Water and heat transport in boreal soils: Implications for soil response to
Zhaosheng Fana,⁎, Jason C. Neffa, Jennifer W. Hardenb, Tingjun Zhangc, Hugo Veldhuisd,
Claudia I. Czimczike, Gregory C. Winstone, Jonathan A. O'Donnellf
aDepartment of Geological Sciences, University of Colorado, Boulder, CO 80305, USA
bU.S. Geological Survey, Menlo Park, CA 94025, USA
cNational Snow and Ice Data Center, Cooperative Institute for Research in Environmental Sciences, University of Colorado, Boulder, CO 80305, USA
dAgriculture and Agri-Food Canada, University of Manitoba, Winnipeg, MB, Canada R3T 2N2
eDepartment of Earth System Science, University of California, Irvine, CA 92697, USA
fDepartment of Biology and Wildlife, University of Alaska, Fairbanks, AK 99775, USA
a b s t r a c ta r t i c l ei n f o
Received 26 October 2010
Received in revised form 3 February 2011
Accepted 7 February 2011
Available online 26 February 2011
Soil water content strongly affects permafrost dynamics by changing the soil thermal properties. However,
the movement of liquid water, which plays animportant role in the heattransport oftemperate soils, has been
under-represented in boreal studies. Two different heat transport models with and without convective heat
transport were compared to measurements of soil temperatures in four boreal sites with different stand ages
and drainage classes. Overall, soil temperatures during the growing season tended to be over-estimated by 2–
4 °C when movement of liquid water and water vapor was not represented in the model. The role of heat
transport in water has broad implications for site responses to warming and suggests reduced vulnerability of
permafrost to thaw at drier sites. This result is consistent with field observations of faster thaw in response to
warming in wet sites compared to drier sites over the past 30 years in Canadian boreal forests. These results
highlight that representation of water flow in heat transport models is important to simulate future soil
thermal or permafrost dynamics under a changing climate.
© 2011 Elsevier B.V. All rights reserved.
A large fraction (approximately 33–37%) of terrestrial carbon (C) is
stored in boreal soils (Ping et al., 2008; Tarnocai et al., 2009b). Over
the last four decades, air temperatures in the boreal region have
increased during winter and spring by 0.2 to 0.3 °C per decade,
accompanied by increased precipitation in autumn and winter (Jones
and Moberg, 2003; Smith and Reynolds, 2005). In some regions,
higher air temperatures are likely triggering increases in net primary
production (Bond-Lamberty et al., 2004) as well as changes in soil
hydrology (Striegl et al., 2007; Walvoord and Striegl, 2007).
Changes in boreal hydrology may result in a positive feedback to
global warming, either via increased emissions of CH4in previously
frozen and now flooded areas or via increased production of CO2from
decomposition or combustion in drained areas. Hydrological changes
might also shift the distribution of peatlands, wetlands, and lakes
(Rapalee et al., 1998; Vitt et al., 2000). While in some areas thawing of
permafrost can promote the development of wetlands and lakes
(thermokarst), in other areas permafrost degradation can lead to
drainage (Jorgenson and Osterkamp, 2005). In central Siberia, 11% of
lakes N40 ha have shrunk or disappeared between 1973 and 1997/
1998 (~93,000 ha of regional lake surface) (Smith et al., 2005).
Degradation of permafrost over the last two decades has been
reported from northern Alaska (Jorgenson et al., 2001), yet the
environmental factors responsible for the loss of permafrost are
debated with evidence for both the direct effects of warmer
temperatures and/or the effects of increasing precipitation (Agafonov
et al., 2004).
Heat transport through soils is a critical factor in determining how
permafrost changes in response to climate. In most of the existing
biophysical models used in boreal settings, pure heat conduction,
which is mainly influenced by the particle contact area (Jury and
Horton, 2004), is assumed to be the main mechanism of heat
transport in the soils. However, heat conduction is not the only
mechanism for heat transport in soils and ecosystems. Cahill and
Parlange (1998) indicated that liquid water and thermally-induced
vapor movement (convection) contributed to ~50% of soil heat flux in
temperate soils. In boreal soils where moisture is highly variable
across space and through seasons, the role of water movement could
be an important factor in seasonal soil energy dynamics and in the
long term response of boreal systems to changes in climate.
Additionally, the convection of latent heat by water vapor movement
Science of the Total Environment 409 (2011) 1836–1842
⁎ Corresponding author. Tel.: +1 303 735 2413; fax: +1 303 492 2606.
E-mail address: firstname.lastname@example.org (Z. Fan).
0048-9697/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
Contents lists available at ScienceDirect
Science of the Total Environment
journal homepage: www.elsevier.com/locate/scitotenv
in response to temperature gradients will release energy through
evaporation–condensation processes, resulting in rapid heat transfer
with only minor changes in soil moisture content (Jury and Horton,
2004; Kane et al., 2001).
Two heat transport models used to describe soil thermal dynamics
were compared in this study. In Model 1, heat was transported by
conduction and convection via the movement of liquid water and
water vapor; in Model 2 heat was transported exclusively by
conduction, a formulation similar to the commonly used soil physical
models for boreal soils, for example, FORFLUX (Zeller and Nikolov,
2000), CLASS (Verseghy, 1991), LPJ-GUESS (Wolf et al., 2008), and
Wetland-DNDC (Zhang et al., 2002).
2. Material and Methods
We studied three black spruce stands (Picea mariana (MILL.) BSP)
on well drained, effectively dry (‘D’) sites that burned in stand-
replacing fires in approximately 1964, 1930 or 1921 (Table 1). The
well drained sites tend to have gravel subsurface substrates that are
highly permeable and tend to inhibit the retention of water and
formation of ice-rich permafrost. The well drained 1964 and 1930
sites (1964D and 1930D) were established in years 2003 and 2002
near Thompson, Manitoba, Canada, while the well drained 1921 site
(1921D) was established in year 2000 and is located within the
Donnelly Flats near Delta Junction, Alaska. We also examined one
black spruce stand on a poorly drained, wet (‘W’) soil that burned in
approximately 1964 (1964W) which is located in Manitoba, Canada
(Table 1) (Litvak et al., 2003; Manies et al., 2003).
2.1. Field Experiments
Field experiments to measuresoiltemperatureandmoisturein the
organic layers and mineral soil for 1964D, 1930D, and 1964W are
presented in the following sections, while Fan et al. (2008) and
Manies et al. (2003) contain field experimental details for 1921D. The
amounts of organic carbon (OC) present in the mineral soils were
similar at all sites and approximately 0.02±0.01 g Ccm−3.
2.1.1. Field Installations
At 1930D site, one soil pit, excavated by shovel to depth of the C
horizon (102–136 cm), was instrumented and backfilled in September
2002, and measured continuously until 2005. At 1964D and 1964W
sites, one similar soil pit was instrumented in May 2003 and measured
until September 2004.
Precision interchangeable thermistors (model EC95H303W
Thermometrics, Edison, NJ, USA) encapsulated in thermally conductive
epoxy were inserted horizontally at 2, 6, 14, 30, 50, 70, and 80 cm for
41, and 55 cm for 1930D. Time domain reflectrometry soil moisture
probes (model CS616, Campbell Scientific Inc., Logan, UT, USA) were
also inserted horizontally at 5, 14, 30, 50, and 70 cm for 1964D, at 5, 25,
35, 45, and 60 cm for 1964W, and at 18, 29, 41, and 55 cm for 1930D.
Temperature and moisture conditions at the soil surface were
monitored with an additional thermistor and a fuel moisture sensor
(‘fuel rod’) (model CS505, Campbell Scientific Inc.).
For monitoring moisture content of near-surface organic soils,
20 cm long ECH2O dielectric constant probes (Decagon Devices, Inc.,
Pullman, WA, USA) were placed 5 cm below the moss surface at the
1964D and1964W sites. ECH2O probes have a reported accuracyof 3%
and resolution of 0.1%, but we found larger errors (10 to 30% moisture
content) when these instruments were calibrated in the lab with soil
blocks from each site. These blocks were saturated and then allowed
to air-dry while moisture content (by volume) and probe output were
monitored. Probe output was plotted against volumetric water
content and a linear fit was determined for each site. Factory
calibrations were used for TDR and thermistors.
All thermistor, TDR, and ECH2O data were collected and recorded
using a data logger (model CR10X, Campbell Scientific Inc.).
2.1.2. Soil Sampling
Detailed methods canbe foundin Manieset al. (2006).At each site,
soil pits were also excavated by shovel to depth of the C horizon (102–
136 cm) in years 2001 to 2004. In addition, small pits were excavated
for describing organic soil layers, generally to depths of 20 to 50 cm.
These excavations were placed 20 m apart along transects about
200 m long. Soils were described according to USDA-NRCS (Soil
Survey Division Staff, 1998) and Canadian (Committee CASC, 1998)
methodologies and sampled by soil horizons. Samples were analyzed
for bulk density and C content. In addition, volumetric soil moisture
content was measured using a variety of tools, including a Soil
Moisture Equipment Corporation (Goleta, CA) Model 0200 core for
deep mineral soils; rectangles of known area (for litter and organic
horizons); and cores of various diameters.
2.1.3. Laboratory Analyses
All soil samples were immediately placed on open shelves in an
isolated room after shipped to U.S.G.S, and allowed to air dry to a
constant weight. Temperature during air-drying ranged from 20 to
30 °C and relative humidity during air-drying ranged from 50 to 60%.
Once air dried, samples were separated into archive and moisture/
analytical splits. Air dry moisture content of both splits was recorded.
Bulk density calculations assumed that the air-to-oven dry moisture
ratio in the entire sample was the same as for the moisture/analytical
2.2. Model Description
The following partial differential equation describes the heat
transport by conduction with phase change, plus by the movement of
Descriptions of the four black spruce forest sitesa,b.
Site descriptions Mass of organic layer
1964D0–1 cm: live Rhytidium sp.
1–7 cm: fibric C with char
Below 7 cm: mineral soil
0–3 cm: live Pleurozium sp.
3–6 cm: dead moss with litter
6–11 cm: fungi, roots, and dead moss
11–15 cm: lightly to moderately decomposed
roots, moss, and wood
15–17 cm: charred and decomposed organic
Below 17 cm: mineral soil
0–2 cm: live moss, lichen, and litter
2–4 cm: dead moss and litter
4–5 cm: lightly decomposed organic matter and
5–11 cm: moderately decomposed organic
Below 11 cm: mineral soil
0–1 cm: live Sphagnum sp.
1–10 cm: dead moss with litter and roots
10–20 cm: lightly decomposed roots and moss
20–25 cm: moderately decomposed organic
25–35 cm: black and brown amorphous organic
35–40 cm: black amorphous organic matter
Below 40 cm: mineral soil
aManies et al. (2006, 2003).
bThe soil horizons were classified based on the method described in Harden et al.
Z. Fan et al. / Science of the Total Environment 409 (2011) 1836–1842
liquid water and water vapor (Cahill and Parlange, 1998; Hansson
et al., 2004; Jones and Moberg, 2003; Lundin, 1985; Saito et al., 2006):
where Cp, Cw, and Cvare volumetric heat capacities (J cm−3K−1) of
soil matrix, liquid water, and water vapor, respectively; T is the soil
temperature (K); t is the time (day); L0is the volumetric latent heat of
vaporization of liquid water (J cm−3); θvis the volumetric fraction
(cm3cm−3) of water vapor; Lfis the latent heat of freezing (J cm−3);
ρiis the density of ice (g cm−3); θiis the volumetric fraction of ice
(cm3cm−3); Ktis the thermal conductivity (J cm−1day−1K−1) of soil
matrix; z is the soil depth (cm); qLand qvare the velocity (cm day−1)
of liquid water and water vapor. Cpand Ktare defined as (Weiss et al.,
Csθs+ Cwθw+ COCθOC+ Ciθi
Cwθw+ COCθOC+ Ciθi
for mineral soil
for organic layer
Ksθs+ Kwθw+ KOCθOC+ Kiθi
Kwθw+ KOCθOC+ Kiθi
for mineral soil
for organic layer
where Cs, COC, and Ciare volumetric heat capacities (J cm−3K−1) of
mineral soil, organic carbon (OC), and ice, respectively; θs, θw, and θOC
are the volumetric fraction (cm3cm−3) of mineral soil, liquid water,
and OC, respectively; Ks, Kw, Kv, Koc, and Ki are the thermal
conductivities (J cm−1day−1K−1) of mineral soil, liquid water,
water vapor, OC, and ice, respectively. Due to the significantly low
volumetric fraction of water vapor and the significantly low
volumetric heat capacity of air, water vapor and air were assumed
to have negligible impacts on the heat capacity and thermal
conductivity of the soil matrix. Also, it was assumed in the model
that the fraction of mineral soil particles in the organic layers was
negligible. The thermal conductivity of dry mineral soil was related to
thesoiltextureandstructure,anddefined asa functionof bulkdensity
(Lu et al., 2007):
Ks= −0:56 1−BD
where BD is the bulk density of mineral soil (g cm−3); PD is the
particle density of mineral soil that is assumed to be 2.65 gcm−3.
The flux densities of liquid water (qL) and water vapor (qv) are
defined as (Cahill and Parlange, 1998; Saito et al., 2006):
and m = 1−1
where λsatis the saturated hydraulic conductivity (cm day−1); Seis
the effective saturation (unitless); m or n is the empirical parameter
that constrains the shape of water retention curve (unitless); h is the
pressure head (cm); θsat and θr are the saturated and residual
moisture content (cm3cm−3), respectively; τ is the tortuosity
(unitless); Dais the diffusivity of water vapor in air (cm2day−1); M
is the molecular weight of water (g mol−1); g is the gravitational
acceleration (cm day−1); R is the universal gas constant; ρsvis satu-
rated vapor density (g cm−3); ρwis the density of water (g cm−3).
The tortuosity (τ), diffusivity of water vapor (Da), and saturated vapor
density (ρsv) are defined as (Saito et al., 2006):
τ =θ7 = 3
Da= 2:12 × 10−5
One of the difficulties of simulating water flow is unstable flow
due to the soil heterogeneity, especially within the organic layers where
water flows quickly through the macropores. The water flow in our
model was assumed to follow kinematic wave approximation (water
flows downward under one unit pressure gradient) for surface organic
layers (i.e., live moss, dead moss, lightly and moderately decomposed
organic matter, and humic OC of dry sites) due to the large proportion of
macropores and lack of wicking of water by feather moss (Wang et al.,
2003). Kinematic wave approximation has been successfully used to
describe the one-dimensional vertical unsaturated subsurface flow
(Germann, 1985), snowmelt water movement (Singh et al., 1997), soil
moisture dynamics (Mdaghri-Alaoui and Germann, 1998), and
hydraulic conductivity. The water flows upward or downward in the
humic OC (i.e., well decomposed organic matter) of wet sites (Devito et
al., 1998; Reeve et al., 2000; Romanowicz et al., 1993) due to the wicking
of water by sphagnum moss (Wang et al., 2003) and in the mineral soil
0for θw= θs
where α is an empirical parameter.
It is extremely difficult to accurately simulate the soil moisture
content in the boreal region due to the complicated OC settings,
overland runoff, precipitation (including rainfall intensity and
duration), evapotranspiration, root water uptake, and latent water
flow. Therefore, we did not simulate the soil water content in this
study. Instead, we used the daily average values of measured soil
moisture content to calculate the daily average values of soil water
pressure heads (using Eq. (10)), which can then be used to calculate
the daily average values of water flow.
Eq. (1) does not include the impacts of heat of wetting on soil
thermal dynamics. Heat of wetting is the heat released during
infiltration when very dry soil (e.g., oven-dried) is “completely
immersed” by soil water (Prunty and Bell, 2005). Heat of wetting is
strongly related to soil water and clay content (Prunty and Bell, 2005).
Higher clay content and lower water content tend to lead to release
more heat of wetting during water infiltration and this factor is
importantfor heattransportin somesoils.However,itis reasonableto
assume that heat of wetting has minor impacts on the soil thermal
dynamics in our study due to the two following reasons. First, heat of
wetting is likely negligible for the deep soil (N6 cm) due to the high
initial water content during infiltration/precipitation (N0.2 cm−3cm3
compared to 0.00–0.05 cm3cm−3initial soil water content used in
Prunty and Bell's, 2005 studies). Second, although the surface soil
might release some amount of heat of wetting during water
infiltration due to the dry soil condition, infiltration/precipitation
events into dry surface soils occur only a handful of times during the
Z. Fan et al. / Science of the Total Environment 409 (2011) 1836–1842
period of our simulations and would have a negligible effect on the
overall heat flux to depth examined in this study.
The thermal conductivity of dry OC for each organic layer (i.e., live
moss, dead moss and lightly decomposed, moderately decomposed,
and well decomposed organic matter) was parameterized based on
the laboratory results of O'Donnell et al. (2009) and assigned to
different sites based on the distribution of each organic layer (see
supplementary material for more details).
The saturated hydraulic conductivity (λsat) of the organic layers is
also strongly related to the structure and components of organic
matter. The surface moss layer has significantly high λsat in
comparison to the deep organic matter (such as mesic and humic).
The saturated hydraulic conductivity and parameters (i.e., α, n, θsat,
and θr) for van Genuchten water retention curve (Eq. (10)) were set
separately for individual organic layers (Price et al., 2008) (see
supplementary material for more details). The saturated hydraulic
conductivity (λsat), α, m, θsat, and θr, for the mineral soil were
estimated using ROSETTA based on the soil texture (Schaap et al.,
2001). For the well drained site, the mineral soil was approximately
comprised of 43.8% sand, 47.4% silt, and 8.9% clay. For the poorly
drained site, the mineral soil was classified as clay according to the
USDA soil texture classification (Soil Survey Division Staff, 1993).
Therefore, the soil texture was assumed to be clay (clay contentN55%)
for the poorly drained site. The mineral soil was assumed to be
homogeneous and physical properties maintained constant through-
out the mineral soil profile.
Eq. (1) can be reduced to the traditional heat transport with only
conduction of sensible heat by setting θv=0, qw=0, and qv=0,
The upper and lower boundary conditions were set to the
measured (or interpolated) daily average soil temperatures at 0 cm
and 50 cm. The soil moisture profile was generated through linear
The thermal properties of frozen soil (e.g., soil thermal conduc-
tivity and heat capacity) were treated in the same manner as the
unfrozen soil and were calculated from Eqs. (2) and (3) based on the
ice content, liquid water content, OC and mineral soil fractions. The
hydraulic properties of frozen soil (e.g., unsaturated soil hydraulic
conductivity) were treated in the same manner as the dry unfrozen
soil. In other words, the freezing has the same impacts as drying in
that the soil matric potential is reduced thus driving the water flow
upward or downward toward the freezing fronts (Hansson et al.,
Eqs. (1) and (11) were solved using a finite difference solver,
CVODE (Soil Survey Division Staff, 1993) which is suitable for stiff and
non-stiff problems, with the backward differentiation formula
integration method and Newton iteration. Both absolute and relative
error tolerances were set to 0.001 as the simulation progresses. The
soil column was discretized equally into 100 soil layers, resulting in a
layer thickness of 0.5 cm during the simulations. The time step size
varied between the predefined minimum and maximum time steps of
0 and 1 day, respectively, until the calculation error was within the
absolute and relative error tolerances.
The soil temperatures for the three dry sites were simulated for a
one-year period, while for the wet site (i.e., 1964W) from 02/2004 to
09/2004. This is due to the incomplete and/or discontinuous
observations on soil temperature and moisture during other time
periods (e.g., for more than six months).
3. Results and Discussion
The comparisons of the calculated residual sum of squares
between Model 1 and Model 2 indicate that Model 1 provided overall
better simulation than Model 2 (Fig. 1). In the growing season (i.e.,
Apr. through Sep.; Fig. 1), Model 2 substantially over-estimated the
soil temperature in the organic layers by approximately 5 °C at the dry
site 1921D. For all sites at 30-cm depth, the soil temperatures
simulated by Model 2 were over 2° warmer than the measured values
for about half the growing season across all sites and by 1–2° for
another 25% of the growing season. The inclusion of heat transport by
water in Model 1 significantly improved the simulation of soil
temperature during the growing season over Model 2 and reduced
the discrepancies between measured and simulated values to be less
than 1 °C. The still large discrepancies at 30 cm and 43 cm depths for
1964W and at 30 cm and 43 cm depths for the winter of 1930D might
be caused by the inaccurately interpolated soil moisture, uncertainties
associated with the estimates of the unsaturated water movement,
and/or absence of heat produced by the biotic decomposition of OC in
the winter (Khvorostyanov et al., 2008).
Comparisons of daily heat flux (DHF) during the growing season
indicated that conduction contributed approximately 93%, 93%, 39%,
and 56% of surface DHF (at 6 cm depths), and 20–28%, 50–67%, 32–
42%, and 96–100% of deep DHF (i.e., at 30 cm and 43 cm depths) for
1964D, 1930D, 1921D, and 1964W, respectively. Liquid water
movement contributed 7%, b1%, 61%, and 44% of DHF at 6 cm depth,
and 72–80%, 33–50%, 58–68%, and 0–4% of DHF at 30 cm and 43 cm
depth during the growing season. Vapor movement contributed ~7%
of surface DHF (during the growing season at 6 cm depth) for 1930D,
which has significantly lower moisture content than the other three
sites, while vapor contributed less than 0.1% of DHF for other soil
depths and sites during both winter and growing seasons. These
results indicated that the impact of water movement on DHF was
significant during the growing season in drier boreal sites whereas
conduction dominated heat transport through the wet site.
Model 1 and Model 2 provided similar simulations for soil
temperatures during the winter season (i.e., Oct. through Mar.;
Fig. 1). The calculated DHF of each component of heat transport (i.e.,
conduction, liquid water, and water vapor) indicated that pure
conduction contributed 89–100%, 89–100%, 92–94%, and 96–100% of
DHF during the winter season for 1964D, 1930D, 1921D, and 1964W,
respectively, suggesting that heat was transported mainly by
conduction when the soil is frozen due to the limited liquid water
heat transport model, we evaluated soil thermal dynamics (and
potential permafrost degradation) under the context of warming
climate. Two warming scenarios projected by 1) the National Center
for Atmospheric Research Parallel Climate Model (NCAR PCM A2) and
2) the Center for Climate System Research/National Institute for
Environmental Studies (CCSR/NIES A2) (data obtained from http://
www.ipcc-data.org) were simulated for the next 90 years (2010–
2100). Several recent studies (Prunty and Bell, 2005; Zeller and
Nikolov, 2000) indicated that the simulated soil temperatures using
soil thermal models were strongly sensitive to the depth of lower
boundary condition. Unless the measured soil temperatures are
available to be used as the lower boundary condition, the depth of
lower boundary should be deep enough (N10 m) to reasonably
simulate the propagations of seasonal, annual, and decadal
temperature signals during the long-term simulations (e.g.,
100 years) (Prunty and Bell, 2005; Zeller and Nikolov, 2000). In our
study, the lower boundary of simulated soil column was set to 20 m.
The layer thickness at the top was 0.01 m and then increased for each
layer by a factor of 1.05, resulting in thickness of 0.93 m at the bottom
of soil column. The soil moisture content below 1.0 m soil depth was
assumed to be at 100% saturation with no water movement based on
the field observations. As a result, the energy flux below 1.0 m soil
depth was exclusively controlled by heat conduction. Below the depth
of 1.0 m, the initial soil temperature increased by a factor of 0.03 °C/m.
The model was first spun up to reach an equilibrium state where the
Z. Fan et al. / Science of the Total Environment 409 (2011) 1836–1842
less than 0.0001 °C through the soil profile. Other soil physical
properties (e.g., organic layer thickness, soil moisture pattern, and
soil thermal conductivities) were assumed to remain unchanged
during the simulation. The upper boundary condition was set to the
measured soil surface temperature at 0-cm depth plus the designed
temperature increase (A2 scenarios from the NCAR PCM and CCSR/
NIES). A time-dependent heat flux condition was assumed to be the
lower boundary condition for the energy equation. The heat flux was
calculated based on the following equation,
J = −Kt
where J is the heat flux density (J cm−2day−1).
Fig. 1. Measured and simulated soil temperatures at three different soil depths (6, 30, and 43 cm) for the four black spruce forest sites (1964D, 1930D, 1921D, and 1964W). “Model 1”
refers to the model where heat was transported by conduction, movement of liquid water, and movement of water vapor. “Model 2” refers to the model where heat was transported
exclusively by conduction.
Z. Fan et al. / Science of the Total Environment 409 (2011) 1836–1842
Both Model 1 and Model 2 indicate that soil temperatures at a
given soil depth linearly increased with the warming climate
regardless of soil drainage condition. However results of Model 1
are strikingly different from Model 2 and indicate that the soil
temperature increase at 30 cm and 90 cm depths was related to
distribution of soil moisture content and physical properties of each
site (Fig. 2). Soil drainage class has been identified in prior work as a
key control on energy transport and a key factor in boreal soil
responses to climate change (Yi et al., 2007). In prior modeling work
however, wet sites with deep OC layers tend to warm more slowly
than dry sites with shallow OC layers, an observation that has been
attributed to the slower conductive heat transport through deeper OC
layers (relative to shallow soils). Evaluated in the context of future
changes in climate, such modeling analyses suggest that drier sites
with shallow soil organic matter layers would be most vulnerable to
warming and associated changes in soil processes (Yi et al., 2007)
whereas the wetter site would be more resistant to permafrost
degradation. In contrast, this study suggests that it is actually the wet
sites with deep OC layers that would experience larger proportional
increases in soil temperature with a given warming scenario when
compared to the drier sites.
The more accurate simulations of Model 2 provide one line of
permafrost degradation under a warming climate. A more substantial
level of support for these conclusions is the recent field observation of
permafrost degradation in Canada. Against a backdrop of rapidly
warming temperatures over the past three decades, Veldhuis et al.
(2002) report that 67% of the poorly drained sites (12 out of 18 sites)
and only 13% of the well drained sites (2 out of 15 sites) underwent
significant changes (loss or degradation) in near surface permafrost
Central Manitoba, Canada. These recent field-based studies combined
with the simulations of future warming strongly suggest that
permafrost under thick OC layers will be more vulnerable to warming
than permafrost under well drained sites with thin OC layers.
The mechanisms that underlie these differences in soil response to
temperature change are complex and involve the combined influence
of conduction and water-related heat transport (and variable
contributions of these two factors across sites). The over-estimation
of soil temperatures in the dry site by conduction-onlymodels andthe
differential response of these sites to warming appear to be related to
the transport of water through soils. The upward gradients of water
potential energy (mainly due to the upward moisture gradient) 1) in
the mineral soils of dry sites and 2) in the humic OC and mineral soils
of wet sites cause significant upward liquid water movement. Since
the temperature of the deeper soil layers is lower than that of the
shallow soil layers, the upward water movement lowers the
temperature of upper soil layers, which also explains why soil
temperatures simulated by the convection Model 1 in summer are
lower than by the conduction-only Model 2. Such upward water
movement is well documented in mineral soils (e.g., Dingman, 1993;
Hansson et al., 2004; Richards, 1941) and has been reported for
1998; Reeve et al., 2000; Romanowicz et al., 1993). The net effect of
this heat transport via water movementis the cooling of surface layers
and appears to be an important contributor to the mismatch between
Model 2 simulations and observed temperatures.
Soil depth and moisture content also play an important role in the
propagation of warming climate through soils. In general, deeper soil
layers experience less increase in soil temperature. Thus atmospheric
warming leads to large vertical gradients in soil temperature. Because
heat transport in the wet sites is dominated by conduction (96–100%
of deep DHF as discussed earlier) due to the limited water movement
soil temperaturein the wet sites than in the dry sites. In drier sites, the
upwardconvection of heatin liquid waterthatresults in the cooling of
surface layers also appears to reduce the vulnerability of these sites to
warming by reducing the energy flux to deeper soil layers.
Permafrost in the boreal region has been warming as a result of
climate warming during the last several decades (Jorgenson et al.,
2001; Lemke et al., 2007; Yoshikawa et al., 2003). This study indicates
simulations of energy (heat) transport into boreal ecosystems and
changes both the quantitative and qualitative predictions of landscape
scale permafrost degradation. These findings, if accurate, have wide
implications for projections of boreal response to climate. A large
fraction (approximately 33–37%) of terrestrial C is stored in northern
the upper 3 m of northern ecosystems (Tarnocai et al., 2009a). If these
sitesalso are the mostvulnerableto warming-induced permafrost loss
as suggested here, then there would also be a higher potential for
destabilization of these large C reserves and release of additional CO2
and CH4from these ecosystems over the coming decades.
This modeling research was supported by the National Institute for
Climate Change Research, U.S. Department of Energy (DOE-NICCR;
grant #: MPC 35UT-01) and the US National Aeronautics and Space
Administration (NASA; grant #: NNX06AE65G). Field research was
supported by the U.S. National Science Foundation, the U.S.
Department of Energy, the U.S. Geological Survey, and the Natural
Resources Canada programs. We thank C.S. Carbone, M. Goulden, S.
Fig. 2. Changes in soil temperature in wet and dry sites at the end of the simulation of
two warming scenarios (NIES/CCSR A2 and NCAR PCM A2) using Models 1 and 2.
Z. Fan et al. / Science of the Total Environment 409 (2011) 1836–1842
Trumbore (University of California at Irvine), K. Manies, and L. Pruett Download full-text
(U.S. Geological Survey) for their assistance in the field and/or
laboratory experiments. Also, we greatly appreciate the helpful
comments from A.D. McGuire (USGS/University of Alaska at Fairbanks
(UAF)) and F. Yuan (UAF).
Appendix A. Supplementary data
Supplementary data to this article can be found online at
Agafonov L, Struck H, Nuber T. Thermokarst dynamics in Western Siberia: insights from
dendrochronological research. Palaeogeogr Palaeoclimatol 2004;209:1–4.
Bond-Lamberty BP, Wang C, Gower ST. Net primary production and net ecosystem
production of a boreal black spruce wildfire chronosequence. Glob Change Biol
Cahill AT, Parlange MB. On water vapor transport in field soils. Water Resour Res
Committee CASC. The Canadian system of soil classification. Ontario, Canada: National
Research Council Canada Research Press; 1998.
Devito KJ, Waddington JM, Branfireun BA. Flow reversals in peatlands influenced by
local groundwater systems. Hydrol Process 1998;11:103–10.
Dingman SL. Physical hydrology. Englewood Cliffs, NJ: Prentice Hall; 1993.
Fan Z, Neff JC, Harden JW, Wickland KP. Boreal soil carbon dynamics under a changing
climate: a model inversion approach. J Geophys Res 2008:113 G04016.
Germann PF. Kinematic wave approach to infiltration and drainage into and from soil
macropores. Trans Am Soc Agric Eng 1985;28:745–9.
Hansson K, Simunek J, Mizoguchi M, Lundin L, Van Genuchten MT. Water flow and heat
transport in frozen soil: numerical solution and freeze–thaw applications. Vadose
Zone J 2004;3:693–704.
Harden JW, Maines KL, Turetsky MR, Neff JC. Effects of wildfire and permafrost on soil
organic matter and soil climate in interior Alaska. Glob Change Biol 2006;12:
Jones PD, Moberg A. Hemispheric and large-scale surface air temperature variations: an
extensive revision and an update to 2001. J Climate 2003;16:206–23.
Jorgenson MT, Osterkamp TE. Response of boreal ecosystems to varying modes of
permafrost degradation. Can J For Res 2005;35:2100–11.
Jorgenson MT, Racine CH, Walters JC, Osterkamp TE. Permafrost degradation and
ecological changes associated with a warming climate in central Alaska. Clim
Jury WA, Horton R. Soil physics. Hoboken, New Jersey: John Wiley & Sons, Inc; 2004.
Kane DL, Hinkel KM, Goering DJ, Hinzman L, Outcalt SI. Non-conductive heat transfer
associated with frozen soils. Glob Planet Change 2001;29:275–92.
Khvorostyanov DV, Krinner G, Ciais P, Heimann M, Zimov SA. Vulnerability of
permafrost carbon to global warming. Part 1: model description and role of heat
generated by organic matter decomposition. Tellus Ser B 2008;60:250–64.
Lemke P, Ren J, Alley RB, Allison I, Carrasco J, Flato G, et al. Observations: change in
snow, ice and frozen ground. In: Solomon S, Qin D, Manning M, Chen Z, Marquis M,
Averyt KB, et al, editors. Climate change 2007: the physical science basis,
contribution of working group I to the fourth assessment report of the
intergovernmental panel on climate change. United Kingdom and New York,
USA: Cambridge University Press; 2007.
Litvak M, Miller S, Wofsy SC. Effect of stand age on whole ecosystem CO2exchange in
Canadian boreal forest. J Geophys Res 2003;108.
Lu S, Ren T, Gong Y, Horton R. An improved model for predicting soil thermal
conductivity from water content at room temperature. Soil Sci Soc Am J 2007;71:
Lundin L. Simulated physical effects of shallow soil heat extraction. Cold Reg Sci Technol
Manies KL, Harden JW, Veldhuis H. Soil data from a moderately well and somewhat
poorly drained fire chronosequence near Thompson, Manitoba, Canada. U.S.
Geological Survey Open File Report 2004–1271; 2006. p. 19.
Manies KL, Harden JW, Yoshikawa K, Randerson J. The effect of soil drainage on fire and
carbon cycling in central Alaska. U.S. Geol. Surv. Prof. Pap. 1678. Reston, Va: U.S.
Geol. Surv; 2003.
Mdaghri-Alaoui A, Germann PF. Kinematic wave approach to flow and moisture during
infiltration and redistribution of water in a structured soil. Hydrol Sci J 1998;43:
O'Donnell JA, Romanovsky VE, Harden JW, McGuire AD. The effect of moisture content
on the thermal conductivity of moss and organic soil horizons from black spruce
ecosystems in Interior Alaska. Soil Sci 2009;174:646–51.
Ping CL, Michaelson GJ, Jorgenson MT, Kimble JM, Epstein H, Romanovsky V, et al. High
stocks of soil organic carbon in the North American Arctic region. Nat Geosci
Price JS, Whittington PN, Elrick DE, Strack M, Brunet N, Faux E. A method to determine
unsaturated hydraulic conductivity in living and undecomposed Sphagnum moss.
Soil Sci Soc Am J 2008;72:487–91.
Prunty L, Bell J. Soil temperature change over time during infiltration. Soil Sci Soc Am J
Rapalee G, Trumbore SE, Davidson EA, Harden JW, Veldhuis H. Soil carbon stocks and
their rates of accumulation and loss in a boreal forest landscape. Glob Biogeochem
Reeve AS, Siegel DI, Glaser PH. Simulating vertical flow in large peatlands. J Hydrol
Richards LA. Hydraulics of water in unsaturated soils. Agric Eng 1941;22.
Romanowicz EA, Siegel DI, Glaser PH. Hydraulic reversals and episodic methane
emissions during drought cycles in mires. Geology 1993;21:231–4.
Saito H, Simunek J, Mohanty BP. Numerical analysis of coupled water, vapor, and heat
transport in the vadose zone. Vadose Zone J 2006;5:784–800.
Schaap MG, Leij FJ, Van Genuchten MT. ROSETTA: a computer program for estimating
soil hydraulic properties with hierarchical pedotransfer functions. J Hydrol
Simunek J, Jarvis NJ, van Genuchten MT, Gardenas A. Nonequilibrium and preferential
flow and transport in the vadose zone: review and case study. J Hydrol 2003;272:
Singh P, Spitzbart G, Hubl H, Weinmeister W. Hydrological response of snowpack under
rain-on-snow events: a field study. J Hydrol 1997;202:1–4.
Smith LC, Sheng Y, MacDonald GM, Hinzman LD. Disappearing arctic lakes. Science
Smith TM, Reynolds RW. A global merged land–air–sea surface temperature
reconstruction based on historical observations (1880–1997). J Climate 2005;18:
Soil Survey Division Staff. Soil Survey Manual. USDA HandbookWashington, DC: USDA;
Soil Survey Division Staff. Keys to soil taxonomy. Washington D.C., USA: National
Resources Conservation Service, U.S. Department of Agriculture; 1998.
Striegl RG, Dornblaser MM, Wickland KP, Raymond PA. Carbon export and cycling by
the Yukon, Tenana, and porcupine rivers, Alaska, 2001–2005. Water Resour Res
Tarnocai C, Canadell JG, Schuur EAG, Kuhry P, Mazhitova G, Zimov S. Soil organic carbon
pools in the northern circumpolar permafrost region. Glob Biogeochem Cycles
Tarnocai C, GCanadell JG, Schuur EAG, Kuhry P, Mazhitova G, Zimov S. Soil organic
carbon pools in the northern circumpolar permafrost region. Glob Biogeochem
Veldhuis H, Eilers RG, Mills GF. Permafrost distribution and soil climate in the Glacial
Lake Agassiz basin in North-central Manitoba, Canada. 17th World Conference on
Soil Science, Bangkok, Thailand; 2002.
Verseghy DL. CLASS—a Canadian land surface scheme for GCMs, I, Soil model. Int J
Vitt DH, Halsey LA, Bauer IE, Campbell C. Spatial and temporal trends in carbon storage
of peatlands of continental western Canada through the Holocene. Can J Earth Sci
Walvoord MA, Striegl RG. Increased groundwater to stream discharge from permafrost
thawing in the Yukon River basin: potential impacts on lateral export of carbon and
nitrogen. Geophys Res Lett 2007;34:L12402.
Wang C, Bond-Lamberty BP, Gower ST. Carbon distribution of a well- and poorly-
drained black spruce fire chronosequence. Glob Change Biol 2003;9:1066–79.
Weiss R, Shurpali NJ, Sallantaus T, Laiho R, Laine J, Alm J. Simulation of water table level
and peat temperatures in boreal peatlands. Ecol Model 2006;192:441–56.
Wolf A, Callaghan TV, Larson K. Future change in vegetation and ecosystem function of
the barents Region. Clim Change 2008;87:51–73.
Yi S, Woo MK, Arain MA. Impacts of peat and vegetation on permafrost degradation
under climate warming. Geophys Res Lett 2007;34:L16504.
Yoshikawa K, Bolton WR, Romanovsky V, Fukuda M, Hinzman L. Impacts of wildfire on
the permafrost in the boreal forests of Interior Alaska. J Geophys Res 2003;108:
Zeller KF, Nikolov NT. Quantifying simultaneous fluxes of ozone, carbon dioxide and
water vapor above a subalpine forest ecosystem. Environ Pollut 2000;107:1-20.
Zhang Y, Li C, Trettin CC, Li H, Sun G. An integrated model of soil, hydrology, and
vegetation for carbon dynamics in wetland ecosystems. Glob Biogeochem Cycles
Z. Fan et al. / Science of the Total Environment 409 (2011) 1836–1842