WETLANDS, Vol. 22, No. 1, March 2002, pp. 100–110 q 2002, The Society of Wetland Scientists
MODELING CARBON ACCUMULATION IN ROCKY MOUNTAIN FENS Rodney A. Chimner1,4, David J. Cooper1,2, and William J. Parton3 1 Graduate Degree Program in Ecology Colorado State University Ft. Collins, Colorado, USA 80523 Department of Earth Resources Colorado State University Ft. Collins, Colorado, USA 80523 2
3
Natural Resources Ecology Laboratory Colorado State University Ft. Collins, Colorado, USA 80523 Present address Institute of Pacific Islands Forestry 1151 Punchbowl Street, Room 323 Honolulu, Hawaii USA 96813 E-mail:
[email protected] 4
Abstract: Despite the importance of peatlands in the global carbon cycle, no widely applicable ecosystem model exists for peatlands. Simulations of three montane fens in Colorado, USA were conducted to test the capabilities of the CENTURY ecosystem model to simulate 1) long-term carbon accumulation and 2) short-term changes in carbon accumulation due to hydrologic changes. The CENTURY model was calibrated to simulate long-term carbon accumulation in two fens for up to 10,000 years by adjusting three variables that represent anaerobic soil conditions. CENTURY was unable to simulate long-term carbon accumulation in a third fen using settings for the two calibrated fens. However, CENTURY correctly simulated total carbon storage by adjusting two of the three anaerobic variables. A sensitivity analysis revealed that carbon accumulation in CENTURY is highly sensitive to anaerobic soil conditions. CENTURY predicted that half of the fen peat is composed of structural root material. The majority of the remaining peat was composed of recalcitrant slow and passive soil organic matter. Precipitation levels were altered to determine if CENTURY could predict the change in carbon accumulation rates due to periodic drier conditions. The simulated drying scenario predicted an average carbon loss of 70 g C m22 yr21 during the 100-year simulation. The loss of carbon occurred despite plant production increasing from an average of 249 g C m22 yr21 to 391 g C m22 yr21. Slightly more than 90% of the carbon lost was from the structural root pool and slow organic matter pool, while there was no carbon loss or a slight net carbon gain in the passive organic matter pool and above-ground structural and metabolic pools. Despite several shortcomings, our results indicate that an ecosystem model, such as CENTURY, can be useful for simulating carbon dynamics in peatlands. Key Words:
peatlands, fens, Colorado, Rocky Mountains, CENTURY, carbon accumulation, modeling
INTRODUCTION
1995, Lappalainen 1996, Clymo et al. 1998), despite occupying only 3% of the land surface (Maltby and Proctor 1996). Peatlands have accumulated carbon over thousands of years, yet this carbon can be oxidized if peatlands are hydrologically modified by drainage ditches or reduced precipitation (Armentano and Menges 1986, Gorham 1991, Silvola et al. 1996, Nyka¨nen et al. 1998, Chimner 2000). The sensitivity of peatlands to changing climate or hydrologic disturbance is especially important in high latitude regions
In a time of changing climate and increasing concentrations of atmospheric greenhouse gases, understanding peatland carbon cycling processes is crucial (Gorham 1991). Peatlands function as long-term sinks of atmospheric CO2 because plant production exceeds decomposition. Peatlands store between 224 and 455 Pg (1 Pg 5 1015g) of carbon, 12–30% of the current global soil carbon pool (Gorham 1991, Botch et al. 100
Chimner et al., MODELING FENS where the greatest concentration of peatlands occur, and where global climate change scenarios predict the greatest temperature increases and hydrologic changes (IPCC 1995). Despite the global importance of peatlands as a major sink or potential source of carbon, they have been neglected in carbon modeling (Maltby and Procter 1996). Concerted efforts have been made to develop models to simulate soil carbon dynamics in many other ecosystems, including grasslands, croplands, and forests, but few ecosystem models exist for peatlands. The most widely used method of modeling peat accumulation is Clymo’s (1984) theoretical mathematical decay model, which back-calculates plant production and decay with dated peat cores and cumulative carbon increases in the profile. It has been used to model longterm peat accumulation in several peatlands (Warner et al. 1993, Charman et al. 1994, Belyea and Warner 1996, Clymo et al. 1998). However, there are inherent drawbacks in using this model. First, it can only be used to explain peat growth after it has happened. Second, it cannot be used to make predictions based on environmental changes, such as in soil or air temperature or hydrologic regime. Predictive, process-based models are necessary to understand the potential effects of climate change, physical disturbances, drainage, or pathways for increased carbon sequestration. There are two approaches for overcoming the lack of peatland models. One approach is to develop new models specifically designed for peatlands (e.g., Frolking et al. 2001), while another approach is to modify existing models for use in peatlands. The objectives of this paper are to test whether the CENTURY ecosystem model could simulate 1) long-term carbon accumulation and 2) short-term changes in carbon storage in response to hydrologic changes in three fens in the Southern Rocky Mountains of Colorado, USA. Although CENTURY was originally developed as a grassland ecosystem model and has never been used to simulate wetlands, CENTURY’s ability to simulate long-term soil carbon dynamics successfully (Kelly et al. 1997, Smith et al. 1997) and its adaptability to many ecosystem types make CENTURY a logical choice to use for attempting to simulate soil carbon accumulation in Colorado sedge-dominated fens. MODEL DESCRIPTION A detailed description of the most recent version of the CENTURY model is presented in Parton et al. (1993), and only a brief summary is provided here. CENTURY is an ecosystem model that simulates soil organic matter (SOM) dynamics on a monthly time step with the major input variables being monthly precipitation, monthly average maximum and minimum
101 air temperatures, soil texture, atmospheric and soil nitrogen inputs, and lignin and nitrogen content of plant material. CENTURY uses a plant submodel to simulate plant production that cycles into the SOM submodel that simulates carbon and nitrogen cycling in the soil (P and S cycling modules are also available). Plant production and carbon cycling are modified by soil temperature and soil wetness, which are simulated in the hydrology and soil temperature submodels. The plant production submodel can simulate grass, tree, agricultural crop, and savanna production. CENTURY simulates grass growth by defining a potential growth (genetic maximum) and decreasing that maximum growth rate by scaling factors. Some of the important scaling factors are soil temperature, soil moisture status, and nutrient availability. Grass is subdivided into readily decomposable material (metabolic) and more resistant material (structural) for live shoots, standing dead shoots, and roots. The SOM submodel uses a series of functional pools to categorize various stages of plant litter and soil organic matter decomposition (Figure 1). Above- and below-ground plant litter are partitioned into either the structural or metabolic pools based on the lignin to nitrogen ratio. The greater the lignin concentration, the greater the partitioning into the structural pool, and the slower the turnover rate compared with the metabolic pool. The surface microbial pool consists of dead microbes and microbial residues that result from decomposing the plant litter and has a very fast turnover time. Soil organic carbon decomposition is simulated via three-pools: active, slow, and passive organic carbon pools (Figure 1). The active pool (SOM1C) consists of microbes and microbial byproducts associated with SOM decomposition and has a turnover time of less than one year. The slow (SOM2C) pool includes plant and microbial products that are biologically resistant to decomposition and has a turnover time in the tens of years. The passive pool (SOM3C) is very resistant to decomposition and includes stabilized biological products that are chemically recalcitrant or physically protected, with a turnover time of hundreds to thousands of years. The modeled turnover time of these pools is a function of the abiotic decomposition factor (DEFAC), which is controlled by soil temperature, nutrient and water availability, and anoxic conditions. CENTURY uses a soil temperature and hydrologic submodel to simulate the soil temperature and moisture status, both of which affect plant growth, decomposition, and nutrient cycling. Soil temperature is calculated as a monthly average of the near soil surface using monthly minimum and maximum air temperatures. Either monthly weather data or mean monthly weather data can be used in CENTURY. CENTURY
102
WETLANDS, Volume 22, No. 1, 2002
Figure 1.
Carbon pools and flows in the CENTURY model (Metherell et al. 1993).
also can run in a stochastic mode that varies monthly precipitation around the mean. CENTURY uses a simple hydrologic model to calculate soil moisture in different soil layers by using precipitation in the weather file as the major water input and calculating monthly evaporation, transpiration, leaching, and stream flow as outputs. Because CENTURY was developed as an upland model, there are no simulated water-table levels. Precipitation occurs as snow when the average monthly air temperature is below 08 C, and snow melts when the average air temperature is above 08 C. Soil water is routed via a bucket method between soil layers by adding water at the top soil layer, with any water above field capacity being passed on to the next soil layer and so on. Water draining below the bottom layer of soil can be lost as fast storm flow, leached into the subsoil and accumu-
late, or released more slowly as streamflow. The default value of six soil layers was used, with the top four layers being 15 cm thick and the bottom two layers being 30 cm thick, for a total depth of 1.2 meters. Anaerobic soil conditions in CENTURY are determined by a ratio of rainfall to potential evapotranspiration (PET). Two parameters, ANEREF1 and ANEREF2, are used to set the occurrence of anaerobic conditions in relation to the rain:PET ratio. When rain: PET is larger than ANEREF2, then full anaerobic conditions occur. When rain:PET is smaller than ANEREF1, then aerobic conditions occur. If the rain: PET ratio is between ANEREF1 and ANEREF2, then partial anaerobic conditions occur. Aerobic conditions do not change decomposition rates, but anaerobic conditions cause decomposition rates to decrease. The more anaerobic the conditions are, the lower the de-
Chimner et al., MODELING FENS
103
Table 1. Characteristics of three study fens used for modeling. Age 5 age of basal peat in 14C calibrated years BP (rounded to nearest 25 years). NPP 5 net primary production. MAT 5 estimated mean annual temperature and MAP 5 estimated mean annual precipitation.
Site Caribou Keystone Zapf’s
Eleva- Peat tion Depth (m) (m) 3400 2920 2725
1.9 2.0 1.3
Age (years)
NPP (g C m22 yr21)
MATa MAPa (8C) (cm)
10,525 9,000 5,000
110b 283c 316c
20.8 0.5 2.7
62 73 50
Adjusted for difference in elevation between site and weather station. Chimner unpublished data. c Average values from sites with similar vegetation (Chimner 2000). a
b
composition rates. The amount that anaerobic decomposition decreases is set by the anaerobic decay modifier ANEREF3, which scales down the maximum decomposition rate (DEFAC) as calculated by CENTURY. METHODS Caribou and Zapf’s fens in Colorado’s Front Range were used for the calibration phase of the modeling (Table 1). Caribou fen is located on Arapaho Pass on the west side of the Continental Divide at 3,400 m elevation. The peat is 190 cm thick and has a basal date of 10,525 years BP. The vegetation is dominated by Carex aquatilis Wahlenberg, C. nigricans Meyer and Eleocharis quinqueflora Boott. Zapf’s fen is at 2725 m on the east side of the Continental Divide and has a peat body 130 cm thick and a basal carbon date of 5,000 years BP. Its vegetation is dominated Carex utriculata (Hartman) Schwartz. Little moss cover occurs at either site. Both peatlands were cored and dated by Jim Benedict in the mid-1980s (Benedict and Maher, unpublished data). Samples every 20 cm through the Caribou fen core were AMS 14C dated, for a total of 10 dates, while Zapf’s fen was bulk carbon 14C dated every 20 cm, for a total of 7 dates. The sites were revisited in the fall of 1996 when two peat cores were collected from each site at the same locations as the cores collected for carbon dating. Each peat core was cut into 5-cm-thick sections, oven-dried for two days at 1058 C, and ashed to determine mineral-free bulk densities (Belyea and Warner 1996). Carbon storage was calculated by multiplying the mineral-free bulk densities by depth. The carbon content of peat varies but has been found to average approximately 50% (Moore 1989, Ovenden 1990, Gorham 1991, Belyea and Warner 1996, Rob-
inson and Moore 1999). Caribou fen had a 10-cmthick mineral sediment lens at 10–20 cm depth that was not included in the analysis. After CENTURY was calibrated using Caribou and Zapf’s fens, a fen located in the West Elk Mountains near Crested Butte in western Colorado was used to test whether CENTURY could predict peat accumulation in an uncalibrated fen. The fen, Keystone Ironbog and here referred to as Keystone fen, was chosen because data were available from a previous analysis of fossil pollen and included carbon dates (Fall 1997a,b). Keystone fen is located at 2,920 m elevation, intermediate in elevation between Zapf’s and Caribou fens (Table 1). The dominant plant species at Keystone are Carex aquatilis, C. utriculata and Eriophorum angustifolium Honckeny (Fall 1997a); in contrast to the other sites, Sphagnum mosses are abundant. Keystone fen has one meter of sedge peat over one meter of woody and highly humified peat (Fall 1997b). A basal peat sample from near the peatland center was originally bulk-dated at 7,100 yr B.P. (Fall 1997a), but this sample appeared to be contaminated with younger carbon. Fall (1997b) calculated that the basal date was approximately 9,000 yr B.P. by AMS-dating wood fragments in the peat body, developing an age-depth relationship, and dating the insoluble fraction of the bulk sample. No bulk density data were available for Keystone fen, so the average bulk density value of Caribou and Zapf’s fen was used (0.25 g/cm3), and carbon content was estimated at 50%. The use of an average bulk density from the other sites to calculate total carbon at Keystone fen might underestimate the total carbon due to the presumably higher bulk density in the lower meter of humified woody peat. However, the total carbon would probably not change drastically due to the generally low bulk densities found in most Colorado peats. MODEL PARAMETERIZATION AND CALIBRATION The CENTURY agroecosystem version 4.0 (Parton et al. 1993) was used in the grassland mode for all simulations. Several adjustments were required to configure CENTURY for peatland modeling. The DRAIN variable, which controls how much water drains from the soil, was set to 0 (1 5 maximum drainage and 0 5 no drainage) to maximize water storage and anaerobic conditions. The baseflow and stormflow variables were also set to 0 to limit the amount of water leaving the sites. CENTURY has no soil texture parameter that can be used to define an organic soil. Sandy soils eliminate texturally mediated carbon transformations in CENTURY. Raich et al. (2000) parameterized CENTURY
104 for Hawaii’s organic soils to be 98% sand, 1% silt, and 1% clay, with a bulk density of 0.15 g cm23 and predicted that little physical stabilization of soil C and little passive soil organic C formation should occur. We used a similar sand texture but used our measured bulk density values of 0.20 g cm23 and 0.30 g cm23 for Caribou and Zapf’s fens and 0.25 g cm23 (the mean of these two fens) for Keystone fen. All simulations were run in the mean weather mode. The Saddle Ridge weather station on Niwot Ridge was used for Caribou fen simulations, although it is 125 m higher in elevation and located in alpine tundra (Table 1). The Nederland weather station (USWS station # 55878) was used for Zapf’s fen simulations; it is 210 m lower in elevation than Zapf’s fen and is located 16 km to the south. The Crested Butte weather station (station # 51959) was used for Keystone fen simulations; it is 216 m lower in elevation than the fen and is 7 km away. Since all three weather stations were at different elevations than the study sites, temperature and precipitation were corrected using lapse rates calculated by Fall (1997a). Mean annual temperature decreases by 6.08 C and mean annual precipitation increases by 22.5 cm for every 1000 meters of elevation gain. All simulations used Niwot Ridge tundra vegetation type, which is calibrated for the wetland grass Deschampsia cespitosa (L.) P. Beauvois (Webber and May 1977, Baron et al. 1994). Although D. cespitosa is a grass and our study sites are dominated by Carex aquatilis, we felt that the vegetation parameters were close enough between the species for the purposes of this simulation. The root-to-shoot ratio was set to 0.5 to reflect the average values measured for two sedge species at several nearby fens, including one of the study fens (Chimner 2000, Chimner unpublished data). The monthly above-ground genetic maximum potential growth rate (PRDX(1)) was left at 150 g C m22. The minimum C:N ratios are 25–35 (PRAMN), and the maximum C:N ratio (PRAMX) is 35–60. Plant production values are not needed for CENTURY because it is simulated, but we tried to be certain that simulated plant production values were close to measured values. Measured plant production values were made only at Caribou fen in 1997 (Chimner unpublished data) (Table 1). No plant production data were available for either Zapf’s fen or Keystone fen. We used plant production data from two other Colorado fens at similar elevations and with similar vegetation to approximate plant production for Zapf’s and Keystone fens. We used NPP values from Green Mt. Pond in 1997 (Chimner 2000) for Zapf’s fen and the average 1997 and 1998 NPP from Big Meadows for Keystone fen (Chimner 2000). CENTURY simulated plant production close to estimated values, and only minor ad-
WETLANDS, Volume 22, No. 1, 2002 justments were made by setting the amount of symbiotic nitrogen fixation (SNFXMF) to 0.002 (g N fixed per g C fixed) to reflect NPP values of Caribou fen and Zapf’s fen. It is unknown how much nitrogen fixation, if any, actually occurs at these sites, but this small amount of nitrogen fixation is reasonable for nitrogen fixation rates in peatlands (Waughman and Bellamy 1980). Default values for nitrogen fixation in other vegetation types in CENTURY range from 0 to 0.0375 (g N fixed per g C fixed). Long-term carbon accumulation rates were calibrated in CENTURY by altering the three anaerobic variables ANEREF1, 2, and 3 and comparing the predicted long-term carbon accumulation rates to measured values. The first step was to determine the value of ANEREF3, which determines the influence of soil anaerobic conditions on decomposition. This was done by creating anaerobic conditions for the entire soil profile at Caribou and Zapf’s fen by setting ANEREF1 and ANEREF2 to 0.1 and 0.2, respectively. The anaerobic decomposition variable ANEREF3 was then adjusted for each Caribou and Zapf’s fens until CENTURY correctly simulated total carbon (SOMTC) measured for each site. To accumulate sufficient carbon, ANEREF3 was lowered to 0.008 and 0.04 for Caribou and Zapf’s fens, respectively. The large difference in ANEREF3 between Caribou and Zapf’s fens indicates that either anaerobic decay rates differ or the frequency of aerobic conditions differ between these peatlands. Because one objective was to determine if CENTURY could be calibrated for several fens using the same variables, we assumed that anaerobic decay processes are similar between peatlands but that real differences occur in the frequency and duration of aerobic conditions. Caribou fen is 700 m higher in elevation, with much greater snowpack and greater summer precipitation, as well as lower air and soil temperatures and evapotranspiration rates. Zapf’s fen is at the low elevation end of peatland distribution in Colorado, with greater evapotranspiration rates and air and soil temperatures and lower precipitation. Having visited all sites, it appears that water tables are relatively stable at Caribou and Keystone fens but may drop by several cm to dm during the summer at Zapf’s fen, which leads to increased frequency and duration of aerobic upper soil horizons. We used these observations to calibrate Zapf’s fen for greater water-table fluctuations. The second step was to determine the values for ANEREF1 and 2, which account for periods of anaerobic conditions. Using the hydrologic information presented above, we set the anaerobic decay variable ANEREF3 of Zapf’s fen to that of Caribou fen (0.008) and altered the duration of anaerobic conditions at Zapf’s fen by altering the ANEREF1 and 2. This cre-
Chimner et al., MODELING FENS
105
Table 2. Default and final values of anaerobic variables used for simulations. Variable
Default
Caribou
Zapf’s
Keystone
ANEREF1 ANEREF2 ANEREF3
1.5 3.0 0.30
0.35 1.1 0.008
0.35 1.1 0.008
0.30 0.95 0.008
ated similar anaerobic decay rates at Zapf’s and Caribou fens, with a seasonal water-table decrease. By altering ANEREF1 and ANEREF2 for Zapf’s fen, the best fit for total carbon was found by changing ANEREF1 from 1.5 to 0.35 and ANEREF2 from 3.0 to 1.1. The final values for ANEREF1–3 for Caribou and Zapf’s fens are given in Table 2. The final simulation for Caribou and Zapf’s fens are shown in Figures 2 and 3, respectively. CENTURY under-predicted carbon accumulation rates for the time period 3,000 to 8,000 years BP for Caribou fen but closely predicted the first 1,000 years and the ending total carbon value (Figure 2). CENTURY over-predicted carbon accumulation rates in Zapf’s fen for the time period 3,000 to 4,500 year BP but also better-predicted the first 1,500 years and end value (Figure 3). Long-term peat accumulation rates were simulated using current mean weather, which does not allow for climate variability over time. Incorporating paleoclimatic conditions might create a more realistic model of long-term carbon accumulation. APPICATION OF THE MODEL We tested the calibrated CENTURY for predicting carbon storage in Keystone fen. Measurements of soil carbon showed that Keystone fen accumulated approximately 200 kg C m22 over 9,000 years. CENTURY predicted that 56 kg C m22 would accumulate and over-predicted plant production at 394 g C m22 yr21
Figure 2. Modeled versus measured total carbon (SOMTC) for Caribou fen, R2 5 0.95.
Figure 3. Modeled versus measured total carbon (SOMTC) for Zapf’s fen, R2 5 0.94.
(Figure 4). This suggests that the ANEREF1 and 2 variables calibrated using Caribou and Zapf’s fens did not produce a successful model for Keystone fen, most likely because CENTURY predicted longer duration periods of aerobic soils than occur. By changing ANEREF1 from 0.35 to 0.3 and ANEREF2 from 1.1 to 0.95 (Table 2), CENTURY could correctly simulate total carbon storage (Figure 4). This indicates that CENTURY is highly sensitive to very small variations in anaerobic soil conditions. An analysis was performed to determine how sensitive total soil carbon and plant production are to variables ANEREF1 and ANEREF2 for Keystone fen. Total soil carbon (Figure 5A) and plant production (Figure 5B) are sensitive to anaerobic variables, especially to ANEREF2. Total soil carbon accumulation was approximately 400 g C m22 when ANEREF2 was set below 0.8 but dropped 25% when ANEREF2 was set at 0.9 and dropped an additional 77% (to ;70 g C m22) when ANEREF2 was set at 1.0. Plant production
Figure 4. Modeled versus measured total carbon (SOMTC) for Keystone fen. Initial simulation using paramterizaions developed from Caribou and Zapf’s fens (dotted line), and final simulation using only paramterizations for Keystone fen (solid line), R2 5 0.99.
106
Figure 5. Sensitivity analysis for Keystone fen (A) total soil carbon pool (SOMTC) and (B) plant production (CPRODA).
had the opposite trend, maintaining 270 g C m22 yr21 when ANEREF2 was set at #0.8 but increasing when ANEREF2 was .0.8. Changing ANEREF1 had little affect on soil carbon accumulation or plant production. Two simulations were run using the final calibrations (Table 2) for each fen to investigate 1) the physical composition of fen peat and 2) the effect of altering precipitation on carbon accumulation rates. Figure 6(a-c) shows carbon stored in different carbon pools as simulated by CENTURY. For all three sites, the below-ground structural pool (STRUCC2) comprised an average of ;55% of the total soil organic carbon pool (SOMTC). The second largest carbon pool is the slow pool (SOM2C), and the third largest is the passive SOM pool (SOM3C). The SOM2C pool is likely larger than the SOM3C pool due to less physically
WETLANDS, Volume 22, No. 1, 2002
Figure 6. Different soil carbon pools for (A) Caribou fen, (B) Keystone fen, and (C) Zapf’s fen using values from Table 2. STRUCC(2) 5 structural belowground material, SOM2C 5 slow organic pool, and SOM3C 5 passive soil organic pool. All other soil pools combined (above-ground structural, above-ground metabolic, surface microbial, below-ground metabolic, and active organic carbon) make up on average less than 2% of the carbon stored and are not shown.
protected SOM forming from the low clay content and the low anaerobic decomposition rates that stops the decomposition process from proceeding to the SOM3C pool. The slow SOM pool is intermediate in decomposability, and the passive SOM pool is the most resistant form of soil carbon composed of physically and chemically resistant material (Figure 1). The slow and passive pools are derived from the most resistant portions of the litter and active SOM pools. The sum of the remaining carbon pools (aboveground structural, aboveground metabolic, surface microbial, belowground metabolic, and active organic carbon) (Figure 1) make up on average of less than 2% of the carbon stored and are not shown in the figure. To explore the usefulness of an ecosystem model for carbon accumulation in peatlands, we altered precipitation levels in CENTURY for Zapf’s fen to de-
Chimner et al., MODELING FENS
Figure 7. (A) Total soil carbon and (B) plant production and total soil mineral nitrogen of Zapf’s fen during 100 years of simulated drier conditions.
termine if CENTURY could predict changes in carbon accumulation rates. This scenario was run for 5,000 years using the mean weather mode, with all years being wet, and the variables set as in Table 2. Then for an additional 100 years, the weather mode was changed to stochastic. This varied monthly precipitation totals based upon monthly standard deviations, which altered the rain:PET ratio and allowed random dry months to occur. Zapf’s fen accumulated 109 kg C m22 over the first 5,000 years, an average of 19.5 g C m22 yr21, but then lost 7 kg C m22 over the next 100 years (Figure 7A). The loss of carbon occurred despite plant production increasing from an average of 249 g C m22 yr21 for the first 5,000 years to 391 g C m22 yr21. Plant production increases were due to increased organic matter mineralization releasing nitrogen (Figure 7B). Slightly more than 90% of the carbon lost was from the below-ground structural (STRUCC2) and slow pools (SOM2C), while there was no carbon loss or a slight net carbon gain in the passive pools and above-ground structural and metabolic pools. In this model run, 38 out of the 100 years had annual precipitation lower than the 5000-year average, but conditions were actually drier than the annual precipitation
107
Figure 8. (A) Rain:PET ratio and (B) ANERB scaling factor for mean weather and a random stochastic year. The lower the value for ANERB scaling factor, the more anaerobic the conditions.
would indicate. Almost every month in the mean weather mode was completely anaerobic, but in the stochastic mode, at least one month in every year was drier than normal, allowing more aerobic conditions to occur (Figure 8). Therefore, even though annual precipitation might be greater than average, there could still be one to several dry months with aerobic conditions. Wetter than average months had no additional effect on anaerobic conditions because the mean weather was already set for maximum anaerobic conditions; anything wetter was still at maximum anaerobic conditions. Therefore, stochastic precipitation could only make things more aerobic in this simulation. DISCUSSION The CENTURY model successfully simulated longterm carbon accumulation in fens for a period up to 10,000 years by calibrating the three variables that represent anaerobic soil conditions. Total soil carbon and, to a lesser extent, plant production were very sensitive to ANEREF2 yet only slightly sensitive to ANEREF1 (Figure 5). ANEREF2 sets the upper soil profile limit of anaerobic conditions. Therefore, soils are anaerobic
108 throughout when the rain:PET ratio is greater than ANEREF2. However, decomposition rates increase as the rain:PET ratio drops below ANEREF2. Total carbon storage and plant production had opposite trends. Perennial anaerobic soils decreased decomposition rates and increased soil carbon accumulation rates. However, decreasing decomposition decreased plant production because nutrients were bound up in the peat. Aerobic conditions increased decomposition rates, increasing the availability of nutrients, especially nitrogen, which increased plant production. The calibration of CENTURY for two Colorado fens did not allow a successful simulation of a third fen with different site variables. This is because the method for calculating the timing, duration, and depth of anaerobic soil conditions is not easily manipulated in CENTURY. CENTURY was developed for nonwetland ecosystems that do not have high water tables and anaerobic soils. The model allows anaerobic conditions to occur only when an excess of precipitation over PET occurs. However, fens are driven by groundwater inflow (Siegel and Glaser 1987) that cannot currently be accounted for in CENTURY. Ground-water inflow to a fen is a complex process controlled by many factors, such as the amount and timing of precipitation amounts, hydraulic conductivities of aquifers and peat, and the size and topography of the watershed. The variability of these factors makes each fen hydrologically distinct, suggesting that it may be impossible to use a single rain:PET ratio to simulate complex hydrologic processes. CENTURY predicted that the majority of fen peat is below-ground structural material, suggesting that little above-ground material is incorporated into the peat. Above-ground decomposition rates are great enough that most structural and metabolic material is oxidized. The frequency and duration of aerobic conditions below-ground is much lower than above-ground, yet the below-ground decomposition rates are still great enough to decompose all belowground metabolic material (METABC2), which can be released as anaerobically produced CH4 (Thomas et al. 1996). If fen peat accumulates primarily from belowground production as CENTURY predicts, then the pattern of peat formation is another important distinction between herbaceous fens, such as were studied here, and poor fens and bogs that are dominated by Sphagnum spp. A large proportion of herbaceous fen biomass and net primary production is below-ground (Reader and Stewart 1972, Francez and Vasander 1995, Saarinen 1996, Chimner 2000), whereas Sphagnum-dominated peatlands have the bulk of their living biomass and net primary production above ground (Reader and Stewart 1972). Bogs are often cited as being created from the top down for this reason (Cly-
WETLANDS, Volume 22, No. 1, 2002 mo 1984, Clymo et al. 1998). Clymo (1984) suggested that carbon is added to the upper peat profile (acrotelm) through net primary production, where it undergoes relatively rapid decomposition, and less than 20% is incorporated into the lower peat profile (catotelm). This framework of peat formation is likely correct for Sphagnum-dominated peatlands but most likely is not correct for many herbaceous fens. Fens with little moss production are built from within the soil, as most root production occurs in the acrotelm and upper catotelm where organic matter is not exposed to high decomposition rates. Because fen peat largely originates as below-ground structural material and slow SOM pool, it is vulnerable to water-table lowering and the development of aerobic conditions that penetrate deep into the soils. The passive pool is resistant to decay under both anaerobic and aerobic soil conditions because it is chemically recalcitrant and possibly physically protected by complexing with clays or other soil particles (Parton et al. 1993). However, the below-ground structural pool is not physically or chemically protected and accumulates only under persistent anaerobic conditions. The simulated drying of Zapf’s fen predicted an average C loss of 70 g m22 yr21. This is similar to the rates of carbon loss (107 and 85 g C m22 yr21) calculated for two fens in Rocky Mountain National Park when water tables dropped below the soil surface during a drought year or were affected by drainage (Chimner 2000). CENTURY also predicted that annual soil respiration (RESP1) increased an average 2.1 times in the drying scenario. This also compares very closely with gas flux data from a water-table experiment in another Rocky Mountain National Park fen (Chimner 2000), which reported that soil respiration increased 2.3 times when the water table dropped beneath the soil surface. Although peatlands in Colorado store only a small fraction of the total carbon stored in peatlands globally (Chimner 2000), the peat accumulation process is similar to many peatlands in other regions. Subsequently, the modeling procedure used in this paper should be valid for other herbaceous fens. It is unknown how CENTURY would fare in simulating Sphagnum- or tree-dominated peatlands, but there is nothing in the vegetation parameters that would prohibit simulating them in CENTURY. However, several modifications could be made to CENTURY that would be beneficial for simulating peatlands and other wetlands in general. A major modification could be made to the hydrologic model. Currently, it is a 2-dimensional model that does not allow for lateral ground-water flow. This is adequate for upland systems and may be valid for ombrotrophic bogs, but it is inadequate for simulating ground-water-dominated wetlands. Unfortunately, there is no easy way to add a ground-water component
Chimner et al., MODELING FENS without simulating the entire watershed. There is also no water-table component in CENTURY, which is a common hydrologic measurement in wetlands. Another option would be to couple a peatland hydrologic model (e.g., PHIM, Guertin et al. 1987) to the CENTURY model. This has the potential to allow simulations of both the hydrologic cycle and carbon cycle. The creation of anaerobic conditions is another area that should be modified to allow better simulation of peatlands. The current method of using the precipitation:PET ratio is unwieldy for ground-water-dominated systems. A better method might be to have anaerobic conditions created by high soil-moisture levels or water-table levels. This would integrate both precipitation and ground-water components. Another modification that would be beneficial for simulating peatlands is to allow peat to physically build itself up. Unlike upland systems that mostly incorporate soil carbon into the mineral matrix, peatlands build themselves up above the underlying mineral soil. The build-up of peat has feedbacks on many other processes that occur in the peatland, such as nutrient acquisition and hydrologic regime, that would have to be accounted for in the hydrology and nutrient submodels. It could then be possible to simulate the thickness of the peatland surface and the amount of carbon stored. It may be quite some time before an ecosystem model could be modified or built to have all these parameters that would realistically simulate peatlands carbon accumulation processes, but it is the direction that we need to take for accurately modeling peatlands. ACKNOWLEDGMENTS We thank Stephen Del Grosso, Melannie Hartman, and Robin Kelly at NREL, Colorado State University for their valuable assistance with the CENTURY model. We also thank Drs. Jim Benedict and Louis Maher for allowing us to use their 14C dates for Caribou and Zapf’s fen. We also thank Drs. Tom Stohlgren, Lee MacDonald, Gene Kelly, Sigrid Resh, and two anonymous reviewers for their helpful reviews. LITERATURE CITED Armentano, T. V. and E. S. Menges. 1986. Patterns of change in the carbon balance of organic soil-wetlands of the temperate zone. Journal of Ecology 74:755–774. Baron, J. S., D. S. Ojima, E. A. Holland, and W. J. Parton. 1994. Analysis of nitrogen saturation potential in Rocky Mountain tundra and forest: implications for aquatic systems. Biogeochemistry 27:61–82. Belyea, L. R. and B. G. Warner. 1996. Temporal scale and the accumulation of peat in a Sphagnum bog. Canadian Journal of Botany 74:366–377. Botch, M. S., K. I. Kobak, T. S. Vinson, and T. P. Kolchugina.
109 1995. Carbon pools and accumulation in peatlands of the former Soviet Union. Global Biogeochemical Cycles 9:37–46. Charman, D. J., R. Aravena, and B. G. Warner. 1994. Carbon dynamics in a forested peatland in north-eastern Ontario, Canada. Journal of Ecology 82:55–62. Chimner, R. A. 2000. Carbon dynamics of Southern Rocky Mountain fens. Ph.D. Dissertation. Colorado State University, Ft. Collins, CO, USA. Clymo, R. S. 1984. The limits to peat bog growth. Philosophical Transactions of the Royal Society of London B 303:605–654. Clymo, R. S., J. Turunen, and K. Tolonen. 1998. Carbon accumulation in peatland. Oikos 81:368–388. Fall, P. L. 1997a. Timberline fluctuations and late Quaternary paleoclimates in the Southern Rocky Mountains, Colorado. Geological Society of America Bulletin 109:1306–1320. Fall, P. L. 1997b. Fire history and composition of the subalpine forest of western Colorado during the Holocene. Journal of Biogeography 24:309–325. Francez, A. J. and H. Vasander. 1995. Peat accumulation and peat decomposition after human disturbance in French and Finnish mires. Acta Ecologica 16:599–608. Frolking, S., N. Roulet, T. Moore, P.J.H. Richard, P. Lafleur, J. Bubier, and P. Crill. 2001. Modeling short-term and long-term carbon accumulation in northern peatlands. International Workshop on Carbon Dynamics of Forested Peatlands: Knowledge Gaps, Uncertainty and Modelling Approaches. Edmonton, Alberta, Canada. Gorham, E. 1991. Northern peatlands: role in the carbon cycle and probable responses to climatic warming. Ecological Applications 1:182–195. Guertin, D. P., P. K. Barten, and K. N. Brooks. 1987. The peatland hydrologic impact model: development and testing. Nordic Hydrology 18:79–100. IPCC Climate Change, 1995. The science of climate change. p. 1– 572. In: J. T. Houghton, L. G. Meiro Filho, B. A. Callander, N. Harris, A. Kattenberg, and K. Maskell (eds.) Contribution of working group I to the Second Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, UK. Kelly, R. H., W. J. Parton, G. J. Crocker, P. R. Grace, J. Klı´r, M. Ko¨rschens, P. R. Poulton, and D. D. Richter. 1997. Simulating trends in soil organic carbon in long-term experiments using the century model. Geoderma 81:75–90. Lappalainen, E. 1996. General review on world peatland and peat resources. p. 53–56. In E. Lappalainen (ed.) Global Peat Resources. International Peat Society and Geological Survey of Finland, Jyska, Finland. Maltby, E. and M. C. F. Proctor. 1996. Peatlands: their nature and role in the biosphere. p. 11–19. In E. Lappalainen (ed.) Global Peat Resources. International Peat Society and Geological Survey of Finland, Jyska, Finland. Metherell, A. K., L. A. Harding, C. V. Cole, and W. J. Parton. 1993. CENTURY Soil organic matter model environment, technical documentation, Agroecosystem version 4.0. U.S. Department of Agriculture, Agricicultural Research Service, Fort Collins, CO, USA. Great Plains System Research Unit Technical Report No. 4. Moore, T. R. 1989. Plant production, decomposition, and carbon efflux in a subarctic patterned fen. Arctic and Alpine Research 21:156–62. Nyka¨nen, H., J. Alm, J. Silvola, K. Tolonen, and P. J. Martikainen. 1998. Methane fluxes on boreal peatlands on different fertility and the effect of long-term experimental lowering of the water table on flux rates. Global Biogeochemical Cycles 12:53–69. Ovenden, L. 1990. Peat accumulation in northern wetlands. Quaternary Research 33:377–386. Parton, W. J., J. M. O. Scurlock, D. S. Ojima, T. G. Gillmanov, R. J. Scholes, D. S. Schimel, T. Kirchner, J. C. Menaut, T. Seastedt, E. Garcia Moya, A. Kamnalrut, and J. L. Kinyamario. 1993. Observation and modeling of biomass and soil organic matter dynamics for the grassland biome worldwide. Global Biogeochemical Cycles 7:785–809. Raich, J. W., W. J. Parton, P. M. Vitousek, A. E. Russell, and R. L. Sanford, Jr. 2000. Analysis of factors regulating ecosystem de-
110 velopment on Mauna Loa using the Century model. Biogeochemistry 51:161–191. Reader, R. J. and J. M. Stewart. 1972. The relationship between net primary production and accumulation for a peatland in southeastern Manitoba. Ecology 53:1021037. Robinson, S. D. and T. R. Moore. 1999. Carbon and peat accumulation over the past 1200 years in a landscape with discontinuous permafrost, northwestern Canada. Global Biogeochemical Cycles 13:591–601. Saarinen, T. 1996. Biomass and production of two vascular plants in a boreal mesotrophic fen. Canadian Journal of Botany 74:934– 938. Siegel, D. I. and P. H. Glaser. 1987. Groundwater flow in a bog-fen complex, Lost River peatland, northern Minnesota. Journal of Ecology 75:743–754. Silvola, J., J. Alm, U. Ahlholm, H. Nyka¨nen, and P. J. Martikainen. 1996. CO2 fluxes from peat in boreal mires under varying temperature and moisture conditions. Journal of Ecology 84:219–228. Smith, P., J. U. Smith, D. S. Powlson, W. B. McGill, J. R. M. Arah, O. G. Chertov, K. Coleman, U. Franko, S. Frolking, D. S. Jen-
WETLANDS, Volume 22, No. 1, 2002 kinson, L. S. Jenson, R. H. Kelly, H. Klien-Gunnewick, A. S. Komarov, C. Li, J. A. E. Molina, T. Mueller, W. J. Parton, J. H. M. Thornley, and A. P. Whitmore. 1997. A comparison of the performance of nine soil organic models using datasets from seven long-term experiments. Geoderma 81:153–225. Thomas, K. L., J. Bernstead, K. L. Davies, and D. Lloyd. 1996. Role of wetland plants in the diurnal control of CH4 and CO2 fluxes in peat. Soil Biology and Biochemistry 28:17–23. Warner, B. G., R. S. Clymo, and K. Tolonen. 1993. Implications of peat accumulation at Point Escuminac, New Brunswick. Quaternary Research 39:245–248. Waughman, G. J. and D. J. Bellamy. 1980. Nitrogen fixation and the nitrogen balance in peatland ecosystems. Ecology 61:1185– 1198. Webber, P. J. and D. E. May. 1977. The magnitude and distribution of belowground structures in the alpine tundra of Niwot Ridge Colorado. Arctic and Alpine Research 9:157–74. Manuscript received 10 January 2001; revisions received 12 June 2001; accepted 19 November 2001.
