Introduction
High-latitude ecosystems are being subjected to rapid changes in climate
(IPCC, 2013) and increases in fire frequency and intensity (Kasischke et al.,
2010), notably in northwestern North America and Alaska (Hinzman et al.,
2005; Ju and Masek, 2016). This will have a wide variety of ecosystem effects
(Alexander and Mack, 2016): in particular, rising temperatures and increasing
fire will likely result in changes in soil temperature and permafrost
degradation (Pastick et al., 2015; Zhang et al., 2015; Genet et al., 2013;
Helbig et al., 2016), with subsequent hydrology changes that will influence
soil greenhouse gas (GHG) fluxes to the atmosphere (Schädel et al.,
2014). Such fluxes are a large component of the global C cycle and could
result in a significant and positive climate feedback (Treat et al., 2015;
Koven et al., 2011; Schaefer et al., 2014).
The magnitude, timing, and form – in particular as methane (CH4) or
carbon dioxide (CO2) – of any such feedback remain highly uncertain
(Schuur et al., 2015). While northern soils hold enormous quantities of
potentially mineralizable soil organic carbon (SOC) (Hugelius et al., 2014),
vegetation and succession dynamics (for example, thermal insulation by
mosses) promote permafrost resilience to even large temperature changes
(Jorgenson et al., 2010; Turetsky et al., 2012). Vegetation type also
influences SOC quality and quantity, with microbial communities (Högberg
et al., 2007), soil respiration (Raich and Tufekcioglu, 2000), and SOC all
linked to aboveground factors such as woody vs. nonwoody stems, deciduous
vs. evergreen canopies, and the presence of nitrogen-fixing plants. A
number of factors may however disrupt these feedbacks between
vegetation type, ground cover, permafrost, and SOC stability. Climate
changes, in particular regional warming and drying, cause vegetation stress
(Ju and Masek, 2016; Barber et al., 2000) and increased mortality.
Conversely, increasing plant productivity in some regions can stimulate the
decomposition of older SOC (Hartley et al., 2012). Climate also drives fire
regime changes, and ecosystem disruption is particularly likely after
intense fires (Johnstone et al., 2010; Genet et al., 2013). Even without
disturbance, the stability of SOC is highly uncertain since it depends on soil
temperature and moisture, the ages of and ratio between the carbon (C) and
nitrogen (N) pools (Weiss et al., 2015; Karhu et al., 2014), and its
protection (whether by organomineral sorption, chemical lability, or
physical location) from capable microorganisms, enzymes, and resources
(Bailey et al., 2012; Schmidt et al., 2011).
Temperature and moisture typically have strong and often interactive
influences on soil GHG emissions. Laboratory incubations, field
observations, and metanalyses have documented changing GHG fluxes with rising temperature (Olefeldt et al., 2013; Davidson and
Janssens, 2006; Hashimoto et al., 2015; Treat et al., 2015). GHG responses
to wetting and thawing dynamics exhibit substantial variability between
studies, probably due to differences in soil type, antecedent conditions,
phase changes, experimental protocols, etc. (Kim et al., 2012). The
common anaerobic conditions following permafrost thaw are expected to lower
CO2 emissions but increase those of CH4 (Treat et al., 2015, 2014),
but emissions from aerobic soils will likely dominate the
permafrost C feedback (Schädel et al., 2016). Decadal warming and drying
trends in Alaska (Bieniek et al., 2014) thus seem likely to increase GHG
emissions from soils, and laboratory incubation experiments are critical in
understanding these dynamics (Elberling et al., 2013).
Most previous studies have focused on surface soils or permafrost soils,
neglecting deep active-layer soils that were identified as subject to strong
effects from a 2-decade warming experiment in the Alaskan Arctic (Sistla
et al., 2013). Such deeper soils have particular characteristics
distinguishing them from both shallow active-layer soils and underlying
permafrost: they are most affected by interannual variability in thaw depth,
potentially flipping the C source–sink status of entire ecosystems (Goulden
et al., 1998; Harden et al., 2012); they are subject to distinctive
freeze–thaw, cryoturbation, and microbial dynamics, which are likely to
change their sensitivity to climate change and feedback potential (Schuur et
al., 2008); and they are known to pose particular problems for accurate
modeling of high-latitude carbon dynamics (Nicolsky et al., 2007). These
soils are likely to be a highly important contributor to future climate
feedbacks, with modeling studies suggesting that one-third of 21st
century climate-induced carbon loss may originate from seasonally frozen
soil carbon (Koven et al., 2015).
The goal of this study was to examine how temperature and moisture control
GHG (CO2 and CH4) emissions from soils sampled from the bottom of
the annual active layer – i.e., directly above permafrost – in an Alaskan
boreal forest. We also aimed to characterize the chemical and structural
properties of these soils following a 100-day incubation at different
temperatures, subjecting some cores to drying treatments. We hypothesized
that (i) CO2 would be the dominant pathway for C loss in these largely
aerobic soils; (ii) soils maintained at field moisture and high
(20 ∘C) temperature would lose more C-CO2 than cores
incubated at 4 ∘C, due to increased aerobic and anaerobic
microbial activity; and (iii) core CH4 fluxes would be small and
sensitive only to temperature, as no anaerobic conditions were imposed on
the cores.
Methods
Field sampling
The field component of this research took place in Caribou-Poker Creeks
Research Watershed (CPCRW), part of the Bonanza Creek LTER
(http://www.lter.uaf.edu/research/study-sites-cpcrw). CPCRW is located in
the Yukon-Tanana Uplands northeast of Fairbanks, AK, a part of the boreal
forest that has seen strong increases in air temperature and forest browning
(Ju and Masek, 2016) over several decades. Annual average air temperature is
-2.5 ∘C, and annual average precipitation 400 mm (Petrone et al.,
2006). The watershed's lowlands and north-facing slopes are dominated by
black spruce (Picea mariana (Mill.) BSP), feathermoss
(Pleurozium schreberi and others), and Sphagnum spp.; the
drier southern slopes tend to be deciduous with a mixture of trembling aspen
(Populus tremuloides Michx.), Alaska paper birch (Betula neoalaskana), and patches of alder (Alnus crispa).
We sampled soils from a southeastern slope (65.1620∘ N,
147.4874∘ W) at CPCRW, in a 60 m transition zone between lowland
Picea mariana and upland Betula neoalaskana, with
significant white spruce (Picea glauca) presence as well. Stand
density in this transition zone was 4060 ± 2310 trees ha-1, with
a basal area of 27.9 ± 7.0 m2 ha-1. The forest was at least
90 years old (see Morishita et al., 2014) according to tree cores taken at
the base of several of the largest white spruce. The soil is characterized as
a poorly drained silt loam, and on average had ∼ 20 cm of organic
material over the mineral soil.
On 3–5 August 2015, 39 soil cores, each 30 cm high by 7.5 cm wide, were taken using a
soil recovery augur (AMS Inc., American Falls, ID). We
sampled from the bottom (within 0–2 cm of permafrost) of the active layer,
which averaged 80 cm depth. Sample points were randomly located in the
transition zone described above and separated by 2–5 m. Cores were kept
cool in the field before being packed in dry ice and shipped to Richland, WA,
within 48–72 h of collection.
Laboratory incubation
In the lab, the soil cores were stored at 4 ∘C for several days
until they were weighed and prepared for incubation. At that point
(11–12 August 2015), three fragmented or otherwise damaged cores were
discarded, and the remaining cores were randomly assigned to one of six
groups (N = 6 in each group). These included two incubation temperatures of 4
and 20 ∘C, following the protocol of a number of previous boreal
incubation studies (Treat et al., 2015). Within each temperature there were
two moisture treatments: one in which soil moisture was maintained at field
conditions (∼ 28 % moisture by volume), and a drought treatment in
which no water was added and cores were allowed to dry down to
∼ 5 % moisture by volume. The fifth group was a 20 ∘C
“controlled drought” one, in which water was added so that the cores'
moisture status would closely match those of the 4 ∘C “drought”
cores, which we anticipated would dry more slowly than their 20 ∘C
counterparts. The final six-core group was used for destructive, pre-incubation
measurements including moisture content, pH, soil carbon and nitrogen, and bulk
density. Subsamples were collected and stored at -20 ∘C for
dissolved organic carbon measurements or air dried for soil C and N (see
below).
On 18 August 2015 cores were placed into one of two growth chambers (Conviron
Control Systems BDW80, Winnipeg, Canada) maintained at 4 and 20 ∘C
temperatures and 70 % relative humidity and allowed to equilibrate for
2 weeks. Starting on 31 August 2015 we measured the cores' mass and GHG
emissions four times in the first week, then twice per week for the first
month, and then once per week for the rest of the 100-day incubation.
Throughout the incubation, cores had a 200 µm mesh screen fit to
the base and were mounted on porous ceramic plates (Soil Moisture Equipment
Corp., Santa Barbara, CA, USA) so that when the plates were placed in
contact with water, water would move up into the cores via capillary action.
The drought cores were mounted on dry plates but were not allowed to drop
below 5 % water content. After each flux measurement, cores received
additional wetting from the top to maintain their desired water status.
For each measurement, a six-core treatment group was connected to a Picarro
A0311 multiplexer that was in turn connected to a Picarro G2301 GHG Analyzer
(Picarro Inc., Santa Clara, CA, USA). Dry CH4 and CO2
concentrations were monitored for 2 min, and this was repeated two to three times
before moving on to a new treatment group. Cores were weighed immediately
after gas measurements. Ambient air was measured between treatment groups
and before starting measurements in a chamber as a check on ambient CO2
conditions and instrument stability.
The incubation experiment concluded on 9 December 2015, following the final
CO2 and CH4 readings. Each soil core was maintained at the
treatment-dependent temperature and moisture content (by mass) until removed
for destructive sampling on 14–18 December 2015. Subsamples were collected
and composited throughout each soil core for dissolved organic carbon
analysis (110 ± 24 g dry mass equivalent) and dry-mass calculations
(∼ 28 g each). The remaining core material was air-dried and separated
into particles (> 2 mm diameter) and soil (≤ 2 mm) using a US
Standard Test Sieve No. 10 (Fisherbrand, Pittsburg, PA, USA). The dry mass
and volume of soil were used in calculations of gravimetric and volumetric
soil moisture content, respectively (Gardner, 1986). Soil volume was
calculated as the total core volume minus the volume of particles > 2 mm
diameter, with the latter determined by water displacement. Air-dried soil
and subsamples stored at -20 ∘C were sent to the Agricultural and
Environmental Services Laboratory at the University of Georgia Extension in
February 2016 for total C, N, and dissolved organic carbon (DOC) testing. Samples
were combusted in an oxygen atmosphere at 1350 ∘C and measured for
gaseous C and N using an Elementar Vario Max CNS (Langenselbold, Germany).
DOC was measured using a Shimadzu 5000 TOC Analyzer (Columbia, Maryland,
USA).
Data and statistical analysis
For each measurement of each sample throughout the 100-day incubation (i.e.,
each gas, core, and date and time), we used the rise in gas concentrations to
calculate a flux rate in ppm s-1 (CO2) or ppb s-1
(CH4), a linear rate of change (δc/δt) based on the
concentration rise from a minimum (up to 10 s after measurement began) to a
maximum (at 10–45 s). Each core's respiration flux (F) was then
calculated as F=δcδtVMPaRT, where V is the core-specific system volume,
M the core dry mass as determined at the end of the incubation,
Pa the atmospheric pressure (101 kPa, the incubation chambers were
∼ 120 m a.s.l.), R the universal gas constant (8.3×10-3 m3 kPa mol-1 K-1), and T the chamber air
temperature (K) at time of measurement. The final respiration rate was
expressed on a soil C basis (µg or ng C g C-1 day-1).
Anomalous data were excluded based on their gas fluxes being more than 5 (for
CO2) or 10 (for CH4) mean absolute deviations (Davies and Gather,
1993) from the treatment mean within a 10-day period, for a given treatment
and temperature. We excluded 172 of 2686 (6.4 %) measurements for this
reason. If the coefficient of variability (CV) of fluxes from any core on a
single day exceeded 140 %, a value chosen based on the distribution of
CVs across all cores, the entire core was excluded for that day (90 data
points, 3.4 %). Other data (4.8 %) were removed because of known
instrument problems, e.g., the analyzer was left running after leaving a
chamber. The final number of valid flux samples from the 100-day incubation
was 2198.
Summary of dissolved organic carbon (DOC), percent C, percent N,
bulk density (BD), and CO2 and CH4 fluxes by treatment. The field
moisture and drought columns summarize (mean ± s.d.) 12 cores,
combining two groups of N = 6 at each incubation temperature, while the
controlled drought and pre-incubation columns are N = 6.
Variable
Field moisture
Controlled drought
Drought
Preincubation
DOC (mg kg-1)
173.62 ± 46.67
165.68 ± 66.46
154.60 ± 57.15
125.43 ± 49.07
C (%)
1.67 ± 1.60
0.87 ± 0.50
0.76 ± 0.60
1.44 ± 1.32
N (%)
0.08 ± 0.08
0.04 ± 0.03
0.03 ± 0.03
0.07 ± 0.06
BD (g cm-3)
0.89 ± 0.18
1.06 ± 0.17
1.08 ± 0.14
1.13 ± 0.29
CO2 (µg C g C-1 day-1)
456.40 ± 543.91
159.77 ± 116.41
97.03 ± 96.38
–
CH4 (ng C g C-1 day-1)
0.10 ± 0.00
0.10 ± 0.00
0.10 ± 0.00
–
The effects of temperature, gravimetric water content, percent C, percent N,
and DOC concentration on instantaneous gas fluxes were evaluated using a
linear mixed-effect model fit by the R function lme in the R
“nlme” package, version 3.1.128. Because the dependent variable (CO2
or CH4 flux) was non-normally distributed, it was transformed using a
natural-logarithm (+0.1 µg C g C-1 day-1 to ensure
all positive fluxes, following Treat et al., 2015) transformation. Soil core
was treated as a random effect in the model. We then performed stepwise model
selection by Akaike's information criterion (AIC) using the stepAIC
function in the R “MASS” package, version 7.3.45. A linear mixed-effect
model was also used to evaluate the effect of treatment on core water
content.
Cumulative respiration for each core and gas was calculated by linearly
interpolating flux rates between measurement dates and summing respired C
over the entire incubation. The effect of temperature and treatment (drought,
controlled drought, or field moisture conditions) on cumulative gas fluxes
was evaluated with a post hoc Tukey's Honest Significant Difference test.
Temperature sensitivity (Q10) was calculated for each gas and treatment
as F2F110T2-T1, where
F1 and F2 are the cumulative gas fluxes (mg C g C-1) at
temperatures T1 and T2 (∘C), respectively.
All data analysis and statistics were performed using R version 3.3.1
(21 June 2016) (R Core Team, 2016). This experiment was run as an
“open experiment” (Bond-Lamberty et al., 2016b) with all analysis code,
data (from raw instrument data to final summaries), diagnostics, etc.,
available at https://github.com/bpbond/cpcrw_incubation. The summarized
flux data backing the main results have been archived under the Digital
Object Identifier 10.6084/m9.figshare.4240436.v1.
Results
The 30 experimental cores had a bulk density of 1.00 ± 0.18
(mean ± sd) g cm-3. Large (> 2 mm) particles, primarily
schist, comprised 41 % ± 11 % of the cores' total mass. Soil
(≤ 2 mm) dry mass was 886 ± 154 g. Sample DOC was
157.93 ± 55.74 mg kg-1. Carbon content was
1.20 % ± 1.19 %, while N content was
0.06 % ± 0.06 %. Mean C : N was 20.7. Neither temperature nor
moisture treatment exerted any significant effect (P > 0.1 for all) on
these highly variable properties (Table 1).
Gravimetric water content was 0.31 ± 0.12 (min 0.19, max 0.77) at the
beginning of the incubation (Fig. 1). “Field moisture” cores were on
average unchanged (0.33 ± 0.13) at the end of the incubation, but both
the drought treatments, which did not differ from each other in their effect
on gravimetric water content (P = 0.880), had declined to
0.06 ± 0.04. Volumetric water content values ranged from
0.29 ± 0.05 (min 0.23, max 0.43) at the beginning of the experiment to
0.15 ± 0.11 (min 0.03, max 0.38) at the end across all cores. Water-filled pore space, assuming a particle density of 2.65 g cm-3, was
22–65 % over all cores, moisture treatments, and temperatures.
Core water content across the course of the incubation
experiment by temperature (left panel 4 ∘C, right panel 20 ∘C) and treatment.
Carbon dioxide fluxes during the incubation ranged from
1.1 µg C g C-1 day-1 to a maximum of
5245.1 µg C g C-1 day-1, with a mean of
248.9 µg C g C-1 day-1 over the 100 days. CH4
rates ranged from 0.00 ng C g C-1 day-1 to a maximum of
1.31 ng C g C-1 day-1, with a mean of
0.06 ng C g C-1 day-1. These means conceal considerable
variability over the course of the incubation (Table 1, Figs. 2 and 3).
Mass-normalized CO2 fluxes over the 100-day
incubation by temperature (4 and 20 ∘C, rows) and treatment
(field moisture, drought, and controlled drought; columns). Error bars show
core-to-core standard deviation. The controlled drought treatment, for
20 ∘C only, was meant to dry cores at roughly the same rate as the
drought cores at 4 ∘C.
In the linear mixed-effect model (AIC = 2992.6), instantaneous CO2
flux was positively correlated with incubation chamber temperature, core
gravimetric water content, and percent soil N (all P < 0.05, and the
latter two P < 0.001; Table 2). Temperature sensitivity decreased
significantly (P < 0.001) over the course of the incubation, while
moisture sensitivity was unaffected by time. Percent C and percent N were
highly correlated (r = 0.99) for these cores. Because percent N was a
slightly stronger predictor, it was retained in the model while percent C was
excluded; see Colman and Schimel (2014). The interaction between water
content and percent N was also highly significant (P < 0.001), although
cores with N > 0.2 % exhibited little relationship between water
content and CO2 flux (data not shown). Instantaneous CH4 fluxes
were positively correlated with percent N, while water content exhibited
significant interactions with percent N and DOC as a predictor (Table A1).
This model had little predictive power (AIC = -10 879.2), however, and
neither temperature nor water content was a significant first-order predictor
of CH4 fluxes.
Mass-normalized CH4 fluxes over the 100-day
incubation by temperature (4 and 20 ∘C, rows) and treatment
(field moisture, drought, and controlled drought; columns). Error bars show
core-to-core standard deviation. The controlled drought treatment, for
20 ∘C only, was meant to dry cores at roughly the same rate as the
drought cores at 4 ∘C.
The cumulative production of C from CO2 (Fig. 4) was over 6 orders of
magnitude higher than that from CH4, with CO2 : CH4 C ratios
ranging from 1.4 million in the 4 ∘C field moisture treatment
to 6.2 million in the 20 ∘C field moisture treatment. Cumulative
evolved CO2 was highly affected by temperature (P=0.003), and
field moisture cores emitted significantly more CO2 than the other
two moisture treatments at both temperatures (P < 0.001 for both, with
no significant interactive effect). There was no difference between fluxes
from the 20 ∘C drought and controlled drought treatments
(P = 0.377). Drought cores' cumulative production was 73 %
(4 ∘C) and 52 % (20 ∘C) lower than the cores kept at
field moisture. Neither temperature (P = 0.200) nor moisture treatment
(mean P = 0.975) was a significant factor in predicting cumulative
CH4 fluxes.
Linear mixed-effect model parameters, testing effects of
temperature (∘C), gravimetric water content (unitless), soil C
(%), soil N (%), and dissolved organic carbon (mg kg-1) on
individual core CO2 fluxes (+0.1 µg C g C-1 day-1);
a colon (“:”) indicates an interaction. Dependent variable has units of
log(µg C g C-1 day-1). Columns include parameter
value,
standard error (SE), degrees of freedom (DF), T statistic, and P value.
Value
SE
DF
T
P
(Intercept)
1.713
0.354
1153
4.839
< 0.001
Temperature
0.046
0.020
26
2.336
0.027
WC_gravimetric
3.496
1.052
1153
3.322
0.001
N_percent
37.976
6.810
26
5.576
< 0.001
Temperature : WC_gravimetric
0.116
0.061
1153
1.905
0.057
Temperature : N_percent
-0.507
0.300
26
-1.690
0.103
WC_gravimetric : N_percent
-37.347
8.425
1153
-4.433
< 0.001
The cumulative flux numbers above result in CO2 temperature sensitivity
(Q10) values of 1.3 and 1.9 for the field moisture and drought
treatments, respectively; the corresponding Q10 values based on
cumulative CH4 were 1.2 and 1.3. Computing Q10 values based on
fluxes normalized by water-filled pore space changed these values only
slightly: to 1.2 and 1.7 for CO2, for the field moisture and drought
treatments, respectively, and 1.1 and 1.2 for CH4.
Discussion
Rises in boreal air temperatures, and unpredictable precipitation changes,
will change fire disturbance regimes, warm and dry many soils, increase
vegetation stress, degrade permafrost, and deepen the active layer (Schuur
et al., 2015), all with uncertain consequences for soil dynamics and GHG
fluxes. In this laboratory experiment we found that CO2 fluxes, but not
CH4 fluxes, from these oxic active-layer mineral soils were sensitive
to temperature and, in particular, moisture.
A number of studies have measured microbial respiration and GHG fluxes very
close to our study site. Morishita et al. (2014) quantified GHG fluxes at
CPCRW and nearby forests and found CO2 production to be correlated with
both temperature and moisture in upland cryosols, consistent with our
results. Waldrop et al. (2010) incubated active-layer and permafrost soils
from Picea mariana sites near Fairbanks, AK, observing aerobic
Q10 values of 9.0 (active layer) and 2.3 (permafrost) from -5 to
5 ∘C, and flux rates of
0.001–0.10 µmol CH4 day-1 g-1
(∼ 0.001–0.133 ng C g C-1 day-1) and
∼ 1–5 µg C-CO2 h-1 g-1
(∼ 2000–10 000 µg C g C-1 day-1),
considerably higher than the CO2 rates observed here. During the first
100 days of an incubation of Fairbanks-area 0–10 cm mineral soils, Neff and
Hooper (2002) observed fluxes of
∼ 55–409 µg C-CO2 g C-1 day-1, in line
with the results here, while Wickland and Neff (2008) reported that
temperature and moisture exhibited interactive effects, of similar magnitude,
on decomposition in P. mariana soils.
A number of synthesis studies have documented dynamics and C feedback
potential of Arctic and boreal soils more generally; comparing to these
results is useful because although the response of soil biota to stresses
such as drought tends to differ between soil types, organisms, and
vegetation, it is often broadly similar across biomes and climatic conditions
(Manzoni et al., 2012). Using two metanalyses of aerobic and anaerobic
permafrost soil incubations, Schädel et al. (2016) showed that C release
was highly sensitive to temperature and that soils released far more
(220–520 %) C under aerobic conditions. Our incubation was fully
aerobic, but its results are consistent with the conclusion that respiration
in the form of CO2 is likely to dominate the high-latitude C feedback,
and that aerobic soils, and the conditions under which currently waterlogged
soils may drain, deserve particular attention. In terms of absolute flux
rates, Treat et al. (2015) reported mean CO2 rates of 47 (all mineral
soils) and 101 (for 20–100 cm
soils) µg C-CO2 g C-1 day-1 from a pan-Arctic
synthesis of anaerobic soil incubations, which is somewhat lower than our aerobic
incubation results. Treat et al. (2014) also found CO2 and CH4
emissions to be strongly correlated with temperature and moisture based on an
incubation of Alaskan peats. Whether climate change makes northern regions
wetter or drier is thus a critical factor affecting the quantity and form of
C release.
The drought treatment imposed in this experiment reduced soil C fluxes by
52–73 %. The importance of this result depends, in part, on the spatial
extent and intensity of precipitation changes across the boreal and Arctic
during this century. There is a detectable anthropogenic influence in high-latitude
precipitation changes (Wan et al., 2015), but these changes are inconsistent:
drier and warmer conditions in boreal Eurasia (Buermann et al., 2014), for
example, but growing season length increases in interior Alaska with no
increase in precipitation (Wendler and Shulski, 2009). This spatial
variability will interact with permafrost thaw dynamics to produce a complex
patchwork of soil moisture changes (Zhang et al., 2012; Watts et al., 2012).
The high uncertainty in this area makes it all the more important to
understand the interactive effects of soil moisture and temperature on
decomposition and GHG emissions (Sierra et al., 2015).
We observed very low but positive CH4 production from these upland
mineral soils. This is in contrast to many field studies that have observed
CH4 uptake (oxidation) in dry boreal sites (Matson et al., 2009;
Schaufler et al., 2010). Anoxic microsites in soil can however provide enough
CH4 production to balance low-level consumption in otherwise aerobic
soils (Kammann et al., 2009). In addition, our results are broadly consistent
with data from 65 studies summarized by Olefeldt et al. (2013), who found
that CH4 emissions were more sensitive to soil temperature in wetter
ecosystems; it would have been a surprise if the little methanogenic activity
in our upland, well-drained soils was temperature-sensitive at all. Methane
was also a far smaller C flux than CO2 from these soils, in particular
at higher temperatures (as CO2 was responsive to temperature, but
CH4 was not). This is true more generally: for example, Treat et
al. (2015)
found a median CO2 : CH4 production ratio of 387 for anaerobic
incubations of boreal soils. This is naturally far lower than our observed
aerobic (and thus high-CO2) ratios, but nonetheless consistent with
them. Thus, we see little opportunity for CH4 to be a significant
contributor to the upland soil C fluxes and climate feedback risk, even
accounting for the 25 times stronger radiative forcing of this gas over a
100-year time horizon (Lee et al., 2012).
Temperature vs. moisture sensitivity for cumulative emissions
The cumulative GHG fluxes (Fig. 4) integrate the entire 100-day incubation,
eliminating the day-to-day variability of instantaneous fluxes and are thus
more generalizable. Our results suggest that moisture limitation could exert
a large effect on CO2 production for deep active-layer soils:
drought cores' cumulative production was 73 % (4 ∘C) and
52 % (20 ∘C) lower than the cores kept at field moisture. This
effect was highly significant, and suggests that moisture limitations could
exert a significant constraint on deep active-layer soils as they slowly
warm. Such moisture constraints are thought to be already exerting effects on
vegetation and soil fluxes at large scales (Ju and Masek, 2016; Bond-Lamberty
et al., 2012), but our understanding of the interactive effects involved is
poor.
Cumulative mass-normalized C fluxes (mg g C-1) over the
incubation by gas (CO2 and CH4, top and bottom panels,
respectively), treatment (columns), and temperature (x axis, ∘C).
Letters within a panel indicate significant differences based on Tukey's
HSD.
The Q10 values observed in this experiment were low (all less than 2.0,
even when controlling for changes in soil moisture). Temperature
sensitivities of ∼ 2 are more typical (Dutta et al., 2006; Schädel
et al., 2016), although the temperature sensitivity of C release can change
over time of incubation (Dutta et al., 2006) and vary between soil fractions
cycling over different time horizons (Karhu et al., 2010; Schädel et al.,
2014). Observed surface CO2 fluxes at this CPCRW site exhibited a
Q10 of 5.1 ± 1.4 over a temperature range of 3.5–15 ∘C
(C. Anderson, personal communication, 2016); however, these surface fluxes were measured over multiple months and
include root respiration preventing any direct comparison. While
increased temperature does not always drive C mineralization rates in forest
mineral soils (Giardina and Ryan, 2000), it is linked with increases in soil
moisture content and can lead to changes in microbial community structure and GHG
fluxes (Xue et al., 2016).
Interestingly, Q10 values were lower in the drought treatment cores, a
mathematical consequence of the fact that drought restricted CO2
respiration more at 4 than at 20 ∘C. There is evidence that climate
warming changes the microbial decay dynamics of soil organic C compounds
generally considered to be stable (Frey et al., 2013; Bond-Lamberty et al.,
2016a). Conditions such as drought can change the amount and quality of DOC
available to microbes (1999), but we observed no DOC changes between
treatments here. Deep active-layer soils store large quantities of soil C
(Mueller et al., 2015) but are not subject to abundant inputs of fresh C from
vegetation. Therefore, the starting quality of the native soil C in active-layer
soils is older, more microbially processed, and dominated by more stable
“heavy” organic C (Karlsson et al., 2011). Thus, it may not be surprising
that these more stable C compounds would be metabolized by processes that
have been reported to be less temperature-sensitive.
Soil nitrogen
Somewhat unexpectedly, percent soil N was very significantly and positively
correlated with both CO2 and CH4 fluxes (Tables 2 and 3). Nitrogen
interacts with microbial respiration via a number of complex, interactive,
and still unclear mechanisms (Luo and Zhou, 2006), including reductions in
belowground plant allocation, shifts in energy source or population of the
saprotrophic community (Saiya-Cork et al., 2002) that leave it less capable
of decomposing recalcitrant compounds, and perhaps abiotic stabilization
mechanisms (Janssens et al., 2010). Metanalyses have generally shown
negative to neutral effects of N deposition on microbial biomass (Treseder,
2008) and respiration (Ramirez et al., 2012) and total soil respiration
across ecosystems and biomes (Janssens et al., 2010; Zhou et al., 2014).
These effect are likely due to several mechanisms involving soil
pH, ligninase enzymes, and phenol oxidase activity (Luo and Zhou, 2006), and
incubation results examining N effects can be highly variable (Lavoie et al.,
2011; Sistla et al., 2012). Some studies have however observed positive
correlations between ambient soil N and microbial respiration. For example,
Weiss et al. (2015) found CO2 production from Siberian Yedoma permafrost
samples to be correlated with both percent C and N, consistent with our
active-layer results (Table 2).
The C : N ratio was not a significant predictor of GHG fluxes in this
study, although this ratio has been found to be important in metanalyses
(Sistla et al., 2012; Schädel et al., 2014). In situ respiration rates
have also been shown to be negatively correlated with C : N at large
spatial scales (Allaire et al., 2012). Percent C and N both varied widely in
our soil cores (Table 1) and were highly correlated with each other, even
though the cores were collected within tens of meters of each other. This
suggests that active-layer SOC response to temperature and moisture may also
be highly spatially variable, even in a mixed-species boreal forest that we
expected, a priori, to provide spatial variation in litter and SOC quality
(Fierer et al., 2005). Spatially explicit analyses of soil biochemistry,
temperatures (Bond-Lamberty et al., 2005), and respiration (Allaire et al.,
2012) are likely necessary to accurately constrain and predict soil fluxes in
this ecosystem.
Limitations and weaknesses
There were weaknesses in our approach and experimental design that should be
considered. Laboratory experiments offer precise control but lack the in
situ nature of field manipulations (Sistla et al., 2013), raising
uncertainties as to what degree their results can be extrapolated. Soils
isolated during incubation may, for example, underestimate temperature
sensitivity of respiration (Podrebarac et al., 2016) or exhibit lag effects
(Treat et al., 2015). It should also be noted that our 100-day incubation was
not long enough to observe slowly cycling soil fractions, which may vary in
their response to experimental manipulation (Karhu et al., 2010).
Nonetheless, the controlled environments of incubations provide an important
way to elucidate the key mechanisms controlling GHG from high-latitude soils
(Schuur et al., 2015).
The soils studied here were from an upland, mixed conifer–deciduous boreal
forest, and care needs to be taken before drawing regional inferences or
inferences about other ecosystem types. We focused on an experimental drought rather
than flooding because of the well-drained nature of the field site: it is
unlikely that the mid-slope forest we sampled in will ever suffer from
thermokarst or excessive soil moisture, but too-dry conditions are a serious
possibility in this relatively low-precipitation ecosystem (Barber et al.,
2000).
Finally, the soils here are not surface-layer soils (where the majority of
microbial activity and C mineralization of labile C takes place); removing
them from in situ conditions (where they are less exposed to O2, for
example) may significantly change the abiotic conditions to which the
microbial community is adapted. However, focusing on the active layer
provides crucial information about the potential loss of C from these soils,
a risk that needs to be well understood since permafrost degradation leads to
expansions in the depth of the active layer across the Arctic.