Article pubs.acs.org/est
Size Does Matter: Importance of Large Bubbles and Small-Scale Hot Spots for Methane Transport T. DelSontro,*,†,⊥ D. F. McGinnis,‡,§ B. Wehrli,† and I. Ostrovsky∥ †
Eawag, Swiss Federal Institute of Aquatic Science and Technology, 6047 Kastanienbaum, Switzerland & Institute for Biogeochemistry and Pollutant Dynamics, ETH Zurich, 8092 Zurich, Switzerland ‡ GEOMAR, Helmholtz Centre for Ocean Research Kiel, Marine Geosystems, 24148 Kiel, Germany § Institute F.-A. Forel, Earth and Environmental Sciences, Faculty of Sciences, University of Geneva, 1227 Geneva, Switzerland ∥ Israel Oceanographic & Limnological Research, Yigal Allon Kinneret Limnological Laboratory, Migdal 14950, Israel S Supporting Information *
ABSTRACT: Ebullition (bubbling) is an important mechanism for the transfer of methane (CH4) from shallow waters to the atmosphere. Because of their stochastic nature, however, ebullition fluxes are difficult to accurately resolve. Hydroacoustic surveys have the potential to significantly improve the spatiotemporal observation of emission fluxes, but knowledge of bubble size distribution is also necessary to accurately assess local, regional, and global water body CH4 emission estimates. Therefore, we explore the importance of bubble size and smallscale flux variability on CH4 transport in and emissions from a reservoir with a bubble-size-calibrated echosounder that can efficiently and economically survey greater areas while still resolving individual bubbles. Using a postprocessing method that resolves bubble density, we found that the largest 10% of the >6700 observed bubbles were responsible for more than 65% of the total CH4 transport. Furthermore, the asymmetry of CH4 ebullition flux distribution and the high spatial heterogeneity of those fluxes suggests that inadvertently omitting emission hot spots (i.e., areas of high flux) could lead to significant underestimations of CH4 emissions from localized areas and potentially from entire water bodies. While the bubble sizes resolved by the hydroacoustic method may provide insight into the factors controlling ebullition (e.g., sediment type, carbon sedimentation), the better resolution of small-scale CH4 emission hot spots afforded by hydroacoustics will bring us closer to the true CH4 emission estimates from all shallow waters, be them lakes, reservoirs, or coastal oceans and seas.
■
INTRODUCTION
The important role of inland water bodies in the carbon cycle is now widely accepted; however, many aspects of aquatic carbon dynamics, such as CH4 production, transport, and emission, are not fully understood.12,13 In fact, the literature on CH4 emissions from inland waters remains discordant, partly due to insufficient measurement methods and partly due to insufficient spatial and temporal coverages. These shortcomings, exasperated by the stochastic nature of ebullition, hinders the accurate quantification of this potentially dominant emission pathway in shallow aquatic systems.2,3,14 Thus, the result is a disjointed understanding of ebullition dynamics in all surface waters, along with a wide range of flux estimates. Ebullition intermittency has been the easiest to demonstrate in high latitude lakes during ice cover.14−16 While the cause for the so-described “patchiness” of ebullition in these particular
Ebullition, the transport of gas via rising bubbles, is the most efficient vertical transport pathway for methane (CH4) in lakes and reservoirs.1 The spatial and temporal variability of CH4 ebullition, however, impedes the accurate quantification of emission rates of this potent greenhouse gas. Measuring ebullition fluxes with floating chambers2 and submerged gas traps3−6 cannot fully resolve the spatiotemporal variability of ebullition because of the number of chambers/traps and deployment times that would be required for adequate temporal and spatial coverage. Recent advances in hydroacoustic processing,7−10 however, overcome some of these limitations and can provide valuable information on both the spatial and temporal heterogeneity of CH4 fluxes and individual bubble sizes.9−11 Ultimately, hydroacoustic surveys deliver small-scale fluxes that not only improve spatial resolution but also offer a more robust understanding of ebullition dynamics by helping elucidate the causes and effects of flux variability. © 2014 American Chemical Society
Received: December 29, 2012 Accepted: December 31, 2014 Published: December 31, 2014 1268
DOI: 10.1021/es5054286 Environ. Sci. Technol. 2015, 49, 1268−1276
Article
Environmental Science & Technology
in a Swiss hydropower reservoir known for high bubbling rates. We detail the hydroacoustic postprocessing method and compare our results to other reported values. Specifically, we investigate how bubble size and other parameters, such as bubble density and water column depth, relate to CH4 fluxes. Finally, we discuss how bubble size and CH4 flux intermittency impacts surface CH4 emission estimates from this reservoir, thus highlighting the importance of large bubbles and localized flux hot spots for CH4 emission estimates from aquatic systems.
lakes may be due to their local environments (i.e., thermokarst or peatland lakes), the fact remains that methods employed on ice-covered lakes are currently the best at constraining the spatial variability of ebullition. In ice-free lakes, chambers and submerged gas traps are the current methods of choice for quantifying ebullitive emissions.2,3,17,18 However, due to their (very) limited coverages, it is difficult with these methods to provide integrated emission assessments that accurately represent the bulk emissions from the water body.3 Missing ebullition hot spots due to insufficient sampling will result in significantly underestimated CH4 fluxes.14,19,20 The hydroacoustic method, e.g., using an echosounder, is a technique that can help overcome these limitations by providing a more accurate ebullition survey, particularly in an effort to locate ebullition hot spots. Although the instrumentation can be more expensive than chambers or gas traps, hydroacoustic methods have been widely used in both aquatic and marine environments. As with other geophysical instrumentation, a single echosounder is sufficient to survey large areas efficiently and can produce highly spatially resolved maps of ebullition activity (and bathymetry).9,20 In addition, high-resolution ebullition time series can be acquired from a moored instrument.7 Scientific echosounders have even progressed to the point of resolving single bubbles, their trajectories, and their rise velocities, which enables the quantification of ebullition flux in relatively shallow (100 m), sediment-released methane bubbles will dissolve before reaching the surface. In shallow waters ( 10 mm or TS > −46 dB) were observed in Lake Wohlen, but as the frequency distribution clearly demonstrates, bubbles larger than −50 dB account for the majority of the gas volume from all surveys (black line, Figure 1a). The cumulative curves for number of bubbles and their volume contribution according to diameter (Figure 1b) shows that >90% of all observed bubbles are less than 10 mm in diameter but only account for ∼35% of the total volume of gas contained within all observed bubbles (dashed line, Figure 1b). This means that the largest 10% of bubble sizes
■
RESULTS AND DISCUSSION The echosounder bubble size calibration results were justifiable and correlated well with previous similar bubble size calibrations (discussed in detail in the Supporting Information); thus, echosounder data were confidently used for ebullition size and flux analysis. The BNS method was compared to two other hydroacoustic postprocessing approaches and shown to be the most reliable postprocessing technique for our conditions (see the Supporting Information for more details). Bubble size distribution, bubble density, and bubble CH4 flux were thus calculated in a total of 556 different segments with an average length of ∼46 m resulting in a spatial analysis of CH4 emissions during seven campaigns. The following sections will use the segment results to describe and discuss the variability in both bubble size and flux within the water column, their relationship to each other and other parameters, and last, how an inadequate sampling resolution of that variability could impact overall CH4 emission estimates. 1270
DOI: 10.1021/es5054286 Environ. Sci. Technol. 2015, 49, 1268−1276
Article
Environmental Science & Technology
Figure 2. (a) Sauter mean diameter (SMD; top axis) of the bubble population per segment correlates better with sediment CH4 ebullition flux (logtransformed) than the weighted mean diameter (WMD; top axis), and both impact flux more than bubble density (BD; bottom axis). (b) Sediment CH4 ebullition flux was not negatively correlated with segment depth as seen in the boxplots (horizontal line is median, red dot is average, whiskers are 25th and 75th percentiles, dots are outliers). (c) There was a positive correlation between weighted mean diameter (WMD) of the bubble population of each segment with the maximum depth of the segment. Red dots are the average per meter depth bin (e.g., 3 m = 2.5−3.5 m), and the lines are the standard deviations. Gray shading shows smoothed maximum and minimum WMD range per depth and dark gray dots are the data.
contributes ∼65% of the total volume of bubble gas. These findings are similar to those of Greinert and Nützel,11 who found that over 50% of the total gas volume in their experiments was from only 7% of the bubbles (>14 mm diameter). Here, we show that even when the bubble population is completely dominated by small bubbles the total volume is supplied by only a few large bubbles, rendering the smallest bubbles almost insignificant for flux calculations. Ebullition Flux Variability. Employing the BNS method with our calibrated echosounder data, we find an average ebullition gas flux of ∼1200 mg m−2 d−1 for all segments. Considering a primary CH4 gas composition in the bubbles of 70 ± 5%,17 the potentially conservative estimate of CH4 ebullition flux from the Lake Wohlen study area is ∼820 mg CH4 m−2 d−1. The fluxes calculated represent the flux of CH4 from the sediment into the water column only. Maximum fluxes from the sediments were in the 103 mg CH4 m−2 d−1 range on all days, except for 1 day which had several segments in the 104 mg CH4 m−2 d−1 flux range (Figure 1c). The high variability across segments, which ranged over 4 orders of magnitude, was expected as transects were subdivided based on bubble and nonbubble densities that were also variable. Zero flux segments were observed on each survey day, but of the 556 total segments analyzed, only 11% of them contained no detectable ebullition. It should be stressed that for the purposes of this study our hydroacoustic surveys were intentionally conducted in a region known for active ebullition. Our ebullition flux estimates compare well with ebullition estimates found with similar and different methods. Using an echosounder, Ostrovsky et al.9 found slightly lower fluxes with a maximum of ∼700 mg CH4 m−2 d−1 in a natural lake of ∼40 m water depth. A few other studies that have used some type of automated system (at the surface or submerged) to directly measure bubble fluxes found ebullition rates similar to those found in our study, including the highest fluxes (>104 mg CH4 m−2 d−1). Varadharajan et al.6 used submerged automated traps in depths up to 25 m in a eutrophic lake of the Northeastern United States and found that the majority of bubble fluxes
(measured at a resolution of 5−10 min) were around 102 mg CH4 m−2 d−1, but events as high as 1.5 × 103 mg CH4 m−2 d−1 were observed. In a Brazilian reservoir, Ramos et al.31 used a surface trap sampling every 5−7.5 min and observed that the majority of fluxes were below 103 mg CH4 m−2 d−1, but several events between 5 × 103 and 1.5 × 104 mg CH4 m−2 d−1 as well as two bubble plumes greater than 2 × 104 mg CH4 m−2 d−1 were recorded. Using submerged automated traps in an impounded river of central Europe, Maeck et al.32 not only showed that ebullition rates were variable and often high, but that flux results varied according to the sampling resolution. In an area of the river known for active bubbling, daily ebullition rates ranged up to ∼4 × 103 mg CH4 m−2 d−1, while hourly and 5 min rates were as high as 104 and >3 × 104 mg CH4 m−2 d−1, respectively. Lake Wohlen is a cross between an impounded river and a reservoir and, unsurprisingly, emits CH4 via ebullition at rates closer to those found in other impounded rivers or reservoirs than those found in natural water bodies. The automated trap methods mentioned above sample the same spot at a high temporal resolution. While our sampling resolution did not consist of multiple high-resolution measurements at one location, our temporal resolution was still high (i.e., short duration) as we were recording from a moving boat (∼1 m s−1). We thus covered a larger area over the course of a day by recording ebullition at each location only once for a short duration. Ultimately, both the high-temporal resolution bubble traps and the high-spatial and short duration echosounding technique seem to converge at the high end of CH4 bubble emission. Relationship between Bubbles and Ebullition Flux. Bubble Density and Size. The two bubble parameters that contribute to the total CH4 flux calculated in a segment are bubble density and the bubble size distribution. Bubble density was almost uncorrelated to flux (r2 = 0.12; triangles, Figure 2a), while a strong relationship existed between bubble size and flux of each segment (r2 = 0.46; black dots, Figure 2a) when the bubble size distribution is represented by the weighted mean 1271
DOI: 10.1021/es5054286 Environ. Sci. Technol. 2015, 49, 1268−1276
Article
Environmental Science & Technology
Figure 3. Four surveys from July 2008 illustrating the weighted mean diameter (WMD) of the bubble population in each segment (represented by the size of the dot) and the sediment ebullition CH4 flux measured in each segment (represented by the color of the dot). Crosses are segments with no observed ebullition. Contours below dots represent depth with darker shades being shallower (see Figure S1 in the Supporting Information for true depths).
the middle range of observed depths (∼5 to 9 m, Figure 2b). We also did not find a negative relationship between depth and ebullition flux to the atmosphere (data not shown). Similarly, there was no apparent relationship between the SMD of each segment and depth (Figure S5, Supporting Information), which would be expected considering that the SMD better reflects the volume of gas present in the segment and thereby ebullition flux. Clearly, the commonly observed relationship between ebullition flux and depth was not present in our study area. The negative depth−flux relationship may not have been as obvious in our study area simply because significant ebullition occurred at all depths; therefore, in order to see a relationship between depth and atmospheric flux the bubble size distribution would have had to be constant with depth so that the bubbles would exchange with the ambient water at rates respective to their rise time and height from the bottom.9,23 We found, however, that the WMD of bubbles (1-m bin averages, red dots, Figure 2c), which reflects the true size distribution of bubbles within a segment, was significantly positively related to maximum depth of the segment (r2 = 0.71, p < 0.05, Figure 2c). Bubble size distribution was thus not constant with depth, and instead sediments from deeper segments generally emitted larger bubbles. Buoyancy theory states that in order for an object to float (i.e., a bubble to rise) the weight of the object must be less than the weight of the displaced volume in which the object lies. As objects, including bubbles, become heavier with depth and higher pressure, bubble volumes must increase in order to counter balance the mass increase and be able to rise. Of course, sediment properties such as cohesiveness also play a role in bubble release, but perhaps in a more unexpected manner as existing fractures may allow for an easier bubble transit through the sediments.35 Ultimately, the increase in WMD with depth that we observed in Lake Wohlen falls within the range of increasing bubble volume with increasing pressure, which in turn likely impacted the observed depth-flux relationship or lack thereof in Lake Wohlen. Since the SMD takes into account the disproportionate volume contribution from larger bubbles,
diameter (WMD) for each segment (WMD segment distribution seen in Figure S4a, Supporting Information). These results imply that it is the size distribution of the bubbles rather than the number of bubbles that are important in terms of overall flux. The significance of bubble size for flux estimates was expected since gas volume increases with bubble diameter and, as already shown, the few large bubbles observed contributed most heavily to the total gas volume (Figure 1a,b). To further support this notion, we found that the relationship between size and flux improves if the Sauter mean diameter (SMD) is calculated instead of WMD (r2 = 0.52; white dots, Figure 2a). In fields of research such as fluid dynamics where the surface area-to-volume ratio (SA:V) is of importance,33 which is the case for spherical gas bubbles, the SMD is often used to represent size distribution instead of an arithmetic mean diameter in order to better incorporate the influence of a relatively low number of large particles (in our case, bubbles). Here, the SMD calculates the diameter of a spherical bubble that would have the same SA:V ratio as the distribution of measured bubbles by giving more weight to the larger ones using the following equation n
SMD =
∑1 d3 n
∑1 d 2
(1)
where d is diameter for each bubble present in the population (see Figure S4b, Supporting Information, for SMD segment distribution). The SMD, however, did not improve the bubble size to flux relationship as much as expected; therefore, additional environmental conditions must also impact ebullition flux from the sediments of Lake Wohlen. Water Column Depth. A common observance in some systems is a negative correlation between water column depth and ebullition flux.2,34 We, however, did not find this in our study area. Rather, we see a very slight peak distribution in which, when sediment ebullition fluxes are binned according to depth (e.g., 3 m bin contains all fluxes from segments with maximum depths between 2.5 and 3.5 m), fluxes are highest in 1272
DOI: 10.1021/es5054286 Environ. Sci. Technol. 2015, 49, 1268−1276
Article
Environmental Science & Technology
Figure 4. (a) Contour plot shows what fraction of the original CH4 will remain in a bubble of a certain initial diameter at the air−water interface when released from a certain depth. For example, a 16 mm bubble would still contain ∼90% of its original CH4 when released from 11 m depth. Inset shows the percent of CH4 reaching the surface of all bubble sizes when released from 10 m depth (∼average depth of Lake Wohlen). (b) Bar plots show the average CH4 sediment ebullition flux (top of black bar) and surface ebullition flux (top of gray bar) when calculated from all fluxes, those less than 104 mg CH4 m−2 d−1, etc. Numbers in bars are amounts of emitted CH4 in tons per day when the corresponding averages are extrapolated to the entire study area (∼0.13 km2).
supersaturation in the sediments;37 therefore, areas of intense methanogenesis are likely to be ebullition hot spots. In order for intense methanogenesis to occur, a significant amount of labile carbon substrate must be present for methanogens to process, which means that carbon sedimentation rates should also be sufficiently high in those areas. Areas of high sedimentation have been shown to be hot spots of ebullition in a larger reservoir20 and an impounded river,19 and high sedimentation rates have already been linked to the high rates of methanogenesis and bubbling in Lake Wohlen.38 Gas volume in sediments as well as ebullition rates have also been shown to relate to sediment characteristics in other lakes.34 The different lines of evidence suggest that organic matter sedimentation rates and sediment properties may drive ebullition rates at quite a small scale, as the immense flux and bubble size variability in this small, ∼0.13 km2 hot spot suggests. Impact of Bubble Size and Flux Variability on Surface CH4 Emissions. Bubble Size. As bubbles rise through the water column, gas exchange occurs between the bubble and the ambient water while the decreasing hydrostatic pressure expands the size of the bubble.23 The hydroacoustic method presented here estimates ebullition flux only from the sediment to the water column; therefore, a discrete bubble dissolution model23 is needed to estimate gas exchange during bubble rise and consequent flux to the atmosphere,9,17 which primarily depends on bubble size and release depth (i.e., depth of the water column). Figure 4a shows the percent of CH4 remaining in a bubble when it reaches the air−water interface based on the release depth (as deep as 12 m in Lake Wohlen) and initial diameter (up to 30 mm in Lake Wohlen) of the rising bubble. As a first estimate, the average bubble diameter (∼6 mm) and average depth (∼10 m) for Lake Wohlen can be used to find that ∼67% of the CH4 contained in Lake Wohlen bubbles will reach the atmosphere. Applying this to the overall average CH4 flux from Lake Wohlen segments (820 mg CH4 m−2 d−1), we find that ∼550 mg CH4 m−2 d−1 could have been released from this study area to the atmosphere at that time. For a more accurate estimate, we calculate bubble dissolution (and hence the fraction of CH4 reaching the atmosphere) per bubble size class (in our case, every 2 dB from −70 to −30 dB for a total of
the lack of relationship between SMD and depth means that larger bubbles are ubiquitous and contribute significantly more to the total flux than is accurately represented by their absolute number or the WMD. In other words, the smaller WMD in shallower depths (Figure 2b) does not adequately reflect the amount of gas contained within the total bubble size distribution because, despite the higher numbers of small bubbles lowering the WMD, the few larger bubbles present still transport most of the gas in those areas. Additional Controls on Ebullition Flux. We have seen that while bubble size does tend to relate to depth in Lake Wohlen, ebullition flux does not exhibit similar trends; therefore, other factors besides bubble size distribution must influence flux. Figure 3 and Figure S6 (Supporting Information) illustrate bubble WMD and sediment ebullition flux spatially on each survey day, which enables the investigations of day-to-day variability as well as any spatial trends. The relationship between size and flux as previously discussed is also seen spatially; in general, higher CH4 flux segments (warmer colors, Figure 3 and Figure S6, Supporting Information) tend to have larger WMDs (size of circles, Figures 3 and Figure S6, Supporting Information) and lower CH4 flux segments (cooler colors) tend to have smaller WMDs (i.e., smaller circles). One exception is July 23 in which the majority of segments had high ebullition fluxes but with WMDs not quite as large as high flux segments on other days. A likely explanation for this is that it was particularly windy on that survey day and perhaps some hydrological conditions, such as wind-induced bottom shear or currents,5,36 enhanced ebullition fluxes mechanically and thus without increasing bubble sizes. A spatial trend was also present in our study area on most of the seven survey days. Low fluxes were typically found in the shallow region near the south shore of the field area as well as in the deepest region to the north that represents the old river channel. Higher fluxes were typically found on the slope between these two regions in depths between 5 and 9 m. This trend was also apparent in the peak distribution of flux versus depth previously described (Figure 2b) and together these results suggest a geomorphic control on CH4 ebullition fluxes. The most necessary constraint for CH4 ebullition is a CH4 1273
DOI: 10.1021/es5054286 Environ. Sci. Technol. 2015, 49, 1268−1276
Article
Environmental Science & Technology 20 bins) per segment in which fluxes were calculated. The result is in fact a higher average surface ebullition flux estimate (∼720 mg CH4 m−2 d−1) that is ∼88% of the total sediment ebullition flux. Bubble dissolution is primarily based on bubble release depth and diameter; therefore the higher surface flux estimate from the per bin per segment calculation is due partially to the variability in release depth among segments but more so to the greater variability in bubble sizes present in each segment. As bubble diameter increases, the SA:V of bubbles decreases (Figure S7, Supporting Information). This implies that bubbles with larger diameters are less efficient at exchanging gases between the bubble interior and surrounding water, and that large bubbles rise faster which permits less contact time with the ambient water.9,23,33,39 The importance of the SA:V to bubble dissolution suggests that the SMD is a more appropriate average bubble size value to use in a bubble gas exchange model for a rough surface flux estimate. The SMD for all 6700+ bubbles observed in Lake Wohlen is 10.1 mm, which is nearly double the WMD (5.9 mm), and reflects 5 times more volume (0.11 mL versus 0.54 mL, assuming a perfect sphere). The exchange model shows that the SMD-sized bubble would release ∼81% of its CH4 to the atmosphere (Figure 4a) and thus ∼665 mg CH4 m−2 d−1 would be emitted from our study area. Using the SMD to estimate surface CH4 ebullition flux provides an estimate much closer to the true flux than using the WMD; therefore the SMD could be used for a fairly accurate first flux estimate as previously shown by McGinnis and Little39 and also indicates that the few large bubbles present in Lake Wohlen are disproportionately more important for surface emissions than the large number of small bubbles. As stated earlier, while bubbles up to 10 mm in diameter make up ∼90% of the total population observed in Lake Wohlen, they supply less than 35% to the total CH4 present in the water column (Figure 1b), and are thus responsible for even less of the CH4 emitted to the atmosphere. From the average depth of Lake Wohlen (10 m), bubbles smaller than ∼10 mm in diameter exchange CH4 more rapidly with the ambient water, therefore emitting much less CH4 to the atmosphere. From 10 mm in diameter and greater, the rate at which bubbles dissolve and exchange gases slows remarkably, thereby naturally presenting an important size threshold to distinguish between small and larger bubbles (Figure 4a, inset). Not only do large bubbles carry exponentially more CH4 with increasing diameter (Figure 1b) but they also dissolve exponentially less during ascent (Figure 4a). Ultimately, larger bubbles are more responsible for atmospheric CH4 release in shallow waters and missing or excluding some of them during a survey could underestimate final CH4 flux results. Flux Variability. The hydroacoustic method provides a range of observed (and possible) sediment CH4 ebullition fluxes when estimated per segment (Figure 1c). When coupled with the dissolution model, atmospheric CH4 fluxes can also be estimated, but the question is how to most accurately upscale emissions to an entire study area, lake, or inundated region. Previous studies have observed a negative relationship between depth and ebullition flux that was subsequently used to upscale surface emissions to an entire water body.2,34 Since no depth− flux relationship was observed in our study area of Lake Wohlen, we could only provide a first rough estimate of flux simply by applying the average of our measurements to the entire study area in question, a method previously used by others with floating chambers40 and submerged traps.4
If the average sediment ebullition flux from the Lake Wohlen hydroacoustic transects (820 mg CH4 m−2 d−1) was applied to the whole study area (only ∼0.13 km2 of the 2.5 km2 lake), we would find a total of 0.108 t CH4 d−1 emitted from the sediments. The associated surface emission estimate (calculated per size bin per segment) would be ∼0.096 t CH4 d−1 (Figure 4b). If, however, the flux variability was not spatially resolved to the extent we have accomplished in our study and some disproportionately important areas of CH4 flux were missed, then the resulting emission estimate could be impacted. For example, missing fluxes greater than 104 mg CH4 m−2 d−1, which are rare and likely easy to miss, would underestimate the atmospheric CH4 emission estimate but only by a small amount (∼0.007 t CH4 d−1 or 8%). Ultimately, if regions emitting over 103 mg CH4 m−2 d−1, which make up only 21% of the segment fluxes observed in the study area, were omitted, then the CH4 ebullition emissions to the atmosphere from this region would be underestimated by >70% when assuming an average ebullition flux (Figure 4b). Similar to the disproportionate contribution of big bubbles to flux, localized areas or hot spots of ebullition contribute disproportionately more to total surface emission estimates when fluxes vary as widely as they do in Lake Wohlen. Hot spots of surface CH4 emission exist even within this small survey area, indicating that mapping small-scale heterogeneity of ebullition is highly important for the accurate assessment of total CH4 flux to the atmosphere from a single water body. These small events or areas of intensive ebullition could easily be missed during routine surveys. On the other hand, if the method employed to calculate ebullition emissions used point measurements (i.e., bubble traps), but from locations with high flux rates, then ebullition emissions would be overestimated. The hydroacoustic method offers the advantage to minimize such risks of sampling bias. Ideally, surveys with transects would be combined with moored observations in order to analyze the temporal (i.e., seasonal, daily, hourly) variability of emissions as well. Implications. When present, ebullition is usually the most dominant emission pathway for CH4 to the atmosphere. The database from Barros et al.41 lists diffusive CH4 emissions from the surfaces of 43 reservoirs that average ∼40 mg CH4 m−2 d−1 with a range from 0 to 260 mg CH4 m−2 d−1. This maximum is equivalent to the median of our flux results (265 mg CH4 m−2 d−1), while the average of our surface ebullition estimates (720 mg CH4 m−2 d−1) is almost three times higher than maximum diffusive fluxes and ∼2 orders of magnitude higher than the average. Bastviken et al.1 reviewed the literature thus far and concluded that reservoirs and natural lakes emit at least an order of magnitude more CH4 via ebullition than diffusion, and our high resolution results also support this fact. The data presented here clearly show that “size does matter” in regards to bubble-mediated transport of CH4. For the large amount of identified bubbles in Lake Wohlen, the upper 10% of the bubble size distribution is transferring over 65% of the total CH4 gas volume to the atmosphere. In addition, a large spatial heterogeneity in surface emissions was present in this small study area of Lake Wohlen such that only 21% of the flux segments (those >103 mg CH4 m−2 d−1), which averaged only a few 10s of meters in size, contributed over 70% of the total surface CH4 emissions. DelSontro et al.20 found that missing large ebullition hot spots like river deltas in a massive lake would negatively impact CH4 emission estimates from that water body. In this study, we have shown that within a localized 1274
DOI: 10.1021/es5054286 Environ. Sci. Technol. 2015, 49, 1268−1276
Article
Environmental Science & Technology
figures. This material is available free of charge via the Internet at http://pubs.acs.org.
study area of a small reservoir there are even smaller emission hot spots present that are disproportionately affecting CH4 transport and emission. Lake/reservoir spatial heterogeneity of CH4 emissions is still poorly understood, which may be due in part to the methods previously used for assessment. Areas of either larger bubbles or higher ebullitive fluxes that are likely present in an actively bubbling water body and significantly contributing to total CH4 emissions make ebullition quantification difficult via traditional point measurement techniques. Since large bubbles are so important to the total gas flux of a system, it is necessary to resolve the entire bubble size distribution to acquire accurate CH4 gas emission estimates. Granted, methods measuring flux more directly such as bubble traps or eddy covariance18,26 do not require bubble size information; however, the effort and time needed to measure an equivalent amount of space that is covered by an echosounder in a single day of monitoring renders at least the trap method severely inefficient. Employing an echosounder and the BNS method along with a dissolution model is fortunately applicable to any body of water, but unfortunately limited to waters deeper than the near-field zone of the echosounder (typically,
Size Does Matter: Importance of Large Bubbles and Small-Scale Hot Spots for Methane Transport T. DelSontro,*,†,⊥ D. F. McGinnis,‡,§ B. Wehrli,† and I. Ostrovsky∥ †
Eawag, Swiss Federal Institute of Aquatic Science and Technology, 6047 Kastanienbaum, Switzerland & Institute for Biogeochemistry and Pollutant Dynamics, ETH Zurich, 8092 Zurich, Switzerland ‡ GEOMAR, Helmholtz Centre for Ocean Research Kiel, Marine Geosystems, 24148 Kiel, Germany § Institute F.-A. Forel, Earth and Environmental Sciences, Faculty of Sciences, University of Geneva, 1227 Geneva, Switzerland ∥ Israel Oceanographic & Limnological Research, Yigal Allon Kinneret Limnological Laboratory, Migdal 14950, Israel S Supporting Information *
ABSTRACT: Ebullition (bubbling) is an important mechanism for the transfer of methane (CH4) from shallow waters to the atmosphere. Because of their stochastic nature, however, ebullition fluxes are difficult to accurately resolve. Hydroacoustic surveys have the potential to significantly improve the spatiotemporal observation of emission fluxes, but knowledge of bubble size distribution is also necessary to accurately assess local, regional, and global water body CH4 emission estimates. Therefore, we explore the importance of bubble size and smallscale flux variability on CH4 transport in and emissions from a reservoir with a bubble-size-calibrated echosounder that can efficiently and economically survey greater areas while still resolving individual bubbles. Using a postprocessing method that resolves bubble density, we found that the largest 10% of the >6700 observed bubbles were responsible for more than 65% of the total CH4 transport. Furthermore, the asymmetry of CH4 ebullition flux distribution and the high spatial heterogeneity of those fluxes suggests that inadvertently omitting emission hot spots (i.e., areas of high flux) could lead to significant underestimations of CH4 emissions from localized areas and potentially from entire water bodies. While the bubble sizes resolved by the hydroacoustic method may provide insight into the factors controlling ebullition (e.g., sediment type, carbon sedimentation), the better resolution of small-scale CH4 emission hot spots afforded by hydroacoustics will bring us closer to the true CH4 emission estimates from all shallow waters, be them lakes, reservoirs, or coastal oceans and seas.
■
INTRODUCTION
The important role of inland water bodies in the carbon cycle is now widely accepted; however, many aspects of aquatic carbon dynamics, such as CH4 production, transport, and emission, are not fully understood.12,13 In fact, the literature on CH4 emissions from inland waters remains discordant, partly due to insufficient measurement methods and partly due to insufficient spatial and temporal coverages. These shortcomings, exasperated by the stochastic nature of ebullition, hinders the accurate quantification of this potentially dominant emission pathway in shallow aquatic systems.2,3,14 Thus, the result is a disjointed understanding of ebullition dynamics in all surface waters, along with a wide range of flux estimates. Ebullition intermittency has been the easiest to demonstrate in high latitude lakes during ice cover.14−16 While the cause for the so-described “patchiness” of ebullition in these particular
Ebullition, the transport of gas via rising bubbles, is the most efficient vertical transport pathway for methane (CH4) in lakes and reservoirs.1 The spatial and temporal variability of CH4 ebullition, however, impedes the accurate quantification of emission rates of this potent greenhouse gas. Measuring ebullition fluxes with floating chambers2 and submerged gas traps3−6 cannot fully resolve the spatiotemporal variability of ebullition because of the number of chambers/traps and deployment times that would be required for adequate temporal and spatial coverage. Recent advances in hydroacoustic processing,7−10 however, overcome some of these limitations and can provide valuable information on both the spatial and temporal heterogeneity of CH4 fluxes and individual bubble sizes.9−11 Ultimately, hydroacoustic surveys deliver small-scale fluxes that not only improve spatial resolution but also offer a more robust understanding of ebullition dynamics by helping elucidate the causes and effects of flux variability. © 2014 American Chemical Society
Received: December 29, 2012 Accepted: December 31, 2014 Published: December 31, 2014 1268
DOI: 10.1021/es5054286 Environ. Sci. Technol. 2015, 49, 1268−1276
Article
Environmental Science & Technology
in a Swiss hydropower reservoir known for high bubbling rates. We detail the hydroacoustic postprocessing method and compare our results to other reported values. Specifically, we investigate how bubble size and other parameters, such as bubble density and water column depth, relate to CH4 fluxes. Finally, we discuss how bubble size and CH4 flux intermittency impacts surface CH4 emission estimates from this reservoir, thus highlighting the importance of large bubbles and localized flux hot spots for CH4 emission estimates from aquatic systems.
lakes may be due to their local environments (i.e., thermokarst or peatland lakes), the fact remains that methods employed on ice-covered lakes are currently the best at constraining the spatial variability of ebullition. In ice-free lakes, chambers and submerged gas traps are the current methods of choice for quantifying ebullitive emissions.2,3,17,18 However, due to their (very) limited coverages, it is difficult with these methods to provide integrated emission assessments that accurately represent the bulk emissions from the water body.3 Missing ebullition hot spots due to insufficient sampling will result in significantly underestimated CH4 fluxes.14,19,20 The hydroacoustic method, e.g., using an echosounder, is a technique that can help overcome these limitations by providing a more accurate ebullition survey, particularly in an effort to locate ebullition hot spots. Although the instrumentation can be more expensive than chambers or gas traps, hydroacoustic methods have been widely used in both aquatic and marine environments. As with other geophysical instrumentation, a single echosounder is sufficient to survey large areas efficiently and can produce highly spatially resolved maps of ebullition activity (and bathymetry).9,20 In addition, high-resolution ebullition time series can be acquired from a moored instrument.7 Scientific echosounders have even progressed to the point of resolving single bubbles, their trajectories, and their rise velocities, which enables the quantification of ebullition flux in relatively shallow (100 m), sediment-released methane bubbles will dissolve before reaching the surface. In shallow waters ( 10 mm or TS > −46 dB) were observed in Lake Wohlen, but as the frequency distribution clearly demonstrates, bubbles larger than −50 dB account for the majority of the gas volume from all surveys (black line, Figure 1a). The cumulative curves for number of bubbles and their volume contribution according to diameter (Figure 1b) shows that >90% of all observed bubbles are less than 10 mm in diameter but only account for ∼35% of the total volume of gas contained within all observed bubbles (dashed line, Figure 1b). This means that the largest 10% of bubble sizes
■
RESULTS AND DISCUSSION The echosounder bubble size calibration results were justifiable and correlated well with previous similar bubble size calibrations (discussed in detail in the Supporting Information); thus, echosounder data were confidently used for ebullition size and flux analysis. The BNS method was compared to two other hydroacoustic postprocessing approaches and shown to be the most reliable postprocessing technique for our conditions (see the Supporting Information for more details). Bubble size distribution, bubble density, and bubble CH4 flux were thus calculated in a total of 556 different segments with an average length of ∼46 m resulting in a spatial analysis of CH4 emissions during seven campaigns. The following sections will use the segment results to describe and discuss the variability in both bubble size and flux within the water column, their relationship to each other and other parameters, and last, how an inadequate sampling resolution of that variability could impact overall CH4 emission estimates. 1270
DOI: 10.1021/es5054286 Environ. Sci. Technol. 2015, 49, 1268−1276
Article
Environmental Science & Technology
Figure 2. (a) Sauter mean diameter (SMD; top axis) of the bubble population per segment correlates better with sediment CH4 ebullition flux (logtransformed) than the weighted mean diameter (WMD; top axis), and both impact flux more than bubble density (BD; bottom axis). (b) Sediment CH4 ebullition flux was not negatively correlated with segment depth as seen in the boxplots (horizontal line is median, red dot is average, whiskers are 25th and 75th percentiles, dots are outliers). (c) There was a positive correlation between weighted mean diameter (WMD) of the bubble population of each segment with the maximum depth of the segment. Red dots are the average per meter depth bin (e.g., 3 m = 2.5−3.5 m), and the lines are the standard deviations. Gray shading shows smoothed maximum and minimum WMD range per depth and dark gray dots are the data.
contributes ∼65% of the total volume of bubble gas. These findings are similar to those of Greinert and Nützel,11 who found that over 50% of the total gas volume in their experiments was from only 7% of the bubbles (>14 mm diameter). Here, we show that even when the bubble population is completely dominated by small bubbles the total volume is supplied by only a few large bubbles, rendering the smallest bubbles almost insignificant for flux calculations. Ebullition Flux Variability. Employing the BNS method with our calibrated echosounder data, we find an average ebullition gas flux of ∼1200 mg m−2 d−1 for all segments. Considering a primary CH4 gas composition in the bubbles of 70 ± 5%,17 the potentially conservative estimate of CH4 ebullition flux from the Lake Wohlen study area is ∼820 mg CH4 m−2 d−1. The fluxes calculated represent the flux of CH4 from the sediment into the water column only. Maximum fluxes from the sediments were in the 103 mg CH4 m−2 d−1 range on all days, except for 1 day which had several segments in the 104 mg CH4 m−2 d−1 flux range (Figure 1c). The high variability across segments, which ranged over 4 orders of magnitude, was expected as transects were subdivided based on bubble and nonbubble densities that were also variable. Zero flux segments were observed on each survey day, but of the 556 total segments analyzed, only 11% of them contained no detectable ebullition. It should be stressed that for the purposes of this study our hydroacoustic surveys were intentionally conducted in a region known for active ebullition. Our ebullition flux estimates compare well with ebullition estimates found with similar and different methods. Using an echosounder, Ostrovsky et al.9 found slightly lower fluxes with a maximum of ∼700 mg CH4 m−2 d−1 in a natural lake of ∼40 m water depth. A few other studies that have used some type of automated system (at the surface or submerged) to directly measure bubble fluxes found ebullition rates similar to those found in our study, including the highest fluxes (>104 mg CH4 m−2 d−1). Varadharajan et al.6 used submerged automated traps in depths up to 25 m in a eutrophic lake of the Northeastern United States and found that the majority of bubble fluxes
(measured at a resolution of 5−10 min) were around 102 mg CH4 m−2 d−1, but events as high as 1.5 × 103 mg CH4 m−2 d−1 were observed. In a Brazilian reservoir, Ramos et al.31 used a surface trap sampling every 5−7.5 min and observed that the majority of fluxes were below 103 mg CH4 m−2 d−1, but several events between 5 × 103 and 1.5 × 104 mg CH4 m−2 d−1 as well as two bubble plumes greater than 2 × 104 mg CH4 m−2 d−1 were recorded. Using submerged automated traps in an impounded river of central Europe, Maeck et al.32 not only showed that ebullition rates were variable and often high, but that flux results varied according to the sampling resolution. In an area of the river known for active bubbling, daily ebullition rates ranged up to ∼4 × 103 mg CH4 m−2 d−1, while hourly and 5 min rates were as high as 104 and >3 × 104 mg CH4 m−2 d−1, respectively. Lake Wohlen is a cross between an impounded river and a reservoir and, unsurprisingly, emits CH4 via ebullition at rates closer to those found in other impounded rivers or reservoirs than those found in natural water bodies. The automated trap methods mentioned above sample the same spot at a high temporal resolution. While our sampling resolution did not consist of multiple high-resolution measurements at one location, our temporal resolution was still high (i.e., short duration) as we were recording from a moving boat (∼1 m s−1). We thus covered a larger area over the course of a day by recording ebullition at each location only once for a short duration. Ultimately, both the high-temporal resolution bubble traps and the high-spatial and short duration echosounding technique seem to converge at the high end of CH4 bubble emission. Relationship between Bubbles and Ebullition Flux. Bubble Density and Size. The two bubble parameters that contribute to the total CH4 flux calculated in a segment are bubble density and the bubble size distribution. Bubble density was almost uncorrelated to flux (r2 = 0.12; triangles, Figure 2a), while a strong relationship existed between bubble size and flux of each segment (r2 = 0.46; black dots, Figure 2a) when the bubble size distribution is represented by the weighted mean 1271
DOI: 10.1021/es5054286 Environ. Sci. Technol. 2015, 49, 1268−1276
Article
Environmental Science & Technology
Figure 3. Four surveys from July 2008 illustrating the weighted mean diameter (WMD) of the bubble population in each segment (represented by the size of the dot) and the sediment ebullition CH4 flux measured in each segment (represented by the color of the dot). Crosses are segments with no observed ebullition. Contours below dots represent depth with darker shades being shallower (see Figure S1 in the Supporting Information for true depths).
the middle range of observed depths (∼5 to 9 m, Figure 2b). We also did not find a negative relationship between depth and ebullition flux to the atmosphere (data not shown). Similarly, there was no apparent relationship between the SMD of each segment and depth (Figure S5, Supporting Information), which would be expected considering that the SMD better reflects the volume of gas present in the segment and thereby ebullition flux. Clearly, the commonly observed relationship between ebullition flux and depth was not present in our study area. The negative depth−flux relationship may not have been as obvious in our study area simply because significant ebullition occurred at all depths; therefore, in order to see a relationship between depth and atmospheric flux the bubble size distribution would have had to be constant with depth so that the bubbles would exchange with the ambient water at rates respective to their rise time and height from the bottom.9,23 We found, however, that the WMD of bubbles (1-m bin averages, red dots, Figure 2c), which reflects the true size distribution of bubbles within a segment, was significantly positively related to maximum depth of the segment (r2 = 0.71, p < 0.05, Figure 2c). Bubble size distribution was thus not constant with depth, and instead sediments from deeper segments generally emitted larger bubbles. Buoyancy theory states that in order for an object to float (i.e., a bubble to rise) the weight of the object must be less than the weight of the displaced volume in which the object lies. As objects, including bubbles, become heavier with depth and higher pressure, bubble volumes must increase in order to counter balance the mass increase and be able to rise. Of course, sediment properties such as cohesiveness also play a role in bubble release, but perhaps in a more unexpected manner as existing fractures may allow for an easier bubble transit through the sediments.35 Ultimately, the increase in WMD with depth that we observed in Lake Wohlen falls within the range of increasing bubble volume with increasing pressure, which in turn likely impacted the observed depth-flux relationship or lack thereof in Lake Wohlen. Since the SMD takes into account the disproportionate volume contribution from larger bubbles,
diameter (WMD) for each segment (WMD segment distribution seen in Figure S4a, Supporting Information). These results imply that it is the size distribution of the bubbles rather than the number of bubbles that are important in terms of overall flux. The significance of bubble size for flux estimates was expected since gas volume increases with bubble diameter and, as already shown, the few large bubbles observed contributed most heavily to the total gas volume (Figure 1a,b). To further support this notion, we found that the relationship between size and flux improves if the Sauter mean diameter (SMD) is calculated instead of WMD (r2 = 0.52; white dots, Figure 2a). In fields of research such as fluid dynamics where the surface area-to-volume ratio (SA:V) is of importance,33 which is the case for spherical gas bubbles, the SMD is often used to represent size distribution instead of an arithmetic mean diameter in order to better incorporate the influence of a relatively low number of large particles (in our case, bubbles). Here, the SMD calculates the diameter of a spherical bubble that would have the same SA:V ratio as the distribution of measured bubbles by giving more weight to the larger ones using the following equation n
SMD =
∑1 d3 n
∑1 d 2
(1)
where d is diameter for each bubble present in the population (see Figure S4b, Supporting Information, for SMD segment distribution). The SMD, however, did not improve the bubble size to flux relationship as much as expected; therefore, additional environmental conditions must also impact ebullition flux from the sediments of Lake Wohlen. Water Column Depth. A common observance in some systems is a negative correlation between water column depth and ebullition flux.2,34 We, however, did not find this in our study area. Rather, we see a very slight peak distribution in which, when sediment ebullition fluxes are binned according to depth (e.g., 3 m bin contains all fluxes from segments with maximum depths between 2.5 and 3.5 m), fluxes are highest in 1272
DOI: 10.1021/es5054286 Environ. Sci. Technol. 2015, 49, 1268−1276
Article
Environmental Science & Technology
Figure 4. (a) Contour plot shows what fraction of the original CH4 will remain in a bubble of a certain initial diameter at the air−water interface when released from a certain depth. For example, a 16 mm bubble would still contain ∼90% of its original CH4 when released from 11 m depth. Inset shows the percent of CH4 reaching the surface of all bubble sizes when released from 10 m depth (∼average depth of Lake Wohlen). (b) Bar plots show the average CH4 sediment ebullition flux (top of black bar) and surface ebullition flux (top of gray bar) when calculated from all fluxes, those less than 104 mg CH4 m−2 d−1, etc. Numbers in bars are amounts of emitted CH4 in tons per day when the corresponding averages are extrapolated to the entire study area (∼0.13 km2).
supersaturation in the sediments;37 therefore, areas of intense methanogenesis are likely to be ebullition hot spots. In order for intense methanogenesis to occur, a significant amount of labile carbon substrate must be present for methanogens to process, which means that carbon sedimentation rates should also be sufficiently high in those areas. Areas of high sedimentation have been shown to be hot spots of ebullition in a larger reservoir20 and an impounded river,19 and high sedimentation rates have already been linked to the high rates of methanogenesis and bubbling in Lake Wohlen.38 Gas volume in sediments as well as ebullition rates have also been shown to relate to sediment characteristics in other lakes.34 The different lines of evidence suggest that organic matter sedimentation rates and sediment properties may drive ebullition rates at quite a small scale, as the immense flux and bubble size variability in this small, ∼0.13 km2 hot spot suggests. Impact of Bubble Size and Flux Variability on Surface CH4 Emissions. Bubble Size. As bubbles rise through the water column, gas exchange occurs between the bubble and the ambient water while the decreasing hydrostatic pressure expands the size of the bubble.23 The hydroacoustic method presented here estimates ebullition flux only from the sediment to the water column; therefore, a discrete bubble dissolution model23 is needed to estimate gas exchange during bubble rise and consequent flux to the atmosphere,9,17 which primarily depends on bubble size and release depth (i.e., depth of the water column). Figure 4a shows the percent of CH4 remaining in a bubble when it reaches the air−water interface based on the release depth (as deep as 12 m in Lake Wohlen) and initial diameter (up to 30 mm in Lake Wohlen) of the rising bubble. As a first estimate, the average bubble diameter (∼6 mm) and average depth (∼10 m) for Lake Wohlen can be used to find that ∼67% of the CH4 contained in Lake Wohlen bubbles will reach the atmosphere. Applying this to the overall average CH4 flux from Lake Wohlen segments (820 mg CH4 m−2 d−1), we find that ∼550 mg CH4 m−2 d−1 could have been released from this study area to the atmosphere at that time. For a more accurate estimate, we calculate bubble dissolution (and hence the fraction of CH4 reaching the atmosphere) per bubble size class (in our case, every 2 dB from −70 to −30 dB for a total of
the lack of relationship between SMD and depth means that larger bubbles are ubiquitous and contribute significantly more to the total flux than is accurately represented by their absolute number or the WMD. In other words, the smaller WMD in shallower depths (Figure 2b) does not adequately reflect the amount of gas contained within the total bubble size distribution because, despite the higher numbers of small bubbles lowering the WMD, the few larger bubbles present still transport most of the gas in those areas. Additional Controls on Ebullition Flux. We have seen that while bubble size does tend to relate to depth in Lake Wohlen, ebullition flux does not exhibit similar trends; therefore, other factors besides bubble size distribution must influence flux. Figure 3 and Figure S6 (Supporting Information) illustrate bubble WMD and sediment ebullition flux spatially on each survey day, which enables the investigations of day-to-day variability as well as any spatial trends. The relationship between size and flux as previously discussed is also seen spatially; in general, higher CH4 flux segments (warmer colors, Figure 3 and Figure S6, Supporting Information) tend to have larger WMDs (size of circles, Figures 3 and Figure S6, Supporting Information) and lower CH4 flux segments (cooler colors) tend to have smaller WMDs (i.e., smaller circles). One exception is July 23 in which the majority of segments had high ebullition fluxes but with WMDs not quite as large as high flux segments on other days. A likely explanation for this is that it was particularly windy on that survey day and perhaps some hydrological conditions, such as wind-induced bottom shear or currents,5,36 enhanced ebullition fluxes mechanically and thus without increasing bubble sizes. A spatial trend was also present in our study area on most of the seven survey days. Low fluxes were typically found in the shallow region near the south shore of the field area as well as in the deepest region to the north that represents the old river channel. Higher fluxes were typically found on the slope between these two regions in depths between 5 and 9 m. This trend was also apparent in the peak distribution of flux versus depth previously described (Figure 2b) and together these results suggest a geomorphic control on CH4 ebullition fluxes. The most necessary constraint for CH4 ebullition is a CH4 1273
DOI: 10.1021/es5054286 Environ. Sci. Technol. 2015, 49, 1268−1276
Article
Environmental Science & Technology 20 bins) per segment in which fluxes were calculated. The result is in fact a higher average surface ebullition flux estimate (∼720 mg CH4 m−2 d−1) that is ∼88% of the total sediment ebullition flux. Bubble dissolution is primarily based on bubble release depth and diameter; therefore the higher surface flux estimate from the per bin per segment calculation is due partially to the variability in release depth among segments but more so to the greater variability in bubble sizes present in each segment. As bubble diameter increases, the SA:V of bubbles decreases (Figure S7, Supporting Information). This implies that bubbles with larger diameters are less efficient at exchanging gases between the bubble interior and surrounding water, and that large bubbles rise faster which permits less contact time with the ambient water.9,23,33,39 The importance of the SA:V to bubble dissolution suggests that the SMD is a more appropriate average bubble size value to use in a bubble gas exchange model for a rough surface flux estimate. The SMD for all 6700+ bubbles observed in Lake Wohlen is 10.1 mm, which is nearly double the WMD (5.9 mm), and reflects 5 times more volume (0.11 mL versus 0.54 mL, assuming a perfect sphere). The exchange model shows that the SMD-sized bubble would release ∼81% of its CH4 to the atmosphere (Figure 4a) and thus ∼665 mg CH4 m−2 d−1 would be emitted from our study area. Using the SMD to estimate surface CH4 ebullition flux provides an estimate much closer to the true flux than using the WMD; therefore the SMD could be used for a fairly accurate first flux estimate as previously shown by McGinnis and Little39 and also indicates that the few large bubbles present in Lake Wohlen are disproportionately more important for surface emissions than the large number of small bubbles. As stated earlier, while bubbles up to 10 mm in diameter make up ∼90% of the total population observed in Lake Wohlen, they supply less than 35% to the total CH4 present in the water column (Figure 1b), and are thus responsible for even less of the CH4 emitted to the atmosphere. From the average depth of Lake Wohlen (10 m), bubbles smaller than ∼10 mm in diameter exchange CH4 more rapidly with the ambient water, therefore emitting much less CH4 to the atmosphere. From 10 mm in diameter and greater, the rate at which bubbles dissolve and exchange gases slows remarkably, thereby naturally presenting an important size threshold to distinguish between small and larger bubbles (Figure 4a, inset). Not only do large bubbles carry exponentially more CH4 with increasing diameter (Figure 1b) but they also dissolve exponentially less during ascent (Figure 4a). Ultimately, larger bubbles are more responsible for atmospheric CH4 release in shallow waters and missing or excluding some of them during a survey could underestimate final CH4 flux results. Flux Variability. The hydroacoustic method provides a range of observed (and possible) sediment CH4 ebullition fluxes when estimated per segment (Figure 1c). When coupled with the dissolution model, atmospheric CH4 fluxes can also be estimated, but the question is how to most accurately upscale emissions to an entire study area, lake, or inundated region. Previous studies have observed a negative relationship between depth and ebullition flux that was subsequently used to upscale surface emissions to an entire water body.2,34 Since no depth− flux relationship was observed in our study area of Lake Wohlen, we could only provide a first rough estimate of flux simply by applying the average of our measurements to the entire study area in question, a method previously used by others with floating chambers40 and submerged traps.4
If the average sediment ebullition flux from the Lake Wohlen hydroacoustic transects (820 mg CH4 m−2 d−1) was applied to the whole study area (only ∼0.13 km2 of the 2.5 km2 lake), we would find a total of 0.108 t CH4 d−1 emitted from the sediments. The associated surface emission estimate (calculated per size bin per segment) would be ∼0.096 t CH4 d−1 (Figure 4b). If, however, the flux variability was not spatially resolved to the extent we have accomplished in our study and some disproportionately important areas of CH4 flux were missed, then the resulting emission estimate could be impacted. For example, missing fluxes greater than 104 mg CH4 m−2 d−1, which are rare and likely easy to miss, would underestimate the atmospheric CH4 emission estimate but only by a small amount (∼0.007 t CH4 d−1 or 8%). Ultimately, if regions emitting over 103 mg CH4 m−2 d−1, which make up only 21% of the segment fluxes observed in the study area, were omitted, then the CH4 ebullition emissions to the atmosphere from this region would be underestimated by >70% when assuming an average ebullition flux (Figure 4b). Similar to the disproportionate contribution of big bubbles to flux, localized areas or hot spots of ebullition contribute disproportionately more to total surface emission estimates when fluxes vary as widely as they do in Lake Wohlen. Hot spots of surface CH4 emission exist even within this small survey area, indicating that mapping small-scale heterogeneity of ebullition is highly important for the accurate assessment of total CH4 flux to the atmosphere from a single water body. These small events or areas of intensive ebullition could easily be missed during routine surveys. On the other hand, if the method employed to calculate ebullition emissions used point measurements (i.e., bubble traps), but from locations with high flux rates, then ebullition emissions would be overestimated. The hydroacoustic method offers the advantage to minimize such risks of sampling bias. Ideally, surveys with transects would be combined with moored observations in order to analyze the temporal (i.e., seasonal, daily, hourly) variability of emissions as well. Implications. When present, ebullition is usually the most dominant emission pathway for CH4 to the atmosphere. The database from Barros et al.41 lists diffusive CH4 emissions from the surfaces of 43 reservoirs that average ∼40 mg CH4 m−2 d−1 with a range from 0 to 260 mg CH4 m−2 d−1. This maximum is equivalent to the median of our flux results (265 mg CH4 m−2 d−1), while the average of our surface ebullition estimates (720 mg CH4 m−2 d−1) is almost three times higher than maximum diffusive fluxes and ∼2 orders of magnitude higher than the average. Bastviken et al.1 reviewed the literature thus far and concluded that reservoirs and natural lakes emit at least an order of magnitude more CH4 via ebullition than diffusion, and our high resolution results also support this fact. The data presented here clearly show that “size does matter” in regards to bubble-mediated transport of CH4. For the large amount of identified bubbles in Lake Wohlen, the upper 10% of the bubble size distribution is transferring over 65% of the total CH4 gas volume to the atmosphere. In addition, a large spatial heterogeneity in surface emissions was present in this small study area of Lake Wohlen such that only 21% of the flux segments (those >103 mg CH4 m−2 d−1), which averaged only a few 10s of meters in size, contributed over 70% of the total surface CH4 emissions. DelSontro et al.20 found that missing large ebullition hot spots like river deltas in a massive lake would negatively impact CH4 emission estimates from that water body. In this study, we have shown that within a localized 1274
DOI: 10.1021/es5054286 Environ. Sci. Technol. 2015, 49, 1268−1276
Article
Environmental Science & Technology
figures. This material is available free of charge via the Internet at http://pubs.acs.org.
study area of a small reservoir there are even smaller emission hot spots present that are disproportionately affecting CH4 transport and emission. Lake/reservoir spatial heterogeneity of CH4 emissions is still poorly understood, which may be due in part to the methods previously used for assessment. Areas of either larger bubbles or higher ebullitive fluxes that are likely present in an actively bubbling water body and significantly contributing to total CH4 emissions make ebullition quantification difficult via traditional point measurement techniques. Since large bubbles are so important to the total gas flux of a system, it is necessary to resolve the entire bubble size distribution to acquire accurate CH4 gas emission estimates. Granted, methods measuring flux more directly such as bubble traps or eddy covariance18,26 do not require bubble size information; however, the effort and time needed to measure an equivalent amount of space that is covered by an echosounder in a single day of monitoring renders at least the trap method severely inefficient. Employing an echosounder and the BNS method along with a dissolution model is fortunately applicable to any body of water, but unfortunately limited to waters deeper than the near-field zone of the echosounder (typically,
