Carbon cycling in boreal wetlands: A comparison of three approaches

. Three independent methods were used to measure net ecosystem production (NEP) in four wetlands near Thompson, Manitoba, Canada. The first method calculated NEP by subtracting heterotrophic respiration from net primary productivity, using both measurements and estimates derived from the literature. The second method used radiocarbon data from cores to derive long-term NEP averaged over the past several decades. The third method used direct measurement of NEP combined with a model to fill in for days with no data. The three methods, with their independently derived uncertainties, all show the same magnitude and pattern of NEP variation across four different wetland types. However, direct measurement yielded distinctly lower estimates of NEP in the most productive sites. Highest NEP (31 - 180 gC m -2 yr -(cid:127)) was observed in the two wetlands with the highest proportion of sedge vegetation. A bog collapse scar and a nutrient-rich fen had NEP values not significantly different from zero. The maximum NEP at sites with intermediate nutrient status is due to slower overall decomposition and is likely associated with greater allocation of production below ground by sedges. The three methods for estimating NEP differ in the effort required, the sources of error, and in the timescale over which they apply. Used in combination, they allow estimation of parameters such as belowground production and the contribution of heterotrophic decomposition to total soil respiration. Using the radiocarbon method, we also derived estimates of the rate of N accumulation in the four wetland types.


Introduction
Boreal wetlands make up a significant portion of the total carbon storage in soils [Gorham, 1991;Schlesinger, 1996]. In Canada these wetlands represent long-term sinks of carbon as they have accumulated since the retreat of the Laurentide Ice Sheet [Harden et al., 1992]. Wetlands of several types have been shown to be currently accumulating carbon [Gorham, 1991;Turunen and Tolonen, 1996] at rates that average about 30 gC m-2 yr-1. These are higher than C accumulation rates of 3-20 g C m '2 yr -• observed in upland boreal forest sites between fires Rapalee et al., 1998]. Little is known about how the carbon balance of wetlands varies with factors such as vegetation or nutrient status, or the evolution of wetlands in zones of discontinuous permafrost regions [Camill and Clark, 1997].
The annual carbon balance for an ecosystem is expressed as net ecosystem production (NEP), the difference between annual net primary production (NPP) (gross photosynthesis (GPP) minus autotrophic respiration (AR)) and heterotrophic respiration (HR):  Woody stem production was estimated at 20% total woody stem biomass [Wallen, 1992]. Bryophyte productivity was measured either by a cranked wire technique [Rochefort et al., 1990] for Sphagnurn mosses found in the bog and poor fen sites and other mosses with a vertical growth pattern or by Velcro markers for prostrate growth forms, such as "brown" mosses (e.g., Scorpidiurn scorpioides) found in the intermediate and rich fen sites and feather mosses. For each technique, linear increments in growth were measured and converted into mass per unit area using the average bulk density of three replicate 103 cm 3 samples for each species.
Bulk density measurements account for stem density differences among species.
Below ground NPP (BNPP) was estimated at 50% of NPP based on Wallen [1992]. BNPP estimated by Wallen ranged from 50 to 80% of NPP; we have chosen the lower estimate because mosses rather than vascular plant dominate NPP at our study sites. Biomass was converted to grams of carbon by using the following ratios: 0.45 for woody biomass and 0.50 for herbaceous biomass [Gower et al., 1997].
Dark CO2 flux measurements were used to measure total ecosystem respiration. (See Method 3 for chamber techniques). CO2 flux was measured every week from April to the end of October 1996. Peat temperature at 5, 10, 20, and 50 cm depth was measured continuously with thermocouples and averaged every hour at each collar location with CR10 and CR7 data loggers (Campbell Scientific, Inc.). An estimate of hourly CO2 flux from May to October was made using the relationship between CO2 flux and temperature at 5 cm derived from the observed correlation between weekly respiration measurements and temperature. Correlation coefficients (r 2) for each site ranged from 0.57 to 0.88 (p < 0.01). We did several measurements over a 24 hour period to assess differences between daytime and nighttime respiration. We found the relationship between 5 cm temperature and CO2 flux over the day-night cycle was consistent with that found over the whole season. Winter CO2 fluxes (November through April) were estimated to be 10% of the April-October flux, based on measurements made in April 1996 at all the sites, and from winter measurements made in Finnish peatlands [Aim et al., 1997].
Autotrophic respiration from roots and aboveground plant components was estimated to be 40% of total ecosystem respiration based on Silvola et al. [1996] and Bhardwaj [1997]. Heterotrophic respiration (60% of annual CO2 flux) was subtracted from NPP to derive NEP for each site. Losses of C from DOC and CH4 flux are not included in the NEP estimates for method 1.

Method 2, Radiocarbon
Radiocarbon may be used to determine the rates of carbon addition and loss in wetlands during recent decades [Trurnbore and ]. Testing of nuclear weapons in the atmosphere in the early 1960s approximately doubled the amount of radiocarbon in atmospheric CO2. With the cessation of testing in 1964, exchange with ocean and terrestrial carbon reservoirs and burning of 14C-free fossil fuels has caused a steady decline in atmospheric •4CO2.
These changes in atmospheric radiocarbon are recorded in plants.
Year-to-year changes in •4C are large enough that the timing of plant growth may be determined to within a year or two over the past 30 years.
To obtain longer-term records of carbon accumulation for analysis using radiocarbon, we collected samples of the upper meter of wetland peat by freeze coring. Freeze coring avoids problems of compaction in the uppermost layers of peat and moss and therefore is more useful for determining bulk density and preserving stratigraphy. A hollow metal rod 1-2 m in length and sharpened at the bottom end was pushed vertically into the peat. Liquid nitrogen was then vented into the hollow tube for 30 to 40 min. Peat frozen to the rod was removed and cut using knives and saws into squares of known volume for bulk density and water content determination. A split of these samples was ground to <100 mesh and analyzed for bulk organic C and N content using a Fisons NA-1500 combustion analyzer. The raw data are available on the BOREAS web site or at the Oak Ridge distributed data archive [Trurnbore et al., 1998].
Samples for radiocarbon analysis were cut as slices of roughly 1 to 4 cm vertical thickness from the freeze core. Macrofossils of moss, sedge, or seeds were selected as representative of the time of growth for the plants preserved at a given depth. Roots, which penetrate deeply into peat layers, were avoided because they probably postdate the organic matter matrix in which they grow.
To remove surface contaminants, macrofossils were washed with a mixture of 0.1N sodium hydroxide and 0.1N sodium pyrophosphate, then rinsed with distilled water, 0.5N HC1, and finally distilled water. The samples were dried, sealed in evacuated quartz tubes with CuO wire, and combusted at 900øC for 2 hours. CO2 evolved during combustion was purified cryogenically and reduced to graphite for measurement by accelerator mass spectrometry (AMS) at the Lawrence Livermore Laboratory Center for AMS. Radiocarbon data are reported in A•4C notation, as the deviation (in parts per thousand) of the 14C/12C ratio of the sample to that of an accepted standard. Samples are corrected to a common/513C of-25%0. Carbon13 analyses were made only on a subset of samples; for the most part we assumed a value of-27%0. A 2 %0 error in the value assumed for/5 13C gives a 4%0 error in the radiocarbon upon correction. The precision of the radiocarbon measurements reported here averaged _+7 %0 in A•4C.
The year of growth for macrofossils was determined by comparing the measured •4C value with the time series of 14C in atmospheric CO2 since 1963 (see Trumbore and Harden, [ 1997], for details). One potential error in this approach is the possibility that a significant fraction of the carbon fixed by mosses and sedges is derived from respired carbon (which is older and therefore likely higher in •4C than the atmosphere for a given year since about 1970). Measurements of radiocarbon in living mosses and plants were made in 1995 and 1996 to assess this effect, which was variable but generally not significant.
Addition of carbon occurs both at the peat surface and through root production by vascular plants. Decomposition losses of C occur not only in the upper 1-2 m accessible to the freeze-coring method but throughout the peat column [Clymo, 1984]. To determine the long-term rate of C accumulation in peat and the contribution of decomposition in deeper peat layers to CO2 loss from each site, we used a modified piston coring device (Livingston corer) to obtain peat cores of up to 4 m depth. We also sampled a frozen section of peat using a permafrost (Sipri) corer on a nearby palsa (frozen peat upland). As with freeze core samples, radiocarbon analyses were made on macrofossils picked from various depth intervals, while bulk density, %C and %N measurements were made on bulk core samples. Negative A t4C values, which indicate that C was fixed prior to 1960, were converted to calendar ages using established calibration curves. These data are reported by Trurnbore and .
To interpret carbon dynamics from carbon and radiocarbon data, we use a simple first-order model of carbon accumulation and decay. The amount of carbon in a given year (C(t)), equals net annual carbon inputs (NPP) minus losses. We assume loss is dominated by decomposition to CO2 and describe it using a first-order decomposition rate constant, k (units of year'•): The solution to equation (4) is: Time (t) is derived by subtracting the year of plant growth, obtained from radiocarbon, from the year of sampling (1995 or 1996). The amount of carbon accumulated between that time and the present is determined by summing the carbon inventory (obtained from the bulk density and carbon content measurements) for all sampling intervals above the depth in question. A plot of C(t) versus years before present was then made, and bestfit values for NPP and k determined (Kaleidograph version 3.0). Reported errors for NPP and k are the 95% confidence interval for the curve-fit parameters. We also attempted to fit curves of C accumulation over the past 35 years using a relation that allows for decomposition rates to decrease linearly with time [Frolking et al., 1996;Harden et al., 1997]. However, the two methods either showed no difference in their ability to fit the data or the formulation with changing decomposition over time failed to fit the data as well as the simpler model (equations (4) and (5)). It is likely that the time span of 35 years is too short for significant changes in k to be discernable.
Our approach determines time from the radiocarbon signature recorded in hand-picked moss and sedge, while bulk density measurements include roots in addition to decomposed mosses and sedges. Hence the values derived for NPP and k are representative of the average of vegetation which may have different times of origin and rates of decay. Our NPP estimate includes both components that probably grew in situ, such as mosses and sedges, and those that grew subsequently, such as sedge and shrub roots. By assuming single NPP and k values for bulk organic matter we are implicitly assuming (1) that the rates of decomposition of roots, moss, and leaves are all similar and (2) the depth distribution of roots (i.e., the ratio of root to moss or leaf biomass at any given depth) remains approximately constant in time. If these assumptions hold, we may subtract direct measurements of ANPP from our radiocarbon-derived value for NPP to obtain an estimate of BNPP.
Carbon dynamics change vertically in the peat profile. Carbon input and decomposition rates in the less decomposed organic matter near the surface of the peat are approximately an order of magnitude greater than those of the permanently water-logged and more decomposed underlying layers [Clyrno, 1984]. To determine NEP with the radiocarbon method, we need to subtract net respiration of not only the carbon in the upper layers, but also the CO: derived from very slow decomposition of the large mass of material that has accumulated since the deglaciation [Clyrno, 1984;Tolonen and Turunen, 1996;Turunen and Tolonen, 1996): where ksh•ow is the decay constant derived from bomb radiocarbon in the upper portion of the peat (normally <70 cm); C.,.naltow is the amount of carbon for which that decomposition constant applies; ka•p is the decomposition constant obtained for the deeper peat layers, which is roughly an order of magnitude slower than that obtained for nearsurface layers , and Ca•p is the carbon inventory over which the slower decomposition rate applies. The choice of Csh.•ow was made using core descriptions, bulk density, and C/N ratio data. We chose the shallow-deep transition as the depth where peat becomes notably more decomposed, which normally coincided with an To establish relationships between NEE and PAR on each sampling day, shrouds with different mesh sizes were used to reduce the light entering the chamber to one half and one quarter full light. An opaque shroud was placed over the chamber to eliminate all light for measuring ecosystem respiration (combined autotrophic and heterotrophic). In each case the chamber was allowed to equilibrate to the new light or dark conditions. Measurements were made within 2 hours either side of solar noon so that the full range of light responses could be measured. Four 2.5 min. sampling runs at different light levels were conducted at each collar location on a weekly basis from mid-April to mid-October 1996. Measurements of PAR and peat temperature were recorded hourly with Campbell Scientific CR10 data loggers. As mentioned above, we made additional measurements of respiration over the 24 hour period several times during the season to ensure that dark respiration values were comparable to nighttime values (corrected for temperature difference). See Bubier et al. [ 1998] for more detail on sampling methods.
A model, using hourly rates of photosynthetically active radiation (PAR) and temperature at 5 cm peat depth, was developed to calculate hourly rates of photosynthesis and respiration and to interpolate between the weekly measurements of CO2 exchange throughout the entire growing season. Net ecosystem exchange of CO2 (NEE) is the difference between gross photosynthesis and respiration: NEE = mT5cm mTseas at x PAR x GPma x _ 10(ST5cm -0.1) x at x PAR + GPma x

Method 1, Measurements of ANPP and Soil Respiration
The mean (s.d.) NEP for the four sites ranges from -3 (56) g C m '2 yr '• in the bog to 164 (145) g C m '2 yr -• in the intermediate fen site (Table 2)

Method 3, Direct NEP Measurements
The poor and intermediate fens have greater carbon accumulation over the season than either the bog or the rich fen. NEP for the bog (3 + 9 g C m '2) and rich fen (13 + 24 g C m '2) are near zero, with lower confidence limits below zero, suggesting that both of these ecosystems could be losing carbon on an annual basis. The poor fen (65 + 47 g C m '2) and intermediate fen (31 + 14 g C m '2) have higher NEP, with lower confidence limits above zero. Reported errors are 95% confidence intervals.  5.1.1. Method 1. The strength of estimating NEP from measurements of ANPP and soil respiration alone is in its overall simplicity. ANPP may be determined from one-time sampling at the end of the growing season, and components of ANPP may be partitioned among different plant components (sedge, moss, shrub). While soil respiration measurements may be time consuming, the methodology for field sampling is simple. The major weakness is that two very important components, B NPP and the fraction of soil respiration that is heterotrophic, must be estimated from previously published work, often from other locations.
Below-ground production for vascular plant species was estimated to be 50% of total production, at the low end of estimates (which range from 50 to 80% of NPP) for northern peatlands [Wallen, 1986] represented by the freeze-coring method. Results for the rich fen are for the portion of that site dominated by Scorpidium moss. Other parts of the fen complex are drier and have Sphagnurn mosses (with higher NEP) instead of brown moss.

Method 3. Direct measurement of NEP offers the
opportunity to derive potentially extrapolatable, process-based relationships among NEE, PAR, temperature, and water table at fine temporal resolution. The relationships offer the potential for spatial and temporal scaling and are useful for predicting changes in C accumulation with changes in temperature and moisture balance on seasonal to interannual timescales. The major disadvantage is that chamber methods are labor intensive. In addition, continuous measurements of PAR, temperature, and moisture are needed at each site. Uncertainties in the interpolation of weekly NEE measurements to derive NEP using hourly measurements of PAR and peat temperature include standard errors in the parameters GPmax, o•, and s. The uncertainties are greatest at maximum temperatures because the respiration-temperature relationship is a log-linear one [Bubier et al., this issue].
Method 3 has the lowest estimates of NEP in the poor and intermediate fen sites. One explanation for this is that the modeling approach we used underestimates maximum photosynthesis (or GPP) during the summer. The interpolation (equation (7)

Differences Between Sites
The upper and lower confidence limits of NEP show that the minimum values for the bog and rich fen are negative for all three methods, indicating that these ecosystems could be net sources of carbon to the atmosphere, although over the last several thousands of years they have been net sinks of carbon (3-5 m of peat accumulation). In the dry late summer period of 1994 when water tables fell to 20-30 cm below the peat surface, these two sites were net sources of carbon based on chamber [Bellisario et al., 1998] [Maimer, 1986], this is counter to observations of slower decomposition in bogs than fens made in the same studies. At the collapse bog site, the average depth to the water table is greater than at the other sites (Table 1). A larger aerobic zone may lead to faster decomposition rates.
Heterotrophic respiration as calculated from method 2 (radiocarbon; equal to kshallowCsha!1o w + kdeepCdeep ) ranges from 46 to 78% (average 60_+14%) of total dark respiration measured at each site (Table 1). This is in accord with the literature-based estimate of 60% heterotrophic respiration used in method 1. Respiration rates are lowest at the bog collapse even though decomposition rates are fast because of overall lower productivity. The apparent rate of annual N accumulation calculated in Table 3 ranges from 0,3 g N m -2 yr '• in the bog up to 3-7 g N m '2 yr '• in the fens. The rate of N accumulation in the bog, which should derive its N mostly from atmospheric N inputs, is close to estimated atmospheric N deposition rates (0.2 to 0.3 g N m '2 yr'•). However, the annual amount of N added as fresh plant residues is significantly larger (1.4 g N m '2 yr'l). If new N inputs are derived from the atmosphere, this difference would indicate that -80% of the N taken up by plants annually comes from decomposition of plant residues.

Calculation of BNPP From ANPP and NPP by Combining Methods
If we assume the NPP values obtained using the radiocarbon method are correct, we may obtain an estimate of belowground production by subtracting the measured values of ANPP (method 1). Figure 4 shows Figure 4. Comparison of estimates of net primary production (NPP) using methods 1 and 2. The breakdown of above-ground net primary production (ANPP) into different components is shown. Total NPP is calculated in method 1 assuming that below-ground production is 50% of ANPP. This value compares favorably with the NPP derived from method 2 (NPP from •4C) except for the poor fen. NPP) is comparable to that obtained from method 2. Within stated errors, NPP values are the same for both methods across sites, which indicates the assumption of BNPP=ANPP made in method I is likely reasonable. The implied BNPP/NPP ratio derived by subtracting measured ANPP (method 1) from total NPP (method 2) ranges from 0.39 to 0.93, with the highest value at the poor fen site. Since -50-70% of the ANPP is in mosses, the woody and sedge/herb plants must allocate a significant portion of their production belowground. Saarinen [1996] found that roots of Carex rostrata (the species found at our intermediate fen site) penetrated deep into peat (230 cm) and accounted for 74% of total production.

Conclusions
Three independent methods yield the same magnitude and pattern of NEP at four wetland sites along a trophic gradient in the BOREAS northern study area. Carbon accumulation rates were greatest at the intermediate and poor fen sites, largely due to slower decomposition rates of C in the upper layers of peat. Carbon accumulation rates were near zero for both a bog (low inputs matched by low outputs) and a rich fen (high productivity and respiration).
Each method used to determine NEP has significant advantages, disadvantages, and sources of uncertainty. The use of more than one method allows for estimation of parameters that are difficult to measure directly, such as the fraction of NPP that is belowground (BNPP) and the fraction of total respiration that is heterotrophic (HR). Because the measurements being compared represent different timescales of averaging (days to months for methods 1 and 3, decades to millennia for method 2), these comparisons must be made cautiously.
Decadal-average nitrogen inputs, losses, and accumulation were also estimated using method 2. Fens accumulate N at least 10 times faster than the bog, which was accumulating N at approximately the same rate as the supply from the atmosphere.