Air-water gas exchange by waving vegetation stems
Abstract
Exchange between wetland surface water and the atmosphere is driven by a variety of motions, ranging from rainfall impact to thermal convection and animal locomotion. Here we examine the effect of wind-driven vegetation movement. Wind causes the stems of emergent vegetation to wave back and forth, stirring the water column and facilitating air-water exchange. To understand the magnitude of this effect, a gas transfer velocity (k600 value) was measured via laboratory experiments. Vegetation waving was studied in isolation by mechanically forcing a model canopy to oscillate at a range of frequencies and amplitudes matching those found in the field. The results show that stirring due to vegetation waving produces k600 values from 0.55 cm/h to 1.60 cm/h. The dependence of k600 on waving amplitude and frequency are evident from the laboratory data. These results indicate that vegetation waving has a nonnegligible effect on gas transport; thus, it can contribute to a mechanistic understanding of the fluxes underpinning biogeochemical processes.
Key Points
- Emergent vegetation is a feature unique to wetlands that couples the atmosphere and water column
- Movement of emergent vegetation stirs the surface water and thereby facilitates air-water gas exchange
- The accuracy of greenhouse gas budgets can be improved by adding this mechanism to the list of gas flux pathways
1 Introduction
1.1 Motivation
Wetlands are highly complex environments that provide a wide variety of ecosystem services, from attenuating wave action to sequestering carbon [Mitsch and Gosselink, 2007; Möller et al., 2014; Ouyang and Lee, 2014]. Wetland functions are becoming of increasing interest as sea level rise compounds with storm activity to threaten coastal areas [Knutson et al., 2010; Stocker et al., 2014]. Yet when it comes to the mechanisms by which wetlands provide these services, there are many areas in which our understanding is lacking. Here we address one of these areas, gas transport across the air-water interface.
Gas transport has both direct and indirect influences on the chemical composition of the water column. A direct effect is the diel emission of methane due to nighttime thermal stirring [Poindexter and Variano, 2013]. An indirect effect is seen in the absorption of oxygen across the air-water interface. The relationship between oxygen concentrations and methane flux has been measured in a number of environments, due to the importance of methane as a greenhouse gas. At sites in the Okavango Delta an increase in surface water-dissolved oxygen from less than 0.2 to 3.6 mg/L has been shown to accompany an order-of-magnitude decrease in diffusive methane flux, presumably due to an increase in aerobic methanotrophy [Masamba et al., 2015]. A change in oxygen levels can also lead to a change in decomposition rates, which in turn affects the rate of peat accretion and the ability of the wetland to maintain its elevation relative to sea level [Miller et al., 2008].
Herein we examine one of the mechanisms influencing gas transport. Our intent is to improve the mechanistic foundations available to predictive modeling efforts. Such improvements can contribute to a better understanding of global carbon cycling [Melton et al., 2012] as well as local gas flux. Gas transport models have the potential to replace the direct measurements of flux currently required by methods accredited by the American Carbon Registry for calculating carbon offsets [Mack et al., 2012].
1.2 Background
While gas transport has been extensively studied in open water environments, such as lakes [MacIntyre et al., 2010] and oceans [Ho et al., 2011], wetlands have unique features that prevent the direct application of other methods and findings [Happell et al., 1995]. Particularly, the presence of “emergent” vegetation greatly alters the system dynamics. Emergent vegetation is rooted underwater but emerges though the air-water interface. Physically, emergent vegetation shields the water surface from wind and sunlight and acts to couple the atmosphere and the water column. This coupling changes the influence of wind from direct shear, as expected on a lake, to damped bursts of momentum and to wind-driven vegetation movement. Previous work has explored the relative importance of these momentum bursts for water column mixing and has shown they are nonnegligible for wetland environments [Tse et al., 2016]. Here we focus on wind-driven vegetation movement and its effect on gas transport.
Wind-vegetation coupling has been studied in a diverse group of disciplines with applications ranging from seed dispersal to computer animation [de Langre, 2008]. Particular attention has been given to the “honami” phenomena, in which waves appear to roll along a canopy of vegetation. It was first documented in Inoue [1955] through observation of wind on wheat fields. Honamis form in terrestrial canopies as shear layer instabilities cause stalks to spring back and forth in a coherent manner; however, the exact generation mechanism is still debated [Finnigan, 1979, 2010; Raupach et al., 1996]. The submerged vegetation counterpart, monami, presents a simpler case due to the restricted scale of the shear layer [Ghisalberti and Nepf, 2002]. For both honami and monami, peaks in the flow velocity spectra occur at the natural frequency of the vegetation, and the frequency of the vegetation oscillations remains at the natural frequency regardless of the flow velocity [Finnigan, 1979; Ghisalberti and Nepf, 2002; Py et al., 2006; Gosselin and de Langre, 2009]. A gap in this research exists for systems with emergent vegetation; however, we work under the hypothesis that the natural frequency would continue to dominate the movement in emergent vegetation.
Numerous studies have explored the natural oscillation frequencies of a variety of vegetation types, using equations for slender rods with varying loading situations [Spatz and Speck, 2002; Brüchert et al., 2003; Speck and Spatz, 2004]. Yet accurately estimating this frequency for real vegetation has proven challenging and requires a combination of video and analytical techniques [Flesch and Grant, 1992; Doaré et al., 2004; Py et al., 2006]. Even for one species, factors such as seasonal senescence, vegetation health, and crowding cause significant variation in morphologies and material properties [Harley and Bertness, 1996; Neumeier, 2005]. For emergent vegetation, the water acts as an additional dampening mechanism, making the frequency a function of the water depth in addition to the vegetation material properties. Due to this complexity and the desire to have generic results that encompass a range of motions, we select a range of frequencies and amplitudes to study, measuring the dependence of gas transfer on both.
2 Methods
2.1 Thin Film Model
(1)
(2)
(3)
(4)
(5)
(6)
The alpha value in equation 6 is dependent on the surface conditions of the interface. It has been found to range from one half for a “clean” surface to two thirds for a no-slip boundary; however, it has been shown that even a low level of surface contamination causes a surface to act similarly to a no-slip boundary in regard to inhibiting gas transport [McKenna and McGillis, 2004]. As wetlands contain a high level of surfactants [Kadlec and Wallace, 2008], the two-thirds exponent seems appropriate.
2.2 Experimental Design
By working in a laboratory, we were able to isolate the mechanism of wind-driven vegetation movement and produce results that are not site specific. While honami motion does not occur in isolation in real wetlands, we isolated it here to judge the relative contribution of this mechanism to the total diffusive flux.
Wind-driven vegetation movement was recreated in a laboratory tank (Figure 1). Plastic tubes acting as emergent vegetation were anchored at the top and bottom in two plates that were separated by 65 cm; the bottom plate was secured to the bottom of the tank, while the top plate sat on two rollers and was oscillated horizontally in one dimension. These tubes have dimensions similar to Schoenoplectus acutus, a common California wetland species known as Tule, with a diameter of 13 mm; they were randomly spaced to give a vegetation density of 2.16 m− 1, which is within the range of naturally grown Tule [Gardner et al., 2001; Miller and Fujii, 2009]. Since we were interested in how the stem acts as a stirring rod at the air-water interface and not in biological interactions, it was not necessary for our laboratory setup to use real vegetation.
A motor was used to move the top plate and therefore the synthetic vegetation back and forth at a set frequency and amplitude. The natural frequency of Tule was measured through an in situ video of a single live stem being plucked and was determined to be approximately 1 Hz. Past measurements of terrestrial alfalfa have shown a range of 0.8 to 1.5 Hz and 1.29 to 1.80 Hz for corn [Flesch and Grant, 1992; Py et al., 2006]. A range of stem-waving frequencies was tested in the laboratory to cover a range of natural conditions. Since the water acts to dampen the motion and lower the frequency [Chen et al., 1976], we tested a range of 0.3–1.2 Hz.
For the frequencies considered, there were three amplitude scenarios: 0.5 cm (small), 0.8 cm (medium), and 1.3 cm (large). These values encompass a range observed from our in situ video and qualitative observations of vegetation waving at a number of wetland sites with a variety of wetland vegetation. The flow in all experiments was laminar, remaining under the Reynolds number, 
, threshold for turbulent wake structures within an array of cylindrical vegetation [Nepf, 1999].
A wave-dampening buffer was added to the perimeter of the study area at the air-water interface. This buffer was meant to prevent the formation of an in-tank seiche and to minimize tank boundary effects. Data collected with and without the buffer are presented.
2.3 Experimental Procedure
Each experiment began by deoxygenating the water using sodium sulfite; cobalt chloride was used as a catalyst. After the reaction was completed, 5 cm were skimmed off the surface to give a final water depth of 25 cm. Skimming this top layer allowed for experimental repeatability but was not expected to produce a surfactant-free surface. Side panels prevented lateral transport to the unvegetated sections of the tank, creating a study area 33 cm wide and 48 cm long.
The frequency for each experiment was determined from videos taken at the beginning and end using ImageJ (available at https://imagej.nih.gov/ij/docs/index.html). The frequency increased over the course of the experiment, which we ascribe to the warming of the motor. The difference between the beginning and ending frequency was kept as small as possible by cooling the motor with an external fan.
As the vegetation moved, oxygen diffused into the water, reequilibrating with the atmosphere. A time series of the dissolved oxygen concentration at the middle of the water column was then recorded using an optical probe (YSI ProODO). An example time series is shown in Figure 2. It was important to use an optical probe, as those which require strong stirring near a membrane (i.e., Clark-type probes) may not have received enough stirring in this flow. From the dissolved oxygen time series, we used the thin film model of gas transport to determine the gas transfer velocity, via equation 3.
Control experiments with no oscillations were also performed in the experimental setup. With no motion, the gas transfer is driven by ambient thermal convection, which is slow and variable. The greatest value of k600 from the no-motion experiments was used as a comparison.
2.4 Statistical Analysis
Each dissolved oxygen time series was divided into subsets, where each subset represents an increase in dissolved oxygen of 0.02 mg/L. This gives n subsets per experiment, where n is always greater than 92. A k600 value was computed for each subset; k600 for the experiment was set as the median across n. The 95% confidence interval was found via bootstrap by resampling from n [Efron and Tibshirani, 1994].
3 Results
Gas transfer velocity is enhanced by the movement of emergent stems. With no motion, the greatest k600 in our experimental setup was 0.4 cm/h, which is the lower bound in Figure 3. In Figure 3, the vertical error bars show the 95% confidence interval, calculated as described in section 6. The horizontal error bars reflect the difference in frequency from the beginning to end of the experiment. In all experiments, the frequency at the end was higher than at the beginning. The choice of the two-thirds exponent, rather than one half discussed in section 4, caused an average decrease of 4.3 ± 1.9% in the resulting k600.
The closed symbols indicate experiments where wave-dampening buffer was placed at the edges of the tank at the air-water interface. The presence of this buffer reduces the tank boundary effects on k600. Linear regression on k600 versus frequency for the small-amplitude case shows that the slopes are the same with and without the buffer but intercepts differ, indicating an amplification of k600 by 0.15 cm/h when the buffer is absent. The full data set is presented and indicates when the buffer setup was used.
For stem-waving frequencies below 0.5 Hz, the results for all amplitude cases are clustered together. Beyond this frequency, the medium- and large-amplitude cases show an increase in gas transfer velocity reaching a maximum of 1.6 cm/h. The small-amplitude case shows little dependence on frequency, remaining less than 1 cm/h at a frequency of 1.25 Hz.
Our results show that for small-amplitude motions, the stem-waving frequency does not greatly alter the gas transfer velocity. At larger amplitudes, k grows with frequency in a manner that seems linear, or slightly more rapid than linear.
3.1 Dimensional Analysis
(7)
(8)
(9)
(10)
(11)
The data confirm the result of equation 8, as seen in Figure 4.
(12)
Here we have confined the analysis to only include data from experiments with the tank buffer. This result highlights the importance of both stem-waving frequency and amplitude. Their product (fa) is the maximum velocity of the vegetation stem as it oscillates. When either the amplitude or frequency is sufficiently large, the gas transfer velocity appears to be linearly proportional to fa.
4 Discussion
Having quantified the effect of stirring by vegetation stems, we can put this mechanism into context with the other drivers of wetland air-water exchange. These drivers include rain, water convection due to differential heating, and wind directly shearing the surface. The latter two drivers have been quantified by Poindexter and Variano [2013], giving k600 values between 0.1 and 4 cm/h depending on conditions. Even with a k600 of 0.6 cm/h, which is at the lower end of this range, Poindexter showed that air-water exchange is an important pathway in the wetland biogeochemical budget. Since vegetation-waving gives k values within the same range considered by Poindexter, we can conclude that it has a nonnegligible effect on wetlands as well.
This effect of waving on the gas transfer velocity is much smaller than the effect of wind would be in the absence of vegetation. The canopy greatly reduces wind speed [Tse et al., 2016], and the small amount of wind shear acting on the surface drives a typical gas transfer velocity around 0.1 cm/h [Poindexter and Variano, 2013]. This value is negligible compared to the gas transfer velocities observed in open waters, which are certainly above 1 cm/h and typically on the order of 10 cm/h [Crusius and Wanninkhof, 2003]. The effect of rain has been quantified by Ho et al. [2004]; taking measurements in a model ocean, results showed that even with ocean-like forcing the presence of rain can increase k600 from 11.2 to 49.7 cm/h. Although these are strikingly different conditions, rain is likely to also be significant in wetlands.
When considering these different drivers, the timing and consistency of the events become important. Rain may be the dominant driver during rain events. The tendency of vegetation waving to be intermittent in time lowers its importance relative to thermal convection, which is driven consistently by the day-night cycle, but the fact that vegetation waving can occur everywhere in a dense wetland makes it more important than direct wind shear, which is greatly weakened away from canopy edges.
Two open questions remain before the effect of vegetation waving on wetland biogeochemistry can be fully described. The first is an understanding of how the different drivers acting on the air-water interface combine with each other. The second is a thorough survey of the frequency and amplitude of the waving and how it varies with wind conditions, canopy density, plant type, and plant life stage. This first question arises from the fact that these drivers of motion naturally occur simultaneously along with additional mechanisms such as animal motion and seiches, whose effect on gas transfer velocity in wetlands is unknown. Until we understand how these individual k values combine, the relative importance of each phenomena is uncertain. That is, there may be nonlinear interactions in which two stirring phenomena amplify or attenuate each other. At the moment, we can only compare the magnitude of different forcings in isolation and suggest that the strongest forcings be included in predictive models for wetland biogeochemical budgets. Our results can also be used to estimate the bias incurred by flux measurements that use static chambers or other methods which block stirring by wind.
To answer the second question, tools for monitoring vegetation motion are needed. We have explored several options and determined that today's accelerometers are insufficient for recording the motion of a single stem. This is because they are either too heavy or not sensitive enough—existing field data suggest that stems' acceleration is on the order of 0.01 m/s2 (or 1 μG), and accelerometers with such fine resolution are currently quite large. Cameras can monitor motion to very fine resolution, as long as an optical path can be found through the canopy. One planar view must be selected because stem motions are too small for accurate stereoscopic 3-D imaging. We think the best camera configuration is a single camera pointed vertically, monitoring the horizontal motion. Different plant stems, or different locations on a single stem, can be monitored simultaneously in a single image. If the monitoring points are clearly labeled, they can be identified by imaging processing in real time, thus greatly reducing the data storage burden that usually accompanies field deployment of cameras. Such data management would be essential, as the intermittent nature of vegetation waving demands continuous monitoring over a long time period. This monitoring would reveal the probability distribution of vegetation-waving frequency and amplitude. It could also indicate the dominant causes of vegetation waving in a wetland. That is, waving will likely be different when generated by gusts in the above-canopy wind field than when it is generated by shear at the canopy top. A range of sites would need to be studied to avoid site-specific results. Factors to consider include but are not limited to the vegetation density, spatial variation of vegetation types, surrounding structures (influencing wind patterns), time of year, and weather conditions. Given these considerations, we conclude that thoroughly measuring vegetation waving in the environment would be a lengthy process, so we must ask whether it is worthwhile.
Herein, we have focused on freshwater wetlands due to their greater production of methane relative to saltmarsh species. Saltmarsh vegetation will likely have different behavior than the Tule measured here, given their different morphology. Specifically, saltmarsh vegetation typically has smaller diameters and greater flexibility than Tule. In terms of flexibility, the maximum stress withstood by stems is an order of magnitude greater for Tule [Groeneveld and French, 1995], than for Spartina alterniflora and Salicornia europaea [Harley and Bertness, 1996], two common saltmarsh species. In terms of size, the diameter of the Tule modeled here was 13 mm, which is larger than the average of 4.3 mm and 3.6 mm determined for healthy and unhealthy Spartina alterniflora, respectively [Feagin et al., 2011]. With smaller stem diameter, the Reynolds number decreases, and motion of vegetation stems is likely to cause a smaller k600 as compared to larger stems. However, other linked factors may alter this effect in the opposite way; for example, there may be increased vegetation density with decreased stem diameter.
5 Conclusions
We have measured the stirring at an air-water interface caused by waving vegetation stems, inspired by the wind-driven waving that occurs in wetlands. The stirring is quantified in terms of a gas transfer velocity, k600. Results suggest that over a wide range of amplitudes and frequencies, vegetation-waving can contribute significantly to vertical fluxes across the air-water interface. These results underscore the value of performing a thorough field survey of vegetation waving over space, time, and vegetation characteristics. These results also suggest that the waving of vegetation should be considered when investigating nonlinear interactions of different phenomena which simultaneously act to stir the air-water interface.
Acknowledgments
The authors would like to thank students Justine Zeni, Samuel Pennypacker, Bobby Kheny, and Seandre Chaw for their assistance in running experiments; Michelle Hummel for research on fieldwork techniques; and Rachel Allen and William Tarpeh for invaluable feedback. This research was made possible in part by a grant from The Gulf of Mexico Research Initiative (SA12-09/GoMRI-006) and by a National Science Foundation Graduate Research Fellowship (DGE-1106400). Two anonymous reviewers provided insightful feedback and inspired additional analysis. Data are stored as time series of dissolved oxygen and MATLAB code for processing it. Data are available by request from MRFM ([email protected]).
