate including the application of tracers, regression modeling of night-time DO, and using hydrology-dependent equations. Formulas that estimate gas exchange coefficients have been developed and calibrated using gaseous tracers, e.g., methyl chloride, (Wilcock, 1988), ethylene (Wanninkhof et al., 1990), propane, SF6, (Clark et al., 1992) and 85Kr (Tsivoglou and Neal, 1976). The reaeration of deoxygenated water (Churchill et al., 1962; Owens et al., 1964) has also been used. Radioactive tracer techniques tend to yield the most accurate k values (Bowie et al., 1985) but have been limited in application due to the potentially hazardous effects of radiation (Bicudo and James, 1989). The use of inert gas tracers is also subject to controversy regarding the accuracy and reproducibility of the data (Bowie et al., 1985). Other approaches use DO changes; Odum and Hoskin (1958) developed a method for calculating k, where the DO deficit is regressed against the rate of change in DO after sunset, the slope of which is k. This method was later recommended by Owens (1974). Chapra and Toro (1991) have also developed an analytical solution for determining k from a single diel DO monitoring station, showing that k is a function of the time lag between solar noon and DO maximum, and day length. One of the most commonly used models, the O’ConnoreDobbins formula (O’Connor and Dobbins, 1956) was developed from the surface renewal theory and the diffusivity of oxygen in natural waters. There are currently more than 34 available formulas that have been proposed to predict k in rivers (McCutcheon, 1989) which vary in their uncertainty and applicability. Comprehensive reviews and critiques of the many formulas have
been provided by Bowie et al. (1985) and McCutcheon (1989). These formulas are usually parameterized in terms of water velocity and mean depth, and sometimes bed slope, and thus the applicability of a chosen approach depends on the similarity between the river under examination and the river used to calibrate the gas transfer relationship (Bennett and Rathburn, 1972). A number of studies have used both DO and stable isotopes of O2 to investigate community metabolism in aquatic systems (Quay et al., 1993, 1995; Parker et al., 2005; Tobias et al., 2007; Venkiteswaran et al., 2007; Poulson and Sullivan, 2010; Wassenaar et al., 2010). The first of these studies assumed steady state for the hydrologic conditions and gas exchange rates (Quay et al., 1995; Wang and Veizer, 2000), but later work quantified river metabolism rates under dynamic conditions (Parker et al., 2005; Tobias et al., 2007; Venkiteswaran et al., 2007). Examined approaches included using DO regression to obtain k (Parker et al., 2005), conducting tracer addition experiments (Tobias et al., 2007), or resolving DO and isotopic mass balances iteratively (Venkiteswaran et al., 2007; Poulson and Sullivan, 2010; Wassenaar et al., 2010). Previous studies have shown that using isotopic techniques for quantifying community metabolism is promising, but requires further evaluation in a variety of aquatic ecosystems, and needs to be compared to traditional techniques. It is of particular interest to investigate the applicability of using the stable isotopes of O2 as a tool to determine k in large, impacted rivers, where tracer techniques may be logistically difficult to use, and the determination of appropriate gas transfer analytical equations is uncertain. The objectives of this study were to i) use stable isotopes of O2, in conjunction with measurement of DO concentration, to quantify k in a large, impacted river, ii) compare k estimates from the isotope technique to other literature approaches and, iii) assess the spatial and temporal variability of k in the Grand River.
2.
Study site: The Grand River watershed
The Grand River watershed, the largest in southern Ontario, covers an area of 6965 km2, and the main channel measures about 290 km in length (Cooke, 2006). The watershed drains agricultural land (76%) and wooded areas (17%) but the central 5% of the watershed is largely urbanized. Most of the basin’s population, over 900,000 people, reside in the cities of Kitchener, Waterloo, Cambridge, Guelph and Brantford (Fig. 1). Currently, the primary indicators of water quality in the Grand River Watershed used by the Grand River Conservation Authority (GRCA) are DO and nutrient concentrations (Cooke, 2006). Nutrient-rich conditions and depletion of DO have been identified as issues of concern in portions of the basin’s watercourses. The river receives organic and nutrient inputs from 26 sewage treatment plants, on-site wastewater treatment systems, and non-point agricultural and urban sources. It also serves as a source of drinking water and supports a significant recreational fishery. The GRCA relies on a dynamic mass balance simulation model (Grand River Simulation Model, GRSM) to predict DO concentrations in the Grand River under various management regimes (e.g., flow
w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 7 8 1 e7 9 0
Fig. 1 e Map of the Grand River Watershed. The West Montrose, Bridgeport and Blair sampling locations are indicated (image obtained and adapted from www. grandriver.ca).
augmentation, inputs from sewage treatment plants). DO predictions derived from the GRSM depend on O2 demand, photosynthesis, and gas exchange (GRCA, 1982). Three sampling locations (Fig. 1) were chosen along the central portion of the Grand River, in 5th to 6th order reaches along the main channel: West Montrose (WM), Bridgeport (BRPT) and Blair (BLR). West Montrose is located upstream of the cities of Kitchener and Waterloo (population ca. 326,500) in a predominately agricultural landscape. Bridgeport, located downstream from WM, is subject to input from the relatively nutrient-laden Canagagigue Creek and the Conestogo River. Blair is located downstream of two major sewage treatment plant outfalls located in Kitchener and Waterloo, and receives urban runoff from the upstream communities. These sites were intended to characterize k at different locations adjacent to GRCA real-time water quality monitoring stations, along a longitudinal gradient of increasing cumulative impact.
3.
Sampling and analytical methods
Sampling was conducted throughout May to August, after sunset to remove the effects of photosynthesis on the DO. For
783
each sampling event, samples were collected every 2e4 h between sunset and sunrise. Samples were manually collected from within the main current of the channel at approximately mid-depth and as close to mid-channel as wade-able conditions allowed. It was assumed that the water column at these locations were well-mixed and representative of homogeneous conditions. In 2003, monitoring was mainly focused on WM, but all three locations were sampled during 2004 and 2005. Water samples were collected for DO and d18OeO2, and temperature was recorded. Dissolved O2 samples were collected in duplicate and titrated within 24 h of collection using the Winkler method (APHA, 1995). The precision of the Winkler analytical method is approximately 0.2 mg L1. Dissolved O2 concentrations were also measured at the time of sample collection using a YSI 556 MPS hand-held meter to serve as supplemental data to the collected DO samples. The d18OeO2 samples were collected, once per location per sample time, in 160 ml Wheaton serum bottles containing 0.3 g of sodium azide as a bactericide. Prior to sampling, the serum bottles were crimp sealed with butyl blue stoppers and then evacuated to <0.001 atm. The evacuated serum bottles were submerged and filled by piercing the septum via an 18 gauge needle. After filling, the needle was removed while the bottle was still submerged to prevent atmospheric 18O from entering. The d18OeO2 samples were prepared for analysis by removing 5 ml of water, inserting 5 ml of He, and shaking to equilibrate the headspace. Headspace gas was analyzed using a continuous flow gas chromatograph isotope ratio mass spectrometer, following Wassenaar and Koehler (1999), with an analytical error of approximately 0.3&.
4.
Model description
To address the variable nature of DO in the Grand River, a dynamic mass balance model similar to the PoRGy model developed by Venkiteswaran et al. (2007) was used. The key potential benefit of using 18OeO2 is that the isotopic signature provides a second signal in addition to DO, that is directly related to the processes consuming and producing DO. There are a number of properties that enable 18O of DO (18OeO2) to be used to quantify metabolism and gas exchange rates in rivers. Atmospheric O2 is isotopically enriched, with a d18O of þ23.5 0.3& relative to standard mean ocean water (Kroopnick and Craig, 1972). During gas dissolution, there is an equilibrium fractionation of approximately 0.7& at 20 C (Kroopnick and Craig, 1972; Wassenaar and Koehler, 1999) resulting in a d18OeO2 of þ24.2&. Photosynthesis produces O2 with the isotopic signature of water as there is no fractionation in the photolysis of water (Guy et al., 1993). However, as the samples were collected after dusk, the effects of photosynthesis were not included in the model presented here. Another key aspect is that respiring organisms consume the isotopically lighter 16O:16O molecules 10e20& more rapidly than 18O:16O molecules (Kiddon et al., 1993) referred to as respiratory fractionation. Therefore, respiration increases the 18 O fraction of the residual DO. Two O2 mass balances (DO and isotopic) were resolved based on the following equations:
784
w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 7 8 1 e7 9 0
dDO ¼ kðDOsat DOÞ CR dt
(2)
CRT ¼ CR20 1:047ðT20Þ
dDO 18:16 O 18 ¼ OGE þ 18 OExch CRð18:16 O ar Þ dt
(3)
where T is temperature, measured in C, k20 and CR20 are k and CR at 20C.
where CR is the community respiration rate (mg L1 h1), k is the gas exchange coefficient (h1), DOsat is the DO concentration under saturated conditions, 18:16O refers to the 18 O:16O of DO, and ar is the isotopic fractionation factor associated with CR. DOsat was calculated based on temperature (Weiss, 1970), and corrected for the altitude (Radtke et al., 1998) at each sampling location. 18OExch describes the mass flux of 18O and 16O due to exchange with the atmosphere, as part of the mass transfer of DO. In undersaturated DO conditions, 18O was assumed to invade the water column more slowly than 16O, due to fractionation associated with solubility and airewater transfer. Similarly, in supersaturated DO conditions, 18O was assumed to preferentially evade the water column, leaving behind residual 16O due the greater solubility of the lighter isotope. Therefore: If (DOsatDO) > 0 then: 18
OGE ¼ k ak ðDOsat DOÞ ð18:16 Oa as Þ
(4)
Else: 18
OGE ¼ k ak ðDOsat DOÞ ð18:16 O=as Þ 18
(5)
16
where 18:16Oa is the ratio of O to O in air, as is the fractionation effect of the dissolution of O2 in water, and ak is the fractionation factor associated with gas transfer. The solubility fractionation of 18OeO2 (as ¼ 1.0007) is due to the slightly different solubility between 18O and 16O, with negligible temperature effects (Benson and Krause, 1984). There is a kinetic fractionation effect associated with airewater transfer of 18OeO2 (ak ¼ 0.9972), as the 18O molecule is slightly larger than 16O (Knox et al., 1992). The temperature effects associated with ak were also assumed to be negligible over the temperatures encountered with respect to the mass balance. According to the equations, if DO is at equilibrium (DOsatDO ¼ 0), 18OGE would be nil. The isotopes of O2 were assumed to continue to diffuse across the airewater interface if d18OeO2 is at disequilibrium even if DO was at saturation. In order to allow for 18O and 16O to exchange across the airewater interface towards equilibrium independently of the bulk exchange of DO, the following equation was used: 18
OExch ¼ k ak DO ½ð18:16 Oair as Þ 18:16 O
(6)
The isotopic signature of atmospheric O2 is assumed to be constant at 23.5& (Kroopnick and Craig, 1972). Isotope values were modeled as ratios but will be reported in units of & in dnotation with respect to Vienna Standard Mean Ocean Water (VSMOW). The effects of groundwater input were assumed to be negligible at the sample locations with respect to the large fluctuations in DO. Both k and CR values were varied with temperature within the model using the following relationships (McCutcheon, 1989; Chapra, 1997): kT ¼ k20 1:024ðT20Þ
(7)
5.
(8)
Model calibration
Modeling was performed using STELLA 7.0.2, with a 15 min time-step using the mass balance equations noted above. The unknowns in the model were CR, k, and ar. Due to the dynamic nature of the system, and because there are three unknowns and only two mass balances, the model could not be solved directly. Therefore, the model was calibrated iteratively to optimize the goodness of fit between modeled and measured data. Model outputs were evaluated by both visually examining the fit of the model to the data, and using root mean square error (RMS) to statistically evaluate goodness of fit. The time zero DO and d18OeO2 points for each modeling event were not included in the RMS calculations, as these points were used to initialize the model at time zero. Iterative simulations were first performed and compared to DO concentration data; model outputs that resulted in RMS values of approximately 0.2 mg L1 or less (similar to the analytical error associated with Winkler titration), supported by visual examination, were used to constrain the potential range in input parameter values. The identified range of k20 and CR20 values were then modeled iteratively with a range of ar, and compared to the observed d18OeO2. Estimates of k20, CR20, and ar that provided an optimal model fit to observed data for both the DO and d18OeO2 were then chosen for rate estimations.
6. Comparison to other gas transfer estimates There are many predictive equations for k, but the most widely used equations (Chapra, 1997) include those by O’Connor and Dobbins (1956), Churchill et al. (1962), and Owens et al. (1964), which are parameterized in terms of mean depth (D) and velocity (V) according to: k20 ¼ z
Vx Dy
(9)
where z, x, and y are the empirically calibrated constants (Table 1). To put the k estimated by the isotope technique in context with estimates from these predictive equations, hydrologic conditions were quantified for each sampling event. Discharge rates were obtained from the GRCA for each of the three sampling locations. Hydrologic data (i.e., discharge, mean depth, mean velocity) for the BLR sampling location were also obtained from stream gauging activities conducted by the GRCA. At the WM location, cross-sectional surveys of the channel were performed just upstream of the sampling location to determine the relationship of mean depth with discharge, and water surface slope was also measured. Mean velocity (V) and depth (D) at WM as a function of the hydraulic geometry and
w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 7 8 1 e7 9 0
785
Table 1 e Selected literature equation coefficientsa for calculating k20 in rivers. Equation
O’Connor and Dobbins (1956) Churchill et al. (1962) Owens et al. (1964) Bennett and Rathburn (1972) a k20 ¼ z
Coefficients z
x
y
3.9 5.026 5.32 5.58
0.5 1 0.67 0.607
1.5 1.67 1.85 1.689
Vx . Dy
discharge at the time of sampling were computed using Manning’s expression (French, 1985). Cross-sectional surveys were not conducted at BRPT; the mean of the hydrologic relationships associated with WM and BLR were used to calculate mean depth and velocity changes based on discharge for the BRPT location. The hydrology data for each sampling location were used to calculate k from other published equations based on mean depth and velocity (Table 1). Regressions using night DO concentration data were also performed to obtain estimates of k (Odum and Hoskin, 1958). The rate of change of DO concentration was regressed against DO deficit (relative to saturation) to obtain the slope of the regression which is an estimate of k.
7.
Results and discussion
7.1.
Gas transfer coefficients in the Grand River
As expected, DO concentrations declined at night due to community respiration, with a concurrent enrichment of d18OeO2 (Figs. 2 and 3, and 4). The associated k values predicted from the isotope technique for the Grand River ranged from 3.6 to 8.6 d1 (Table 2). Overall, the non-steady state model appears to satisfactorily reproduce the changes in both DO and d18OeO2. RMS errors for the DO model output were near, or less than, the analytical error of 0.2 mg L1 associated with Winkler titrations. There was greater error associated with d18OeO2 model output, with a particularly poor model fit to d18OeO2 for BLR July 29e30, 2005. During this sampling event, DO concentrations were very low, falling to less than 2 mg L1. Under the extremely low DO conditions, the model may not have been able to adequately represent CR or the fractionation processes as constant rates are assumed for both. There also may have been additional analytical error associated with d18OeO2 due to the low levels of DO in the water sample. There appear to be greater RMS values associated with the BRPT and BLR as compared to WM. The high error in d18OeO2 model fit at BRPT on August 30e31, 2004, appears to be attributable to one anomalous data point, potentially due to inadequate mixing at the sampling site or analytical error. Community respiration rates ranged from approximately 10e47 mg O2 L1 d1, with associated ar values ranging from 0.971 to 0.996. Most values were between 0.980 and 0.989, which are within range of other literature values (Kiddon et al., 1993; Quay et al., 1995; Tobias et al., 2007; Venkiteswaran et al., 2007; Poulson and Sullivan, 2010), as discussed elsewhere (Jamieson, 2010).
Fig. 2 e Observed and model-predicted DO and d18OeO2 in the Grand River at West Montrose.
The model uses one rate of CR20 and k20, over the modeled time period, and short-term variability in either of these terms beyond what is attributable to the temperature correction could potentially lead to deviations from the modeled DO and 18 OeO2 output. This study was undertaken as a single-station approach, common to other DO modeling studies (Servais et al., 1984; Edwards and Meyer, 1987; Livingstone, 1991; Pearson and Crossland, 1996; Wilcock et al., 1998; Williams et al., 2000). As a result, the findings obtained for a particular location are an integrated measure of processes occurring in the watershed upstream from that point. Gas transfer coefficients at BRPT were roughly half of the estimates for WM and BLR, which exhibited similar k. Gas exchange in rivers is mainly influenced by physical attributes of the channel. At WM and BLR, the river upstream consists of main channel, with no confluences with other major tributaries or reservoirs located immediately upstream, which may have led to similar k values noted among these locations. However, BRPT is located downstream from the Conestoga river confluence, one of the major tributaries in the watershed, as well as a floodplain area that formerly served as a gravel mining site. Gas transfer upstream of BRPT may be influenced by Conestoga river input, or other channel
786
w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 7 8 1 e7 9 0
Fig. 3 e Observed and model-predicted DO and d18OeO2 in the Grand River at Bridgeport.
morphological conditions which could exhibit lower k values such as deep zones. The upstream gravel floodplain could potentially be an area of groundwater discharge of a magnitude great enough to affect the DO budget, which was not accounted for in the model.
The fit of the model to observed data was generally poorest at BLR, as compared to WM and BRPT. Error in associated model fits may be due to short-term temporal variation in upstream processes that are not accounted for in the model, as constant rates of CR20 were assumed. BLR is located downstream of two major sewage treatment plants that would have a daily pattern of discharge which could cause small temporal variability in model parameters. Varying inputs of DO consuming substances (e.g., organic matter, ammonium), or even aerated effluent, could cause short-term variability in observed DO and d18OeO2. Community respiration rates tend to be greater at BLR than upstream, with lower DO minima at night. As DO concentration declines at BLR, an associated decline in the rate of CR and changes in isotopic fractionation could also be occurring as the system approaches hypoxic conditions. The range in k observed in the Grand River (3.6e8.6 d1) is well within ranges measured in other similarly sized rivers. Wilcock (1988) found k values of 6.34e8.54 d1 obtained from gas tracer experiments for a river with mean annual discharge and velocity of 30 m3 s1, and 0.8 m s1. Using the isotopic approach similar to the present study, Venkiteswaran et al. (2007) reported a k value of 3.96 d1 in the South Saskatchewan River, which was characterized as having a mean depth of 0.65 m and mean velocity of 0.21 m s1 O’Connor and Dobbins (1956) estimated re-aeration k values ranging from 0.03 to 4.8 d1 in rivers with mean depths of 0.27e11 m, where shallower rivers exhibited greater k values. Gas transfer coefficients in the Grand River were relatively constant over changes in mean depth and velocity during sampling events, with virtually no relationship between discharge and k (R ¼ 0.49). The development of a mean depth and velocity dependant equation specific to Grand River conditions would be inappropriate based on the findings in the current study as there was not a strong relationship between k and hydrologic conditions. Other studies that have derived an empirical predictive k equation have analyzed data sets covering k values that range by two to three orders of magnitude (Bowie et al., 1985). The lack of variability k in the Grand River is likely due to the shape of the river channel, which tends to be wide and shallow at all three locations. As discharge rates increase, both mean depth and velocity increased at similar rates; increases in depth and velocity would effectively cancel one another, since k tends to be proportional to velocity divided by depth (Melching and Flores, 1999). Genereux and Hemond (1992) measured k in a first order stream using a petroleum gaseous tracer technique, and found that measured k was relatively constant with increases in flow. Melching and Flores (1999) found no relationship between k and discharge in 493 reaches in 166 streams across the U.S that exhibited discharges from <0.01 to >100 m3 s1 Wassenaar et al. (2010) also noted that k values varied little seasonally and longitudinally in the South Saskatchewan River despite variations in discharge of approximately 100e500 m3 s1.
7.2. Fig. 4 e Observed and model-predicted DO and d18OeO2 in the Grand River at the Blair.
Comparison to other approaches
Gas transfer coefficient estimates obtained from the isotope modeling approach were found to be lower than those
787
w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 7 8 1 e7 9 0
Table 2 e Fitted model input parameter results obtained from DO concentration and d18OeO2 night-time data modeling. Sampling event WM
BRPT
BLR
May 28/03 Jun 24/03 Jul 28/03 Aug 13/03 Jul 27/05 Jul 9/03 May 11/04 Jun 7/04 Aug 30/04 Jun 7/04 Aug 30/04 Jul 29/05
Model ka (d1)
Model CRa (mg L1 d1)
Model ar
RMS DO (mg L1)
RMS d18OeO2 (&)
7.4 6.0 7.0 6.7 5.3 3.6 3.8 3.6 3.6 8.6 7.2 6.0
22.8 21.4 23.0 20.6 23.8 10.1 13.8 13.7 11.5 43.2 38.2 46.8
0.988 0.987 0.986 0.983 0.990 0.986 0.980 0.984 0.971 0.989 0.989 0.996
0.03 0.12 0.04 0.04 0.16 0.26 0.18 0.35 0.05 0.09 0.09 0.19
0.10 0.73 0.72 0.54 0.27 0.21 0.77 0.91 3.21 0.53 0.29 2.04
a Corrected to 20 C.
predicted by the other approaches (Table 3 and Fig. 5). The isotopic model-derived k was, on average, 5.2 d1 less than k calculated from equations. The night-time regression technique (Odum and Hoskin, 1958) was in closest agreement with isotopic model-derived k. Discrepancies between the k found by regression and the isotopic model are likely due to the fact that the dual model also fits the change in d18OeO2, as well as error that would occur in averaging the temperature and DO deficit over the intervals between samples. Water temperature in the Grand River was not constant during the sampling events, resulting in varying DOsat conditions and CR rates during the regression time periods. The temperature declined by between 0.8 and 4.7 C during a given night-time sampling event, with an average decline of 2.7 C. A decline in 2.7 C corresponds to a change in DOsat of approximately 0.5 mg L1, and a change in CR of about 1.6 mg O2 L1 d1 (given a mean CR rate of 24 mg O2 L1 d1). This variation may be enough to cause a change in k estimation using a regression technique, which assumes constant DOsat and CR.
In theory, night-time regression should predict an accurate k, based on the DO deficit and dDO/dt under steady state CR, temperature, and k conditions. To test varying temperature effects on k estimation using the regression technique, DO output data were generated for 15 min time steps using STELLA 7.0.2, assuming the model conditions used for the BLR August 30, 2004 and WM August 13, 2003 sampling events. The DO regression technique was then used to calculate k from the theoretical curve (reported in Jamieson, 2010). Changes in temperature over the regression time period appear to cause a consistent over-prediction in k using the night-time regression versus the isotopic model. In the BLR and WM simulation trials tested, the over-predictions were 1.6 and 1.1 d1 for respective changes in temperature of 1.2e3.8 C, which are similar to the average over-prediction of night-time regressed k compared to isotope model derived k obtained from the sampling events (Table 3). This suggests that the k values estimated from the isotope modeling technique are likely representative of in situ conditions and that the
Table 3 e Comparison of model derived k to selected literature approaches. D (m) V Model ka O’Connor Churchill et al. Owens et al. Bennett and DO Regressiona Sampling Q (m s1) (d1) and Dobbins (1962) (1964) Rathburn (1972) (Odum and Hoskin, event (m3 s1) (1956) 1958) WM May 28/03 Jun 24/03 Jul 28/03 Aug 13/03 Jul 27/05
6.8 5.6 6.0 12.7 5.6
0.42 0.37 0.39 0.58 0.37
0.50 0.47 0.48 0.63 0.47
7.4 6.0 7.0 6.7 5.3
10.1 11.9 11.1 7.0 11.9
10.7 12.4 11.6 7.9 12.4
16.6 20.2 18.6 10.7 20.2
15.9 18.9 17.5 10.6 18.9
9.6 12.1 8.1 7.7 8.0
BRPT Jul 9/03 May 11/04 Jun 7/04 Aug 30/04
8.2 73.1 11.7 18.7
0.40 1.07 0.47 0.58
0.41 1.01 0.48 0.58
3.6 3.8 3.6 3.6
9.9 3.5 8.4 6.7
9.5 4.5 8.5 7.2
15.9 4.7 13.2 10.1
15.3 5.0 12.8 10.1
5.0 2.9 9.5 2.5
BLR Jun 7/04 Aug 30/04 Jul 29/05
14.9 22.4 11.0
0.41 0.46 0.37
0.39 0.47 0.33
8.6 7.2 6.0
9.3 8.6 10.0
8.5 8.6 8.7
14.7 13.5 15.9
14.2 13.1 15.3
9.0 8.1 6.0
a k corrected to 20 C.
788
w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 7 8 1 e7 9 0
Fig. 5 e Plot of k values as predicted by other selected published approaches versus modeled k values in the Grand River.
deviation from night-time regression estimates lies in: 1) the paucity of data points collected to perform night-time regression analyses and 2) the varying temperature. In addition, Odum and Hoskin (1958) used hourly DO concentration data to obtain estimates of k, CR, and GPP. Data of this resolution were only available for the WM August 15, 2003 sampling event. Of the empirical equations, the closest k estimates were from the O’Connor and Dobbins (1956) equation, followed by Bennett and Rathburn (1972), and then Owens et al. (1964). O’Connor and Dobbins (1956) developed a theoretical equation, based on fundamental turbulence parameters, and tested it over a wide range of depth and velocity conditions. As a result, the equation tends to have the widest applicability for a range of flow conditions, although it is most appropriate for moderate to deep streams with moderate to low velocities (Chapra, 1997). Bennett and Rathburn (1972) also found that O’Connor and Dobbins (1956) was the formula that best fit the entire range of existing gas transfer data sets that they reviewed, and the Churchill et al. (1962) formula provided the best fit to natural stream data. The standard error of estimate was reported to range from 44 to 61% when comparing equation estimates developed by Melching and Flores (1999) to direct k results from tracer-gas techniques. Rathburn (1977) reported standard errors of estimate ranging from 3.0 to 7.2 d1 when comparing results using the equations addressed in the current study (Table 3) to tracer derived estimates, using a composite of results of five US rivers. Rathburn (1977) found that Churchill et al. (1962) had the lowest error, followed by O’Connor and Dobbins (1956), Bennett and Rathburn (1972), and Owens et al. (1964). According to Rathburn (1977), k values calculated from empirical equations tend to be greater than values obtained by tracer techniques. The discrepancy was partially attributed to possible effects of wind or substances in a stream (e.g., detergents, pollutants from sewage treatment plants). The Grand River receives nutrient loading from both point and non-point sources along the main channel at both locations studied, has high macrophyte biomass, possibly
resulting in depressed k coefficients. Parkhurst and Pomeroy (1972) noted that the presence of impurities results in a more stable surface film, thus reducing turbulence at the surface, leading to a decrease in k as compared to pure water. The authors cited research conducted by Kehr (1938) which showed that sewage additions as low as 0.5 % to tap water reduced k by 10 %. In the current study, the calculation of k in the Grand River using the empirical equations were dependent on the assumed hydrologic conditions associated with the sampling location of interest. Characteristics of the sampling locations chosen were accessible, shallow depths for ease of monitoring and sample collection. Upstream differences in hydrologic characteristics may lead to k rates that are not reflective of the conditions observed at the sampling location due to time of travel, leading to discrepancy between k obtained by the isotopic approach and that from the literature equations examined. For instance, in the literature equations noted, deeper stream channels or slower moving waters would lead to lower k estimates, more in line with that obtained by the isotopic method or DO regression method. These findings highlight the importance of determining site-specific values of k as the literature derived estimates varied by three-fold. The potential for over-estimation of k in large, impacted rivers should be taken into consideration when making watershed management decisions, as it could lead to error in respiration estimates, and under-prediction of DO deficits. Gas transfer coefficient values were relatively constant at each location under hydrologic conditions that were mainly representative of non-storm and summer lowflow periods. A greater range in k might be observed under a wider range of hydrologic conditions. However, from a water quality management perspective, low-flow conditions are the times of highest concern where gas transfer would be a key factor in mitigating DO deficits due to O2 demand. Groundwater influences in the Grand River were also assumed to be negligible in the DO mass balance given the high level of daily DO concentration amplitude observed; this assumption may not be appropriate for some locations. Further investigation into the importance of groundwatersurface water interactions as a model addition may be warranted in areas where groundwater discharge greatly influences the DO regime. Additional work to assess the robustness of the isotope approach under varying hydrologic regimes and locations is also needed, along with comparison to direct measurement techniques (e.g., tracers) to assess the accuracy of the technique in greater detail. However, the isotope modeling technique appears to offer a promising advantage in simulating changes in DO regimes as compared to the available literature. The dual mass balance provided a constraint on the potential range of input parameters to optimize fit to observed data, leading to a more direct estimation of k in large, productive systems.
Acknowledgments Funding for this research was supported by the Natural Sciences and Engineering Research Council, and the Ontario Centres of Excellence. We thank Richard Elgood for laboratory
w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 7 8 1 e7 9 0
research support and would like to acknowledge Dr. Len Wassenaar and Dr. Ramon Aravena for their critical input during the research process. We are also grateful for the help provided by field and laboratory assistants throughout this study, including: Stuart Elson, Brianna Kelly, Kevin Maurice, Paul Malone, Matt Mittler, Ashley Smyth, David Snider, Simon Thuss, and Andrea Wojtyniak.
references
American Public Health Association (APHA), 1995. Standard Methods for the Examination of Water and Wastewater, nineteenth ed. American Public Health Association, Inc., New York. Bennett, J.P., Rathburn, R.E., 1972. Reaeration in Open-channel Flow. U.S. Geological Survey. Professional Paper 737. Benson, B.B., Krause, D.J., 1984. The concentration and isotopic fractionation of oxygen dissolved in fresh water and seawater in equilibrium with the atmosphere. Limnology and Oceanography 29, 620e632. Bicudo, J.R., James, A., 1989. Measurement of reaeration in streams: comparison of techniques. Journal of Environmental Engineering 115, 992e1010. Bowie, G.L., Mills, W.B., Porcella, D.B., Campbell, C.L., Pagenkopf, J.R., Rupp, G.L., Johnson, K.M., Chan, P.W.H., Gherini, S.A., Chamberlain, C.E., 1985. Rates, Constants, and Kinetic Formulations in Surface Water Modeling. U.S. Environmental Protection Agency, ORD, Athens, GA, ERL. EPA/600/3e85/040. Chapra, S.C., 1997. Surface Water-quality Modeling. McGraw-Hill Companies, Inc, New York, NY, 844 p. Chapra, S.C., Di Toro, D.M., 1991. Delta method for estimating primary production, respiration, and reaeration in streams. Journal of Environmental Engineering, American Society of Civil Engineering 117, 640e655. Churchill, M.A., Elmore, H.L., Buckingham, R.A., 1962. The prediction of stream reaeration rates. Journal of the Sanitary Engineering Division, American Society of Civil Engineering SA 4, 1. Proc. Paper 3199. Clark, J.F., Simpson, H.J., Smethie Jr., W.M., Toles, C., 1992. Gas exchange in a contaminated estuary inferred from chlorofluorocarbons. Geophysical Research Letters 19, 1133e1136. Cooke, S., 2006. Water Quality in the Grand River: A Summary of Current Conditions (2000-2004) and Long Term Trends. Grand River Conservation Authority, Cambridge, ON, 88 p. http:// www.grandriver.ca/Water/2006_WaterQuality_complete.pdf. Danckwerts, P.V., 1951. Significance of liquid-film coefficients in gas absorption. Industrial and Engineering Chemistry 43, 1460e1467. Edwards, R.T., Meyer, J.L., 1987. Metabolism of sub-tropical low gradient blackwater river. Freshwater Biology 17, 251e263. French, R.H., 1985. Open-channel Hydraulics. McGraw-Hill, USA, 739 p. Genereux, D.P., Hemond, H.F., 1992. Determination of gas exchange rate constants for a small stream on Walker Branch Watershed, Tennessee. Water Resources Research 28, 2365e2374. Grand River Conservation Authority (GRCA), 1982. Grand River Basin Management Study. Grand River Implementation Committee, Grand River Conservation Authority, Cambridge, ON, 8 p. http://www.grandriver.ca/Water/2007_WaterQuality_ boardreport.pdf. Guy, R.D., Fogel, M.L., Berry, J.A., 1993. Photosynthetic fractionation of the stable isotopes of oxygen and carbon. Plant Physiology 101, 37e47. Jamieson, T.S. 2010. Quantification of Oxygen Dynamics in the Grand River Using a Stable Isotope Approach. Ph.D. thesis, University of Waterloo, ON, Canada. 198 p.
789
Kehr, R.W., 1938. Measures of natural oxidation in polluted streams. IV. Effect of sewage on atmospheric reaeration rates under stream flow conditions. Sewage Works Journal 10, 228e240. Kiddon, J., Bender, M.L., Orchardo, J., Caron, D.A., Goldman, J.C., Dennett, M., 1993. Isotopic fractionation of oxygen by respiring marine organisms. Global Biogeochemical Cycles 7, 679e694. Knox, M., Quay, P.D., Wilbur, D.O., 1992. Kinetic isotopic fractionation during air-water gas transfer of O2, N2, CH4, and H2. Journal of Geophysical Research 97, 20335e20343. Kroopnick, P., Craig, H., 1972. Atmospheric oxygen: isotopic composition and solubility fractionation. Science 175, 54e55. Livingstone, D.M., 1991. The diel oxygen cycle in three subalpine Swiss streams. Archiv fu¨r Hydrobiologie 120, 457e479. McCutcheon, S.C., 1989. Water Quality Modeling: Volume I Transport and Surface Exchange in Rivers. CRC Press, Inc, Boca Raton Fla, 334 p. Melching, C.S., Flores, H.E., 1999. Reaeration equations derived from U.S. geological survey database. Journal of Environmental Engineering 125, 407e414. O’Connor, D.J., Dobbins, W.E., 1956. Mechanism of reaeration in natural streams. American Society of Civil Engineering Transactions 86, 35e55. Odum, H.T., Hoskin, C.M., 1958. Comparative studies on the metabolism of marine waters. Publications of the Institute Marine Sciences, University of Texas 5, 15e46. Owens, M., 1974. Methods of measuring production rates in running water. In: Vollenweider, R.A. (Ed.), A Manual on Methods of Measuring Primary Production in Aquatic Environments. IBP Handbook 12. Blackwell Scientific Publishers, Oxford, pp. 111e119. Owens, M., Edwards, R.W., Gibbs, J.W., 1964. Some reaeration studies in streams. International Journal of Air and Water Pollution 8, 469e486. Parker, S.R., Poulson, S.R., Gammons, C.H., Degrandpre, M.D., 2005. Biogeochemical controls on diel cycling of stable isotopes of dissolved O2 and dissolved inorganic carbon in the Big Hole River, Montana. Environmental Science and Technology 39, 7134e7140. Parkhurst, J.D., Pomeroy, R.D., 1972. Oxygen absorption in streams. Journal of the Sanitary Engineering Division, American Society of Civil Engineering 98, 101e124. Pearson, N., Crossland, N.O., 1996. Measurement of community photosynthesis and respiration in outdoor artificial streams. Chemosphere 32, 913e919. Poulson, S.R., Sullivan, A.B., 2010. Assessment of diel chemical and isotope techniques to investigate biogeochemical cycles in the upper Kalamath River, Oregon, USA. Chemical Geology 269, 3e11. Quay, P.D., Emerson, S., Wilbur, D.O., Stump, C., Knox, M., 1993. The d18O of dissolved O2 in the surface waters of the Subarctic Pacific: a tracer of biological productivity. Journal of Geophysical Research 98, 8447e8458. Quay, P.D., Wilbur, D.O., Richey, J.E., Devol, A.H., Brenner, R., Forsberg, B.R., 1995. The 18O:16O of dissolved oxygen in rivers and lakes in the Amazon Basin: determining the ratio of respiration to photosynthesis rates in freshwaters. Limnology and Oceanography 40, 718e729. Radtke, D.B., White, A.F., Davis, J.V., Wilde, F.D., 1998. National Field Manual for the Collection of Water-quality Dataedissolved Oxygen. In: U.S. Geological Survey Techniques of Water-Resources Investigations, Book 9. chap. A6, sect. 6.2. Rathburn, R.E., 1977. Reaeration coefficients of streams e state of the art. Journal of the Hydraulics Division, American Society of Civil Engineering 103, 409e424. Servais, P., Debecker, E., Billen, G., 1984. Annual cycle of gross primary production and respiration in the Viroin River (Belgium). Hydrobiologia 111, 57e63. Tobias, C.R., Bohlke, J.K., Harvey, J.W., 2007. The oxygen-18 isotope approach for measuring aquatic metabolism in
790
w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 7 8 1 e7 9 0
high-productivity waters. Limnology and Oceanography 51, 1439e1453. Tsivoglou, E.C., Neal, L.A., 1976. Tracer measurements of reaeration: III. Predicting the reaeration capacity of inland streams. Journal of the Water Pollution Control Federation 48, 2669e2689. Venkiteswaran, J.J., Wassenaar, L.I., Schiff, S.L., 2007. Dynamics of dissolved isotopic ratios: a transient model to quantify primary production, community respiration, and air-water exchange in aquatic ecosystems. Oecologia 153, 385e398. Wang, X., Veizer, J., 2000. Respiration-photosynthesis balance of terrestrial aquatic ecosystems, Ottawa area, Canada. Geochimica et Cosmochimica Acta 64, 3775e3786. Wanninkhof, R., Mulholland, P.J., Elwood, J.W., 1990. Gas exchange rates for a first-order stream determined with deliberate and natural tracers. Water Resources Research 26, 1621e1630. Wassenaar, L.I., Koehler, G., 1999. An on-line technique for the determination of the d18O and d17O of gaseous and dissolved oxygen. Analytical Chemistry 71, 4965e4968.
Wassenaar, L.I., Venkiteswaran, J.J., Schiff, S.L., Koehler, G., 2010. Aquatic community metabolism response to municipal effluent inputs in rivers quantified using diel d18O values of dissolved oxygen. Canadian Journal of Fisheries and Aquatic Sciences 67, 1232e1246. Weiss, R.F., 1970. Solubility of nitrogen, oxygen and argon in water and seawater. Deep-Sea Research 17, 721e735. Wilcock, R.J., 1988. Study of river reaeration at different flow rates. Journal of Environmental Engineering, American Society of Civil Engineering 114, 91e103. Wilcock, R.J., Nagels, J.W., McBride, G.B., Collier, K.J., 1998. Characterisation of lowland streams using a single-station diel curve analysis model with continuous monitoring data for dissolved oxygen and temperature. New Zealand Journal of Marine and Freshwater Research 32, 67e79. Williams, R.J., White, C., Harrow, M.L., Neal, C., 2000. Temporal and small-scale variations of dissolved oxygen in the Rivers Thames, Pang, and Kennet, UK. Science of the Total Environment 251/252, 497e510.
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Using stable isotopes of dissolved oxygen for the determination of gas exchange in the Grand River, Ontario, Canada Terra S. Jamieson a,*, Sherry L. Schiff a, William D. Taylor b a b
Department of Earth Sciences, University of Waterloo, 200 University Avenue West, Waterloo, ON, Canada N2L 3G1 Department of Biology, University of Waterloo, 200 University Avenue West, Waterloo, ON, Canada N2L 3G1
article info
abstract
Article history:
Gas exchange can be a key component of the dissolved oxygen (DO) mass balance in
Received 17 August 2012
aquatic ecosystems. Quantification of gas transfer rates is essential for the estimation of
Received in revised form
DO production and consumption rates, and determination of assimilation capacities of
21 October 2012
systems receiving organic inputs. Currently, the accurate determination of gas transfer
Accepted 1 November 2012
rate is a topic of debate in DO modeling, and there are a wide variety of approaches that
Available online 21 November 2012
have been proposed in the literature. The current study investigates the use of repeated measures of stable isotopes of O2 and DO and a dynamic dual mass-balance model to
Keywords:
quantify gas transfer coefficients (k) in the Grand River, Ontario, Canada. Measurements
Dissolved oxygen
were conducted over a longitudinal gradient that reflected watershed changes from agri-
Gas exchange
cultural to urban. Values of k in the Grand River ranged from 3.6 to 8.6 day1, over
Impacted river
discharges ranging from 5.6 to 22.4 m3 s1, with one high-flow event of 73.1 m3 s1. The k
Stable isotopes
values were relatively constant over the range of discharge conditions studied. The range in discharge observed in this study is generally representative of non-storm and summer low-flow events; a greater range in k might be observed under a wider range of hydrologic conditions. Overall, k values obtained with the dual model for the Grand River were found to be lower than predicted by the traditional approaches evaluated, highlighting the importance of determining site-specific values of k. The dual mass balance approach provides a more constrained estimate of k than using DO only, and is applicable to large rivers where other approaches would be difficult to use. The addition of an isotopic mass balance provides for a corroboration of the input parameter estimates between the two balances. Constraining the range of potential input values allows for a direct estimate of k in large, productive systems where other k-estimation approaches may be uncertain or logistically infeasible. ª 2012 Elsevier Ltd. All rights reserved.
1.
Introduction
Oxygen is considered to be a key parameter for the evaluation of water quality in aquatic ecosystems. In turbulent streams,
gas exchange can be an important component of the dissolved oxygen (DO) balance, mediating the effects of DO production and consumption due to chemical and biological processes by driving concentrations toward atmospheric equilibrium. Two
* Corresponding author. Present address: Agriculture and Agri-Food Canada, 1791 Barrington Street Suite 405, P.O. Box 248, Halifax, NS, Canada B3J 2N7. Tel.: þ1 902 407 7663. E-mail addresses: [email protected] (T.S. Jamieson), [email protected] (S.L. Schiff), [email protected] (W.D. Taylor). 0043-1354/$ e see front matter ª 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.watres.2012.11.001
782
w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 7 8 1 e7 9 0
theories are widely used to describe gas transfer in natural waters: the two-film model, mainly applied to standing waters, and the surface-renewal model, commonly used in flowing waters (Chapra, 1997). The surface-renewal theory, assumes that in an agitated liquid, as in the Grand River, turbulence extends to the surface and therefore no laminar boundary layer exists (Danckwerts, 1951). That is, it assumes that the surface is continually being replaced with fresh liquid from a mixed vessel. The gas transfer velocity for O2 entering or leaving the water column is then dependant on the liquid surface renewal rate, as well as the diffusion rate of the gas in water (Chapra, 1997). For surface renewal models, the mass flux of O2 in rivers is generally assumed to be a product of the deviation of DO from saturation in the water column, multiplied by a mass-transfer coefficient: GE ¼ kðDOsat DOÞ
(1)
Where GE is gas exchange, the mass flux of O2 in DO (concentration time1), k is the gas transfer coefficient, and DOsat is the concentration of DO at atmospheric equilibrium. For rivers, the k value is typically treated as a first order rate constant expressed on a time1 basis, where the water column is assumed to be well-mixed with a homogeneous DO concentration. Determining k is essential for modeling DO in aquatic environments; i.e., estimating rates of production and consumption and determining the assimilation capacity of systems receiving organic inputs. Currently, the accurate determination of k is a topic of debate, and there are a wide variety of approaches for determining gas exchange rate including the application of tracers, regression modeling of night-time DO, and using hydrology-dependent equations. Formulas that estimate gas exchange coefficients have been developed and calibrated using gaseous tracers, e.g., methyl chloride, (Wilcock, 1988), ethylene (Wanninkhof et al., 1990), propane, SF6, (Clark et al., 1992) and 85Kr (Tsivoglou and Neal, 1976). The reaeration of deoxygenated water (Churchill et al., 1962; Owens et al., 1964) has also been used. Radioactive tracer techniques tend to yield the most accurate k values (Bowie et al., 1985) but have been limited in application due to the potentially hazardous effects of radiation (Bicudo and James, 1989). The use of inert gas tracers is also subject to controversy regarding the accuracy and reproducibility of the data (Bowie et al., 1985). Other approaches use DO changes; Odum and Hoskin (1958) developed a method for calculating k, where the DO deficit is regressed against the rate of change in DO after sunset, the slope of which is k. This method was later recommended by Owens (1974). Chapra and Toro (1991) have also developed an analytical solution for determining k from a single diel DO monitoring station, showing that k is a function of the time lag between solar noon and DO maximum, and day length. One of the most commonly used models, the O’ConnoreDobbins formula (O’Connor and Dobbins, 1956) was developed from the surface renewal theory and the diffusivity of oxygen in natural waters. There are currently more than 34 available formulas that have been proposed to predict k in rivers (McCutcheon, 1989) which vary in their uncertainty and applicability. Comprehensive reviews and critiques of the many formulas have
been provided by Bowie et al. (1985) and McCutcheon (1989). These formulas are usually parameterized in terms of water velocity and mean depth, and sometimes bed slope, and thus the applicability of a chosen approach depends on the similarity between the river under examination and the river used to calibrate the gas transfer relationship (Bennett and Rathburn, 1972). A number of studies have used both DO and stable isotopes of O2 to investigate community metabolism in aquatic systems (Quay et al., 1993, 1995; Parker et al., 2005; Tobias et al., 2007; Venkiteswaran et al., 2007; Poulson and Sullivan, 2010; Wassenaar et al., 2010). The first of these studies assumed steady state for the hydrologic conditions and gas exchange rates (Quay et al., 1995; Wang and Veizer, 2000), but later work quantified river metabolism rates under dynamic conditions (Parker et al., 2005; Tobias et al., 2007; Venkiteswaran et al., 2007). Examined approaches included using DO regression to obtain k (Parker et al., 2005), conducting tracer addition experiments (Tobias et al., 2007), or resolving DO and isotopic mass balances iteratively (Venkiteswaran et al., 2007; Poulson and Sullivan, 2010; Wassenaar et al., 2010). Previous studies have shown that using isotopic techniques for quantifying community metabolism is promising, but requires further evaluation in a variety of aquatic ecosystems, and needs to be compared to traditional techniques. It is of particular interest to investigate the applicability of using the stable isotopes of O2 as a tool to determine k in large, impacted rivers, where tracer techniques may be logistically difficult to use, and the determination of appropriate gas transfer analytical equations is uncertain. The objectives of this study were to i) use stable isotopes of O2, in conjunction with measurement of DO concentration, to quantify k in a large, impacted river, ii) compare k estimates from the isotope technique to other literature approaches and, iii) assess the spatial and temporal variability of k in the Grand River.
2.
Study site: The Grand River watershed
The Grand River watershed, the largest in southern Ontario, covers an area of 6965 km2, and the main channel measures about 290 km in length (Cooke, 2006). The watershed drains agricultural land (76%) and wooded areas (17%) but the central 5% of the watershed is largely urbanized. Most of the basin’s population, over 900,000 people, reside in the cities of Kitchener, Waterloo, Cambridge, Guelph and Brantford (Fig. 1). Currently, the primary indicators of water quality in the Grand River Watershed used by the Grand River Conservation Authority (GRCA) are DO and nutrient concentrations (Cooke, 2006). Nutrient-rich conditions and depletion of DO have been identified as issues of concern in portions of the basin’s watercourses. The river receives organic and nutrient inputs from 26 sewage treatment plants, on-site wastewater treatment systems, and non-point agricultural and urban sources. It also serves as a source of drinking water and supports a significant recreational fishery. The GRCA relies on a dynamic mass balance simulation model (Grand River Simulation Model, GRSM) to predict DO concentrations in the Grand River under various management regimes (e.g., flow
w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 7 8 1 e7 9 0
Fig. 1 e Map of the Grand River Watershed. The West Montrose, Bridgeport and Blair sampling locations are indicated (image obtained and adapted from www. grandriver.ca).
augmentation, inputs from sewage treatment plants). DO predictions derived from the GRSM depend on O2 demand, photosynthesis, and gas exchange (GRCA, 1982). Three sampling locations (Fig. 1) were chosen along the central portion of the Grand River, in 5th to 6th order reaches along the main channel: West Montrose (WM), Bridgeport (BRPT) and Blair (BLR). West Montrose is located upstream of the cities of Kitchener and Waterloo (population ca. 326,500) in a predominately agricultural landscape. Bridgeport, located downstream from WM, is subject to input from the relatively nutrient-laden Canagagigue Creek and the Conestogo River. Blair is located downstream of two major sewage treatment plant outfalls located in Kitchener and Waterloo, and receives urban runoff from the upstream communities. These sites were intended to characterize k at different locations adjacent to GRCA real-time water quality monitoring stations, along a longitudinal gradient of increasing cumulative impact.
3.
Sampling and analytical methods
Sampling was conducted throughout May to August, after sunset to remove the effects of photosynthesis on the DO. For
783
each sampling event, samples were collected every 2e4 h between sunset and sunrise. Samples were manually collected from within the main current of the channel at approximately mid-depth and as close to mid-channel as wade-able conditions allowed. It was assumed that the water column at these locations were well-mixed and representative of homogeneous conditions. In 2003, monitoring was mainly focused on WM, but all three locations were sampled during 2004 and 2005. Water samples were collected for DO and d18OeO2, and temperature was recorded. Dissolved O2 samples were collected in duplicate and titrated within 24 h of collection using the Winkler method (APHA, 1995). The precision of the Winkler analytical method is approximately 0.2 mg L1. Dissolved O2 concentrations were also measured at the time of sample collection using a YSI 556 MPS hand-held meter to serve as supplemental data to the collected DO samples. The d18OeO2 samples were collected, once per location per sample time, in 160 ml Wheaton serum bottles containing 0.3 g of sodium azide as a bactericide. Prior to sampling, the serum bottles were crimp sealed with butyl blue stoppers and then evacuated to <0.001 atm. The evacuated serum bottles were submerged and filled by piercing the septum via an 18 gauge needle. After filling, the needle was removed while the bottle was still submerged to prevent atmospheric 18O from entering. The d18OeO2 samples were prepared for analysis by removing 5 ml of water, inserting 5 ml of He, and shaking to equilibrate the headspace. Headspace gas was analyzed using a continuous flow gas chromatograph isotope ratio mass spectrometer, following Wassenaar and Koehler (1999), with an analytical error of approximately 0.3&.
4.
Model description
To address the variable nature of DO in the Grand River, a dynamic mass balance model similar to the PoRGy model developed by Venkiteswaran et al. (2007) was used. The key potential benefit of using 18OeO2 is that the isotopic signature provides a second signal in addition to DO, that is directly related to the processes consuming and producing DO. There are a number of properties that enable 18O of DO (18OeO2) to be used to quantify metabolism and gas exchange rates in rivers. Atmospheric O2 is isotopically enriched, with a d18O of þ23.5 0.3& relative to standard mean ocean water (Kroopnick and Craig, 1972). During gas dissolution, there is an equilibrium fractionation of approximately 0.7& at 20 C (Kroopnick and Craig, 1972; Wassenaar and Koehler, 1999) resulting in a d18OeO2 of þ24.2&. Photosynthesis produces O2 with the isotopic signature of water as there is no fractionation in the photolysis of water (Guy et al., 1993). However, as the samples were collected after dusk, the effects of photosynthesis were not included in the model presented here. Another key aspect is that respiring organisms consume the isotopically lighter 16O:16O molecules 10e20& more rapidly than 18O:16O molecules (Kiddon et al., 1993) referred to as respiratory fractionation. Therefore, respiration increases the 18 O fraction of the residual DO. Two O2 mass balances (DO and isotopic) were resolved based on the following equations:
784
w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 7 8 1 e7 9 0
dDO ¼ kðDOsat DOÞ CR dt
(2)
CRT ¼ CR20 1:047ðT20Þ
dDO 18:16 O 18 ¼ OGE þ 18 OExch CRð18:16 O ar Þ dt
(3)
where T is temperature, measured in C, k20 and CR20 are k and CR at 20C.
where CR is the community respiration rate (mg L1 h1), k is the gas exchange coefficient (h1), DOsat is the DO concentration under saturated conditions, 18:16O refers to the 18 O:16O of DO, and ar is the isotopic fractionation factor associated with CR. DOsat was calculated based on temperature (Weiss, 1970), and corrected for the altitude (Radtke et al., 1998) at each sampling location. 18OExch describes the mass flux of 18O and 16O due to exchange with the atmosphere, as part of the mass transfer of DO. In undersaturated DO conditions, 18O was assumed to invade the water column more slowly than 16O, due to fractionation associated with solubility and airewater transfer. Similarly, in supersaturated DO conditions, 18O was assumed to preferentially evade the water column, leaving behind residual 16O due the greater solubility of the lighter isotope. Therefore: If (DOsatDO) > 0 then: 18
OGE ¼ k ak ðDOsat DOÞ ð18:16 Oa as Þ
(4)
Else: 18
OGE ¼ k ak ðDOsat DOÞ ð18:16 O=as Þ 18
(5)
16
where 18:16Oa is the ratio of O to O in air, as is the fractionation effect of the dissolution of O2 in water, and ak is the fractionation factor associated with gas transfer. The solubility fractionation of 18OeO2 (as ¼ 1.0007) is due to the slightly different solubility between 18O and 16O, with negligible temperature effects (Benson and Krause, 1984). There is a kinetic fractionation effect associated with airewater transfer of 18OeO2 (ak ¼ 0.9972), as the 18O molecule is slightly larger than 16O (Knox et al., 1992). The temperature effects associated with ak were also assumed to be negligible over the temperatures encountered with respect to the mass balance. According to the equations, if DO is at equilibrium (DOsatDO ¼ 0), 18OGE would be nil. The isotopes of O2 were assumed to continue to diffuse across the airewater interface if d18OeO2 is at disequilibrium even if DO was at saturation. In order to allow for 18O and 16O to exchange across the airewater interface towards equilibrium independently of the bulk exchange of DO, the following equation was used: 18
OExch ¼ k ak DO ½ð18:16 Oair as Þ 18:16 O
(6)
The isotopic signature of atmospheric O2 is assumed to be constant at 23.5& (Kroopnick and Craig, 1972). Isotope values were modeled as ratios but will be reported in units of & in dnotation with respect to Vienna Standard Mean Ocean Water (VSMOW). The effects of groundwater input were assumed to be negligible at the sample locations with respect to the large fluctuations in DO. Both k and CR values were varied with temperature within the model using the following relationships (McCutcheon, 1989; Chapra, 1997): kT ¼ k20 1:024ðT20Þ
(7)
5.
(8)
Model calibration
Modeling was performed using STELLA 7.0.2, with a 15 min time-step using the mass balance equations noted above. The unknowns in the model were CR, k, and ar. Due to the dynamic nature of the system, and because there are three unknowns and only two mass balances, the model could not be solved directly. Therefore, the model was calibrated iteratively to optimize the goodness of fit between modeled and measured data. Model outputs were evaluated by both visually examining the fit of the model to the data, and using root mean square error (RMS) to statistically evaluate goodness of fit. The time zero DO and d18OeO2 points for each modeling event were not included in the RMS calculations, as these points were used to initialize the model at time zero. Iterative simulations were first performed and compared to DO concentration data; model outputs that resulted in RMS values of approximately 0.2 mg L1 or less (similar to the analytical error associated with Winkler titration), supported by visual examination, were used to constrain the potential range in input parameter values. The identified range of k20 and CR20 values were then modeled iteratively with a range of ar, and compared to the observed d18OeO2. Estimates of k20, CR20, and ar that provided an optimal model fit to observed data for both the DO and d18OeO2 were then chosen for rate estimations.
6. Comparison to other gas transfer estimates There are many predictive equations for k, but the most widely used equations (Chapra, 1997) include those by O’Connor and Dobbins (1956), Churchill et al. (1962), and Owens et al. (1964), which are parameterized in terms of mean depth (D) and velocity (V) according to: k20 ¼ z
Vx Dy
(9)
where z, x, and y are the empirically calibrated constants (Table 1). To put the k estimated by the isotope technique in context with estimates from these predictive equations, hydrologic conditions were quantified for each sampling event. Discharge rates were obtained from the GRCA for each of the three sampling locations. Hydrologic data (i.e., discharge, mean depth, mean velocity) for the BLR sampling location were also obtained from stream gauging activities conducted by the GRCA. At the WM location, cross-sectional surveys of the channel were performed just upstream of the sampling location to determine the relationship of mean depth with discharge, and water surface slope was also measured. Mean velocity (V) and depth (D) at WM as a function of the hydraulic geometry and
w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 7 8 1 e7 9 0
785
Table 1 e Selected literature equation coefficientsa for calculating k20 in rivers. Equation
O’Connor and Dobbins (1956) Churchill et al. (1962) Owens et al. (1964) Bennett and Rathburn (1972) a k20 ¼ z
Coefficients z
x
y
3.9 5.026 5.32 5.58
0.5 1 0.67 0.607
1.5 1.67 1.85 1.689
Vx . Dy
discharge at the time of sampling were computed using Manning’s expression (French, 1985). Cross-sectional surveys were not conducted at BRPT; the mean of the hydrologic relationships associated with WM and BLR were used to calculate mean depth and velocity changes based on discharge for the BRPT location. The hydrology data for each sampling location were used to calculate k from other published equations based on mean depth and velocity (Table 1). Regressions using night DO concentration data were also performed to obtain estimates of k (Odum and Hoskin, 1958). The rate of change of DO concentration was regressed against DO deficit (relative to saturation) to obtain the slope of the regression which is an estimate of k.
7.
Results and discussion
7.1.
Gas transfer coefficients in the Grand River
As expected, DO concentrations declined at night due to community respiration, with a concurrent enrichment of d18OeO2 (Figs. 2 and 3, and 4). The associated k values predicted from the isotope technique for the Grand River ranged from 3.6 to 8.6 d1 (Table 2). Overall, the non-steady state model appears to satisfactorily reproduce the changes in both DO and d18OeO2. RMS errors for the DO model output were near, or less than, the analytical error of 0.2 mg L1 associated with Winkler titrations. There was greater error associated with d18OeO2 model output, with a particularly poor model fit to d18OeO2 for BLR July 29e30, 2005. During this sampling event, DO concentrations were very low, falling to less than 2 mg L1. Under the extremely low DO conditions, the model may not have been able to adequately represent CR or the fractionation processes as constant rates are assumed for both. There also may have been additional analytical error associated with d18OeO2 due to the low levels of DO in the water sample. There appear to be greater RMS values associated with the BRPT and BLR as compared to WM. The high error in d18OeO2 model fit at BRPT on August 30e31, 2004, appears to be attributable to one anomalous data point, potentially due to inadequate mixing at the sampling site or analytical error. Community respiration rates ranged from approximately 10e47 mg O2 L1 d1, with associated ar values ranging from 0.971 to 0.996. Most values were between 0.980 and 0.989, which are within range of other literature values (Kiddon et al., 1993; Quay et al., 1995; Tobias et al., 2007; Venkiteswaran et al., 2007; Poulson and Sullivan, 2010), as discussed elsewhere (Jamieson, 2010).
Fig. 2 e Observed and model-predicted DO and d18OeO2 in the Grand River at West Montrose.
The model uses one rate of CR20 and k20, over the modeled time period, and short-term variability in either of these terms beyond what is attributable to the temperature correction could potentially lead to deviations from the modeled DO and 18 OeO2 output. This study was undertaken as a single-station approach, common to other DO modeling studies (Servais et al., 1984; Edwards and Meyer, 1987; Livingstone, 1991; Pearson and Crossland, 1996; Wilcock et al., 1998; Williams et al., 2000). As a result, the findings obtained for a particular location are an integrated measure of processes occurring in the watershed upstream from that point. Gas transfer coefficients at BRPT were roughly half of the estimates for WM and BLR, which exhibited similar k. Gas exchange in rivers is mainly influenced by physical attributes of the channel. At WM and BLR, the river upstream consists of main channel, with no confluences with other major tributaries or reservoirs located immediately upstream, which may have led to similar k values noted among these locations. However, BRPT is located downstream from the Conestoga river confluence, one of the major tributaries in the watershed, as well as a floodplain area that formerly served as a gravel mining site. Gas transfer upstream of BRPT may be influenced by Conestoga river input, or other channel
786
w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 7 8 1 e7 9 0
Fig. 3 e Observed and model-predicted DO and d18OeO2 in the Grand River at Bridgeport.
morphological conditions which could exhibit lower k values such as deep zones. The upstream gravel floodplain could potentially be an area of groundwater discharge of a magnitude great enough to affect the DO budget, which was not accounted for in the model.
The fit of the model to observed data was generally poorest at BLR, as compared to WM and BRPT. Error in associated model fits may be due to short-term temporal variation in upstream processes that are not accounted for in the model, as constant rates of CR20 were assumed. BLR is located downstream of two major sewage treatment plants that would have a daily pattern of discharge which could cause small temporal variability in model parameters. Varying inputs of DO consuming substances (e.g., organic matter, ammonium), or even aerated effluent, could cause short-term variability in observed DO and d18OeO2. Community respiration rates tend to be greater at BLR than upstream, with lower DO minima at night. As DO concentration declines at BLR, an associated decline in the rate of CR and changes in isotopic fractionation could also be occurring as the system approaches hypoxic conditions. The range in k observed in the Grand River (3.6e8.6 d1) is well within ranges measured in other similarly sized rivers. Wilcock (1988) found k values of 6.34e8.54 d1 obtained from gas tracer experiments for a river with mean annual discharge and velocity of 30 m3 s1, and 0.8 m s1. Using the isotopic approach similar to the present study, Venkiteswaran et al. (2007) reported a k value of 3.96 d1 in the South Saskatchewan River, which was characterized as having a mean depth of 0.65 m and mean velocity of 0.21 m s1 O’Connor and Dobbins (1956) estimated re-aeration k values ranging from 0.03 to 4.8 d1 in rivers with mean depths of 0.27e11 m, where shallower rivers exhibited greater k values. Gas transfer coefficients in the Grand River were relatively constant over changes in mean depth and velocity during sampling events, with virtually no relationship between discharge and k (R ¼ 0.49). The development of a mean depth and velocity dependant equation specific to Grand River conditions would be inappropriate based on the findings in the current study as there was not a strong relationship between k and hydrologic conditions. Other studies that have derived an empirical predictive k equation have analyzed data sets covering k values that range by two to three orders of magnitude (Bowie et al., 1985). The lack of variability k in the Grand River is likely due to the shape of the river channel, which tends to be wide and shallow at all three locations. As discharge rates increase, both mean depth and velocity increased at similar rates; increases in depth and velocity would effectively cancel one another, since k tends to be proportional to velocity divided by depth (Melching and Flores, 1999). Genereux and Hemond (1992) measured k in a first order stream using a petroleum gaseous tracer technique, and found that measured k was relatively constant with increases in flow. Melching and Flores (1999) found no relationship between k and discharge in 493 reaches in 166 streams across the U.S that exhibited discharges from <0.01 to >100 m3 s1 Wassenaar et al. (2010) also noted that k values varied little seasonally and longitudinally in the South Saskatchewan River despite variations in discharge of approximately 100e500 m3 s1.
7.2. Fig. 4 e Observed and model-predicted DO and d18OeO2 in the Grand River at the Blair.
Comparison to other approaches
Gas transfer coefficient estimates obtained from the isotope modeling approach were found to be lower than those
787
w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 7 8 1 e7 9 0
Table 2 e Fitted model input parameter results obtained from DO concentration and d18OeO2 night-time data modeling. Sampling event WM
BRPT
BLR
May 28/03 Jun 24/03 Jul 28/03 Aug 13/03 Jul 27/05 Jul 9/03 May 11/04 Jun 7/04 Aug 30/04 Jun 7/04 Aug 30/04 Jul 29/05
Model ka (d1)
Model CRa (mg L1 d1)
Model ar
RMS DO (mg L1)
RMS d18OeO2 (&)
7.4 6.0 7.0 6.7 5.3 3.6 3.8 3.6 3.6 8.6 7.2 6.0
22.8 21.4 23.0 20.6 23.8 10.1 13.8 13.7 11.5 43.2 38.2 46.8
0.988 0.987 0.986 0.983 0.990 0.986 0.980 0.984 0.971 0.989 0.989 0.996
0.03 0.12 0.04 0.04 0.16 0.26 0.18 0.35 0.05 0.09 0.09 0.19
0.10 0.73 0.72 0.54 0.27 0.21 0.77 0.91 3.21 0.53 0.29 2.04
a Corrected to 20 C.
predicted by the other approaches (Table 3 and Fig. 5). The isotopic model-derived k was, on average, 5.2 d1 less than k calculated from equations. The night-time regression technique (Odum and Hoskin, 1958) was in closest agreement with isotopic model-derived k. Discrepancies between the k found by regression and the isotopic model are likely due to the fact that the dual model also fits the change in d18OeO2, as well as error that would occur in averaging the temperature and DO deficit over the intervals between samples. Water temperature in the Grand River was not constant during the sampling events, resulting in varying DOsat conditions and CR rates during the regression time periods. The temperature declined by between 0.8 and 4.7 C during a given night-time sampling event, with an average decline of 2.7 C. A decline in 2.7 C corresponds to a change in DOsat of approximately 0.5 mg L1, and a change in CR of about 1.6 mg O2 L1 d1 (given a mean CR rate of 24 mg O2 L1 d1). This variation may be enough to cause a change in k estimation using a regression technique, which assumes constant DOsat and CR.
In theory, night-time regression should predict an accurate k, based on the DO deficit and dDO/dt under steady state CR, temperature, and k conditions. To test varying temperature effects on k estimation using the regression technique, DO output data were generated for 15 min time steps using STELLA 7.0.2, assuming the model conditions used for the BLR August 30, 2004 and WM August 13, 2003 sampling events. The DO regression technique was then used to calculate k from the theoretical curve (reported in Jamieson, 2010). Changes in temperature over the regression time period appear to cause a consistent over-prediction in k using the night-time regression versus the isotopic model. In the BLR and WM simulation trials tested, the over-predictions were 1.6 and 1.1 d1 for respective changes in temperature of 1.2e3.8 C, which are similar to the average over-prediction of night-time regressed k compared to isotope model derived k obtained from the sampling events (Table 3). This suggests that the k values estimated from the isotope modeling technique are likely representative of in situ conditions and that the
Table 3 e Comparison of model derived k to selected literature approaches. D (m) V Model ka O’Connor Churchill et al. Owens et al. Bennett and DO Regressiona Sampling Q (m s1) (d1) and Dobbins (1962) (1964) Rathburn (1972) (Odum and Hoskin, event (m3 s1) (1956) 1958) WM May 28/03 Jun 24/03 Jul 28/03 Aug 13/03 Jul 27/05
6.8 5.6 6.0 12.7 5.6
0.42 0.37 0.39 0.58 0.37
0.50 0.47 0.48 0.63 0.47
7.4 6.0 7.0 6.7 5.3
10.1 11.9 11.1 7.0 11.9
10.7 12.4 11.6 7.9 12.4
16.6 20.2 18.6 10.7 20.2
15.9 18.9 17.5 10.6 18.9
9.6 12.1 8.1 7.7 8.0
BRPT Jul 9/03 May 11/04 Jun 7/04 Aug 30/04
8.2 73.1 11.7 18.7
0.40 1.07 0.47 0.58
0.41 1.01 0.48 0.58
3.6 3.8 3.6 3.6
9.9 3.5 8.4 6.7
9.5 4.5 8.5 7.2
15.9 4.7 13.2 10.1
15.3 5.0 12.8 10.1
5.0 2.9 9.5 2.5
BLR Jun 7/04 Aug 30/04 Jul 29/05
14.9 22.4 11.0
0.41 0.46 0.37
0.39 0.47 0.33
8.6 7.2 6.0
9.3 8.6 10.0
8.5 8.6 8.7
14.7 13.5 15.9
14.2 13.1 15.3
9.0 8.1 6.0
a k corrected to 20 C.
788
w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 7 8 1 e7 9 0
Fig. 5 e Plot of k values as predicted by other selected published approaches versus modeled k values in the Grand River.
deviation from night-time regression estimates lies in: 1) the paucity of data points collected to perform night-time regression analyses and 2) the varying temperature. In addition, Odum and Hoskin (1958) used hourly DO concentration data to obtain estimates of k, CR, and GPP. Data of this resolution were only available for the WM August 15, 2003 sampling event. Of the empirical equations, the closest k estimates were from the O’Connor and Dobbins (1956) equation, followed by Bennett and Rathburn (1972), and then Owens et al. (1964). O’Connor and Dobbins (1956) developed a theoretical equation, based on fundamental turbulence parameters, and tested it over a wide range of depth and velocity conditions. As a result, the equation tends to have the widest applicability for a range of flow conditions, although it is most appropriate for moderate to deep streams with moderate to low velocities (Chapra, 1997). Bennett and Rathburn (1972) also found that O’Connor and Dobbins (1956) was the formula that best fit the entire range of existing gas transfer data sets that they reviewed, and the Churchill et al. (1962) formula provided the best fit to natural stream data. The standard error of estimate was reported to range from 44 to 61% when comparing equation estimates developed by Melching and Flores (1999) to direct k results from tracer-gas techniques. Rathburn (1977) reported standard errors of estimate ranging from 3.0 to 7.2 d1 when comparing results using the equations addressed in the current study (Table 3) to tracer derived estimates, using a composite of results of five US rivers. Rathburn (1977) found that Churchill et al. (1962) had the lowest error, followed by O’Connor and Dobbins (1956), Bennett and Rathburn (1972), and Owens et al. (1964). According to Rathburn (1977), k values calculated from empirical equations tend to be greater than values obtained by tracer techniques. The discrepancy was partially attributed to possible effects of wind or substances in a stream (e.g., detergents, pollutants from sewage treatment plants). The Grand River receives nutrient loading from both point and non-point sources along the main channel at both locations studied, has high macrophyte biomass, possibly
resulting in depressed k coefficients. Parkhurst and Pomeroy (1972) noted that the presence of impurities results in a more stable surface film, thus reducing turbulence at the surface, leading to a decrease in k as compared to pure water. The authors cited research conducted by Kehr (1938) which showed that sewage additions as low as 0.5 % to tap water reduced k by 10 %. In the current study, the calculation of k in the Grand River using the empirical equations were dependent on the assumed hydrologic conditions associated with the sampling location of interest. Characteristics of the sampling locations chosen were accessible, shallow depths for ease of monitoring and sample collection. Upstream differences in hydrologic characteristics may lead to k rates that are not reflective of the conditions observed at the sampling location due to time of travel, leading to discrepancy between k obtained by the isotopic approach and that from the literature equations examined. For instance, in the literature equations noted, deeper stream channels or slower moving waters would lead to lower k estimates, more in line with that obtained by the isotopic method or DO regression method. These findings highlight the importance of determining site-specific values of k as the literature derived estimates varied by three-fold. The potential for over-estimation of k in large, impacted rivers should be taken into consideration when making watershed management decisions, as it could lead to error in respiration estimates, and under-prediction of DO deficits. Gas transfer coefficient values were relatively constant at each location under hydrologic conditions that were mainly representative of non-storm and summer lowflow periods. A greater range in k might be observed under a wider range of hydrologic conditions. However, from a water quality management perspective, low-flow conditions are the times of highest concern where gas transfer would be a key factor in mitigating DO deficits due to O2 demand. Groundwater influences in the Grand River were also assumed to be negligible in the DO mass balance given the high level of daily DO concentration amplitude observed; this assumption may not be appropriate for some locations. Further investigation into the importance of groundwatersurface water interactions as a model addition may be warranted in areas where groundwater discharge greatly influences the DO regime. Additional work to assess the robustness of the isotope approach under varying hydrologic regimes and locations is also needed, along with comparison to direct measurement techniques (e.g., tracers) to assess the accuracy of the technique in greater detail. However, the isotope modeling technique appears to offer a promising advantage in simulating changes in DO regimes as compared to the available literature. The dual mass balance provided a constraint on the potential range of input parameters to optimize fit to observed data, leading to a more direct estimation of k in large, productive systems.
Acknowledgments Funding for this research was supported by the Natural Sciences and Engineering Research Council, and the Ontario Centres of Excellence. We thank Richard Elgood for laboratory
w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 7 8 1 e7 9 0
research support and would like to acknowledge Dr. Len Wassenaar and Dr. Ramon Aravena for their critical input during the research process. We are also grateful for the help provided by field and laboratory assistants throughout this study, including: Stuart Elson, Brianna Kelly, Kevin Maurice, Paul Malone, Matt Mittler, Ashley Smyth, David Snider, Simon Thuss, and Andrea Wojtyniak.
references
American Public Health Association (APHA), 1995. Standard Methods for the Examination of Water and Wastewater, nineteenth ed. American Public Health Association, Inc., New York. Bennett, J.P., Rathburn, R.E., 1972. Reaeration in Open-channel Flow. U.S. Geological Survey. Professional Paper 737. Benson, B.B., Krause, D.J., 1984. The concentration and isotopic fractionation of oxygen dissolved in fresh water and seawater in equilibrium with the atmosphere. Limnology and Oceanography 29, 620e632. Bicudo, J.R., James, A., 1989. Measurement of reaeration in streams: comparison of techniques. Journal of Environmental Engineering 115, 992e1010. Bowie, G.L., Mills, W.B., Porcella, D.B., Campbell, C.L., Pagenkopf, J.R., Rupp, G.L., Johnson, K.M., Chan, P.W.H., Gherini, S.A., Chamberlain, C.E., 1985. Rates, Constants, and Kinetic Formulations in Surface Water Modeling. U.S. Environmental Protection Agency, ORD, Athens, GA, ERL. EPA/600/3e85/040. Chapra, S.C., 1997. Surface Water-quality Modeling. McGraw-Hill Companies, Inc, New York, NY, 844 p. Chapra, S.C., Di Toro, D.M., 1991. Delta method for estimating primary production, respiration, and reaeration in streams. Journal of Environmental Engineering, American Society of Civil Engineering 117, 640e655. Churchill, M.A., Elmore, H.L., Buckingham, R.A., 1962. The prediction of stream reaeration rates. Journal of the Sanitary Engineering Division, American Society of Civil Engineering SA 4, 1. Proc. Paper 3199. Clark, J.F., Simpson, H.J., Smethie Jr., W.M., Toles, C., 1992. Gas exchange in a contaminated estuary inferred from chlorofluorocarbons. Geophysical Research Letters 19, 1133e1136. Cooke, S., 2006. Water Quality in the Grand River: A Summary of Current Conditions (2000-2004) and Long Term Trends. Grand River Conservation Authority, Cambridge, ON, 88 p. http:// www.grandriver.ca/Water/2006_WaterQuality_complete.pdf. Danckwerts, P.V., 1951. Significance of liquid-film coefficients in gas absorption. Industrial and Engineering Chemistry 43, 1460e1467. Edwards, R.T., Meyer, J.L., 1987. Metabolism of sub-tropical low gradient blackwater river. Freshwater Biology 17, 251e263. French, R.H., 1985. Open-channel Hydraulics. McGraw-Hill, USA, 739 p. Genereux, D.P., Hemond, H.F., 1992. Determination of gas exchange rate constants for a small stream on Walker Branch Watershed, Tennessee. Water Resources Research 28, 2365e2374. Grand River Conservation Authority (GRCA), 1982. Grand River Basin Management Study. Grand River Implementation Committee, Grand River Conservation Authority, Cambridge, ON, 8 p. http://www.grandriver.ca/Water/2007_WaterQuality_ boardreport.pdf. Guy, R.D., Fogel, M.L., Berry, J.A., 1993. Photosynthetic fractionation of the stable isotopes of oxygen and carbon. Plant Physiology 101, 37e47. Jamieson, T.S. 2010. Quantification of Oxygen Dynamics in the Grand River Using a Stable Isotope Approach. Ph.D. thesis, University of Waterloo, ON, Canada. 198 p.
789
Kehr, R.W., 1938. Measures of natural oxidation in polluted streams. IV. Effect of sewage on atmospheric reaeration rates under stream flow conditions. Sewage Works Journal 10, 228e240. Kiddon, J., Bender, M.L., Orchardo, J., Caron, D.A., Goldman, J.C., Dennett, M., 1993. Isotopic fractionation of oxygen by respiring marine organisms. Global Biogeochemical Cycles 7, 679e694. Knox, M., Quay, P.D., Wilbur, D.O., 1992. Kinetic isotopic fractionation during air-water gas transfer of O2, N2, CH4, and H2. Journal of Geophysical Research 97, 20335e20343. Kroopnick, P., Craig, H., 1972. Atmospheric oxygen: isotopic composition and solubility fractionation. Science 175, 54e55. Livingstone, D.M., 1991. The diel oxygen cycle in three subalpine Swiss streams. Archiv fu¨r Hydrobiologie 120, 457e479. McCutcheon, S.C., 1989. Water Quality Modeling: Volume I Transport and Surface Exchange in Rivers. CRC Press, Inc, Boca Raton Fla, 334 p. Melching, C.S., Flores, H.E., 1999. Reaeration equations derived from U.S. geological survey database. Journal of Environmental Engineering 125, 407e414. O’Connor, D.J., Dobbins, W.E., 1956. Mechanism of reaeration in natural streams. American Society of Civil Engineering Transactions 86, 35e55. Odum, H.T., Hoskin, C.M., 1958. Comparative studies on the metabolism of marine waters. Publications of the Institute Marine Sciences, University of Texas 5, 15e46. Owens, M., 1974. Methods of measuring production rates in running water. In: Vollenweider, R.A. (Ed.), A Manual on Methods of Measuring Primary Production in Aquatic Environments. IBP Handbook 12. Blackwell Scientific Publishers, Oxford, pp. 111e119. Owens, M., Edwards, R.W., Gibbs, J.W., 1964. Some reaeration studies in streams. International Journal of Air and Water Pollution 8, 469e486. Parker, S.R., Poulson, S.R., Gammons, C.H., Degrandpre, M.D., 2005. Biogeochemical controls on diel cycling of stable isotopes of dissolved O2 and dissolved inorganic carbon in the Big Hole River, Montana. Environmental Science and Technology 39, 7134e7140. Parkhurst, J.D., Pomeroy, R.D., 1972. Oxygen absorption in streams. Journal of the Sanitary Engineering Division, American Society of Civil Engineering 98, 101e124. Pearson, N., Crossland, N.O., 1996. Measurement of community photosynthesis and respiration in outdoor artificial streams. Chemosphere 32, 913e919. Poulson, S.R., Sullivan, A.B., 2010. Assessment of diel chemical and isotope techniques to investigate biogeochemical cycles in the upper Kalamath River, Oregon, USA. Chemical Geology 269, 3e11. Quay, P.D., Emerson, S., Wilbur, D.O., Stump, C., Knox, M., 1993. The d18O of dissolved O2 in the surface waters of the Subarctic Pacific: a tracer of biological productivity. Journal of Geophysical Research 98, 8447e8458. Quay, P.D., Wilbur, D.O., Richey, J.E., Devol, A.H., Brenner, R., Forsberg, B.R., 1995. The 18O:16O of dissolved oxygen in rivers and lakes in the Amazon Basin: determining the ratio of respiration to photosynthesis rates in freshwaters. Limnology and Oceanography 40, 718e729. Radtke, D.B., White, A.F., Davis, J.V., Wilde, F.D., 1998. National Field Manual for the Collection of Water-quality Dataedissolved Oxygen. In: U.S. Geological Survey Techniques of Water-Resources Investigations, Book 9. chap. A6, sect. 6.2. Rathburn, R.E., 1977. Reaeration coefficients of streams e state of the art. Journal of the Hydraulics Division, American Society of Civil Engineering 103, 409e424. Servais, P., Debecker, E., Billen, G., 1984. Annual cycle of gross primary production and respiration in the Viroin River (Belgium). Hydrobiologia 111, 57e63. Tobias, C.R., Bohlke, J.K., Harvey, J.W., 2007. The oxygen-18 isotope approach for measuring aquatic metabolism in
790
w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 7 8 1 e7 9 0
high-productivity waters. Limnology and Oceanography 51, 1439e1453. Tsivoglou, E.C., Neal, L.A., 1976. Tracer measurements of reaeration: III. Predicting the reaeration capacity of inland streams. Journal of the Water Pollution Control Federation 48, 2669e2689. Venkiteswaran, J.J., Wassenaar, L.I., Schiff, S.L., 2007. Dynamics of dissolved isotopic ratios: a transient model to quantify primary production, community respiration, and air-water exchange in aquatic ecosystems. Oecologia 153, 385e398. Wang, X., Veizer, J., 2000. Respiration-photosynthesis balance of terrestrial aquatic ecosystems, Ottawa area, Canada. Geochimica et Cosmochimica Acta 64, 3775e3786. Wanninkhof, R., Mulholland, P.J., Elwood, J.W., 1990. Gas exchange rates for a first-order stream determined with deliberate and natural tracers. Water Resources Research 26, 1621e1630. Wassenaar, L.I., Koehler, G., 1999. An on-line technique for the determination of the d18O and d17O of gaseous and dissolved oxygen. Analytical Chemistry 71, 4965e4968.
Wassenaar, L.I., Venkiteswaran, J.J., Schiff, S.L., Koehler, G., 2010. Aquatic community metabolism response to municipal effluent inputs in rivers quantified using diel d18O values of dissolved oxygen. Canadian Journal of Fisheries and Aquatic Sciences 67, 1232e1246. Weiss, R.F., 1970. Solubility of nitrogen, oxygen and argon in water and seawater. Deep-Sea Research 17, 721e735. Wilcock, R.J., 1988. Study of river reaeration at different flow rates. Journal of Environmental Engineering, American Society of Civil Engineering 114, 91e103. Wilcock, R.J., Nagels, J.W., McBride, G.B., Collier, K.J., 1998. Characterisation of lowland streams using a single-station diel curve analysis model with continuous monitoring data for dissolved oxygen and temperature. New Zealand Journal of Marine and Freshwater Research 32, 67e79. Williams, R.J., White, C., Harrow, M.L., Neal, C., 2000. Temporal and small-scale variations of dissolved oxygen in the Rivers Thames, Pang, and Kennet, UK. Science of the Total Environment 251/252, 497e510.