Warmer spring conditions increase annual methane emissions from a boreal peat landscape with sporadic permafrost

About a fifth of the global wetland methane emissions originate from boreal peatlands, which represent an important land cover type in boreal landscapes in the sporadic permafrost zone. There, rising air temperatures could lead to warmer spring and longer growing seasons, changing landscape methane emissions. To quantify the effect of warmer spring conditions on methane emissions of a boreal peat landscape in the sporadic permafrost zone of northwestern Canada, we analyzed four years (2013–2016) of methane fluxes measured with the eddy covariance technique and long-term (1951–2016) meteorological observations from a nearby climate station. In May, after snowmelt was complete, mean air temperatures were more than 2 °C warmer in 2013, 2015, and 2016 than in 2014. Mean growing season (May–August) air temperatures, in contrast, differed by less than 1 °C over the four years. Warmer May air temperatures caused earlier wetland soil warming, with temperatures rising from ~0 °C to >12 °C 25 to 40 days earlier and leading to ~6 °C warmer mean soil temperatures between May and June. However, from July to August, soil temperatures were similar among years. Mean May to August and annual methane emissions (6.4 g CH4 m−2 and 9.4 g CH4 m−2, respectively) of years with warmer spring (i.e. May) temperatures exceeded emissions during the cooler year by 20%–30% (4.5 g CH4 m−2 and 7.2 g CH4 m−2, respectively). Among years with warmer springs, growing season methane emissions varied little (±0.5 g CH4 m−2). The observed interannual differences are most likely caused by a strong soil temperature control on methane fluxes and large soil temperature differences during the spring. Thus, in a warming climate, methane emissions from waterlogged boreal peat landscapes at the southern limit of permafrost are likely to increase in response to more frequent occurrences of warm springs.


Introduction
Boreal peatlands contribute about 20% of the global wetland methane emissions (Bridgham et al 2013, Saunois et al 2016, which mainly control interannual variability in global atmospheric methane concentrations (Bousquet et al 2006). At the same time, methane emissions from wetlands are the largest source of uncertainty in the global methane budget (Saunois et al 2016). In a warming climate, boreal landscape methane emissions are expected to increase with wetland extent (e.g. Helbig et al 2017a) and warming soil temperatures (e.g. Olefeldt et al 2013, Turetsky et al 2014, while varying in their response to changing hydrological conditions (e.g. Moore et al 2011). Peatlands in the sporadic permafrost zone of northwestern Canada store large amounts of organic carbon (C) (Tarnocai et al 2009), which could be partly released to the atmosphere as methane in a warming climate (e.g. Moore et al 1998). For example, thaw-induced wetland expansion in the sporadic permafrost zone has been shown to increase landscape methane emissions exerting a warming effect on climate (Helbig et al 2017a).
In boreal peat landscapes at the southern limit of permafrost, spatial variability in growing season methane emissions is strongly controlled by the heterogeneity in water table position across the landscape (Bubier et al 1995, Sabrekov et al 2014. In contrast, temporal variability in methane emissions is strongly controlled by wetland soil temperature seasonality, when the water table is close to the peat surface. With a shallow methane oxidation zone between water table and peat surface, methane emissions to the atmosphere are primarily driven by temperature-controlled methane production in the anaerobic zone (Shannon and White 1994, Waddington et al 1996, Christensen et al 2003, Treat et al 2007, Malhotra and Roulet 2015, Helbig et al 2017a. In a changing climate, wetland soil temperature regimes will likely be modified by warming air temperatures and altered snowpack dynamics (Maurer andBowling 2014, Zhao et al 2016). Warmer wetland soils may also promote vascular plant productivity and consequently lead to higher root exudation and labile organic C supply for methanogens (Bergman et al 2000, Rustad et al 2001, Aerts et al 2006, Chanton et al 2008, Dorodnikov et al 2011. For example, Prater et al (2007) have shown that large methane emissions in recently thawed wetlands are partly driven by the high productivity of the colonizing herbaceous vascular plants. These herbaceous plants with aerenchymous tissue additionally facilitate gas exchange between roots, shoots, and leaves, and act as conduits between the rhizosphere and the atmosphere. And as methane is efficiently transported through air-filled aerenchyma, it can bypass oxidizing soil layers (Shannon et al 1996).
The first few weeks after the completion of snowmelt may be a particularly important period for soil temperature dynamics and thus for methane emissions: After the removal of the insulating snow layer (e.g. Zhang 2005), wetland soils often remain close to 0 • C despite positive air temperatures as ground ice in the top soil layers delays soil warming by consuming melt energy (e.g. Hayashi et al 2007). Once ground ice melted in spring, high thermal heat conductivity of saturated peat soils allows rapid soil warming (O'Donnell et al 2009). Warmer spring air temperatures in North America (e.g. Wang et al 2011) may therefore lead to warmer soil temperatures and increased spring methane emissions, but their magnitude and the effect on growing season and annual methane emissions remains unknown. Supported by long-term meteorological observations, we analyzed four years of methane emissions from a boreal peat landscape in the sporadic permafrost zone of northwestern Canada to examine the effect of interannual variations in air and soil temperature seasonality on methane emissions.

Study site
We examined methane fluxes from a boreal peat landscape located in the southern portion of the Scotty Creek watershed (61 • 18 ′ 29 ′′ N; 121 • 17 ′ 57 ′′ W) in the sporadic permafrost zone of the Canadian Taiga Plains (figure 1). The continental subarctic climate is characterized by a mean annual air temperature of −2.8 • C and an annual total precipitation of 388 mm  at Fort Simpson Airport, which is located ca. 60 km north of Scotty Creek (WMO ID: 71946; Environment Canada 2016). The Scotty Creek landscape comprises approximately equal proportions of coniferous forests, which are found on forested permafrost peat plateaus with oxic soil conditions, and mostly saturated permafrost-free wetlands, mainly collapsescar bogs (figure 1). The deep peat soils (>3 m) have mean organic C stocks of 167 ± 11 kg C m −2 (n = 3; Pelletier et al 2017). Thawing permafrost at Scotty Creek is replacing forested permafrost peat plateaus with collapse-scar bogs and channel fens (e.g. Quinton et al 2011, Baltzer et al 2014, Helbig et al 2016a. Landscape methane emissions at Scotty Creek originate almost exclusively from permafrost-free wetlands, while the black spruce (Picea mariana)dominated permafrost peat plateaus appear to be negligible sources or sinks of methane (Helbig et al 2017a). Collapse-scar bogs are characterized by a shallow water table that drops from the moss surface shortly after snowmelt to about 10 cm to 30 cm below the moss surface at the end of the growing season (Helbig et al 2017a). Collapse-scar bog vegetation mainly consists of Sphagnum balticum and magellanicum and ericaceous shrubs (Chamaedaphne calyculata, Andromeda polifolia, and Vaccinium oxycoccos). In the wettest bog sections, Scheuchzeria palustris is abundant and its well-developed aerenchyma and allocation of recent photosynthates to roots enhances methane emissions (Shannon et al 1996, Dorodnikov et al 2011.

Long-term meteorological observations
Long-term monthly air temperature and precipitation data was obtained from the nearest climate station at Fort Simpson Airport in order to compare 2013-2016 site measurements to long-term climate normals. From year 1951 through 2016, we derived growing season (average of monthly May to August air temperatures) and May air temperatures and total annual (October to September) and growing season precipitation. In the southern Taiga Plains, May is usually the month of snowmelt completion and the first month with positive monthly mean air temperatures (figure S1 available at stacks.iop.org/ERL/12/115009/mmedia). For the Scotty Creek watershed, the month of May is thus defined as spring in this study.

Eddy covariance and supporting measurements
Half-hourly methane fluxes from the boreal peat landscape were measured using the eddy covariance technique. Between 15 May 2013 and 16 December 2016, an eddy covariance system was deployed on the top of a triangular 15 m tower structure. The system consisted of an open-path LI-7700 CH 4 analyzer (LI-COR Biosciences, Lincoln, NE, USA), a CSAT3A sonic anemometer (Campbell Scientific Inc., Logan, UT, USA), and an open-path EC150 CO 2 /H 2 O gas analyzer (Campbell Scientific Inc.). Between 2013 and 2015, the LI-7700 was removed from the flux tower due to electrical power limitations between September and October (at the end of each measurement period) and reinstalled before the beginning of snowmelt (between March and April). We checked the performance of the LI-7700 CH 4 analyzer using a zero-and a span-gas (Ultra Zero Ambient Air and 2.02 ± 0.1 ppm; Praxair Canada Inc, Mississauga, ON, Canada) at the beginning and end of each measurement period, and once or twice during growing season. Methane fluxes were calculated using the EddyPro software (version 6.1.0, LI-COR Biosciences). A more detailed description of instrumental setup, post-processing, and quality control is given by Helbig et al (2017a). Overall, 42% of half-hourly methane fluxes passed quality control (2013: 31%; 2014: 46%; 2015: 42%; 2016: 44%). Gaps in methane fluxes were filled using a marginal distribution sampling algorithm (Reichstein et al 2005). To estimate annual emissions, we assumed a constant daily winter methane flux between November and April. We derived the winter flux estimate (0.006 ± 0.006 g CH 4 m −2 day −1 [±1 standard deviation]) as the average of non-gapfilled wintertime methane fluxes between 2013 and 2016 (i.e. soil temperature at 32 cm depth below the moss surface < 0.2 • C). We also assessed if flux footprint contributions from forested permafrost peat plateaus and collapse-scar bogs differed among study years using a 2D footprint modelling approach (Kljun et al 2004, Kljun et al 2015. The mean collapse-scar bog contribution to flux footprints was 53% with individual years ranging from 52% to 54%. Forested permafrost peat plateaus contributed on average 47% with individual years ranging from 45% to 48%. Interannual variability in methane emission was therefore not affected by flux footprint composition. More details on the flux footprint modelling approach can be found in Helbig et al (2016b).
Wetland soil temperature and water table position ca. 100 m north of the flux tower in a collapse-scar bog within the eddy covariance flux footprint were examined as environmental controls on methane emissions. Air temperature (at 2 m; HC2-S3, Rotronic AG, Bassersdorf, Switzerland), rainfall (TE25WS, Texas Instruments, Dallas, TX, USA), and snow depth (using a sonic ranger; SR50A, Campbell Scientific, Logan, UT, USA) were analyzed to characterize the coupling between soil thermal and moisture dynamics and meteorological conditions. Between 2014 and 2016, soil temperature at 32 cm below the moss surface (just beneath the water table) was measured with a type T thermocouple (Omega Engineering, Stamford, CT, USA). Water table position in the centre of the bog was recorded with a vented pressure transducer (OTT PLS, Mellingen, Switzerland). In 2013, water table position relative to the moss surface was measured in the same collapse-scar bog using a HOBO U20 Water Level Logger (Onset Computer Corporation, Bourne, MA, USA). To approximate soil temperatures in 2013 for our analysis, we used soil temperatures measured at 30 cm depth (109 Temperature Probe, Campbell Scientific Inc.) from a collapse-scar bog located ∼700 m from the flux tower. A linear regression between soil temperatures in both bogs in 2014 (r 2 = 0.98; RMSE = 0.9 • C; n = 7838) was used to create the soil temperature time series for 2013. We estimated the relative importance of monthly soil temperature and water table position in a linear model of

Vegetation productivity and greenness index
To relate vegetation productivity to methane emissions, we derived daily gross primary productivity (g C m −2 day −1 ) of the collapse-scar bog ( . The green chromatic coordinate was calculated from the red-green-blue color channel information from the digital photographs and processed (e.g. outlier filtering) using methods implemented in the phenopix R package (Filippa et al 2016).

Climate and meteorological conditions
In the southern Taiga

Water table dynamics
Interannual variability in water table position between May and June was small with mean positions ranging from −10 cm (2013) to −6 cm (2016), with −8 cm in 2014. Mean May to August deviations from the four-year average water table position ranged from −0.6 cm (2013) to +1.4 cm (2016). In July to August, mean water table deviations were slightly more pronounced ranging from −4.5 cm (2016) to +4.7 cm (2015). Despite considerable interannual variability in precipitation (see above), maximum interannual differences in water table position never exceeded 20 cm at any time of the year.

Soil temperature dynamics
During the snow-cover period from December to April, wetland soil temperatures at 32 cm depth remained close to 0 • C despite large variations in air temperature from −40 • C to +10 • C (figure S3). In the three warmer years (i.e. 2013, 2015, and 2016), rapid soil warming from 0 • C to >12 • C occurred earlier than in 2014. Consequently, mean soil temperatures between May 15 and June 30 were about 6 • C warmer in 2013, 2015, 2016 (9.8 • C ± 0.5 • C [±1 standard deviation]) than in 2014 (3.8 • C). The earliest soil warming occurred in 2016, but was followed by up to 2.5 • C cooler soil temperatures later in May compared to 2013 and 2015 (figure 3). The interannual soil temperature differences were less pronounced during the four study years in July (13.2 • C-14.5 • C) and August (13.5 • C-14.0 • C) and therefore May and June appear to have the largest interannual soil temperature variability.

Methane emissions from the boreal peat landscape
Summertime methane emissions characteristic for Scotty Creek (> 0.05 g CH 4 m −2 day −1 ) were reached several weeks earlier (by the end of June) in years with warmer May air and soil temperatures than in 2014 ( figure 3(a)). During the growing season, daily methane fluxes continued to increase before peaking in July and August at about 0.08 g CH 4 m −2 day −1 in the warmer years and at only 0.06 g CH 4 m −2 day −1 in 2014. At the start of the snow-cover period, methane fluxes dropped again close to pre-snowmelt methane emissions in each year. Total methane emissions between May 15 and August 31 were not significantly different among years with warm May temperatures (2013: 6.2 [6.0-6.6] g CH 4 m −2 with 95% confidence interval [CI]; 2015: 6.5 [6.4-6.9] g CH 4 m −2 ; and 2016: 6.4 [6.3-6.8] g CH 4 m −2 ). However, for the same period in 2014, total methane emissions were significantly smaller with 4.5 [4.3-4.7] g CH 4 m −2 (i.e. about 25% smaller). The same pattern continued through to October 31, with total May to October methane emissions in the warm years exceeding the 2014 emissions by about 30%. About half of the observed interannual differences in May to October methane emissions accumulated between May and June, while the remaining differences accumulated between July and October (total May to October differences: +2.4 g CH 4 m −2 in 2015 and +1.9 g CH 4 m −2 in 2016). Similarly, annual methane emissions in 2014 were about 20%-25% smaller than in 2015 and 2016 (2014: 7.2 g CH 4 m −2 , 2015: 9.6 g CH 4 m −2 , 2016: 9.1 g CH 4 m −2 ).

Soil temperature controls on methane emissions
Monthly methane emissions were significantly correlated with monthly mean soil temperature at 32 cm and explained 90% of the variance in monthly methane emissions (2013-2016; r 2 = 0.90; p < 0.001; n = 28, figure 4). In three out of four years ( [2016]. No significant correlation with soil temperature was observed in 2013 (p = 0.21; n = 4), most likely due to a smaller sample size and less variation in temperature than in the other three years. In contrast, monthly wetland water table position was not correlated with monthly methane emissions (p = 0.14; n = 28; figure S4). A multiple linear regression model with soil temperature (p < 0.001) and water table position (p = 0.03) explained only 1% more of the variance in monthly methane emissions (adjusted R 2 : 0.91) than soil temperature alone. The relative importance of water table position in the model was only 2.1% (95% CI: 0.5%-15.8%) compared to 97.9% (95% CI: 84.1%-99.5%) for soil temperature.

Vegetation productivity
Similar to the reduction in methane emissions in 2014, gross primary productivity of the wetland was smaller in 2014 than in the warmer years ( figure 5). During the warmer years, a more rapid increase in gross primary productivity in the early growing season resulted in peak gross primary productivity rates exceeding those in 2014 by about 1 g C m −2 day −1 . Only in early September, did gross primary productivity rates become similar again. There were also interannual differences in green chromatic coordinate (GCC), with a consistently lower GCC between May and August 2014 compared to the warmer years 2013 and 2015.

Methane flux response to soil temperature and hydrology
We show that warmer May air and soil temperatures significantly increase annual landscape methane emissions by 20%-25% in a boreal peat landscape with sporadic permafrost. Increases in annual methane emissions are mainly caused by the large interannual variability in spring and early summer soil temperatures. Despite similar end-of-winter soil temperature profiles (figures S6 and S7), the interannual differences in wetland soil temperatures (at 32 cm below the moss surface) exceeded 10 • C in May, with cool May air temperatures in 2014 contributing to delayed soil thaw and warming. However, interannual differences in ground ice content in wetlands may have additionally modified soil warming rates in spring (e.g. Hayashi et al 2007). Soil temperature exerts a strong control on methane emissions in waterlogged boreal peat landscapes. For example, Helbig et al (2017a) showed that large differences in summertime methane emissions between forested permafrost peat plateaus and wetlands are reflected in up to 10 • C cooler soils (at 32 cm below the moss surface) on the plateaus compared to the wetland.
At Scotty Creek, the wetland water table position remained always within about 30 cm from the moss surface (figure 2) despite large variability in total annual precipitation (table 1). Thus, the water table likely did not drop below the rooting depth of herbaceous vascular plants, which usually extends to about 30 cm below the moss surface in peatlands (e.g. Pelletier et al 2017, Strack et al 2006 and can extend to more than 60 cm in collapse-scar bogs (Finger et al 2016). When water table positions are close to the moss surface early in the growing season, any delay in soil warming is likely to result in lower methane emissions. Thus, warmer May air temperatures and an earlier soil warming lead to increased annual methane emissions.

Methane flux response to vegetation productivity
Large July and August methane emissions were only observed in years with warmer May temperatures despite similar average soil temperatures in July and August in all years. The observed differences in late growing season methane emissions are reflected in interannual differences in wetland vegetation productivity and phenology (figure 5). Earlier onset of soil warming likely enables faster vegetation growth (Natali et al 2012, Peichl et al 2014, Zhao et al 2016 and leads to a 'greener' (i.e. more abundant) vegetation, more plant biomass, and higher peak gross primary productivity rates (figures 5 and S5). Increased availability of labile organic C and a higher biomass of aerenchymous plants with larger wetland vegetation productivity likely explains larger late growing season methane emissions in the years with warmer spring seasons. Small mean growing season methane uptake rates on forested permafrost peat plateaus with lower soil moisture content typically range between −0.2 and −1 mg CH 4 m −2 day −1 (Bubier et al 2005, Savage et al 1997 and are therefore unlikely to explain interannual variability in methane emissions. Plant productivity rapidly increased during warm springs (figure 5) and may have rapidly minimized substrate limitation of methane production and boosted methanogen activity. Metabolic activity of methanogens has been shown to increase exponentially with temperature (e.g. Yvon-Durocher et al 2014). In many wetlands, methane emissions show a similar exponential increase with soil temperature (e.g . Hartley et al 2015, Helbig et al 2017a). Here, we did not observe exponentially increasing, but linearly increasing methane emissions with soil temperatures on a monthly time-scale (figure 4). However, if summer soil temperatures become warmer in a changing climate, water table position remains close to the peat surface, and labile organic C supply is maintained, monthly methane emissions may deviate from the observed linear relationship. The methane emission response to warming spring air and soil temperatures may therefore change with continued warming.

Methane flux response to soil warming in warm spring seasons
Methane production rates at low soil temperatures (<5 • C) are strongly temperature-limited (Treat et al 2014). However, differences in annual and growing season methane emissions between years with warmer springs were small, despite differences in the timing of spring soil warming. The earliest soil warming among the four study years in 2016 was followed by soil cooling in late May (figure 3). A return to stronger temperature-limitation of methane production could explain the attenuated response of methane emissions to the early onset of soil warming. Also, vegetation productivity may not have immediately responded to the rapid and short-term warming (∼four days) in May 2016 and the following return to lower soil temperatures may have further limited vegetation productivity. As a result, low production of root exudates may have limited the supply of labile organic C for methane production. In contrast, later onset of spring soil warming in 2015 was accompanied by increasing vegetation productivity (figure 5).
In peatlands, water table positions above the moss surface can reduce methane emissions (Pelletier et al 2007) and tend to limit plant productivity due to prevailing anoxic soil conditions (e.g. Sulman et al 2012). After snowmelt in 2016, water table position was above the moss surface and higher than in the other study years (figure 2). The high water table position may have contributed to the attenuated response of methane emissions to the early soil warming. Thus, our results suggest that warmer spring air temperatures and earlier soil thaw lead to larger growing season and annual methane emissions. However, spring methane emissions may be less sensitive to soil warming, if followed by intermittent cooler periods or if accompanied by flooding.
Here, we show that warmer spring temperatures can increase annual methane emissions by 20% to 25%. Thus, in a changing climate, an increase in long-term methane emissions from boreal peat landscapes in the sporadic permafrost zone can be expected with warmer spring conditions. Additionally, methane emissions in the southern Taiga Plains and similar regions in North America are projected to increase with thaw-induced wetland expansion (Lara et al 2016, Turetsky et al 2008, Helbig et al 2017a). In a warming climate with warmer spring temperatures (Wang et al 2011), thawing boreal peat landscapes along the southern limit of permafrost may therefore become increasingly important for the global methane budget.