This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier’s archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/copyright
Author's personal copy
Estuarine, Coastal and Shelf Science 83 (2009) 342–348
Contents lists available at ScienceDirect
Estuarine, Coastal and Shelf Science journal homepage: www.elsevier.com/locate/ecss
Turbidity limits gas exchange in a large macrotidal estuary Gwenae¨l Abril*, Marc-Vincent Commarieu, Aldo Sottolichio, Patrice Bretel, Fre´de´ric Gue´rin 1 Laboratoire Environnements et Pale´oenvironnements Oce´aniques (EPOC), UMR CNRS 5805, Universite´ Bordeaux1, Talence, France
a r t i c l e i n f o
a b s t r a c t
Article history: Received 21 October 2008 Accepted 5 March 2009 Available online 13 March 2009
In estuaries, the gas transfer velocity (k) is driven by a combination of two major physical drivers, wind and water current. The k values for CO2 in the macrotidal Gironde Estuary were obtained from 159 simultaneous pCO2 and floating chamber flux measurements. Values of k increased with wind speed and were significantly greater when water currents and wind were in opposing directions. At low wind speeds ( 0.2 g L1) had a significant role in attenuating turbulence and therefore gas exchange. This result has important consequences for modeling water oxygenation in estuarine turbidity maxima. For seven low turbidity estuaries previously described in the literature, the slope of the linear regression between k and wind speed correlates very well with the estuary surface area due to a fetch effect. In the Gironde Estuary, this slope follows the same trend at low turbidity (TSS < 0.2 g L1), but is on average significantly lower than in other large estuaries and decreases linearly with the TSS concentration. A new generic equation for estuaries is proposed that gives k as a function of water current velocity, wind speed, estuarine surface area and TSS concentration. Ó 2009 Published by Elsevier Ltd.
Keywords: gas exchange turbulence wind current turbidity
1. Introduction A better understanding of the gas transfer velocity (k) in estuaries is crucial for quantification of their contribution to the contemporary global carbon cycle (Frankignoulle et al., 1998; Raymond and Cole, 2001; Borges, 2005) and to global greenhouse gas budgets (Bange, 2006). In addition, estuaries often experience oxygen deficits in turbidity maxima (Parker et al., 1994; Mitchell et al., 1999; Uncles et al., 1997), and a precise parameterization of k is necessary for water oxygenation modeling (Thouvenin et al., 1994; Kremer et al., 2003a; Vanderborght et al., 2002). In estuaries, k is highly variable over time and space, and is influenced by local meteorological and hydrological conditions (Kremer et al., 2003a; Borges et al., 2004a). Further research efforts are necessary in order to understand the environmental and physical factors that control the rate of gas transfer and its variability in estuaries. In addition, it is necessary to develop adequate generic equations adapted to
* Corresponding author. E-mail address:
[email protected] (G. Abril). 1 Present address: Laboratoire des Me´canismes et Transferts en Ge´ologie, UMR UPS/CNRS/IRD 5563, OMP, 14, Avenue Edouard Belin, F-31400 Toulouse, France. 0272-7714/$ – see front matter Ó 2009 Published by Elsevier Ltd. doi:10.1016/j.ecss.2009.03.006
these environments that can compute k from wind speed, water current velocity and other measurable parameters. For sparingly soluble gases like CO2 and oxygen, the gas transfer velocity at the water–air interface depends on turbulence at the aqueous boundary layer. In lakes and in the ocean, wind is the major forcing factor generating turbulence at the air–water interface, and k is often parameterized as a function of wind speed (Wanninkhof, 1992). In streams, turbulence is mostly generated by the friction at the bottom due to water flow, and k can be parameterized as a function of water current velocity, water depth and bed roughness (O’Connor and Dobbins, 1958; Wanninkhof et al., 1990). Estuaries, and particularly macrotidal estuaries, are unique environments in which turbulence is simultaneously generated by wind forcing and boundary friction (at the surface and at the bottom) due to tidal currents (Zappa et al., 2003, 2007; Borges et al., 2004a,b). In estuaries, k versus wind speed plots show significant scatter; also, with the exception of quiescent embayments (Kremer et al., 2003a), k is generally greater than in oceans or lakes because water current also contributes to k and to its variability (Marino and Howarth, 1993; Zappa et al., 2003, 2007; Borges et al., 2004a,b). In addition, k increases faster with wind speed in large estuaries than in small estuaries due to a fetch effect, showing that k–wind speed relationships are site-specific (Kremer et al., 2003a; Borges et al., 2004a; Gue´rin et al., 2007). Borges et al. (2004a) analyzed a large
Author's personal copy
G. Abril et al. / Estuarine, Coastal and Shelf Science 83 (2009) 342–348
dataset of different estuaries and proposed that linear and sitespecific models of k versus wind speed are the most appropriate, as the Y-intercept of the equation is largely influenced by the water current and the slope of the equation is influenced by the surface area of the estuary. Thus, the establishment of k–wind speed parameterizations in estuaries requires large data sets that integrate the effects of the water current velocity and the wind speed over adequate ranges. In the present paper, we report on a large (n ¼ 159) dataset of concurrent water pCO2 and CO2 fluxes and ancillary parameters (wind, current and suspended solids) measured in the Gironde, a large, macrotidal and highly turbid estuary. This is the first time gas exchange has been documented in such a large and highly turbid system. Our dataset allows us to verify whether the previous findings in smaller systems also apply to large systems, and to investigate the potential impact of turbidity on gas exchange. The synthesis of these new and previously published data allows us to propose a new generic equation that can be applied to a large array of estuarine features to estimate gas exchange rates. 2. Methods Data were gathered in the macrotidal Gironde Estuary (SW France, Fig. 1A) during two 7-day cruises in May and November 2005, at a single station at the center of the estuary. The Gironde is a large (500 km2), macrotidal (tidal range 1.7–5.1 m) estuary, with a typical depth of 7–10 m, and characterized by high turbidity (fine clay minerals) in the water column. The total suspended solid (TSS) concentration in the area where we conducted this experiment varies between 0.05 g L1 and 2 g L1 at the surface, with a w1m-thick turbid layer (50 g L1) at the bottom. Highly concentrated benthic layers with TSS concentrations of 150–300 g L1 (fluid
343
mud) often deposit during the neap tides in this area (Abril et al., 1999; Sottolichio et al., 2001). Both field experiments covered neap and spring tides. Water depth, current velocity, temperature, salinity, turbidity, CO2 partial pressure (pCO2), air temperature and wind speed were all measured from a research vessel anchored at two points, alternating between the eastern and western channels of the Gironde Estuary (Fig. 1B). Air pCO2 and CO2 fluxes across the water–air interface were measured from a rubber boat close by the research vessel. Water depth and current were measured using an ADCP at 600 kHz. The profiler used was a bottom-tracking model mounted in a floating buoy attached 20 m from the hull of the vessel. The transducers were 40 cm below the surface and facing downwards. The instrument continuously recorded instantaneous velocities at 1 Hz frequency, covering cells of 50 cm height. Mean vertical current profiles were obtained by averaging instantaneous velocities over 10 min. The current was relatively homogenous from the water surface to 1 m above the bottom, and decreased in the last meter. We used depth-averaged velocities for the data analysis, which give a good approximation of currents forced by the tide in the absence of significant salinity stratification. Water was continuously pumped 50 cm below the surface to record the temperature, salinity and turbidity once per minute with a YSIÒ 6920 multiparameter sonde. The turbidity sensor was calibrated on eight water samples taken from the water circuit during the experiment in the range of 0–3 g L1, by comparing the sonde readings to TSS values determined by filtration onto 0.7-mm glass fiber filters (r2 ¼ 0.98, n ¼ 8). Water was also pumped to an equilibrator connected to an LI7000 LicorÒ CO2 analyzer, as described in Frankignoulle et al. (2001). Water pCO2 was recorded every minute. Air temperature and wind speed (U10 in meters per second) and direction were measured at 10 m height on a mast of the ship
Fig. 1. Map of the Gironde Estuary (A) and bathymetry of the study area (B). Black circles in (B) indicate the location of the two anchor stations in the eastern and western channels. Bathymetry (in meters) is relative to the lowest water level at spring tides under low river water runoff. (B) Shows how wind speeds were sorted by direction along the axis of the estuary: when the wind direction was in quadrant 1, it was considered against the current during the flood tide and with the current during the ebb tide; when the wind direction was in quadrant 3, it was against the current during the ebb tide and with the current during the flood tide; and when the wind direction was in quadrants 2 and 4, it was considered perpendicular to the current.
Author's personal copy
344
G. Abril et al. / Estuarine, Coastal and Shelf Science 83 (2009) 342–348
(YoungÒ instruments model 12005). For all parameters, 10-min averages concomitant to flux chamber measurements were used for data analysis. Water current and wind directions were compared by sorting the data into four 90 quadrants around the longitudinal axis of the estuary (Fig. 1B). CO2 fluxes were measured with a floating chamber, similar to the one described in Frankignoulle (1988), connected to an LI820 LicorÒ gas analyzer. Air was dried with magnesium perchlorate. Fluxes were calculated from the slope of pCO2 versus time, taking into account the effect of air temperature. Measurements lasted 5– 7 min (20–28 pCO2 data points) and were performed from a small rubber boat that was left drifting to limit creation of artificial turbulence. As discussed by Kremer et al. (2003b), a chamber that moves relative to the surface water would disrupt the aqueous boundary layer and artificially enhance gas exchange. Although in environments with low water current, but high wind, it is preferable to anchor the chamber to avoid its movement due to the wind, in macrotidal estuaries where water current can reach 2 m s1, drifting with water masses is necessary (Frankignoulle et al., 1996). Between each flux measurement, the boat was repositioned upstream of the research vessel in order to always stay at a distance of less than 500 m. Air pCO2 was obtained when the chamber system was lifted up. The floating chamber technique has been criticized by several authors (e.g., Raymond and Cole, 2001), one major criticism being that the chamber eliminates wind stress. Following the theoretical and experimental arguments of Frankignoulle (1988), Kremer et al. (2003b) and Borges et al. (2004a), as well as a recent field comparison with the eddy covariance technique in a tropical lake (Gue´rin et al., 2007), we interpret chamber flux data as within an error range of 20%, as long as measurements are short (4 m s1, k600 was 15.1 7.5 cm h1 (SD; n ¼ 11) when wind and current were in the same direction and 30.7 9.2 cm h1 (n ¼ 21) when they were in the opposite direction. When wind was opposed to current, k600 was correlated to wind speed; however, this was not the case when wind and current were in the same direction (Fig. 2A; Table 3). When data were sorted by TSS ranges (low, 0.8 g L1, n ¼ 52), we obtained three linear correlations with significantly different equations (Fig. 2C; Table 3). The slope of the regression was significantly lower at high TSS than at intermediate TSS and the Y-intercept was significantly lower at intermediate TSS than at low TSS (Table 3). In addition, when considering the entire dataset, k600 was negatively correlated to TSS concentration (Fig. 2D). When comparing k600 values within 2 m s1 wind speed intervals, k600 was significantly greater with low TSS than with intermediate or high TSS, in most situations of wind speeds >2 m s1 (Table 2). Choosing other TSS ranges significantly changed the correlation parameters and confidence intervals, but did not change the general trend that k600 significantly decreases with TSS concentration. When wind speed was less than 1 m s1, k600 was positively correlated to the water current velocity, following an exponential trend in the 50–150 cm s1 range of water current velocity (Fig. 3). Borges et al. (2004a) reported that in the microtidal Randers Fjord, k600 was 1.2 0.7 for wind speeds below 1 m s1, which is close to the Yintercept of the exponential regression in the present study (1.8 0.5 and Fig. 3A). Observed k600 values at high water current velocities were greater than those predicted by the O’Connor and Dobbins (1958) formulation. When the k600 data were ranked according to four ranges of water current velocities, we obtained four linear regressions (Fig. 2B) with non-significantly different slopes but significantly different Y-intercepts (Table 3). The Y-intercepts calculated for each interval increased with the average water current velocity, and followed the exponential equation derived from individual points at low wind speeds relatively well (Fig. 3B). Despite more dispersion and variability, the entire Gironde dataset, as well as the data grouped according to TSS concentration, also fit well with the relationship of k600 Y-intercept and current velocity. 4. Discussion
3. Results 4.1. Combined wind–current effects As summarized in Table 1, the data obtained during the two cruises cover a wide range of wind speeds (0–9 m s1), water current velocities (0.1–1.7 m s1), TSS (0.05–2.2 g L1) and k600
Zappa et al. (2007) have shown that, in shallow coastal and estuarine environments where wind forcing, water currents and
Author's personal copy
G. Abril et al. / Estuarine, Coastal and Shelf Science 83 (2009) 342–348
A
60
B
wind/current direction opposing
60 current velocity cm.s-1 0-50 80-120
with
50-80
K600 (cm.h-1)
K600 (cm.h-1)
perpendicular
40
20
>120
40
20
0
0 0
2
4
6
8
0
2
Wind speed (m.s-1)
C
345
4
6
D
60 TSS Range g.L-1
60 wind speed m.s-1
< 0.2
K600 (cm.h-1)
0.2-0.8 > 0.8
40
8
Wind speed (m.s-1)
0-2
4-6
2-4
>6
40
20
20
0
0 0
2
4
6
8
0.0
0.5
1.0
1.5
2.0
2.5
TSS (g.L-1)
Wind speed (m.s-1)
Fig. 2. (A) Relationship between k600 and wind speed at 10 m height in the Gironde Estuary, with data sorted according to wind and current directions (see Fig. 1): dark squares, wind opposing current; open squares, wind with current; crosses, wind perpendicular to current. (B) Relationship between k600 and wind speed with data sorted according to four ranges of water current velocity, from white to dark squares: 0–50 cm s1, 50–80 cm s1, 80–120 cm s1 and >120 cm s1. (C) Relationship between k600 and wind speed with data sorted according to three TSS ranges: low, 0.8 g L1 (dark). (D) Distribution of k600 as a function of TSS concentration, with data sorted according to four wind speed ranges; the line corresponds to the linear regression of the entire dataset, significant at p < 0.001).
bottom interactions all contribute to gas exchange, k is better correlated to the turbulent kinetic dissipation rate in the surface water than to the wind speed. These authors also reported that, in the Parker River Estuary, k was enhanced by a factor of 2 when water and wind flowed in the opposite direction compared to when they flowed in the same direction. In the Gironde Estuary, the enhancement of gas exchange when current and wind were in opposite directions was also close to a factor of 2 at wind speeds >4 m s1 (Fig. 2A; Table 2), which suggests that this finding can be extrapolated to other tidal systems. When data were ranked according to water current intensity instead of its direction relative to the wind, slopes of linear regressions were not significantly different, but Y-intercepts were (Fig. 2B; Table 3). This suggests, first, that the gas exchange enhancement at opposite wind/current directions and its limitation at the same wind/current directions cancel each other out when integrated over the tidal cycle and, second, that the average water current velocity affects the intercept of the k–wind regression but has a limited effect on the slope. This latter assumption is consistent with the observations of Borges et al. (2004a) in the Randers and Scheldt estuaries, and with the trend shown for the Gironde in Fig. 3B. Because Y-intercept values of linear regressions are generally driven by the data at low wind speeds, the observed differences in Y-intercepts between groups of data is due to the current velocity (Fig. 3).
In small rivers and streams, the friction of the current at the bottom is the primary process controlling gas exchange (O’Connor and Dobbins, 1958; Wanninkhof et al., 1990; Melching and Flores, 1999). As shown in Fig. 3A, the O’Connor and Dobbins (1958) Table 2 Statistical analysis (Wilcoxon signed rank test) of k600 values in 2 m s1 wind speed intervals according to wind and current direction and TSS concentrations. Asterisks indicate when k600 values are significantly different from those immediately on their left, at ***p < 0.001, **p < 0.01 or *p < 0.1. Wind speed (m s1)
0–2
2–4
4–6
>6
k600 (cm h1)
Average SD n Average SD n Average SD n Average SD n
Wind with current
13.7 3.6 6 16.2 9.5 10 14.9 7.5 8 18.9 4.8 3
Wind opposing current
5.4* 1.1 2 22.6 9.9 4 23.2** 3.4 8 35.2* 8.78 13
TSS concentration ranges (g L1) 0.8
12.3 3.1 7 29.1 7.1 12 23.7 10.4 6 40.4 8.3 10
11.1 4.3 11 11.6** 5.6 11 18.0 4.2 10 21.2** 6.3 20
9.8 5.8 12 14.7*** 5.9 15 21.4** 9.1 26 31.5** 8.5 9
Author's personal copy
346
G. Abril et al. / Estuarine, Coastal and Shelf Science 83 (2009) 342–348
Table 3 Regression functions between k600 (cm h1) and wind speed at 10 m height (m s1) for various data groups ranked according to wind and current direction, water current velocity and TSS concentration. Pairs of regression functions were compared statistically (two-tailed tests), comparing first the differences between slopes and, when slopes were not significantly different, testing the differences between Y-intercepts. Asterisks indicate when slopes or Y-intercepts were significantly different from those immediately above, at ***p < 0.001, **p < 0.01 or *p < 0.1. # indicates when slopes or Y-intercept were not significantly different (p > 0.1).
All data Data ranked according to wind/current direction Data ranked according to current velocity (cm s1)
Data ranked according to TSS concentration (g L1)
Opposing With 0–50 50–80 80–120 >120 0–0.2 0.2–0.8 >0.8
Slope
Y-intercept
r2
p
n
2.98 0.33 4.18 0.73 0.89 0.81 4.58 0.67 3.91 0.66# 3.33 0.55# 3.79 1.10# 4.04 0.72 3.55 0.43# 2.01 0.33**
8.21 1.61 6.01 4.02 12.42 3.19 2.23 3.20 4.53 3.01** 7.49 2.33** 11.92 4.78* 11.28 3.34 5.10 2.03*** 7.33 1.68
0.33 0.57 0.04 0.63 0.53 0.41 0.40 0.49 0.51 0.42
