| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a USDA-ARS, P.O. Box E, Fort Collins, CO 80522, USA
b Agriculture and Agri-Food Canada, 2560 Hochelaga Blvd., Sainte-Foy, Québec, Canada G1V 2J3
* Corresponding author (glhutch{at}lamar.colostate.edu)
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONAPPROACHRESULTS AND DISCUSSIONRECOMMENDATIONS AND PERSPECTIVEREFERENCES |
|---|
5% despite that headspace concentration rose more than 70% within 2 h. Larger errors occurred for chamber designs less well matched to the soil respiration rate they were intended to measure, but if such serious design deficiencies are avoided, the method offers a simple inexpensive means for obtaining multiple reliable time-integrated estimates of soil respiration, even at remote locations.
Abbreviations: AFP, air-filled porosity (m3 m-3 soil) • D, binary molecular diffusion coefficient of CO2 in air • Fa, estimate of CO2 flux density at the soil-atmosphere boundary obtained from the amount of CO2 absorbed by the alkali trap in a non-flow-through steady-state chamber • Fc, CO2 flux density at the soil-atmosphere boundary • Fo, depth-integrated rate of subsurface CO2 production expressed in units of flux density • FT, flow-through • NFT, non-flow-through • NSS, non-steady-state • SS, steady-state
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONAPPROACHRESULTS AND DISCUSSIONRECOMMENDATIONS AND PERSPECTIVEREFERENCES |
|---|
The rate of gaseous CO2 absorption by an alkali trap is proportional to the CO2 concentration to which it is exposed. In a perfectly designed NFT-SS chamber, the absorption rate is equal to the surface flux, Fc, when headspace CO2 is equal to the ambient level. In that situation, no change would occur in concentrations of the gas either above or below the soil surface during the period of deployment, Fc would remain equal to the underlying rate of CO2 production, and the chamber would provide an accurate estimate of soil respiration. However, achieving and maintaining such a balance is difficult, if not impossible, and the CO2 concentration of both headspace and subsurface air often rises following chamber deployment until the rate of CO2 absorption becomes equal to Fc (e.g., Rochette et al., 1997). The result is an underestimate of soil respiration—first, because the increase represents CO2 production not accounted for in the alkali trap, and second, because the elevated concentrations support CO2 losses by leakage through imperfect chamber seals and by lateral diffusion beneath the chamber walls. In contrast, headspace and subsurface CO2 concentrations may decline following chamber deployment if Fc is particularly small. The result in this case is an overestimate of soil respiration, because the decline represents absorbed CO2 not produced during the deployment period, and because the reduced concentrations support CO2 gain by the chamber system via leakage and subsurface lateral diffusion.
The amount of increase (or decrease) in headspace CO2 concentration depends not only on the magnitude of the soil respiration rate, but also on the efficiency of CO2 absorption by the alkali trap, the effective diffusivity and storage coefficient of the gas in underlying soil, and its rate of exchange (leakage) between the chamber system and its surroundings. In a recent exhaustive review of existing literature regarding NFT-SS chambers, Rochette and Hutchinson (2003) concluded that (i) the optimal strength of the alkali solution is 0.5 to 1.0 M, (ii) the alkali trap should have total capacity approximately three times greater than the amount of CO2 expected to be emitted during the deployment period, (iii) a 20% ratio of exposed alkali trap area to emitting soil surface area provides good absorption efficiency in many situations, but can be altered when needed to keep headspace CO2 concentration as close as possible to the ambient level, (iv) the chamber should be nonvented and should have good seals that minimize CO2 exchange between the chamber and its surroundings, and (v) the deployment period should be at least 12 and preferably 24 h to minimize measurement bias due to the initial nonsteady-state condition, as well as bias due to chamber-induced temperature disturbances that often differ in algebraic sign between day and night (and thus tend to at least partially cancel across a 24-h deployment). Despite the decades of experience and scores of studies summarized by this review, however, many questions remain regarding the accuracy and functioning of NFT-SS chambers, as well as the optimal protocol for their use.
|
APPROACH |
|---|
|
TOPABSTRACTINTRODUCTIONAPPROACHRESULTS AND DISCUSSIONRECOMMENDATIONS AND PERSPECTIVEREFERENCES |
|---|
Briefly, we assumed that the soil was bare and had uniform properties with pH of 6.5, total porosity of 0.5, AFP of 0.3, tortuosity computed from the equation of Sallam et al. (1984), and effective diffusivity equal to the three-way product of tortuosity, AFP, and the binary molecular diffusion coefficient of CO2 in air (D). Total porosity and AFP are understood to have units of m3 m-3 soil. Properties for CO2 were chosen at 20°C and 100 kPa air pressure: D = 0.160 cm2 s-1 (Lide, 1996), ambient atmospheric concentration = 375 µmol mol-1, Henry's Law coefficient for dissolution in water = 0.935 (Lindsay, 1979), and the dissociation constant of dissolved CO2 (H2CO3) = 10-6.40 (Sillén and Martell, 1964). Equilibrium between gas phase and solution phase CO2 and among dissolved species in soil water and in the alkali solution was assumed to occur instantaneously. The CO2 was generated by a constant (or 24-h sinusoidal) zero-order source term with magnitude that decreased exponentially (with 10-cm relaxation depth) from a maximum at the soil surface to zero at the impermeable bottom of the simulated domain 50 cm below the surface (Hutchinson et al., 2000).
The nonvented chamber headspace (30-cm diameter x 20-cm height) was assumed to be convectively mixed above a 0.5-cm diffusively mixed atmospheric interfacial layer, the depth of which was not altered by deployment of the chamber. Unless otherwise specified, the convection supported mixing equivalent to a tenfold increase in D. Chamber walls were assumed to be inserted 5 cm into the soil, except when studying the effect of changes in this parameter. Wall thickness was 0.25 cm except near the bottom, which we assumed was beveled to provide a cutting edge, as is often done to ease insertion and minimize the resulting potential for soil compaction. Finally, the cylindrical alkali container (5-cm internal depth x 0.2-cm walls) was mounted in the horizontal center of the chamber with its bottom 2 cm above the soil surface. Its internal diameter was 13.4 cm when the ratio of exposed alkali surface area to emitting soil surface area was 0.20, and liquid depth (regardless of trap area) was 0.5 cm. The diffusively mixed atmospheric interfacial layer immediately above the alkali solution surface also had a 0.5-cm depth, and the convection in overlying air surrounded by vertical walls of the trap was assumed to support mixing equivalent to a threefold increase in D (compared with a tenfold increase in the bulk headspace). Justification for all these rather arbitrary assumptions is examined in detail in a later section. To distinguish between the two interfacial layers in this paper we refer to them as the soil interfacial layer and the alkali interfacial layer.
Controllers of NFT-SS chamber performance that we investigated included the magnitude and diurnal periodicity of the soil respiration rate, the ratio of exposed alkali surface area to emitting soil surface area (hereafter abbreviated as the alkali:chamber area ratio), the efficiency of convective transport from the top of the soil interfacial layer to the top of the alkali interfacial layer, depth of the latter layer, height of the alkali trap above the soil surface, soil AFP, soil pH (which impacts the equilibrium between dissolved CO2 and CO-3/HCO-3 in soil water), depth of chamber wall insertion into the soil, and the leakiness of chamber seals. For convenience of the reader, the values assumed for these variables are listed at the top of all figures that report simulation results; the variable under examination in each case is identified by bold type. Except in the simulations reported in Fig. 2, the surface flux prior to and including the moment of chamber deployment was defined as the steady-state CO2 flux density under the assumed conditions (equal to the depth-integrated rate of CO2 production), which we designated Fo. Chamber performance was examined in the top graph of Fig. 1 , 4, 5, and 6 by normalizing the instantaneous simulated flux into the chamber (Fc) with respect to Fo and plotting the result as a function of time after deployment. Deviations of the resulting trace from a horizontal straight line at Fc/Fo = 1 then represent a measure of chamber-induced perturbation of the pre-deployment CO2 exchange rate. The bottom graph in these three-panel figures shows the mean CO2 concentration of the headspace as a function of time, while the bar graphs represent CO2 accumulated by the alkali trap after several time periods (expressed as flux density, Fa); like Fc, Fa was normalized with respect to Fo. The solid curves in the top and bottom graphs of these figures identify the level at which the variable under investigation was held constant in other simulations while another was varied to study its influence on NFT-SS chamber performance.
|
|
|
RESULTS AND DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONAPPROACHRESULTS AND DISCUSSIONRECOMMENDATIONS AND PERSPECTIVEREFERENCES |
|---|
Focusing first on Fig. 1a, note that for the three larger rates of soil respiration, the increases in headspace CO2 were substantial (e.g., +72, +238, and +734% at 24 h), but the corresponding chamber-induced changes in Fc and Fa were much smaller (-4.0, -6.0, and -7.1% for Fc and -4.9, -7.3, and -8.8% for Fa, both at 24 h). This observation differs sharply from the situation for NSS and FT-SS chambers deployed for relatively short periods. For these systems, the fractional chamber-induced error in the measured flux is often of the same order as the fractional change in their headspace concentrations (Hutchinson et al., 2000). Such contrasting behavior is a consequence of the presence of a CO2 sink and the use of much longer deployment times for NFT-SS chambers. This combination permits establishing a new near-steady-state condition with a similar surface flux despite a very different CO2 concentration profile. Data in the top and bottom graph panels of Fig. 1a suggest that the new near-steady-state condition was achieved in 1 to 4 h, as headspace CO2 reached a level where the sum of its rate of absorption by the alkali trap and its rate of leakage from the chamber system approached Fo. For NSS chamber systems, lack of a CO2 sink ensures that adjustment of the subsurface gas diffusion gradient seriously lags rapid changes in headspace CO2 concentration throughout a typical deployment period (2–60 min). Exceptions include rare situations where CO2 concentrations rise enough to negate the assumption of a constant production rate or support leakage approaching the rate of production.
For the smallest value of Fo in Fig. 1a, the changes in concentrations and fluxes had the opposite algebraic signs. At 24 h, for example, headspace CO2 decreased by 60%, while Fc and Fa increased by 12 and 15%, respectively. Compared with the three largest soil respiration rates, the smaller fractional change in headspace CO2 resulted in disproportionately large fractional changes in Fc and Fa, suggesting that chamber-induced measurement errors resulting from declining CO2 concentration may exceed (on a relative basis) those spawned by rising concentrations. Thus, while the ideal NFT-SS chamber design causes no change in CO2 concentrations, it may be preferable to err slightly on the side of underdesigning rather than to chance overdesigning the CO2 absorption efficiency of the alkali trap if the range of fluxes to be measured is known to include mostly only very small values. Conversely, if the approximate magnitude of the soil respiration rates to be measured is unknown (or highly variable), it may instead be preferable to overdesign the chamber to do the best possible job measuring the larger fluxes that have greatest influence on the integrated estimate of respiration by the system under study.
Fortuitously, the chamber design selected for our simulations approached the ideal design for measuring the second smallest soil respiration rate in Fig. 1a (0.025 mg C m-2 s-1). At 24 h, mean headspace CO2 concentration decreased only 10% from 375 to 337 µmol mol-1, while Fc and Fa changed by -0.1 and +0.1%, respectively. The simultaneous decline in Fc and chamber concentration appears inconsistent with the other four cases in Fig. 1a, but it occurs only because means in the bottom panel of Fig. 1a mask the variability in CO2 across the chamber headspace. For example, the concentration at the chamber's centerline (at 24 h) was 279 µmol mol-1 at the height of the top of the alkali container, but 404 µmol mol-1 atop the 0.5-cm soil interfacial layer. Because CO2 concentration was assumed uniform (at 375 µmol mol-1) across this distance at the time of chamber deployment, the lower concentration actually increased nearly 8% instead of decreasing by 10%, thereby explaining the negative change in Fc.
During the period from 4 to 24 h in Fig. 1a there was very little change in headspace CO2 concentration or Fc, and the latter remained distinctly different than Fo for each of the five soil respiration rates. Because this group of simulations included no pathway for leakage of CO2 from the chamber headspace directly to the surrounding atmosphere, the differences between Fc and Fo during this period must have resulted largely from loss of CO2 by lateral diffusion beneath chamber walls that extended only 5 cm beneath the soil surface. Modeling studies of NSS chambers by Healy et al. (1996) and Hutchinson et al. (2000) indicated that the magnitude of this loss is strongly dependent on the depth of chamber wall insertion into the soil, which we showed was also true for NFT-SS chambers via the results reported in Fig. 1b. In this case, we assumed that chamber side walls extended to the impermeable bottom of the simulated domain 50 cm below the soil surface. The result was that Fc/Fo and Fa/Fo ratios continued to trend steadily toward unity throughout the 24-h period included in the graph, because the alkali trap was the only available sink for CO2 produced beneath the chamber. It is logical that despite the marked difference in Fc/Fo ratios in Fig. 1a vs. 1b, the difference in headspace CO2 concentrations in the two cases was relatively small. At 24 h, for example, the alkali trap's absorption rate in Fig. 1a amounted to between 93 and 112% of Fo, so only small changes in headspace CO2 were required to guide the concentration-dependent absorption rate toward 100% when lateral diffusion was precluded (Fig. 1b).
Without exception, the deviations of Fa from Fo in Fig. 1a and 1b were slightly greater than the corresponding deviations of Fc from Fo. The reason for this observation is that Fc is an instantaneous flux that reflects only the conditions prevailing at the time for which it is plotted in the graphs, while Fa is an integrated (or cumulative) flux based on the total amount of CO2 absorbed by the alkali trap since the time of chamber deployment. Thus, all bars in both center panel graphs include effects of the initial nonsteady-state condition that immediately followed chamber deployment, but those effects were increasingly diluted as the deployment period lengthened.
At 4-h deployment time, differences among Fa/Fo ratios for the five different soil respiration rates were similar for both chamber wall insertion depths, but they changed in different ways with time. Differences among the five ratios continually decreased in Fig. 1b as the corresponding values of Fc/Fo became more and more alike. In contrast, differences among the five series of simulated Fc/Fo ratios in Fig. 1a remained nearly constant after the first 4 h, so their corresponding Fa/Fo ratios also remained more different than in Fig. 1b. The Fc/Fo and Fa/Fo ratios in Fig. 1b would continue to approach unity if the deployment period were lengthened, but in Fig. 1a they would not, because some CO2 would continue to be lost via lateral diffusion beneath the chamber walls.
Additional simulations summarized in Fig. 2
indicated that NFT-SS chamber performance was nearly the same whether Fo was assumed to be constant or to follow a sinusoidal diurnal cycle with the same mean value (0.050 mg C m-2 s-1), but that was true only for a 24-h deployment period. The solid curve in the bottom panel of the figure represents assumed values for Fo when the sinusoidal CO2 production rate varied from one-third greater to one-third smaller than its 24-h mean value. Such a twofold change might occur, for example, if Q10 = 2 for soil respiration (Rochette and Hutchinson, 2003) and the depth of soil responsible for most CO2 production cooled by 10°C from its daily maximum (typically late afternoon) to its daily minimum (typically early morning) 12 h later. Other curves in the figure show instantaneous (not cumulative) values of Fa for four simulated 24-h chamber deployment periods commencing when Fo was at its minimum value (6 h), at its maximum value (18 h), rising through its mean value (12 h), or declining through its mean value (0 h). To establish the initial soil CO2 concentration gradient for these simulations, we first ran the model for 24 h with no chamber present, instead of assuming a steady-state initial gradient as was done for constant Fo. Other simulation conditions were the same as in Fig. 1a.
Note that in each case, 2 to 3 h was required for headspace and subsurface CO2 concentrations to increase enough that the alkali trap's absorption rate came into approximate balance with the CO2 production rate. Then, the four curves merged and followed the sinusoidal production rate, but with smaller amplitude and a lag of 1 to 2 h. The table in the center panel of Fig. 2 shows that cumulative measurement error for 12-h deployments ranged from -18.1 to +6.8% when Fa was compared with the 24-h mean value of Fo and from -11.9 to +4.0% when Fa was compared with the corresponding 12-h mean value of Fo. More importantly, the error ranged only from -4.4 to -5.2% for 24-h deployments—about the same as for constant Fo in Fig. 1a (-4.8%). In addition to justifying our simplifying assumption of constant Fo in the other simulations reported here, these data provide strong support for the recommendation by Rochette and Hutchinson (2003) that NFT-SS chambers should be deployed for a full 24-h diurnal cycle to minimize measurement error due to the effects of daily temperature change on the magnitude of soil respiration. It is important to realize, however, that during cool seasons and in soil shaded by clouds, vegetation, or plant residue, the disparity between daily maximum and minimum Fo is often smaller than the two-fold difference assumed in Fig. 2, so the potential for measurement error due to diurnal temperature variation is correspondingly smaller.
Efficiency of CO2 Absorption by the Alkali Trap
The performance of a NFT-SS chamber depends not only on the magnitude of the soil respiration rate, but also on factors that determine the efficiency of CO2 absorption by the alkali trap. Here, we examine assumptions that were used to define absorption efficiency in the simulations reported in Fig. 1 and 2 and elsewhere in this paper.
Alkali:Chamber Area Ratio
The increase in headspace CO2 concentration during simulated NFT-SS chamber deployments, as well as the fractional deviation of both Fc and Fa from Fo, all generally declined as the alkali:chamber area ratio increased from 0.05 to 0.40. There was, however, significant interaction between these relations and the magnitude of soil respiration. Figure 3
shows how the 24-h Fa/Fo ratio and mean headspace concentration varied as a function of the alkali:chamber area ratio for each of the five simulated soil respiration rates. For the two largest respiration rates, NFT-SS chamber-induced measurement error (i.e., the fractional deviation of Fa from Fo) was very nearly inversely proportional to the alkali:chamber area ratio. For example, when Fo was 0.100 mg C m-2 s-1, doubling the area ratio from 0.10 to 0.20 resulted in approximately halving that error from -14.5 to -7.3%, and for Fo = 0.250 mg C m-2 s-1 redoubling the area ratio from 0.20 to 0.40 again nearly halved the error from -8.8 to -5.1%. The behavior of headspace CO2 concentration and Fc was similar. For example, in the first case the elevation of mean headspace CO2 was pared from +1943 to +892 µmol mol-1 (Fig. 3b), and the deviation of Fc from Fo declined from -11.8 to -6.0% (data not shown); corresponding changes in the second case were +2752 to +1277 µmol mol-1 and -7.1 to -4.2%, respectively, all at 24 h.
|
The performance of chambers with the two smallest area ratios was unacceptable for the three largest soil respiration rates, as was the smallest area ratio for the 0.025 mg C m-2 s-1 respiration rate. However, the improvement gained by increasing the ratio above 0.20 was much smaller in absolute terms. In practice, the value of that improvement must be weighed against the increased risk of alkali spillage and other logistical problems associated with large traps. Moreover, large area ratios increase the risk of reducing headspace CO2 substantially below ambient levels when the unknown flux to be measured is small, thus triggering the greater measurement errors shown in Fig. 3 for the smallest soil respiration rate. The logistical problems may be reduced by using a baffled trap, by using a sponge to absorb the alkali solution and increase its contact area with headspace air (Nakadai et al., 1993; Bekku et al., 1997), or by replacing the alkali solution with granular soda lime (Zibilske, 1994). Problems associated with using a NFT-SS chamber having an alkali trap too large for the flux to be measured can be minimized by using preliminary surveys to establish the range of fluxes that might be encountered.
Apparently, a ratio near 0.20 exhibits much of the improved performance of larger area ratios while retaining smaller risk of reducing headspace CO2 below ambient levels. This conclusion is supported by experimental results of Ewel et al. (1987), Norman et al. (1992), and Rochette et al. (1992), all of whom noted that for a given alkali:chamber area ratio, the underestimation of Fo by Fa responded nonlinearly as Fo increased. For these reasons, and because a ratio near 0.20 was used in many recent experimental studies (e.g., Gupta and Singh, 1977; Rochette et al., 1997), we chose that value for simulations performed to investigate the effects of other parameters on the efficiency of CO2 absorption by an alkali trap.
Headspace Air Mixing Rates
The data in Fig. 4a
suggest that NFT-SS chamber performance is also strongly dependent on the headspace air mixing rate, because of its effect on the time required for CO2 transport from the soil surface to the alkali trap. The slowest mixing rate included in the graph (D x 1) assumed no convection (i.e., molecular diffusion only), while the greatest rate assumed sufficient convection to achieve homogeneous headspace concentration, which we defined as that attained by arbitrarily increasing D by a factor of 104. The similarity in chamber performance data for the D x 102 and D x 104 mixing rates confirms that the distribution of headspace CO2 was effectively uniform in both cases. Note also that a relatively small difference in the combined rates of convective and diffusive mixing separated the performance curves for diffusively vs. homogeneously mixed chamber types; for example, the D x 3 mixing rate overcame more than 60%, and the D x 10 rate nearly 90%, of the difference between the two.
|
The alkali trap in a NFT-SS chamber is normally suspended above the soil surface and has minimal (or no) contact with the chamber walls and lid, so it is at least partially isolated from solar radiation, surfaces directly impacted by the wind, differences between soil and air temperatures, and other factors responsible for the temperature and pressure gradients that drive convective mixing of the bulk headspace. On the other hand, using formulae summarized by Campbell (1977), Hutchinson et al. (2000) calculated that only a very small temperature gradient is required to support convective gas transport that is threefold greater than diffusion. We concluded that even in the relatively stable air inside the walls of an alkali trap, free convection is likely the dominant mixing mechanism, but that the mixing is probably less efficient than in the bulk headspace outside the trap. On this basis, we chose a D x 3 mixing rate for the enclosed air in simulations performed to investigate the effects of other parameters on the alkali trap's efficiency of CO2 absorption. Data in Fig. 4b suggest that overall chamber performance is less sensitive to changes in the mixing rate of this enclosed air than to changes in the mixing rate of bulk headspace air shown on the same scales in Fig. 4a, but that conclusion depends, of course, on the dimensions of the alkali trap and the chamber in which it is located.
In an attempt to compare our chosen mixing rates with values encountered in a typical field situation, we computed CVs among selected groups of simulated vs. measured chamber concentrations. For the first, we used 49 concentrations obtained from a vertical transect through the convectively mixed headspace midway between the alkali trap radius and the chamber radius. The result was inversely proportional to the assumed headspace mixing rate, increasing from 0.006% at the D x 104 rate to 0.6% at the D x 102 rate and 6% at the D x 10 rate after a 2-min simulated deployment with Fo = 0.050 mg C m-2 s-1. The CVs were positively correlated with the magnitude of Fo, amounting to 2, 6, and 17% when Fo was 0.010, 0.050, and 0.250 mg C m-2 s-1, respectively (all with D x 10 bulk headspace mixing and D x 3 alkali trap air mixing).
To obtain a comparable CV for measured concentrations, we used second-order regression to remove the temporal trend from 99 concentrations measured at 1.5-s intervals in a FT-NSS chamber, then computed the CV among these adjusted concentrations. The result ranged from 0.03 to 2.2% in seven separate chamber placements on sandy loam amended with variable rates of pig slurry, for which the measured soil respiration rates ranged from 0.008 to 0.6 mg C m-2 s-1 (P. Rochette, 2000, unpublished data). From the relation of these CVs to the measured flux, we estimated a value of 0.2% when Fo was 0.050 mg C m-2 s-1. Because considerable integration likely occurred during sampling and analysis, we view this value as at least not inconsistent with the 6% value that characterized the mixing rate used in our simulations. Note that if we had instead interpreted this comparison as evidence that our chosen D x 10 bulk headspace mixing rate should be increased, Fig. 4a suggests that such an increase would have only a small influence on simulated chamber performance.
Although it appears that our CV calculations compare spatial variation (simulated case) with temporal variation (measured case), we reasoned that because of convective mixing in the latter, the parcel of air passing the fixed sampling point probably arose from a different chamber location at each sampling time. Thus, the variation with time might also be viewed as spatial in character, especially after removing the temporal trend. Finally, it is noteworthy that the CV for simulated concentrations varied little with changes in the mixing rate of air enclosed by walls of the alkali trap (5.5, 5.9, and 6.5% at D x 1, D x 3, and D x 10, respectively), so these calculations provided little corroboration of the value we chose for this variable.
We gained some confidence in that choice from the data of Lieth and Ouellette (1962) who reported that after completely sealing their chamber, it took 12 h for the enclosed alkali trap to absorb ambient CO2 from the air trapped inside. Applying assumptions that yielded the solid curves in Fig. 4, we estimated that for our chamber only 1 to 2 h should be required, but after adjusting chamber and trap dimensions according to the data and drawings provided by Lieth and Ouellette (1962), the reduction in headspace CO2 was only 76% after 12 h—in reasonable agreement with their report. For comparison, <5 h was required for similar absorption of the captured CO2 with a D x 10 mixing rate in air enclosed by walls of the alkali trap, and with D x 1 mixing only 65% had been absorbed at 24 h.
In contrast to conclusions drawn from the earlier analysis of Fig. 4a, performance of the NFT-SS chamber used by Lieth and Ouellette (1962) was only very weakly dependent on the rate of air mixing in the bulk headspace. For the D x 3, D x 10, and D x 102 rates, for example, simulated 12-h absorption of ambient CO2 trapped when the chamber was sealed amounted to 74, 76, and 77%, respectively. The reason for this small range is that the alkali trap was tall (15 cm, compared with 5 cm for our chamber) and had small diameter (5 cm, compared with 13.4 cm for our chamber), which boosted the transport limitation imposed by its enclosed air while diminishing the relative importance of that imposed by bulk headspace air outside the trap. Specifically, for our chamber design convectively mixed air inside and outside the alkali trap contributed 44 and 30%, respectively, of the total resistance to CO2 transport from its point of emission at the soil surface to its carbonate form in the alkali trap, but for the design used by Lieth and Ouellette (1962) the contributions were 84 and 4%, respectively. At 24 h, the cumulative Fa/Fo ratio for the latter chamber was only 0.30 (using their dimensions, but other assumptions that yielded the solid curves in Fig. 4), which may at least partially explain why the soil respiration rates reported by these authors are among the smallest in existing literature.
Surface-Atmosphere Interfacial Layers
As specified in the Approach section, we assumed that both the soil surface and alkali solution surface were overlain by a 0.5-cm layer of still air through which gas transport depended entirely upon diffusion. Assigning the same thickness to both layers required further assuming that the tendency for interfacial layer depth to increase with the roughness of the underlying surface (Stull, 1988) was offset by its tendency to vary inversely with the turbulence, or mixing efficiency, in overlying air. Simulation data in Fig. 4c demonstrate that for our chamber design, changes in the thickness of the alkali interfacial layer over the range from 0.1 to 2 cm had substantially smaller influence on Fc, Fa, and headspace CO2 than changes in either air mixing rate presented on the same scales in Fig. 4a and 4b. Its limited influence gave us confidence that the potential for error in our assumptions about the depth of this interfacial layer did not significantly bias our analysis of NFT-SS chamber performance.
Because of its greater surface area, the influence of the soil interfacial layer on chamber performance was even smaller, and it did not represent an important transport limitation compared with the three transport zones described above. For the range of assumptions used in the simulations reported in Fig. 1 and 3, it accounted for only 1 to 5% of the total resistance to CO2 transport. As illustrated for the two air mixing rates, however, the CO2 absorption efficiency of an alkali trap with dimensions different than we used in our simulations will exhibit different sensitivities to changes in the depth of either interfacial layer. All four transport zones within the chamber headspace interact to determine the alkali trap's CO2 absorption efficiency (and thus, NFT-SS chamber performance), but when one is particularly limiting, variation in the others has diminished relative influence.
Finally, we performed additional simulations to determine if the relative influence of the four transport zones was significantly altered by the alkali trap's mounting height within the chamber. Anderson (1982) recommended placing the trap 2 cm above the soil surface, but we found (using assumptions that yielded the solid curves in Fig. 4) that the cumulative Fa/Fo ratio at 24 h varied only from 0.951 to 0.944 when the bottom of our trap was positioned at various heights ranging from 1 to 10 cm above the soil surface. Apparently, an alkali trap's mounting height has relatively little influence on its CO2 absorption efficiency.
Soil Transport Properties
It has long been recognized that the functioning and accuracy of all chamber types depend not only on their physical dimensions and the properties of headspace air, but also on subsurface transport properties that determine soil gas diffusion rates and storage coefficients. For example, the effects of changes in soil AFP and pH are summarized in Fig. 5a and 5b
, respectively.
|
2 h under the conditions assumed in Fig. 5a), the concentrations and fluxes reached new near-steady-state values that depended on the rate of CO2 loss via lateral diffusion beneath chamber sidewalls inserted only 5 cm into the soil. Because that, too, was a function of soil gas diffusivity, accuracy of the resulting flux estimates remained strongly dependent on AFP. Note that Fa for the NFT-SS chamber design used in our simulations represented an unacceptable estimate of Fo at 0.5 AFP, but a very good estimate of Fo at 0.1 AFP (bar graph in Fig. 5a).
When lateral diffusion loss was precluded (the two simulations with chamber walls inserted to the 50-cm depth in the top graph of Fig. 5a), Fc returned asymptotically toward Fo at a rate that also depended strongly on the effective gas diffusivity. For example, when soil AFP was 0.5, the Fc/Fo ratio declined initially to 0.86, then returned to 0.99 at 6 h; in contrast, when AFP was 0.3, the initial drop in the Fc/Fo ratio was smaller (minimum value: 0.92), but 12 h was required for it to return to 0.99. Because of this difference in recovery time, the 24-h Fa/Fo ratio (data not shown) was actually slightly greater at 0.5 AFP, despite that its initial deviation from unity was almost double that at 0.3 AFP. The potential for large losses of CO2 by lateral diffusion emphasizes the need for deep wall insertion when making measurements in soil that is dry or highly porous, as well as in litter or other media with large AFP.
CO2 Solubility and Soil pH
Hutchinson et al. (2000) reported that chamber performance differs with the water solubility of the gas being measured, but no one has examined how that response varies as a function of soil pH when the gas actually reacts with the water. Figure 5b indicates that because of the relatively low solubility of CO2 in water, there was only a small transient difference in simulated CO2 fluxes and concentrations depending on whether the water solubility of the gas was considered. When the potential for reaction of the dissolved CO2 with water was also included, further degradation of chamber performance became a function of soil pH. At pH 6.5, the amount of HCO-3 resulting from this reaction was similar (in molar units) to the amount of dissolved CO2, so the additional decline in chamber performance was also similar in magnitude and was again limited to the early part of the deployment period. However, each unit increase in soil pH resulted in a tenfold increase in the concentration of HCO-3 in equilibrium with the constant amount of dissolved CO2 in soil water, so at the higher pH values typical of saline or calcareous soils, this reaction had a substantial impact on chamber performance.
Note that including these processes in our simulations had little, if any, effect on the magnitude of the postdeployment near-steady-state values of Fc and headspace CO2; instead, they changed only the time required to reach this new near-steady-state condition. Hutchinson et al. (2000) explained that greater water solubility (and in this case, greater reaction of the diffusing gas with water) increases the total amount of CO2 stored in a given soil volume at each value of its concentration in soil air. Because of that increase, a larger quantity of CO2 must be redistributed to overcome headspace concentration feedback effects on the gradient driving CO2 diffusion between its subsurface source and the soil-atmosphere boundary. Thus, greater storage results in poorer chamber performance, because it engenders greater dependence on the comparatively slow rate of gas diffusion in soil.
CO2 Exchange Between the Chamber and its Surroundings
Lateral Diffusion beneath the Chamber Walls
Figure 6a
summarizes how performance of the NFT-SS chamber used in our simulations differed as a function of the depth that its sidewalls were inserted into the soil. When inserted to the impermeable bottom of the simulated domain 50 cm beneath the surface, there was no opportunity for gas exchange between the chamber and its surroundings, but Fc had not returned all the way to its predeployment steady-state value even at 24 h. The reason is that because of the slow rate of gas diffusion in soil, the CO2 in deep soil air was still increasing to the level required to support a surface flux equal to Fo at the new elevated concentration of CO2 in the chamber headspace. When the walls were inserted <50 cm, some CO2 was lost from the chamber system via lateral diffusion. As a result, the return of Fc toward Fo slowed dramatically and would eventually stop when that concentration-dependent loss became equal to the difference between Fo and the rate of CO2 absorption by the alkali trap. The effect of lateral diffusion loss on chamber performance was increasingly delayed at greater wall insertion depths, but the delay was <4 h for even a 25-cm insertion.
|
10 000 µmol mol-1) when the largest soil respiration rate was combined with the smallest alkali:chamber area ratio in Fig. 3. Finally, for the chamber design used by Lieth and Ouellette (1962), which had only a 3-cm wall insertion and a tall alkali trap with small diameter, Fa underestimated Fo by 70% for a 24-h deployment. The greater potential importance of this pathway for CO2 escape from NFT-SS chambers compared with other chamber types results from their much longer deployment time, which allows feedback effects of the chamber to reach greater soil depths.
Leaky Chamber Seals
Like the exchange of CO2 by lateral diffusion beneath chamber sidewalls, its exchange through imperfect seals between chamber components (or other above-surface leakage pathways) may have smaller or greater impact on the performance of NFT-SS chambers than other chamber types. To simulate this impact, we adopted the approach used by Hutchinson and Livingston (2001) who assumed that (i) the seal between the chamber top and its permanently installed base was an imperfect closed-cell foam gasket (0.25 cm wide by 0.25 cm high) with its bottom edge 2.5 cm above the soil surface, and (ii) leakage through the foam seal could be characterized by an effective diffusivity estimated by the same formula used for soil. Just as they did, we set AFP equal to the total connected porosity, which we varied from 0.001 to 0.1 (not including isolated, unconnected pores).
Applying assumptions that yielded the solid curves in Fig. 4, the 24-h underestimate of Fo by Fa was 4.80, 4.83, 5.19, and 9.12% when foam porosity was 0, 0.001, 0.01, and 0.1, respectively (Fig. 6b). The leakage-induced increase in measurement error was only about half that reported for NSS chambers by Hutchinson and Livingston (2001). However, for NFT-SS chamber designs not so well matched to the soil respiration rate they were deployed to measure, the deviation of headspace CO2 from the ambient level, and thus the potential for error due to leakage, were correspondingly larger. For example, when the largest respiration rate was combined with the smallest alkali:chamber area ratio in Fig. 3, the underestimate at 0.1 seal porosity increased to 44% (compared with 9.1% when Fo = 0.050 mg C m-2 s-1 and the alkali:chamber area ratio was 0.2).
Mean headspace CO2 concentration varied relatively little with changes in seal porosity (bottom graph in Fig. 6b), just as it was only weakly dependent on chamber wall insertion depth (bottom graph in Fig. 6a). As a result, there was little interaction between the effects of these two exchange pathways on NFT-SS chamber performance until their sum became rather large. For conditions in the above example from Fig. 3 (Fo = 0.250 mg C m-2 s-1 and alkali:chamber area ratio = 0.05), the 14% underestimate of Fo by Fa for zero foam porosity and 50 cm chamber wall insertion increased to 16% when foam porosity was changed from 0 to 0.01, to 27% if insertion depth was instead reduced from 50 to 5 cm, and to 29% when the two changes were combined. These 2, 13, and 15% absolute increases in the underestimate demonstrate the lack of interaction noted above, but when chamber seal porosity was 0.1 instead of 0.01, the analogous increases were less additive (22, 13, and 30%, respectively). More importantly, about one-half of the measurement error in the first scenario, and about one-third in the second, was due to headspace and subsurface CO2 accumulation that was, of course, not avoided by even perfect chamber seals and infinite chamber wall insertion. This observation again emphasizes the importance of choosing a NFT-SS chamber design having an alkali:chamber area ratio that is appropriate for the soil respiration rate to be measured; that is, a design that minimizes the difference between headspace and ambient CO2 concentrations.
|
RECOMMENDATIONS AND PERSPECTIVE |
|---|
|
TOPABSTRACTINTRODUCTIONAPPROACHRESULTS AND DISCUSSIONRECOMMENDATIONS AND PERSPECTIVEREFERENCES |
|---|
Often, the best approach to a given research objective or monitoring requirement may be a combination of measurements by different chamber types. For example, process studies intended to identify or characterize the sources and controls of soil–atmosphere CO2 exchange are usually best addressed by short-term measurements employing an accurate and portable CO2 analyzer that yields real-time data, but generalizing the resulting conclusions across time and space may be accomplished more efficiently by NFT-SS chamber measurements. The distinguishing feature of the latter approach that makes it well suited to this task is that it represents a simple inexpensive means for obtaining multiple time-integrated measurements, even at remote locations. The same feature makes it a particularly useful tool for studying soil respiration in complex ecosystems with large temporal and spatial variability (Janssens and Ceulemans, 1998; Franzluebbers et al., 2002).
This important advantage of NFT-SS chambers is a byproduct of their long deployment time, which allows ample time for the diffusion gradient of CO2 to adjust to any difference between the subsurface rate of production of the gas and its rate of absorption by the chamber's alkali trap. For these two rates to approach and remain near equality, the chamber's seals must be tight and its walls may need to be inserted to greater depth than for other chamber types. Otherwise, the production and absorption rates may differ significantly due to leakage and/or subsurface lateral gas diffusion. Both these pathways for CO2 loss (gain), as well as measurement error attributable to accumulation (depletion) of the gas within and beneath the chamber, become more problematic with increasing departure from ambient CO2 concentration. Thus, the alkali trap must be designed so that its absorption rate at that concentration matches Fo as closely as possible.
The decision whether to vent a NFT-SS chamber is not so straightforward. If not vented, the chamber is susceptible to the same errors spawned by unavoidable short-term or cyclical disturbances in pressure, volume, or temperature that were described for NSS chambers by Hutchinson and Livingston (2001) and Davidson et al. (2002). However, the quantity of headspace CO2 lost (gained) during the one-time chamber placement and chamber sampling disturbances simulated in Fig. 1 of Hutchinson and Livingston (2001) will have diminished importance when compared with the total amount of CO2 emitted from soil during the typically long deployment period of a NFT-SS chamber. Moreover, most repetitive negative disturbances in pressure, volume, or temperature are likely to be offset by similar positive disturbances during the long period of deployment. In response to each such disturbance in a nonvented chamber, a small amount of headspace air will move upward or downward across the soil surface by mass flow, but most of the CO2 in that air will eventually be absorbed by the alkali trap rather than lost from the chamber system.
Similarly, air expelled from the headspace of a vented chamber during a small increase in its temperature or pressure is captured within the vent tube and then returned to the headspace when its temperature or pressure declines again (Hutchinson and Mosier, 1981). However, if a disturbance-induced change in the volume of enclosed air exceeds the internal volume of the vent tube, there is opportunity for headspace dilution by ambient air. Such a large disturbance becomes increasingly more likely as the deployment period lengthens, because of potentially significant changes in mean air temperature or barometric pressure. We believe that although the weight of evidence strongly supports including a vent tube in the design of all NSS chambers typically deployed for 1 h or less (Hutchinson and Livingston, 2001, 2002), a vent has less advantage and greater disadvantage in a NFT-SS chamber deployed for 24 h, so we recommend its elimination.
Other elements of optimal NFT-SS chamber design not examined in our simulations can be inferred from previous descriptions for NSS chambers. For example, the effects of chamber height on the headspace CO2 accumulation rate, of chamber radius on CO2 loss (gain) by lateral diffusion, of the kinetics or distribution of the gas source, and of chamber-induced changes in atmospheric mixing processes operating near the soil surface are discussed in recent reviews by Livingston and Hutchinson (1995), Hutchinson and Livingston (2002), Rochette and Hutchinson (2003), and references cited therein. These citations may also be consulted for information regarding one- vs. two-component chamber designs, for other details regarding chamber construction, and for acceptable deployment protocol, including suggestions for avoiding site disturbance or pressure disturbance during chamber placement and sampling, guidelines for applying soil amendments, excluding or augmenting natural precipitation, and avoiding changes in the microclimate of soil inside a chamber's permanently installed collar. On the basis of the data in Fig. 2, we propose adding to existing guidelines that NFT-SS chamber measurements with duration <24 h will have different bias depending on the time of day chosen for deployment, so this chamber type is not well-suited for comparing daytime with nighttime soil respiration rates (e.g., Grahammer et al., 1991). The method is probably also not the best choice for studying variability in soil respiration rates, because a NFT-SS chamber designed to measure the mean rate at a particular site slightly overestimates rates smaller than the mean and slightly underestimates rates greater than the mean, thereby masking natural variability.
Unfortunately, a single ideal NFT-SS chamber design and deployment protocol that is applicable in all (or even most) situations is nonexistent. The best design in any given situation is instead a compromise based on thoughtful consideration of all the influencing factors described in this paper and elsewhere. On the other hand, striving for the perfect design in every measurement situation is neither practical nor required; for example, applying assumptions that yielded the solid curves in Fig. 4, 5, and 6, measurement error associated with the chamber used in our simulations was only 5% despite that its headspace concentration increased >70% within 2 h of deployment. Such measurement error is no greater than (sometimes less than) that associated with other chamber types (e.g., Nay et al., 1994; Healy et al., 1996; Pedersen, 2000; Davidson et al., 2002). Apparently, by avoiding serious mismatches between Fo and the alkali trap's CO2 absorption rate at ambient concentration, the careful user can capture the principal advantage of this chamber type (i.e., that it allows time for adjustments in the CO2 diffusion gradient to bring the alkali trap's absorption rate of the gas into approximate balance with its subsurface rate of production) while minimizing the potential for expression of its principal disadvantage (i.e., that it also allows time for substantial net CO2 gain or loss via leakage and/or lateral diffusion).
We freely concede that our conclusions are based primarily on a simplified and idealized model for which assumed values of some input parameters were chosen somewhat arbitrarily. However, we would be just as quick to argue that a twofold (or greater) change in the magnitude of the values most in question would have little impact on our perspective; for example, changing the alkali interfacial layer depth from 0.5 cm to 0.2 or 1.0 cm inconsequentially altered the estimate of overall measurement error from -4.8% to -4.0 or -6.1%, respectively (Fig. 4c). Certainly, we do not propose that the measurement error will always be this small, or that our model results don't embody significant uncertainty. NFT-SS chamber methods are susceptible to error from a large number of sources, but these sources are not independent and their effects are often negatively correlated. For example, in the discussion of Fig. 4 and 6 we pointed out that when one factor is particularly limiting to NFT-SS chamber performance, variation in the others has diminished relative influence.
Our confidence in the simulation model as a whole escalated substantially when we matched its output with data from Rochette et al. (1992)( 1997) and P. Rochette (1992–1994, unpublished data) in Fig. 7
. They compared NFT-SS with multiple FT-NSS chamber estimates of soil respiration on consecutive days at 15 paired sites on sandy loam in each of two different years and at 15 paired sites on an organic soil in each of three different years. Deployment times were 11 h and 2 min, respectively; for more detailed site descriptions and methodological information, see Rochette et al. (1997). Fortunately, they also measured the headspace CO2 concentration in each NFT-SS chamber 5 h after its deployment. Data in Fig. 7 indicate that after changing the dimensions and deployment times of our modeled chamber and trap to match the apparatus and procedures used by Rochette et al. (1992)( 1997) and P. Rochette (1992–1994, unpublished data), the model-predicted relation between Fa and headspace CO2 was indistinguishable from their measurements.
|
Finally, our simulations provide some potentially useful new tools for evaluating previous and future comparisons of NFT-SS chambers with other chamber types. For example, data in Fig. 6a might be used to examine the often-cited laboratory investigations of Nay et al. (1994) for possible error due to lateral diffusion beneath chamber sidewalls inserted only 2.5 cm into a foam slab with relatively large diffusivity. Also, lack of agreement among chamber types in comparisons like the ones that yielded some of the data in Fig. 7 might be explained by the phenomenon described in Fig. 2, if the amount and timing of soil temperature change during each deployment period were known. When evaluating any such intercomparison, it is also important to remember that (i) investigators often do not have the same experience and expertise with all measurement techniques being compared, so the different systems may not be equally optimized, and (ii) the unperturbed soil respiration rate is usually unknown (especially in field studies), making it difficult to know which method most accurately estimates it. With these limitations and the conclusions drawn from our simulations in mind, we think that careful examination of previous intercomparisons might confirm our belief that the potential weaknesses of NFT-SS chamber systems are often overstated. Thus, we also think that this chamber type should always be included among the methods considered for addressing research objectives and monitoring requirements for which it is suited.
Received for publication February 5, 2002.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONAPPROACHRESULTS AND DISCUSSIONRECOMMENDATIONS AND PERSPECTIVEREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  R. T. Venterea and J. M. Baker Effects of Soil Physical Nonuniformity on Chamber-Based Gas Flux Estimates Soil Sci. Soc. Am. J., August 20, 2008; 72(5): 1410 - 1417. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. P. Livingston, G. L. Hutchinson, and K. Spartalian Trace Gas Emission in Chambers: A Non-Steady-State Diffusion Model Soil Sci. Soc. Am. J., August 3, 2006; 70(5): 1459 - 1469. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| The SCI Journals | Agronomy Journal | Crop Science | |||
| Journal of Natural Resources and Life Sciences Education |
Vadose Zone Journal | ||||
| Journal of Plant Registrations | Journal of Environmental Quality |
The Plant Genome | |||
