SEARCH

SEARCH BY CITATION

Keywords:

  • Alaska;
  • boreal;
  • climate change;
  • peatland;
  • photosynthesis;
  • respiration;
  • winter

ABSTRACT

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. ACKNOWLEDGEMENTS
  8. REFERENCES
  9. Supporting Information

Boreal wetlands hold vast stocks of soil carbon (C), which may be vulnerable to changes in climate. In southcentral Alaska, wetlands of the Kenai Lowlands have experienced a warming and drying trend that has led to woody vegetation encroachment into herbaceous wetlands. We examined whether predicted higher rates of gross ecosystem photosynthesis (GEP) would be offset by higher rates of ecosystem respiration (ER) in woody wetland communities. We measured net ecosystem exchange (NEE) in four communities along a hydrologic and vegetation gradient during (1) a warm and dry growing season, (2) a more typical cool and wet growing season and (3) the intervening winter. We fit simple GEP and ER models to our data and estimated annual NEE for each community using hourly measurements of photosynthetically active radiation and air temperature. We found that woody communities exhibited greater GEP than herbaceous communities under cool and moist conditions, but more similar GEP under warm and dry conditions. Woody communities also showed greater ER than herbaceous communities during all seasons, outpacing GEP during the warm and dry growing season. On an annual basis, we estimated that herbaceous communities were either net sinks or approximately CO2 neutral, ranging from −132·8 to 4·7 g CO2–C m−2 y−1. In contrast, woody communities were sources of CO2 to the atmosphere, ranging from 78·8 to 181·7 g CO2–C m−2 y−1. Our results suggest that the initial encroachment of woody vegetation into herbaceous wetlands will lead to a substantial loss of C, particularly if conditions continue to become warmer and drier. Copyright © 2011 John Wiley & Sons, Ltd.


INTRODUCTION

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. ACKNOWLEDGEMENTS
  8. REFERENCES
  9. Supporting Information

Boreal and subarctic peatlands hold vast stores of soil carbon (C), with global estimates ranging from 2·7 to 4·6 × 1011 metric tonnes (Gorham, 1991; Turunen et al., 2002; Vasander and Ketunnen, 2006). Peat accumulation occurs where net production exceeds losses through decomposition, leaching and disturbance. Consistently cool and anoxic soil conditions inhibit decomposition (Clymo et al., 1998), allowing peat accumulations up to several metres deep (e.g. Turunen et al., 2002; Berg et al., 2009). As northern latitudes experience increasing temperatures and changing precipitation patterns (ACIA, 2004; IPCC, 2007), these large soil C stocks may be released to the atmosphere through increased decomposition. The wet, acidic and cool soils of many boreal wetlands also inhibit woody plant establishment and growth. As these soils dry and warm, more productive woody communities often replace sedge-dominated and moss-dominated communities (Jukaine and Laiho, 1995; Riutta et al., 2007). Hence, while warming and drying may increase autotrophic and heterotrophic respiration, replacement of herbaceous communities with woody vegetation may lead to the development of a multi-layered canopy and substantial increases in photosynthesis as leaf area index increases. Vegetation change in response to wetland drying may, therefore, partially or completely offset C losses expected from wetland drying alone.

The CO2 budget of northern wetlands depends upon environmental conditions. The organic soils of northern wetlands are susceptible to increased decomposition, causing increased CO2 emissions in warmer (Alm et al., 1999; Bubier et al., 2003; Blodau et al., 2007; Cai et al., 2010) and drier conditions (Chimner and Cooper, 2003; Chivers et al., 2009). Boreal and sub-arctic areas have experienced substantial warming over the previous century (Serreze et al., 2000; ACIA, 2004; IPCC, 2007). Most research suggests that warming is expected to reduce the CO2 sink strength of boreal environments through reduced photosynthesis (drought induced reductions in stomatal conductance) and/or increased respiration (Bunn et al., 2007; Soja et al., 2007). However, higher temperatures may also increase boreal forest carbon storage through a lengthening of the growing season and greater productivity of woody vegetation, a growth form that is often limited by cold and/or waterlogged soils. Bronson et al. (2009) documented that warmer soil and air temperatures led to earlier bud burst and greater shoot lengths in black spruce (Picea mariana) forests. Other evidence for increased growth in drying peatlands comes from a Finnish study, which observed higher rates of peat accumulation and C storage in peatlands drained for silviculture than in undrained peatlands (Minkkinen et al., 2002). Recent work by Strack et al. (2009) indicated that short-term drought stress did not release large stocks of soil C, and that climate change may not lower peatland water tables enough to do so.

Complicating the relationships, the presence of shrubs and/or trees in low-statured ecosystems leads to greater accumulation and retention of snow in winter (Sturm et al., 2001; Sullivan, 2010). Deeper snow is generally associated with warmer soils, which lead to greater ecosytem respiration rates and greater rates of winter C loss (Welker et al., 2000; Schimel et al., 2004; Schimel et al., 2006). Nevertheless, given the strong positive relationship between leaf area index and photosynthesis in northern ecosystems (Street et al., 2007), increased growth and a change in species composition driven by a warming climate could lead to rates of photosynthesis high enough to exceed the concurrent increase in both autotrophic and heterotrophic respiration (Flanagan and Syed, 2011).

The Kenai Lowlands of south central Alaska hold large stocks of soil C, which may be susceptible to the rising temperatures and reduced precipitation observed in recent decades (Serreze et al., 2000; Klein et al., 2005; Berg et al., 2009). The Kenai Lowlands contain approximately 2100 km2 of peatlands (Gracz et al., 2008) that are approximately 1–7 m deep (Berg et al., 2009). A wide range of changes have been documented on the Kenai Peninsula in response to recent climate change, including treeline advance (Dial et al., 2007), increased insect infestations (Berg et al., 2006; Boucher and Mead, 2006), reduced open water (Klein et al., 2005) and replacement of herbaceous communities with woody vegetation (Berg et al., 2009). A regional warming and drying trend (Klein et al., 2005; Dial et al., 2007; Berg et al., 2009) may be responsible for increased rates of peat decomposition in the Kenai Lowlands. Gracz et al. (2008) note that many Kenai Peninsula organic soils mapped as fibric soil series (undecomposed) in the early 1970s are now being mapped as hemic soil series (partially decomposed). A shift from fibric to hemic soil series is suggestive of substantial decomposition in peatlands of the Kenai Lowlands since the 1970s. However, as peatlands have warmed and dried, more productive woody plant species such as Betula nana L. and Picea mariana have encroached (Klein et al., 2005; Berg et al., 2009).

In this study, we made monthly measurements of the light-response of ecosystem CO2 exchange in four communities along a hydrologic gradient during two contrasting growing seasons (cool and moist vs warm and dry). Measurements were also undertaken in the intervening snow-covered season, as recent work has shown that winter respiration may be an important and climate-sensitive component of annual C budgets in boreal ecosystems (Vogel et al., 2005; Sullivan et al., 2010). Models of ecosystem photosynthesis and respiration were fitted to measurements of CO2 flux, temperature and photosynthetically active radiation [photosynthetic photon flux density (PPFD)] data to predict seasonal and annual C budgets for each of the four communities along the hydrologic gradient. We had three principal questions (1) Is greater photosynthesis in the drier, woody communities sufficient to offset increased rates of respiration, yielding only small differences in annual CO2 budgets across the zones? (2) Does photosynthesis in woody communities respond more strongly to temporal variation in climate than in herbaceous communities? (3) Does the presence of shrubs and small trees trap snow, increasing winter rates of CO2 efflux through the snowpack? Given the large area of the Kenai Lowlands and the compelling evidence of rapid vegetation succession, we feel the answers to our questions have important implications for the global climate system.

METHODS

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. ACKNOWLEDGEMENTS
  8. REFERENCES
  9. Supporting Information

Study site

The Kenai Peninsula of south central Alaska is characterized by a maritime climate of rain forests, mountains, glaciers and icefields in the east, and by permafrost-free wetlands and boreal forest lowlands subject to a continental climate in the west. Approximately 2100 km2 of the 9400 km2 Kenai Lowlands are peatlands, nearly all of which are fens (Gracz et al., 2008). A detailed description of the Kenai Lowlands can be found in Berg et al. (2009) study. The study site was a poor fen located 0·8 km southwest of the Swanson River, within the Kenai Lowlands (60·71°N, 150·89°W, 45 m asl). A basal radiocarbon date at northern boundary of the site (minimum age of 7288 calibrated years before present) indicated that this portion of the fen began accumulating peat after the Holocene Thermal Maximum. The 2·4 m of peat was underlain by fine subangular gravels and preserved sedge bases. Mean annual temperature (1949–2009) at the Kenai Airport (WRCC, 2010a) was 1·3 °C, with the warmest mean monthly temperature in July (12·7 °C) and coldest in January (−10·9 °C). Mean annual precipitation (1949–2009) was 484 mm. Soil dry bulk density was 41·6 kg m−3, and total C and N pools were 15·4 and 0·6 kg m−2, respectively.

We identified an area with a hydrologic gradient leading from open water through progressively drier communities: seasonally inundated bryophytes, graminoid vegetation in continuously saturated soils with near-surface water table, dwarf shrubs in seasonally saturated soils and low stature evergreen trees in saturated to moist soils. Five study plots were established parallel to the ecotones and perpendicular to the hydrologic gradient in each of the four community types (moss, sedge, shrub and small tree) for a total of 20 plots. On average, communities were 12 m apart. Plots were established approximately 1 m apart, with the exception of the small tree community, where plots 1 and 2 were 1 m apart, plots 2 and 3 were 3 m apart and plots 3 through 5 were 0·5 m apart. The plots themselves were 70 × 70 cm, which matched the footprint of our CO2 flux chamber. Boardwalks installed on the north side of the study plots minimized disturbance from monthly trips to the site. Ground covering mosses (Sphagnum and Polytrichum spp.) dominated both the moss and sedge communities, with relatively low and sparse canopies of buckbean (Menyanthes trifoliata), sedges (Carex pluriflora, C. chordorrhiza and C. rotundata) and cottongrass (Eriophorum angustifolium). Bog rosemary (Andromeda polifolia), dwarf birch (Betula nana) and black spruce (Picea mariana) dominated the upper canopies of the shrub (mean canopy height 4·5 ± 0·5 cm) and small tree communities (9·0 ± 0·5 cm), with ground covering mosses (Sphagnum and Polytrichum spp.). The small tree community had pronounced microtopography, where woody species grew on hummocks and herbaceous species in hollows.

The study site was within the 33 500 ha footprint of a 1969 wildfire (DeVolder, 1999) and we were able to find one piece of charred wood at the study site. Increment cores collected from mature black spruce trees, over 10 m from the small tree community study plots, showed an age range of 24–65 years indicating that the fire was not stand-clearing. Peatland vegetation generally recovers to pre-fire condition within decades of a surface fire, with no long-term effect on vegetation development (Kuhry, 1994). This is supported by Klein et al. (2005) who found no difference in the rate of wetland vegetation change over a 50 year period between burned and un-burned areas of the Kenai Lowlands. Sirois and Payette (1989) reported 80% of black spruce seedlings established within 17 years of a lowland boreal fire; thus the majority of black spruce in the small tree community would be 23 years or older if recovering from the 1969 fire. We harvested 12 black spruce (16·0 to 106·5 cm tall) from a sparsely forested area adjacent to the study site; ages ranged from 8–52 years. Using these data to create an age-height regression (age = 0·36*height + 6·16, r2 = 0·69, n = 12, p < 0·01), we estimated black spruce in small tree plots were 11–25 years old. We believe that the lack of a stand-clearing fire, the passage of 40 years since fire, and the relatively young age of black spruce in the small tree community indicate that vegetation zonation at the study site reflected the hydrologic gradient and not recovery from fire.

Microclimate

In early June 2008, we installed a 3 m high meteorological (met) station at the study site, controlled by a CR1000 datalogger (Campbell Scientific, Logan UT) and powered by a 20 W solar panel and a 12 V 100 amp-hour battery. The station recorded incoming photosynthetically active radiation (PPFD, LI-COR LI-190SB), liquid precipitation (Texas Electronics TE-525L), air temperature and relative humidity at 2 m (Campbell Scientific CS-215, housed within a radiation shield), wind speed and direction at 3 m (05103 R.M. Young) and soil temperature and soil water content (SWC) at 10 cm depth in each community (Campbell Scientific CS-107 and CS-616). The station scanned sensors every 30 s and logged hourly averages beginning 19 June 2008 for atmospheric variables and 24 June 2008 for soil temperature and SWC. We estimated early 2008 air temperature and PPFD by regressing met station data against data from a remote automated weather station approximately 2·3 km to the northeast [Swanson River, Alaska Remote Automated Weather Station (RAWS), 60·73°N, 150·87°W, 85 m asl; air temperature r2 = 0·96, PPFD r2 = 0·95]. Additionally, during each sampling session, handheld sensors were used to make four measurements of soil temperature (10 cm depth, VWR Scientific) and soil water content (0–12 cm depth, Campbell Scientific Hydrosense) adjacent to each plot.

CO2 flux measurements

We measured mid-day CO2 exchange during the 2008 and 2009 growing seasons (28 May, 20 June, 9 July, 7 August, 17 September and 24 October 2008; 17 May, 11 June, 7 July, 20 August and 2 October 2009) in five plots in each of the four communities. Only three plots were measured per community on 11 June 2009 due to inclement weather. We also measured CO2 efflux through the snowpack during the winter of 2008/2009 (26 November and 14 December 2008; 3 January, 16 February and 24 March 2009). We present measurements from the atmospheric perspective, where positive values indicate CO2 release from biota to the atmosphere and negative values indicate CO2 uptake by the biotic community.

We made growing season measurements using a closed system: a custom-designed 70 × 70 × 40 cm clear acrylic chamber fitted with an infrared gas analyzer (LI-6400 Portable Photosynthesis System, LI-COR Biosciences, Lincoln, NE). The LI-6400 was calibrated according to manufacturer specifications prior to each sampling session. Internal chamber fans ensured mixing, a vent ensured pressure equilibrium between the chamber and ambient air, and temperature and PPFD sensors recorded the chamber environment. The chamber fit atop a clear acrylic base, sealed by a strip of closed-cell foam. A clear vinyl skirt ran from the base to the ground, and was sealed to the ground surface with a heavy chain.

Each sampling session included five measurements of CO2 flux at each of the 20 plots, corresponding to five different light levels, one clear chamber measurement of net ecosystem exchange (NEE), one dark chamber measurement of ecosystem respiration (ER), and three measurements of NEE at intermediate light levels through varying meshes of shade cloth (reducing light levels by approximately 30%, 50% and 70%). We lifted the chamber to restore ambient conditions prior to measuring flux at each light level. For each 55 s measurement, we recorded CO2 concentrations and chamber microclimate data every second. The CO2 data exhibited considerable variability within the first 10 s, thus only the final 45 s were used to calculate CO2 fluxes. For each of the five light levels, we fit a line to the relationship between CO2 concentration and time to estimate a slope (dC/dt, µmol CO2 s−1). We calculated NEE (µmol CO2 m−2 s−1) using the equation described by Williams et al. (2006). The chamber volume was corrected for plot-level microtopography by measuring the distance between the ground surface and the base over a grid of 50 points.

Comparison of microclimate conditions indicates that our measurements were made across a range of conditions that closely matches the range of conditions observed during the TGS at the study site (Table 1). One notable exception is that PPFD recorded within our chamber exceeds the range observed at the site. This is a consequence of the orientation of the sensor within the chamber with respect to the position of the sun. The woody plots, in particular, are not level and some are south-facing. When these plots were measured on sunny days, they experienced a greater PPFD than measured on the met station, where the sensor is installed in a level position.

Table 1. Range of climate data during sampling events and thermal growing seasons. PPFD sampling range recorded within chamber (60-second means), all other values recorded at on-site met station (hourly means).
 Air temperaturePPFDMoss SWCSedge SWCShrub SWCSmall tree SWC
(°C)(µmol photons m−2 s−1)(%)(%)(%)(%)
  1. PPFD, photosynthetic photon flux density, SWC, soil water content, TGS, thermal growing season.

2008 Sampled range9·5–19·70–1834·139·7–41·339·5–41·224·5–42·427·3–42·6
2008 TGS range−5·7–21·250–1687·039·3–41·539·0–41·424·3–42·427·1–42·6
2009 Sampled range5·3–27·20–1692·539·4–41·627·2–41·723·3–27·223·5–27·4
2009 TGS range−6·9–28·10–1665·019·4–42·018·9–41·920·2–27·619·7–28·1

Modelling NEE, ER and gross ecosystem photosynthesis

Light response curves

The photosynthetic rate of each plot varied seasonally and with changing light levels. A plot-specific light response curve was modelled for each sampling date, allowing us to compare NEE and gross ecosystem photosynthesis (GEP) without the confounding effect of varying light levels. Light response curves were modelled for each month and plot (Proc NLIN, SAS 9·2) as rectangular hyperbolas (Street et al., 2007) using measured NEE and PPFD,

  • display math

where inline image is a fitted estimate of ER (µmol CO2 m−2 s−1), amax is the light saturated photosynthetic rate (µmol CO2 m−2 s−1) and ks is the half-saturation constant (µmol photons m−2 s−1). We used a coefficient of determination (r2) to characterize the goodness-of-fit for each regression. Data with regression r2 < 0·8 were individually examined. We discarded one light level from a May 2009 sedge plot due to a chamber leak, one sedge plot from September 2009 due to a chamber leak at all light levels and all flux data from October 2008 when patchy snow cover left little confidence in the accuracy of the closed chamber technique. The discarded October 2008 data indicated little flux activity in general. Over 97% of the remaining 191 regressions had r2 > 0·8 and extended beyond a PPFD of 500 µmol photons m−2 s−1.

To compare NEE and GEP between communities and over time without the confounding effect of changing light levels, the parameter estimates for each light response curve were used to predict NEE at a PPFD of 500 µmol photons m−2 s−1 (NEE500), the highest light level consistently observed during measurements. Because NEE = ER + GEP, we calculated GEP at a light level of 500 µmol photons m−2 s−1 (GEP500) as the difference between NEE500 and inline image. We recognize that standardizing our NEE and GEP estimates to a common light level disregards error in the fit of the rectangular hyperbola. To examine the fit of the rectangular hyperbola to our NEE data, we regressed modelled against measured flux from one randomly selected measurement per plot in May 2008, revealing a strong linear relationship (n = 20, r2 = 0·97, p < 0·01) with slope (0·97) and intercept (0·04) not significantly different from 1 and 0, respectively. Given this close relationship, we felt the benefits of comparing fluxes at a common light level exceeded the drawbacks associated with error in the fit of the hyperbolas. Given that our GEP estimates were derived from model-fitting, we decided to use inline image for consistency, rather than our measured values of ER. Regressing inline image against measured ER for July 2008 plot data showed a very strong linear relationship (inline image = 1·01*ER − 0·008, n = 20, r2 = 0·98, p < 0·01) with no evidence of a bias. The strength of this relationship gave us confidence in using inline image for further modelling and hypothesis testing.

Estimating winter respiration (ER)

We estimated CO2 efflux through the winter snowpack using the diffusion gradient technique (e.g. Fahnestock et al., 1999; Welker et al., 2000; Sullivan, 2010). A hollow stainless steel probe with a perforated tip was plumbed with 3·2 mm tubing attached to a LI-800 CO2 analyzer (LI-COR Biosciences, Lincoln, NE), which was calibrated and allowed to warm-up for 2·5 h before each sampling session. The LI-800 was equipped with a pump (850 ml/min) downstream of the optical bench. A digital multi-meter was wired to the analyzer output channels to read CO2 concentration [CO2]. Two replicates were collected at each plot by inserting the probe to the base of the snowpack and drawing air through the analyzer until [CO2] stabilized. We characterized snowpack density and temperature at two locations per community; the averages of these measures were used in flux calculations. We sampled snowpack density at 10 cm intervals along the vertical wall of a snow pit with a stainless-steel RIP 1 density cutter (Elder et al., 1991; Snowmetrics, Fort Collins, CO) and a 600 g capacity spring scale with 5 g resolution. We measured snow temperature at 10 cm intervals and at the ground surface with a dial stem thermometer.

We applied Fick's Law of Diffusion to estimate CO2 flux from the snowpack to the atmosphere (Musselman et al., 2005),

  • display math

where snowpack porosity (unitless) is θ = 1 – ρsi, [ρs = measured snowpack density (g l−1) and ρi = ice density = 917 g l−1], τ is snowpack tortuosity (estimated as θ1/3; Millington, 1959), DCO2 is the diffusion constant (0·1381 × 10−4 m2 s−1) for CO2 through air at standard temperature and pressure (STP), P0/RT0 is the molecular density (44·613 mol m−3) of CO2 at STP, T is measured snowpack temperature (K), ΔC is the difference in [CO2] between the atmosphere and the subnivean, and z is snowpack depth (m).

Modelling respiration as a function of temperature

Simple models of ER and GEP were used to interpolate between our data points. It should be noted that our goal was not to provide objective tests of model performance. Preliminary analysis showed that ecosystem respiration was more closely correlated with air temperature than with soil temperature, and that a logistic model better described the relationship than a Q10 model, as observed in recent studies (Sullivan et al., 2010). To estimate respiration rates of each community through 2008 and 2009, ER was modelled (Proc Model, SAS 9·2) as a logistic function of air temperature

  • display math

where T is the site air temperature measured at the met station, and ER comprises both inline image from light response curves and ER measured in winter months. Regressing predicted against observed ER for one randomly selected measurement per plot showed a strong linear relationship (Predicted = 0·76*Observed + 0·44, n = 20, p < 0·01, r2 = 0·74; Figure 1A), suggesting the logistic respiration model was a useful tool for interpolating ER, which was confirmed by regressing predicted against observed ER for all plot measurements (Predicted = 0·80*Observed + 0·47, n = 291, p < 0·01, r2 = 0·81; Figure 1B). We estimated 2008 and 2009 hourly ER per community by applying the logistic model with community-specific parameters to hourly air temperature data from the met station.

image

Figure 1. (A) Measured versus modelled ecosystem respiration (ER) for one randomly selected measurement per plot (n = 20) and (B) for all plot measurements (n = 291); open symbols indicate winter and closed symbols growing season ER values. (C) Measured versus modelled gross ecosystem photosynthesis (GEP) for one randomly selected measurement per plot (n = 20) and (D) for all plot measurements (n = 954). Note that, in panels A and C, random selection of date and one measurement per plot were performed to meet assumption of independent observations. Solid lines represent regression equations, dashed lines represent ideal relationships (slope = 1, intercept = 0).

Download figure to PowerPoint

Modelling GEP as a function of LAI and PPFD

We defined the beginning of the thermal growing season (TGS) as the first day following a span of 5 days with mean daily temperatures (MDT) >5 °C, and the end of the TGS as the first day following a span of 5 days with MDT <5 °C.

To estimate GEP at each plot, we applied the model of Shaver et al. (2007):

  • display math

where amaxl is the light-saturated photosynthetic rate per unit leaf area (µmol CO2 m−2 leaf s−1), k is the Beer's law light extinction coefficient (m2 ground m-2 leaf), E0 is the initial slope of the light response curve (µmol CO2 µmol-1 photons), PPFD is the light level at the top of the canopy, and leaf area index (LAI) is leaf area index. We calculated GEP600 for each plot measurement (n = 197) and then estimated a corresponding LAI from a linear relationship between LAI and GEP600 developed across a wide range of low arctic ecosystems (Street et al., 2007).

We then used these LAI estimates with the corresponding GEP and PPFD measurements to parameterize the Shaver et al. (2007) GEP model (Proc Model, SAS 9·2). We then applied the Shaver et al. (2007) model with site-level parameters to hourly estimates of LAI for each plot, allowing us to interpolate GEP between data points. Regressing predicted against observed GEP for one randomly selected measurement per plot showed a very strong linear relationship (Predicted = 0·94*Observed − 0.06, n = 20, p < 0·01, r2 = 0·90; Figure 1 C), suggesting that the Shaver et al. (2007) model was a useful tool for interpolating GEP, which was confirmed by regressing predicted against observed GEP for all plot measurements (Predicted = 0·94*Observed − 0·16, n = 291, p < 0·01, r2 = 0·93; Figure 1D).

Outside of the TGS, we set LAI equal to zero. During the TGS, we generated daily LAI estimates by linear interpolation (Proc Expand, SAS 9·2) between chamber measurement dates. We estimated 2008 and 2009 hourly GEP per plot by applying the Shaver et al., (2007) GEP model to daily plot LAI estimates and hourly met station PPFD data. Hourly GEP and ER estimates were combined to estimate hourly NEE for each plot, allowing us to calculate NEE on each day of 2008 and 2009 for each community.

Applying the Shaver et al. (2007) model with estimated LAI involves several assumptions. First, that a regression of GEP against LAI for low arctic plants (Street et al., 2007) is appropriate for a boreal setting. Second, generating LAI estimates by inverting the regression of Street et al. (2007) gives an estimate of effective LAI, not actual LAI. We believe this is advantageous, as effective LAI integrates unquantified parameters that are likely to affect photosynthesis, such as variations in temperature, leaf nitrogen concentrations and stomatal conductance during the TGS. We recognize the circularity of using GEP as both an independent variable to estimate effective LAI and as model output. Therefore, we reiterate the point that our goal was not to test the model, but to use it as a tool to interpolate between measurement dates. In order to gauge confidence in our GEP predictions, we compared GEP estimates from the Shaver et al. (2007) model to GEP estimates from a linear interpolation of plot-specific light response curves. Results were nearly identical, leaving us confident that using effective LAI as an input to the Shaver et al. (2007) model was a useful tool to interpolate our GEP data.

Hypothesis testing

Using the rectangular hyperbola fit for each plot measurement allowed us to compare dates and communities without confounding variation in light levels. We used two-way repeated measures ANOVA with Tukey post hoc tests (alpha = 0·05) to examine the effects of community and sampling date on NEE500, GEP500 and Log10(inline image+1), as well as the effects of community and year on annual NEE (Proc Mixed, SAS 9·2). All t values presented are from Tukey post hoc pairwise comparisons.

RESULTS

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. ACKNOWLEDGEMENTS
  8. REFERENCES
  9. Supporting Information

Microclimate

Overall, 2008 was cool and wet compared with 2009 (Table 2). Although monthly growing season temperatures in 2008 were very similar to the 60 year means (1949–2009), 2009 monthly temperatures were generally >1 SD above the 60 year means. At the study site, mean growing season air temperature was 0·4 °C higher in 2009 (10·5 °C) than in 2008 (10·1 °C), with the highest monthly means occurring in July of both 2008 (12·1 °C) and 2009 (14·2 °C). In addition to being warmer, the 2009 Thermal Growing Season (TGS) was a week longer: 7 May to 1 October in 2008 and 30 April to 1 October in 2009.

Table 2. Long-term (1949–2009) 2008 and 2009 monthly mean temperatures (inline image± 1 SD, °C) recorded at the Kenai Airport, Kenai, Alaska [WRCC, 2010a]. Monthly water balance (precipitation – potential evapotranspiration) at nearby Swanson River RAWS during 2008 and 2009 growing seasons. Penman–Montieth potential evapotranspiration was calculated following Allen et al. (1998) using complete daily data (with estimated missing values).
 MayJuneJulyAugustSeptember
  1. RAWS, remote automated weather station.

Air temperature (°C)
20087·210·013·012·49·9
20098·411·414·113·09·8
1949–20096·8 (±1·3)10·4 (±1·1)12·7 (±1·1)12·2 (±0·9)8·4 (±1·2)
Water balance (mm H2O)
2008−78·22−55·2531·22−12·9784·03
2009−69·12−64·06−70·56−9·05−5·52
      

July 2009 precipitation (18 mm) was one-quarter that of July 2008 (80 mm), and September 2009 precipitation (73 mm) was nearly half that in September of 2008 (135 mm). Swanson River RAWS (WRCC, 2010b) precipitation data also indicated that July 2009 precipitation (25 mm) was nearly 1 SD below the mean of the 17 year data record, whereas July 2008 precipitation (94 mm) was >1 SD greater than 17 year mean. Monthly water balance (precipitation – potential evapotranspiration) calculations for the Swanson River RAWS indicated that July and September were each water deficits in 2009 and surpluses in 2008 (Table 2).

Soil temperature differences across community types varied by season. During the growing season, soils in the herbaceous communities were about 2 °C warmer than in woody communities, with soil temperatures peaking in August of each year. In contrast, winter (November 2008–March 2009) soil temperatures were about 0·5 °C warmer in woody communities than in herbaceous communities. October and April mean soil temperatures were very similar among all communities. The 2008 and 2009 growing season soil temperatures were identical for moss and sedge communities (moss 8·8 °C, sedge 7·7 °C). In contrast, shrub and small tree growing season soil temperatures were about 0·7 °C warmer in 2009 than 2008 (shrub2008 6·0 °C, shrub2009 6·7 °C, small tree2008 6·3 °C, small tree2009 7·0 °C). As expected, moss and sedge communities consistently had the wettest soils. Shrub community soils were consistently the driest, which we believe was an artefact of sensor placement in the small tree community. SWC measured with a handheld Hydrosense during each sampling session indicated that small tree community hummocks were consistently drier than the shrub community. The small tree community SWC sensor for the met station was placed at 10 cm depth in a relatively level area within the community, and was not necessarily representative of hummock conditions. Regardless, we believe that the SWC sensor provides valuable data for comparing trends in SWC over time within the small tree community. Growing season mean SWC in herbaceous communities were very similar in 2008 and 2009 (approximately 40%). In contrast, mean SWC in shrub and small tree communities were 4%–5% lower in 2009 than 2008 (shrub2008 28%, shrub2009 24%, small tree2008 31%, small tree2009 26%). Snow depths were consistently greater in woody communities than in herbaceous communities, with the greatest depths measured in January of 2009 (moss 33 ± 0·2, sedge 34 ± 1·0, shrub 37 ± 1·0 and small tree 40 ± 2·0 cm). March of 2009 snow depth in the Kenai Lowlands (NRCS, 2009; Kenai Moose Pens SNOTEL, 60·73 º N 150·48 º W, 91 m asl; 46 cm) was very similar to the 30-year mean (43 cm).

CO2 flux measurements

NEE500, GEP500, inline image and ER comparisons

Overall, we detected two main trends in our monthly growing season measurements of CO2 exchange (Figure 2). First, both GEP500 and inline image were generally greater in woody communities than in herbaceous communities. Second, the warmer, drier weather in 2009 increased ER and reduced GEP more in woody than in herbaceous communities. These trends led to significant differences in NEE500 across community types, sampling dates and years (Table 3). In May of both years, all community types were either weak CO2 sources or showed negligible net CO2 fluxes. During this early season period, herbaceous communities were less of a CO2 source than woody communities. During the cool, wet 2008 growing season, all community types gradually became net CO2 sinks and remained either CO2 sinks or showed a neutral CO2 balance through the September measurements. In July 2008, the early season pattern was reversed and the woody communities became much stronger CO2 sinks than the herbaceous communities. During 2008, herbaceous communities gained more or lost less CO2 than woody communities near the margins of the growing season, whereas woody communities were greater CO2 sinks during the mid-summer. During the warm, dry conditions of 2009, the pattern was largely reversed. Variation in NEE500 across community types in May 2009 was very similar to that observed in 2008, with herbaceous communities acting as lesser CO2 sources than woody communities. In contrast with measurements made in June 2008, all communities were acting as relatively strong CO2 sinks in June 2009. During July 2009, all communities, with the exception of the moss community, were acting as relatively strong CO2 sources, with the greatest CO2 losses from the woody communities. Measurements made in August and September of 2009 showed patterns similar to 2008, with either weak CO2 sink activity or neutral CO2 budgets in all communities and a tendency toward greater uptake and/or reduced CO2 losses in the herbaceous communities.

image

Figure 2. Comparison of seasonal variation in inline image, GEP500, and NEE500 rates by community (μmol CO2 m−2 s−1, ± 1 SE) for 2008 and 2009.

Download figure to PowerPoint

Table 3. Main (date, community) and interaction (date*community) effects of two-way repeated measures ANOVA on GEP500, inline image, NEE500 and winter ER flux rates, all p values <0·01.
F-statisticsGEP500Log10(inline image+1)NEE500Winter ER
  1. ANOVA, analysis of variance; GEP, gross ecosystem photosynthesis; NEE, net ecosystem exchange; ER, ecosystem respiration; DF, Degrees of Freedom; F, F-statistic.

 DFFDFFDFFDFF
Date9, 13534·699, 135116·49, 135161·64, 6413·6
Community3, 1619·23, 16109·83, 165·23, 1632·1
Date*community27, 1352·127, 1353·427, 1359·712, 6413·8

During the growing season, monthly estimates of inline image and GEP500 showed significant differences both among communities and between years (Table 3). GEP500 rates were consistently lowest in the moss community and highest in the small tree community, generally following the trend moss < sedge < shrub < small tree. All community GEP500 rates were similarly low in May of both 2008 and 2009, and woody communities had higher GEP500 rates than herbaceous communities in September of both 2008 and 2009. In contrast, June through August GEP500 rates were quite different in 2008 and 2009. During the cool, moist 2008, each community GEP500 rate increased to a maximum in July and August, when woody community GEP500 rates were 50%–100% greater than herbaceous community rates. During the warm, dry 2009, however, woody communities reached maximum GEP500 rates in June and maintained these levels through August. Herbaceous community GEP500 rates continued to increase, reaching peak rates very similar to woody community rates in July and August of 2009. Although moss and sedge peak GEP500 rates were greater in 2009 than in 2008, shrub and small tree peak GEP500 rates were substantially lower in 2009 than 2008.

As with GEP500, inline image was consistently lowest in the moss and highest in the small tree community. inline image rates generally peaked in August of 2008 and in July of 2009, and peak inline image rates were 50%–100% higher in 2009 than in 2008. Unlike GEP500, July 2009 inline image rates were substantially higher in woody communities than herbaceous communities. ER rates were between one and two orders of magnitude less during the snow-covered season (November–April), ranging from 0·00 to 0·30 µmol CO2 m−2 s−1. Woody communities generally showed greater winter ER rates than herbaceous communities, following the same between community trend as inline image. The differences between community ER rates in February and March, however, were striking, shrub and small tree community ER rates (0·05–0·28 µmol CO2 m−2 s−1) were 5–10 times higher than moss and sedge community rates (0·00–0·02 µmol CO2 m−2 s−1).

Model-based interpolations and annual NEE comparisons

Herbaceous and woody communities contrasted in their seasonal responses to the differing years of 2008 and 2009. For moss and sedge communities, the 2009 growing season was longer and allowed for greater net CO2 uptake. During the TGS of both years, moss and sedge communities displayed net CO2 sink activity, a sink that became stronger during the warm, dry 2009 growing season (Figure 3). In contrast, the woody communities showed a longer period of sink activity during the cool, wet 2008 growing season. Our model-based interpolations suggest that, while both woody communities first became daily net sinks 22 days earlier in 2009 than in 2008, 2009 sink activity was generally less strong, with an earlier and more rapid reversal to net CO2 sources (Figure 3). Outside of the 2008 and 2009 TGS, moss and sedge communities lost substantially less CO2 to the atmosphere than shrub and small tree communities (Figure 3). Integrating the seasonal curves gave the annual CO2 balance of each community (Table 4). Annual NEE (g CO2–C m−2) estimates for 2008 and 2009 indicated that the moss community was a CO2 sink, the sedge community was approximately CO2 neutral, and shrub and small tree communities were sources of CO2 to the atmosphere. The strength of the CO2 source in the shrub and small tree communities was apparently much greater during 2009. In both years, the estimated CO2 source strength of the small tree community (170·7 and 181·7 g CO2–C in 2008 and 2009, respectively) far surpassed the estimated CO2 sink strength of the moss community (−59·0 and −132·8 g CO2–C in 2008 and 2009, respectively). Outside of the TGS, integrated NEE estimates indicated winter respiration followed the pattern moss < sedge < shrub < small tree, with more CO2–C respired over the 2008 winter than the 2009 winter. During both 2008 and 2009, herbaceous community winter respiration (moss2008 = 40·9, sedge2008 = 67·7, moss2009 = 46·6, sedge2009 = 76·3 g CO2–C) was half that of woody communities (shrub2008 = 113.4, small tree2008 = 134.8, shrub2009 = 127·4, small tree2009 = 151·9 g CO2–C). Winter respiration alone was enough to offset any carbon sequestered during the growing season for woody communities in both 2008 and 2009, and for the sedge community in 2008 (i.e. fluxes of equal or greater magnitude). Growing season NEE estimates for the moss community were approximately 3–4 times greater than winter respiration estimates.

image

Figure 3. Daily net ecosystem exchange (NEE) estimates per community (g CO2–C m−2). Daily NEE mean (black) ± standard error (grey) by community (NEE = ER + GEP). ER values estimated using logistic respiration model with community-specific parameter estimates and hourly air temperatures (n = 1); GEP values estimated using the Shaver et al. (2007) model with plot-specific LAI estimates (n = 5) and hourly PAR; hourly flux values then summed over each day. Vertical dashed lines give start and end dates of thermal growing seasons (TGS).

Download figure to PowerPoint

Table 4. Integrated annual NEE estimates per community (g CO2–C m−2, inline image± 1 SE, n = 5 plots). There were significant differences between communities (two-way repeated measures ANOVA – Tukey post hoc comparisons: year F1, 16 = 0·2, p = 0·68; community F3, 16 = 39·0, p < 0·01; year*community F3, 16 = 3·2, p = 0·05). With the exception of shrub and small tree communities (t16 = −1·79, p = 0·31), the annual net exchange was significantly different for each pair-wise community comparison (t16 < −4·09, p < 0·004).
 Annual NEE (g CO2–C m−2)
  1. NEE, net ecosystem exchange; ANOVA, analysis of variance.

 MossSedgeShrubSmall tree
2008−59·0 (±24·8)4·7 (±22·5)78·8 (±30·5)170·7 (±39·6)
2009−132·8 (±11·0)−28·1 (±21·2)138·7 (±17·4)181·7 (±53·5)

DISCUSSION

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. ACKNOWLEDGEMENTS
  8. REFERENCES
  9. Supporting Information

Wetlands in the Kenai Lowlands are shifting from herbaceous (moss and sedge) to woody (shrub and small tree) communities during a regional warming and drying trend. The combination of environmental (warming and drying) and ecological (vegetation succession) changes could lead to increased CO2 emissions through greater respiration rates, or increased CO2 sequestration through greater rates of photosynthesis. Our measurements of CO2 exchange along a hydrologic gradient during a cool, moist and a warm, dry year indicate that woody communities (shrub and small tree) were net sources of CO2, whereas herbaceous communities (moss and sedge) were net CO2 sinks. The differences in NEE between woody and herbaceous communities were amplified in the warm, dry year when compared with a more normal cool, wet year. GEP was greater in the woody communities than in the herbaceous communities, due to their greater leaf area and canopy photosynthetic capacity (Street et al., 2007). Woody community GEP rates were depressed during unusually warm and dry conditions in July and August of 2009, presumably due to partial stomatal closure in response to the combination of dry soils and a warm, dry atmosphere. Growing season ER was consistently greater in woody than in herbaceous communities and substantially greater in the warmer, drier 2009 than in 2008 for all communities. Similar to our results, Cai et al. (2010) measured growing season CO2 flux in a Canadian boreal peatland and found that both GEP and ER increased under warm and dry conditions, but increases in ER far outweighed increases in GEP. Bubier et al. (2003) measured NEE of a Canadian boreal peatland during wet and dry years in various communities, including bog hummocks and hollows, sedge dominated fens, and along vegetation and hydrology gradients. GEP was similar among communities and declined during dry years, whereas ER was substantially higher in dry communities and increased during dry years.

There are several possible explanations for greater ER in drier woody community soils, and particularly in the warm 2009 growing season. Drier woody community soils, combined in 2009 with warmer temperatures, may have increased ER rates through either decreased SWC increasing the aerobic profile of soils (Silvolva et al., 1996), and/or increased soil temperature increasing physiological activity, irrespective of water table position (Updegraff et al., 2001; Lafleur et al., 2005). Greater GEP in the shrub and small tree communities likely provides substrates for higher rates of autotrophic and heterotrophic soil respiration (Bardgett et al., 2008; Heimann and Reichstein, 2008), further increasing ER rates in woody communities. Introduction of labile carbon to new depths in the soil profile can stimulate microbial decomposition of more recalcitrant compounds, leading to even further increases in heterotrophic respiration (Fontaine et al., 2007; Bardgett et al., 2008; Heimann and Reichstein, 2008).

Our results, which show woody communities were greater C sources to the atmosphere, particularly under warm and dry conditions, contrast with those of a recent study in a rich fen in Alberta, where it appears that warming, drying and vegetation succession have led to an increase in C sequestration (Flanagan and Syed, 2011). Flanagan and Syed (2011) used the eddy covariance method to examine inter-annual variation in CO2 exchange of a treed fen. They found that both photosynthesis and respiration increased under warm and dry conditions, and those current rates of NEE are much greater than historical rates at their study site. It is possible that our study shows a different pattern than observed in Alberta, because the two studies were focused on different stages of succession. Although our study examined the initial process of directional woody vegetation encroachment, Flanagan and Syed (2011) examined a later stage of succession, the process of establishment and recruitment of new trees within an area already occupied by trees. Given the contrasting results of the two studies, we hypothesize that changes in NEE over time with vegetation succession may be decidedly non-linear. Because our study was carried out along the invasion front, it is possible that roots of more mature black spruce could be present beneath the plots in our small tree community. Inclusion of GEP estimates for these more mature trees could affect our conclusion that warming, drying and woody vegetation encroachment lead to a substantial loss of C in the Kenai Lowlands. In order to evaluate this possibility and our hypothesis that more mature black spruce communities beyond the dry end of our hydrologic gradient may be weaker C sources or even C sinks, we made a set of simple calculations based on stand density, tree size and tree age in two zones of increasingly mature black spruce beyond our small tree community.

Most individuals in the small tree community stood <40 cm in height, but there were several taller mature trees (from 60 to 360 cm in height) near enough to our study plots (i.e. within a distance of twice their height) that their roots could be present beneath our plots. We estimated the potential contribution of these trees to GEP by estimating their biomass (Bond-Lamberty et al., 2002), converting biomass to grammes of C by multiplying by 0·5, dividing biomass by age (Gower et al., 2001) and adding this estimate to our annual GEP estimates. Using this rough estimate of productivity outside of our plots reduced the strength of the C source in the shrub community by 2·3 g CO2–C m–2 to 76·5 in 2008 and 136·4 g CO2–C m–2 in 2009. The same approach reduced the strength of the small tree community C source by 42·2 g CO2–C m–2 to 128·5 in 2008 and 139·5 g CO2–C m−2 in 2009. The results of this exercise clearly show that, although the presence of roots from mature black spruce beneath our small tree plots could inflate our estimates of the C source, it would not alter our main conclusion that warming, drying and woody vegetation encroachment lead to a substantial loss of C at our study site.

In order to further evaluate the potential for non-linear changes in NEE over time, as the vegetation community develops into a more mature black spruce forest, we made similar estimates of annual productivity in a more established black spruce community (individual tree height 18–360 cm tall), in a drier area 17 m beyond our small tree community. Assuming that ecosystem respiration and understory productivity were similar to that measured in the small tree community, this more mature black spruce community would have been a source of 142·1 and 153·1 g CO2–C m−2 in 2008 and 2009. These analyses suggest that our study plots likely spanned the most dramatic shift in NEE, between the sedge and shrub communities, and that shifts in NEE become more gradual as black spruce increase in abundance. Nevertheless, our estimates of C loss from the woody plots and the adjacent black spruce communities suggest that, if current rates are maintained, a large proportion of the soil C stock could be lost to the atmosphere over the next century. This finding is consistent with results of an experiment in a sub-arctic peatland, where 1 °C of warming induced 60% greater spring respiration rates, and 52% greater summer respiration rates (Dorrepaal et al., 2009). Although both surface (upper 25 cm) and subsurface peat (25–50 cm) exhibited increased respiration rates, the majority (69%) of the increase in respiration came from subsurface peat (Dorrepaal et al., 2009). Increased respiration rates were sustained through the 8 years of study by Dorrepaal et al. (2009), indicating that large pools of soil carbon in sub-arctic peatlands are susceptible to long-lasting, warming-induced losses to the atmosphere. Their results suggest that increased heterotrophic respiration in drying Kenai peatlands will likely not exhaust labile organic compounds in a short-lived pulse of CO2 emissions, but instead lead to a long lasting conversion of peat into atmospheric CO2.

It is likely that development of a more mature black spruce forest will partially offset the large losses of C observed in our study. Eddy covariance studies in Canadian black spruce forests under various permafrost regimes indicate that, although intermediate-age forests are relatively strong sinks, mature black spruce forests range from weak sources to weak sinks (Bergeron et al., 2007; Coursolle et al., 2007; Dunn et al., 2007; Flanagan and Syed, 2011; Goulden et al., 2011). Meanwhile, Kane and Vogel (2009) showed that C stored in black spruce tissues can be as high as 8·5 kg m−2 in fully-stocked stands. It is likely that continued warming, drying and black spruce forest development at our site would lead to a shift in the C stock from soils to black spruce trees and that this shift might partially, but probably not fully, compensate for the loss of soil C. The C stocks and fluxes of forests on the permafrost-free Kenai Peninsula have never been assessed and are an important future research objective.

During the winter, snow depths, soil temperatures and ER rates were greater in the woody than in the herbaceous communities. Winter ER made up 20%–24% of annual respired CO2 in our study, consistent with the 22% reported by Larsen et al. (2007) for a study of a subarctic heath in northern Sweden, the 21%–33% reported by Alm et al. (1997) for sedge lawns and ericaceous hummocks in a Finnish boreal peatland, and the 23% reported by Kim et al. (2007) for Picea mariana, Sphagnum, and feathermoss permafrost forests of interior Alaska. Winter ER rates were an order of magnitude lower than growing season ER rates. However, these low but consistent CO2 emissions over the long winters are an important component of annual CO2 budgets. Although snow depths were deeper and soil temperatures were warmer during winter in the woody communities, the differences in respiration were greater than can be explained by the rather small differences in soil temperature. Instead, we hypothesize the greater rates of winter respiration in the woody communities are largely attributable to greater aeration and a larger biomass of respiring organisms.

In order to understand the net radiative forcing of drying wetlands on the Kenai Lowlands, we must also understand changes in surface energy exchange related to encroaching woody vegetation (e.g. Lohila et al., 2010), and how CH4 emissions change with warming and drying soils and accompanying vegetation succession. Methane emissions are greatest in waterlogged soils, and experimental water table drawdown generally reduces CH4 release (Updegraff et al., 2001; Strack and Waddington, 2007; Turetsky et al., 2008). The role of sedges in methane release, however, seems to vary depending upon the interaction between sedge species (Ström et al., 2005) and water table heights (Joabsson and Christensen, 2001; Strack et al., 2006). As methane has 25 times the warming potential of CO2 over a 100 year horizon (IPCC, 2007), future work in the Kenai Lowlands should seek to quantify the role of CH4 in herbaceous and woody community greenhouse gas exchange.

Wetland drying results in a shift from herbaceous to woody communities in the Kenai Lowlands (Berg et al., 2009; see also Riutta et al., 2007). Continued warming and drying will likely lead to further invasion of herbaceous communities by woody communities. Although woody communities have higher GEP rates than the herbaceous communities they replace, increased GEP will likely be more than offset by increased ER (both autotrophic and heterotrophic). We expect that warming and drying on the Kenai Lowlands will result in conversion of ecosystems that have long been CO2 sinks into CO2 sources. Given the volume of peat stored in the Kenai Lowlands, this change will result in a substantial loss of C from terrestrial ecosystems to the atmosphere. Our estimates of the substantial differences in CO2 flux between woody and herbaceous communities, particularly under warm and dry conditions, underscore the need for further work to understand the complex relationships between a changing climate and vegetation succession as determinants of greenhouse gas fluxes from peatlands of the Kenai Lowlands to the atmosphere.

ACKNOWLEDGEMENTS

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. ACKNOWLEDGEMENTS
  8. REFERENCES
  9. Supporting Information

This project was supported by a US Fish and Wildlife Service Challenge Cost Share Program grant and grants from the National Science Foundation to JMW (ARC-0612534) and PFS (ANT-0528748 and ARC-0909155). We thank the Kenai National Wildlife Refuge and staff for assistance with this project. The comments of an anonymous reviewer substantially improved an early draft of this article.

REFERENCES

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. ACKNOWLEDGEMENTS
  8. REFERENCES
  9. Supporting Information

Supporting Information

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. ACKNOWLEDGEMENTS
  8. REFERENCES
  9. Supporting Information

Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
eco_274_sm_figure.docWord document98KSupporting Information

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.