Introduction

Changes in atmospheric greenhouse gas (GHG) concentrations have been linked to large-scale earth system changes such as warming global temperatures, sea level rise, ocean acidification, and changes in snowpack dynamics (IPCC 2013). In 2011, carbon dioxide (CO2), methane (CH4), and nitrous oxide (N2O) concentrations in the atmosphere reached 391 ppm, 1803 ppb, and 324 ppb, respectively, which represent an increase of 40, 150, and 20 % over preindustrial levels (IPCC 2013). Although anthropogenic emissions of GHG are unequivocally tied to these changes, natural systems such as forests have the potential to affect GHG concentration in the atmosphere (Beier et al. 2015). Forest ecosystems store a globally estimated 1146 Pg of carbon, with roughly two thirds of this amount belowground (Dixon et al. 1994). Thus, a thorough characterization of the amount of GHG produced or consumed at the soil–atmosphere interface in forested systems is needed to fully assess the potential of forested environments to regulate GHG concentration in the atmosphere.

Although several studies have documented GHG dynamics at the soil–atmosphere interface in forested environments (Bowden et al. 1993; Castro et al. 1995; Smith et al. 2003; Riveros‐Iregui et al. 2007; Ullah and Moore 2011), watershed-scale estimates of GHG emissions are too few to allow for robust estimates of global GHG emission at the soil–atmosphere interface across a variety of forested systems (IPCC 2007; Hashimoto 2012). Furthermore, developing GHG budgets at the watershed scale is complicated by the known spatial and temporal heterogeneity of soil biogeochemical processes such as aerobic respiration (CO2 source), nitrification and denitrification (N2O source and sink), methane oxidation (CO2 source, CH4 sink), methanogenesis (CO2 sink, CH4 source), and the dependence of GHG fluxes on these processes (Hedin et al. 1998; Naiman et al. 2010).

Some studies nevertheless tackle the issue of GHG budget estimates at the watershed scale. For instance, Pacific et al. (2010) and Riveros-Iregui and McGlynn (2009) related GHG fluxes at the watershed scale to landscape position as a strategy to better predict GHG emission across space. In a companion study, Gomez et al. (in review) showed the importance of local landscape geomorphic characteristics in regulating GHG fluxes at the watershed scale. Other studies, such as Hill et al. (2004) and Gold et al. (2001), relate landscape geomorphic characteristics to soil biogeochemical processes (e.g., denitrification), which influence GHG fluxes at the soil–atmosphere interface. Many other studies have also linked landscape geomorphic characteristics (e.g., topography, contributing area, surficial geology) to soil biogeochemical processes, water quality, and soil organic carbon content (Hill 2000; Groffman et al. 2009; Vidon and Hill 2006; Vidon et al. 2010; Jacinthe et al. 2012). Because these variables affect GHG emission (Riveros-Iregui and McGlynn 2009; Pacific et al. 2010; Gomez et al. in review), quantifying GHG emission for a variety of landscape geomorphic classes and developing watershed GHG budgets based on landform classes is a promising avenue for research as a tool to advance the inventory and assessment of natural GHG emissions at the watershed scale.

Within this context, geographic information systems (GIS) are valuable tools to classify watersheds based on landscape hydrogeomorphic classes and subsequently assess a variety of ecosystem processes. In particular, topographic indices, which are often assessed using GIS, are increasingly used to model ecosystem processes such as hydrology, biogeochemical activity, evapotranspiration, erosion, and sedimentation (Quinn et al. 1995). Baker et al. (2001) also utilized GIS models to increase our ability to predict nutrient export via the analysis of the topographic influence on the hydrologic conditions of a study area. Rosenblatt et al. (2001) demonstrated the value of using GIS data to identify various geomorphic settings that were assessed for nitrate removal potential in riparian zones. With respect to topographic indices in particular, their use is especially valuable in classifying the landscape into a variety of geomorphic or hydrogeomorphic classes. For instance, the topographic wetness index (TWI) defined as TWI = ln (a/Tan b) (where “a” is the accumulated upslope contributing area per unit contour length and “b” is the local slope angle) is often used to assess the general distribution of soil moisture at the watershed scale (Moore et al. 1991) and is widely used in hydrological and biogeochemical studies (Dosskey and Qiu 2010; Walter et al. 2002; Ogawa et al. 2006; Grabs et al. 2012; Ruhoff et al. 2011). Within a given climate zone, a high wetness index indicates an area that has a high propensity to be moist relative to other locations with lower TWI values.

Our objective in this study is to estimate greenhouse gas (GHG) fluxes and characterize their spatial and temporal heterogeneity within a small-forested watershed of the US Northeast. To this end, we measured GHG fluxes (N2O, CO2, CH4) using a sampling design stratified by landscape position, on 22 occasions over a 3-year period, capturing a range of seasonal climatic and flow conditions. To extrapolate GHG point measurements to the watershed scale, we calculated a very high-resolution TWI (based on a <1-m digital elevation model) to derive landscape hydrogeomorphic classes (e.g., wetlands vs. riparian zones vs. hillslopes), and then used the TWI classes to estimate GHG fluxes and their spatiotemporal patterns at the watershed scale. As a GHG budget at the soil–atmosphere interface is developed, we discuss which areas of the landscape contribute the most to GHG fluxes in CO2-equivalent units (CO2eq), both seasonally and on an annual basis, and discuss the relative importance of soil–atmosphere GHG efflux relative to global above ground–below ground carbon sequestration by live vegetation.

Materials and methods

Study area

The Archer Creek Watershed is a 135-ha forested catchment located in the Adirondack region of northern New York State (Fig. 1). Climate is cool, moist, and continental (Shepard et al. 1989). Between the years of 1941–2007, the mean annual temperature in this area was 5.0 °C, the mean annual precipitation was 1046 mm, and the mean annual snowfall was 303 cm (Mitchell et al. 2009). Precipitation in 2011 was 47 % greater than the 30-year average, 8 % less than the 30-year average in 2012, and 12 % greater than the 30-year average in 2013. Snowfall was less than the 30-year average for the entire study period with 3, 52, and 40 % less snowfall occurring in 2011, 2012, and 2013, respectively. Precipitation and air temperature were recorded at the National Deposition Program/National Trends Network Site (ID NY20) located in the Huntington Wildlife Forest approximately 2.0 km from the study area. Stream discharge for Archer Creek (15 min. interval) was determined for the outlet of Archer Creek watershed where stage height is recorded as part of the ongoing long-term monitoring program of the watershed described in Mitchell et al. (2009).

Fig. 1
figure 1

Experimental site location and sampling locations within the Archer Creek Watershed (135 ha). Numbers 1–10 correspond to the sampling locations as indicated on the figure

Full size image

From a geomorphological standpoint, the Archer Creek catchment has an average slope of 11 %, a total relief of 225 m, and is located within the Anorthosite Massif igneous intrusion, which is composed of up to 90 % calcium-rich plagioclase feldspar (McHale et al. 2002; Mitchell et al. 2009). The surficial geology consists of glacial till, which was deposited by the Wisconsin Glacier approximately 14,000 years ago. The till is composed of approximately 75 % sand, less than 10 % clay content, and includes many boulders and cobblestones (Christopher et al. 2006). The forest soils are classified as Becket–Mundal series sandy loams (coarse-loamy, isotic frigid, Oxyaquic Haplorthods). Upland soils are approximately 1 m thick with localized areas of deeper soil. The O-horizon and Bs-horizon overlay the igneous bedrock and are generally 5 and 80 cm thick, respectively. Wetlands and valley-bottoms located in the Archer Creek Watershed are composed of Greenwood Mucky Peats deposits 1 to 5 m deep (Somers 1986). The lowland overstory in the Archer Creek Watershed is composed of eastern hemlock (Tsuga Canadensis (L.) Carr), yellow birch (Betula alleghaniensis Britt.), speckled alder (Alnus incana (L.) Moench), red spruce (Picea rubens Sarg.), and balsam fir (Abies balsamea (L.) Mill). American beech (Fagus grandifolia Ehrh.), sugar maple (Acer saccharum Marsh.), red maple (Acer rubrum L.), yellow birch, and eastern white pine (Pinus strobus L.) dominate the upland environments (Christopher et al. 2006).

Greenhouse gas sampling and analysis

Ten sampling locations corresponding to the variety of landscape hydrogeomorphic classes found in the watershed (palustrine wetland, lowlands, riparian zones, upper and lower hillslopes, and headwater wetlands; Fig. 1) were selected throughout the Archer Creek Watershed for GHG flux measurements. Although this approach targets specific areas of the watershed for sampling, it is designed to account for all major landscape hydrogeomorphic classes in the watershed that other sampling techniques (e.g., random or stratified sampling design approach) may miss. The lowland was located in a gently sloping area near the palustrine wetland. The riparian zone locations were located immediately adjacent to headwater streams, while the headwater wetlands were located near the source of the two streams used in this study (Fig. 1). The upper hillslope sites were located on generally steep terrain (>20 % slope) while the lower hillslope sites were located downstream of the headwater sites on more moderate slopes (<20 % slope). At each of these sites, greenhouse gas (CO2, CH4, N2O) sampling occurred 19 times during the 2011, 2012, and 2013 snow-free seasons, with an additional 3 winter sampling days in January, February, and March of 2013.

Gas samples for GHG flux calculations at the soil–atmosphere interface were collected using six open top static chambers per site (closed only during sampling), for a total of 60 chambers (Fig. 1). Static chambers were made of white PVC material and consisted of two parts: a bottom section (37 cm height and 27 cm diameter) inserted 5 cm into the ground, and an airtight lid fitted with a gas sampling port installed only during sampling to close the chamber (Jacinthe and Dick 1997). Chambers were only inserted 5 cm into the ground and left untouched during the study period to minimize the impact of chamber installation on shallow root systems. At the time of sampling, soil temperature was recorded at each sampling location 5 cm below the ground surface using a portable Hanna® pH/ORP/temperature meter (Model HI 9125, Hanna Instrument Inc., Woonsocket, Rhode Island, USA). When sampling, the chambers were closed with the lid and three headspace gas samples were extracted over a period of 50 min. Air samples (∼15 mL) were stored in evacuated vials (10 mL) fitted with gray butyl rubber septa. GHG fluxes (F) were computed as follows:

$$ F=\left(dC/dt\right)*\left(V/A\right)*k $$
(1)

where dC/dt is the rate of change in GHG concentration inside the chamber (mass GHG m3 air per min), V is the chamber volume (m3), A is the area circumscribed by the chamber (m2), and k is a unit conversion factor (1440 min/day) (Jacinthe and Lal 2004). Winter sampling was carried out following the same protocol, except that snow was removed by hand from each chamber immediately before the airtight lid was placed onto the chamber. The snow was returned into the chamber immediately after sampling. Samples were kept out of the light during transport and storage periods and were usually analyzed within 2 weeks of collection.

Carbon dioxide, CH4, and N2O concentrations were analyzed utilizing a Shimadzu GC-2014® gas chromatograph (Shimadzu Corporation, Kyoto, Japan) equipped with a flame ionizing detector and an electron capture detector interfaced with a CombiPal® autosampler (LEAP Technologies, Carrboro, North Carolina). The stationary phase consisted of a Porapak Q® precolumn (90 cm long) and Hayesep D® analytical columns (180 cm long). The gas chromatograph (GC) was calibrated using standard gases obtained from Alltech (Deerfield, Illinois). To prevent moisture buildup in the GC columns, the GC oven was heated (150 °C) for 15 min for every 20 samples.

Carbon dioxide, CH4, and N2O fluxes at each site were determined by taking the median flux of the six chambers for each sampling date. The CO2eq flux for CH4, CO2, and N2O fluxes were calculated according to the global warming potential (GWP) values for the 100-year time horizon reported by the International Panel on Climate Change (IPCC 2013). Because CO2 is the baseline, it is has a GWP of 1. The GWP for CH4 and N2O are 25 and 298, respectively (IPCC 2013). Seasonal GHG flux calculations were based on astronomical seasons.

We tested for the importance of sampling time on daily soil GHG flux estimates by conducting four intensive sampling campaigns (5-h intervals over a 24-h period) on September 6 and October 20, 2012, and June 6 and July 16, 2013, at one chamber per site (for logistical reasons) at sites RZ1, LH1, RZ2, LH2, HW2, and UH2 (Fig. 1). Carbon dioxide, CH4, and N2O fluxes varied over a 24-h period; however, sampling time did not consistently affect flux estimates. Fluxes are therefore reported on a per day basis as opposed to a per hour basis.

Topographic analyses and geomorphic classification of the landscape

A 1-meter resolution bare earth digital elevation model (DEM), derived from discrete LiDAR data collected in 2009, was used to assess the topographic attributes of the Archer Creek Watershed. The topographic wetness index (TWI) was calculated across the watershed using the following equation:

$$ \mathrm{T}\mathrm{W}\mathrm{I}= \ln \left(\mathrm{a}/ \tan \kern0.5em \mathrm{b}\right) $$
(2)

where “a” is the upslope contributing area per unit contour length and “b” is the local slope angle (Beven and Kirkby 1979; Dosskey and Qiu 2010). ArcGIS Model Builder was used to implement the TWI in a stepwise fashion using raster calculator functions. The default algorithm—D8 single flow direction—was used to calculate the flow direction from each cell, and then the flow accumulation for each grid cell was calculated. The D8 single flow direction method apportions all flow from a cell toward the neighboring cell in the steepest downslope direction of one of the eight cardinal directions (Holmgren 1994). A 5 m × 5 m moving average was implemented to smooth, or generalize, the TWI grid, which was then used to classify the watershed into discrete geomorphic classes.

For our study, we used four landscape classes based on TWI thresholds and visual confirmation in the field: wetland, lower hillslope, upper hillslope, and riparian zone. Stream locations were also confirmed by GPS coordinates taken with a Trimble GeoXH GPS unit and each GPS point was accurate to within <0.1 m. The classes based on TWI values ranged from 0 to 1.67, 1.68 to 3.99, 4.00 to 4.70, and 4.71 to 10.70 for the upper hillslope class, the lower hillslope, the riparian zone class, and the wetland class, respectively. Because the small headwater wetlands had GHG fluxes between those of the riparian zone sites and lower hillslope sites (see the “Results” section) and represented less than 0.5 % of the watershed, they were treated as “riparian zones “or “lower hillslopes” based on their TWI values for landscape geomorphic classification purposes. The lowland site was binned in the lower hillslope category for GHG budget purposes based on TWI values. Finally, when more than one site was found in one category, the arithmetic mean of the each site’s median flux was used for assigning a GHG flux to each geomorphic class.

Uncertainty analysis

In order to better characterize the uncertainty introduced by scaling up plot-level estimates to the entire landscape, we used a resampling (i.e., Monte-Carlo) procedure to estimate the variance around the daily landscape-level GHG flux in CO2 equivalent. For each combination of season and geomorphic class, we resampled the dataset consisting of the daily median flux values (gCO2-eq m−2 day−1) and weighted the resulting mean value by the estimated total acreage under that geomorphic class. We ran this resampling procedure 1000 times, after which we were able to calculate the mean and standard deviation of the resulting sampling distribution (n = 1000) for each season × geomorphic class combination. Finally, we added together the daily area-weighted flux values of each of the four geomorphic classes to estimate the total daily flux from the entire watershed within each season and took the mean of the four seasons to calculate the average daily flux from the entire watershed throughout the entire year. Analyses were done using the sample function in R version 3.0.2 (R Core Team 2013).

Results

Site hydrology and CO2, N2O, and CH4 fluxes across locations

A detailed description of hydrological conditions and associated CO2, N2O and CH4 emissions are provided in Gomez et al. (in review). Briefly, the 60 static chambers used in this study were sampled across a range of hydrological conditions ranging from high flow to low flow (measured at the Archer Creek watershed outlet) across all four seasons (3 winter dates, 5 spring dates, 9 summer dates, 5 autumn dates) (Fig. 2). Carbon dioxide fluxes exhibited a strong seasonal pattern (Fig. 3). Regardless of location, winter CO2 fluxes were significantly smaller and less variable (0.02 to 1.34 g C-CO2 m−2 day−1) than during the remainder of the year (−0.77 to 6.87 g C-CO2 m−2 day−1). Across locations, the lowland, LH1, UH1, and LH2 sites had mean CO2 fluxes greater than 2 g C-CO2 m−2 day−1 over the study period, while mean CO2 fluxes at the wetland and RZ1 site were less than 1 g C-CO2 m−2 day−1. Other sites (HW1, RZ2, HW2, and UH2) had intermediate CO2 flux values between 1 and 2 g C-CO2 m−2 day−1.

Fig. 2
figure 2

Daily discharge for years 2011, 2012, and part of 2013 at the Archer Creek Watershed outlet. Labels D1-D22 correspond to the dates when sampling for greenhouse gas emission occurred throughout the watershed. The dashed line indicates the mean annual flow at the watershed outlet

Full size image
Fig. 3
figure 3

Median CO2 (a), CH4 (b), and N2O (c) fluxes recorded at each sampling location over the study period. Labels D1-D22 correspond to the sampling dates as described in Fig. 2. Acronyms used in the legend correspond to the sampling locations as described in Fig. 1

Full size image

Methane fluxes were highest at the wetland site (range 2 to 1293 mg C m−2 day−1), with the largest fluxes occurring in the summer and early autumn (July through October). Positive CH4 fluxes were also observed at the RZ1 site for 18 of the 22 sampling dates; however, CH4 fluxes were negative on most dates at all other locations. When sites were compared, the average study period CH4 flux at the wetland site (330 mg C m−2 day−1; Relative Standard Deviation 127 %) was two orders of magnitude higher than the average CH4 flux at any other site. Nitrous oxide fluxes did not exhibit a strong seasonal pattern or clear differences between locations. On most dates, N2O fluxes ranged from −0.5 to 0.5 mg N m−2 day−1.

Average CO2-equivalent fluxes of GHG across landscape positions

Average CO2eq GHG efflux (sum of g CO2eq for each of CO2, CH4, and N2O) at each site varied among seasons (Table 1). Soil CO2eq GHG emissions at each sampling location were consistently the lowest during the winter season and the highest during summer. Across all locations, the wetland site (palustrine wetland) had the highest CO2eq fluxes in the summer (20.56 g CO2eq m−2 day−1) and the third lowest CO2eq flux in the winter (0.78 g CO2eq m−2 day−1). The mean CO2eq GHG efflux for the entire study period (n = 22) ranged between 2.98 g CO2eq m−2 day−1 (RZ1 riparian zone) and 13.27 g CO2eq m−2 day−1 (wetland site) across the study sites (Fig. 4). Carbon dioxide was the primary contributor to the positive CO2eq fluxes at each of the ten sampling locations except at the wetland site where CH4 contributed 83 % of average CO2eq fluxes over the study period. The CO2 flux comprised more than 100 % of the net positive CO2eq GHG flux at the lowland, HW1, and UH1 sites owing to negative CO2eq fluxes of N2O and/or CH4 at these sampling locations. During the study period N2O flux expressed in CO2eq ranged between −0.34 to 0.38 g CO2eq m−2 day−1 across the landscape. The N2O flux contributions to the net daily CO2eq fluxes were very small (∼1 % CO2eq emissions in most cases) compared to the contributions of CH4 fluxes in the wetland or CO2 fluxes at all other locations to total CO2eq emissions (Fig. 4).

Table 1 Mean CO2eq flux for the wetland, lowland, RZ1—riparian zone (RZ1), LH1—lower hillslope (LH1), HW1—headwater wetland (HW1), UH1—upper hillslope (UH1), RZ2—riparian zone (RZ2), LH2—lower hillslope (LH), HW2—headwater wetland (HW), and UH2—upper hillslope (UH2) sites for each season of the year
Full size table
Fig. 4
figure 4

Mean CH4, CO2, and N2O fluxes in CO2eq for each sampling location (see N for sampling location acronym definition) over the study period

Full size image

Topographic and geomorphic classification of the Archer Creek Watershed

The slope of the land surface in the Archer Creek watershed ranged between 0 and 77° (Fig. 5). The TWI index ranged from 0 (least propensity to be saturated) to 15.88 (greatest propensity to be saturated) with an average of 1.87 (±2.12) (Fig. 5). Following the 5 m × 5 m moving average smoothing procedure described in the “Materials and methods” section, the generalized (smoothed) TWI values across the watershed ranged between 0 and 10.70. The total geomorphic areal contributions were 46.6, 72.5, 6.83, and 7.62 ha for the upper hillslope, lower hillslope, riparian area, and wetland categories, respectively (Fig. 5). These geomorphic areas were adjusted to account for the estimated stream area of 12,075 m2 (i.e., stream areas were excluded from geomorphic classes). The lower hillslope class comprised the largest portion of the watershed area (54 %) followed by the upper hillslope (35 %), wetland (5.7 %), riparian area (5 %), and stream (1 %) classes.

Fig. 5
figure 5

Various stages of the landscape classification procedure based on the hydrogeomorphic characteristics of the Archer Creek watershed: a hillshade view of the topography of the watershed and sampling site location (white dots), b slope, c topographic wetness index (TWI); and d generalized geomorphic classes of the watershed. Numbers in parenthesis indicate the surface area of each land class

Full size image

Area-weighted CO2eq GHG budgets for the Archer Creek Watershed

Extrapolating average seasonal CO2eq fluxes recorded for each season for each geomorphic class on an areal basis resulted in estimated daily winter, spring, summer, and autumn GHG fluxes at the soil–atmosphere interface of 3.60 × 106, 9.65 × 106, 1.58 × 07, and 1.11 × 107 g CO2eq day−1, respectively (Table 2). When the average daily CO2eq fluxes for each landscape class were combined, the watershed CO2eq efflux due to CO2 was at least an order of magnitude greater than the CO2eq efflux due to CH4 and N2O fluxes across the four seasons. The Archer Creek Watershed was a net CH4 sink (−3.28 × 104 g CO2eq day−1) during the winter season, but was a net source for all other seasons. When seasonal averages are considered across all land classes, the annual GHG budget for the entire watershed was 1.01 × 107 g CO2eq day−1 or 3.68 Gg CO2eq year−1. The summer, fall, winter, and spring seasons accounted for 40, 27, 9, and 24 % of the annual CO2eq budget from the Archer Creek Watershed.

Table 2 Winter, spring, summer, and autumn daily GHG flux for the entire Archer Creek watershed (135 ha) in CO2eq (left column), and daily CH4, CO2, and N2O fluxes in CO2eq for each landscape hydrogeomorphic class and each season (columns 3, 4, 5)
Full size table

Overall, the influence of CO2 flux across the landscape was the greatest, with CO2 flux contributing approximately 92 % of the net GHG efflux from the Archer Creek Watershed on a daily basis. Methane emission comprised 6.94 % of the net average daily CO2eq efflux across the watershed, while N2O fluxes comprised just over 1 % of the annual net CO2eq emissions in the whole watershed. The upper and lower hillslopes (which comprise 90 % of the Archer Creek Watershed Area) were combined CH4 sinks on the order of −4.56 × 105 g CO2eq day−1, partially offsetting the combined positive CH4 efflux of 6.80 × 105 g CO2eq day−1 from the riparian and wetland areas.

Uncertainty analysis for area-weighted CO2eq GHG budgets

The mean daily carbon flux from the Archer Creek watershed plus two standard errors (i.e., standard deviations of the sampling distribution) was equal to 1.01 × 107 ± 1.05 × 106 g CO2eq day−1 (Fig. 6); in other words (according to the central limit theorem), >95 % of our resampled values fell within 10 % of the initial point estimate. Compared to the annual mean, estimated uncertainty was greatest during the summer (2 SE = 2.75 × 106 g CO2eq day−1) and least during the winter season (2 SE = 8.67 × 105 g CO2eq day−1). The uncertainty associated with daily total flux measurements differed across geomorphic class (Fig. 7). The least certain contribution to the annual landscape carbon flux was associated with the lower hillslope region during the summer (2 SE = 2.23 × 106 g CO2eq day−1), which is the largest category (54 % of the watershed area) as well as the category with the greatest variability of methane fluxes as compared to carbon dioxide and nitrous oxide fluxes. In the summer, when methane fluxes are highest, uncertainty estimates around the wetland fluxes (2 SE = 8.59 × 105 g CO2eq day−1) are more than an order of magnitude greater than those around the fluxes from the riparian zones (2 SE = 7.72 × 104 g CO2eq day−1).

Fig. 6
figure 6

Total daily carbon fluxes in CO2eq for Archer Creek watershed on an annual and seasonal basis. Error bars represent two standard errors as calculated as part of the uncertainty analysis using Monte-Carlo analysis (n = 1000)

Full size image
Fig. 7
figure 7

Total daily carbon fluxes in CO2eq for each geomorphic class in Archer Creek watershed on a seasonal basis. Error bars represent two standard errors as calculated as part of the uncertainty analysis using Monte-Carlo analysis (n = 1000)

Full size image

Discussion

Impact of landscape position on CO2, N2O, CH4, and total CO2 equivalent

As discussed in details in Gomez et al. (in review), the GHG fluxes measured in this study were consistent with previously published results (Bowden et al. 1993; Riveros-Iregui and McGlynn 2009; Ullah and Moore 2011; Jacinthe et al. 2012; Morse et al. 2012; Turetsky et al. 2014). Similarly, when converted to CO2 equivalent, our results are consistent with other studies. For instance, Jungkunst et al. (2008) reported CO2 equivalent fluxes between 8.9 and 13.2 g CO2eq m−2 day−1 in a mineral hydromorphic soil in a cool-humid forested watershed dominated by Norway Spruce in Germany. Jungkunst et al. (2008) also reported CO2eq fluxes across a range of induced moisture conditions with ranges of 6.3 to 20.4 g CO2eq m−2 day−1 and 6.0 to 19.1 g CO2eq m−2 day−1 in a mineral soil with peat layers and in a purely organic soil, respectively. When converted to CO2 equivalent, the CH4, N2O, and CO2 fluxes reported by Ullah and Moore (2011) in forested catchments in Quebec, Canada were 5.1, 4.5, and 4.0 g CO2eq m−2 day−1 at an Old Growth Upland, Old Growth riparian, and a wetland site, respectively. Although these CO2eq flux magnitudes were generally smaller than those reported in our study (Table 1), the degree of influence of each gas on the net CO2eq flux was similar to those in our study (Fig. 4).

The importance of CH4 emissions in driving total CO2eq emissions at the wetland site is consistent with the highly reduced conditions often observed in wetlands, which promote methanogenesis and high rates of CH4 emission (Groffman and Pouyat 2009; Morse et al. 2012). Our findings also suggest that the CO2eq flux of riparian areas, which are areas in the landscape that are often saturated and may promote reduced conditions, were influenced largely by CO2 fluxes rather than CH4 (Fig. 4). This suggests that anaerobic conditions did not persist for sufficiently long periods of time at these locations to promote CH4 production as observed at the wetland site. Generally, the same areas associated with high CO2 flux are CH4 sinks, though these sinks account for less than 1 % of the net positive CO2eq flux in this study (Fig. 4).

In terms of seasonality, the highest net CO2eq fluxes at each of the sampling sites throughout the year were recorded during the summer months, which were associated with high CO2 fluxes across the watershed, and high CH4 fluxes at the wetland site. These elevated fluxes can be attributed to higher soil temperatures and increased belowground microbial and root respiration (Davidson et al. 1998; Bowden et al. 1998). Autumn and spring net CO2eq GHG fluxes at each site were similar except at the wetland site where autumn CO2eq flux was 17.50 g CO2eq m−2 day−1 and that of the spring was 3.40 g CO2eq m−2 day−1. Lower spring CO2 fluxes at the wetland site were likely due to the wetland remaining frozen late into spring (personal observation) and high water levels induced by spring snow melt, which would limit gas diffusion (Davidson et al. 1998). This variation in GHG fluxes between the winter and summer seasons has implications for the calculation of annual GHG flux estimates. For instance, by sampling for greenhouse gas fluxes only during growing seasons, important fluctuations in greenhouse gas emissions due to seasonal trends in moisture patterns, temperature patterns, respiration (both autotrophic and heterotrophic), and photosynthesis may be missed (Rosenbaum et al. 2012; Carrara et al. 2004; Monson et al. 2006). If these temporal changes in GHG fluxes are not measured, total CO2eq GHG budgets will not provide accurate estimates of the forest soil contribution to GHG fluxes to the atmosphere on an annual basis. Our data suggests that sampling only during the growing season and extrapolating to the entire year would result in overestimates of GHG fluxes to the atmosphere.

Pros and cons of using the TWI method to identify landscape geomorphic classes and validation of land class estimates

Freer et al. (1997) indicate that knowledge of the hydrologic and biogeochemical functions of the watershed is necessary in order to implement the TWI method. Rodhe and Seibert (1999) also indicate that TWI results are often validated using stream hydrographs or by validating the predicted saturated overland flow or wetland occurrence in the landscape. Thus, TWI approaches offer a tool to generalize landscape analysis across location, but they should be validated in the field. In this study, we used ground survey measurements and selected GPS points (<1 m accuracy) to ensure that the TWI results accurately represented the moisture patterns in the watershed. In general, the TWI method gave an accurate portrayal of hydrological conditions in the watershed, with highest TWI values (indicating the highest propensity to be saturated) at known stream locations and lowest TWI values (indicating least propensity to be saturated) at known steep hillslope areas.

Although the TWI method is a useful tool to help identify locations in the landscape with a high propensity to be saturated, several assumptions were made when applying the TWI methodology to our watershed: (1) soil hydraulic conductivity and soil depth are uniform across the watershed; (2) topography represents the underlying geologic gradient; (3) local slope angles are representative of the downward hydraulic gradient; and (4) the resolution of the topographic data are high enough to sufficiently capture changes in topographic conditions. In addition to the fact that previous studies have shown that topography can be a useful predictor for soil moisture potential in mountainous regions (Nippgen et al. 2011; Riveros-Iregui and McGlynn 2009), field observations indicated the presence of very shallow soil above glacial till (confining layer for groundwater) in Archer Creek Watershed. It is therefore likely that the Archer Creek Catchment topography reflects the subsurface geologic gradient. Furthermore, our study used high resolution (1 m) topographic data in order to best capture small-scale topographic changes. High resolution Lidar data has been shown to be a good representation of the actual ground surface, especially in comparison to digital elevation models (DEMs) derived from topographic contour maps (Vaze et al. 2010). Therefore, we believe that in our study, the local slope angle is captured at a high resolution and represents an accurate downward hydraulic gradient. Nevertheless, it is important to note that the use of high resolution topographic data (∼1 m) tend to overemphasize local variations in topography and the overemphasis can be propagated in the analysis of topographic wetness indexes, skewing realistic flow and moisture patterns (Buchanan et al. 2013). To address this limitation, a DEM smoothing procedure was used in our study (see the “Materials and methods” section). Smoothing, or generalization, has been shown to reduce the effect of local topographic anomalies and moderate excessive flow convergence propagated by algorithms used in GIS analysis (Buchanan et al. 2013). The resulting smoothed TWI grid therefore represents a more realistic representation of soil moisture patterns than when implementing the TWI directly on high-resolution data.

The TWI values found in this study fell between 0 and 16, which is consistent with ranges reported in a study by Ruhoff et al. (2011) for a variety of landscapes. The estimated area for the wetland (5 % total surface area) using the TWI procedure described herein is consistent with wetland area estimates previously established in Archer Creek (4 %; Bischoff et al. 2001). This value is also consistent with the overall surface area of wetlands (4 %) in the Adirondack Park (Homer et al. 2007). Assuming a 2-m wide buffer along the approximate stream length of 12,075 m in the Archer Creek watershed, riparian areas would include 3.2 % of the Archer Creek Watershed. We found a slightly higher riparian zone area using the TWI method (5 %), which could be due to wider riparian areas in level terrain along certain stretches of the stream channel. Elsewhere, Riveros-Iregui and McGlynn (2009) estimate riparian areas to be 2 % of a 393-ha forested watershed on Montana, USA. In that study, the authors classified the remainder of the watershed as hillslopes.

Consequently, in spite of the assumptions needed to implement a TWI methodology to identify land classes, this approach provides an effective tool to identify major hydrogeomorphic land classes across multiple watersheds. Given comparable results with estimations of land class areas in Archer Creek and the greater Adirondacks Park offered in other studies (see references above), we believe that the TWI landscape classification methodology proposed here is a viable and easily scalable alternative to extensive field surveying or manual analysis of topographic maps to identify dominant landscape geomorphic classes at the watershed scale.

Applications of the TWI method to GHG budget estimation at the watershed scale

From a soil–atmosphere GHG budget estimation perspective, this TWI landscape classification offers the advantage of accounting for small areas of the watershed (e.g., wetlands) that have a disproportionate effect on total GHG emissions. As an alternative to this approach, a random sampling design or a systematic sampling design approach would have limited capacity to systematically account for fluxes in areas of the landscape that have a small footprint but large GHG fluxes. Another benefit of this TWI method over a traditional random or stratified sampling design is that it decreases the number of sampling locations needed to accurately capture the heterogeneity of the hydrogeomorphic conditions typically found in a forested watershed.

When the GHG budget at the soil–atmosphere interface was calculated using seasonal averages extrapolated to an areal basis, the lower hillslopes were the greatest influence on the net CO2eq watershed GHG efflux, largely because these hillslopes encompass the largest proportion of the Archer Creek Watershed and have high CO2eq fluxes (Table 2). Wetlands were a greater CO2eq GHG source than riparian areas were, even though the Wetland class had a smaller area, because of high CH4 emissions during the summer and autumn seasons (Table 2). These results are consistent with higher temperatures and increased biological activity in the soil during the summer season (Bowden et al. 1998; Davidson et al. 1998). Low winter GHG fluxes can be explained by reduced soil microbial activity and depressed autotrophic respiration (Atkin et al. 2000). Frozen soils also likely limited the soil–atmosphere fluxes during the winter.

Over the entire study period, the net GHG flux across the whole Archer Creek watershed (135 ha) was 1.01 × 107 or 3.68 Gg CO2eq year−1, 99 % of which comes from CO2 and CH4 emissions as N2O represents ∼1 % CO2eq emissions in most cases (see the “Results” section). Although fluxes of GHG (mainly CO2 and CH4 fluxes) to the atmosphere at the soil–atmosphere interface are only one of the components of a full C budget for the watershed, when put in context with the C sink provided by live vegetation, these fluxes provide new insights into the potential role of forests at regulating GHG emission to the atmosphere. According to Birdsey (1992), live trees in the US Northeast accumulate between 1400 and 2000 lbs C acres−1 year−1, which converts to an approximate 0.21–0.30 Gg C year−1 sequestration rate for vegetation in the entire Archer Creek watershed. By comparison, we estimated a 1.0 Gg C year−1 flux from the soil to the atmosphere for the entire Archer Creek Watershed. If we assume the mature forest (>80 year old) approximates the Birdsey (1992) average, C emissions from soil exceed annual C sequestration by 3–5 times. However, recent growth rates based on forest inventory data from Archer Creek (Beier, unpublished data) indicate a lower net annual sequestration rate than Birdsey (1992), which suggests an even larger disparity between fluxes. A full ecosystem C budget at the watershed scale, including estimates of soil C storage and dissolved organic carbon losses to streams, should be the focus of a future study; however, our current results suggest that Archer Creek watershed could be a net source of greenhouse gases in the landscape.

Conclusion

This study reports GHG emission as CO2, CH4, N2O, and CO2eq fluxes across a range of landscape hydrogeomorphic classes (upper hillslope, lower hillslope, wetland, and riparian zone), and demonstrate a methodology using the topographic wetness index (TWI) to classify the landscape in distinct landscape geomorphic classes for the extrapolation of point measurements to the watershed scale, using a small forested catchment in the US Northeast. Results showed clear differences in both the average study period and average seasonal fluxes at upper hillslope, lower hillslope, wetland, and riparian zone geomorphic landscape positions within the 135-ha Archer Creek Watershed. Results also identified the TWI landscape classification method presented in this study as a viable and scalable tool to identify dominant landscape geomorphic classes in forested watersheds of the US Northeast. When used to estimate GHG fluxes across the entire surface area of the watershed, this TWI landscape classification method offers the possibility of systematically accounting for small areas of the watershed (e.g., wetlands) that have a disproportionately large effect on total GHG emissions. When compared to estimates of average C uptake rates by forest trees in the US Northeast, results suggest that soil CO2eq efflux at the Archer Creek Watershed exceeds annual carbon sequestration by vegetation. A more comprehensive ecosystem C budget is needed to assess whether the forest watershed is a net source or sink of greenhouse gases.