Remote Sensing of Environment 175 (2016) 349–358
Contents lists available at ScienceDirect
Remote Sensing of Environment
journal homepage: www.elsevier.com/locate/rse
Widespread permafrost vulnerability and soil active layer increases over
the high northern latitudes inferred from satellite remote sensing and
process model assessments
Hotaek Park a, Youngwook Kim b,⁎, John S. Kimball b
a
b
Institute of Arctic Climate and Environment Research, JAMSTEC, Yokosuka 237-0061, Japan
Numerical Terradynamic Simulation Group, College of Forestry & Conservation, The University of Montana, Missoula, MT 59812, USA
a r t i c l e
i n f o
Article history:
Received 4 June 2015
Received in revised form 23 December 2015
Accepted 31 December 2015
Available online 15 January 2016
Keywords:
Permafrost
Active layer thickness
Freeze/thaw
FT-ESDR
CHANGE
a b s t r a c t
Permafrost extent (PE) and active layer thickness (ALT) are important for assessing high northern latitude (HNL)
ecological and hydrological processes, and potential land–atmosphere carbon and climate feedbacks. We
developed a new approach to infer PE from satellite microwave remote sensing of daily landscape freeze–thaw
(FT) status. Our results document, for the first time, the use of satellite microwave FT observations for monitoring
permafrost extent and condition. The FT observations define near-surface thermal status used to determine
permafrost extent and stability over a 30-year (1980–2009) satellite record. The PE results showed similar
performance against independent inventory and process model (CHANGE) estimates, but with larger differences
over heterogeneous permafrost subzones. A consistent decline in the ensemble mean of permafrost areas
(− 0.33 million km2 decade−1; p b 0.05) coincides with regional warming (0.4 °C decade−1; p b 0.01), while
more than 40% (9.6 million km2) of permafrost areas are vulnerable to degradation based on the 30-year PE
record. ALT estimates determined from satellite (MODIS) and ERA-Interim temperatures, and CHANGE
simulations, compared favorably with independent field observations and indicate deepening ALT trends
consistent with widespread permafrost degradation under recent climate change.
© 2016 Elsevier Inc. All rights reserved.
1. Introduction
Climate change is causing widespread warming in the permafrost
zone, defined as permanently frozen ground and occupying approximately one quarter of the Northern Hemisphere land area (Payette,
Delwaide, Caccianiga, & Beaucjemin, 2004; Romanovsky, Smith, &
Christiansen, 2010; Zhang, Olthof, Fraser, & Wolfe, 2014). Warming of
permafrost influences multiple interactive properties affecting land–
atmosphere water, energy and trace gas exchange, including active
layer thickness (ALT) defined as the maximum depth of seasonal
thawing in soil layers overlying permafrost (Hayes et al., 2014;
Vaughan et al., 2013). The freeze/thaw (FT) signal observed from satellite microwave remote sensing captures abrupt shifts in landscape dielectric properties between predominantly frozen and non-frozen
conditions (Kimball, McDonald, Running, & Frolking, 2004; Kim,
Kimball, McDonald, & Glassy, 2011). The relatively coarse (~25-km resolution), but near-daily FT observations from available global satellite
environmental data records provide for effective high northern latitude
(HNL) regional monitoring of the timing and duration of frozen and
non-frozen seasons, which may be interactive with underlying soil
and permafrost conditions; however, finer scale properties, including
⁎ Corresponding author.
E-mail address: [email protected] (Y. Kim).
http://dx.doi.org/10.1016/j.rse.2015.12.046
0034-4257/© 2016 Elsevier Inc. All rights reserved.
vegetation composition, organic litter layers, subsurface drainage,
snow cover, and topography may be dominant factors influencing
permafrost distribution and condition at local scales (Duguay,
Zhang, Leverington, & Romanovsky, 2005; Zhang et al., 2014;
Podest, McDonald, & Kimball, 2014).
Regional permafrost extent (PE) is difficult to determine from direct
ground surveys of permafrost features (e.g., digging, core drilling, and
temperature measurements in boreholes) due to the extensive PE domain, high costs and inconsistent sampling (Duguay et al., 2005). Consequently, indirect methods for regional PE estimation have been
employed using near-surface indicators of underlying permafrost inferred from remote observations of vegetation cover, ground thermal
(air and surface temperatures) and hydrological parameters (snow
depth and soil moisture) (Minsley et al., 2012; Nguyen, Burn, King, &
Smith, 2009; Panda, Prakash, Jorgenson, & Solie, 2012). Land surface
modeling is commonly used to estimate PE and related subsurface processes at coarse spatial resolution (Burke, Kankers, Jones, & Wiltshire,
2013; Gruber, 2012; Lawrence, Slater, & Swenson, 2012; Park et al.,
2013). Coarse regional patterns of PE can also be observed from sparse
climate stations or model reanalysis of interpolated station observations
(Zhang et al., 2014). Satellite optical–infrared (IR) remote sensing has
been used to infer PE from empirical analyses of various surface indicator observations, including vegetation, land cover, thermokarst ponds
and terrain (Morrissey, Strong, & Card, 1986; Panda et al., 2012), and
350
H. Park et al. / Remote Sensing of Environment 175 (2016) 349–358
image classification techniques (Leverington & Duguay, 1997; Nguyen
et al., 2009; Yoshikawa & Hinzman, 2003). However, there are limitations in identifying presence/absence of permafrost due to optical–IR
sensor constraints under low light levels and persistent cloud cover conditions characteristic of HNL landscapes. Alternatively, microwave remote sensing is well-suited for identifying permafrost features (Dean
& Morrissey, 1988; Granberg, 1994; Yoshikawa & Hinzman, 2003) and
degradation (Granberg, 1994; Strozzi, Kaab, & Frauenfelder, 2004),
with minimal negative impacts from solar illumination, cloud and atmosphere aerosol contamination. However, to date, little is known regarding the potential for satellite microwave remote sensing to be
used for regional PE characterization and monitoring.
The motivation for this study was to develop an effective approach
for HNL assessment and monitoring of PE using available satellite microwave remote sensing observations of landscape FT status; here an
existing global satellite microwave FT Earth System Data Record (FTESDR; Kim, Kimball, Zhang, & McDonald, 2012, Kim, Kimball, Glassy
and McDonald, 2014) is used to infer subsurface permafrost conditions
and PE. The PE results derived from the FT-ESDR inputs were evaluated
against other independent PE estimates from a land surface model
(CHANGE; Park et al., 2011) and a static regional inventory based permafrost map (Brown, Ferrians, Heginbottom, & Melnikov, 2014). To
our knowledge, this is the first study to estimate PE over the HNL domain using satellite passive microwave remote sensing based FT observations. The FT-ESDR derived PE also defined the domain for estimating
ALT and associated changes in the ground thermal regime overlying
permafrost using a previously developed empirical method (Zhang
et al., 2005) and ancillary surface temperature inputs from both satellite
thermal IR remote sensing and global model reanalysis data. These results were evaluated against alternative ALT estimates determined
from sparse soil temperature station observations, global model reanalysis, and independent process model simulations. The resulting PE and
ALT estimates were determined over a 30-year study period (1980–
2009) at 25-km spatial resolution and yearly time step, and used to
evaluate regional patterns and recent trends in HNL permafrost and active layer conditions.
2. Data and methods
2.1. Microwave remote sensing based estimation of permafrost extent
We used a global landscape FT-ESDR derived from daily (ascending
and descending orbit) 37 GHz, vertically polarized brightness temperature observations from calibrated SMMR (scanning multi-channel microwave radiometer) and SSM/I (special sensor microwave imager)
satellite sensor records (Kim, Kimball, Glassy and McDonald, 2014);
the FT-ESDR was used to estimate PE over the HNL domain, including
all vegetated land area poleward of 45°N. Grid cells characterized by
predominantly barren land, permanent ice and snow, and open water
bodies are excluded from the FT classification (Kim et al., 2012). The
FT-ESDR provides a daily measure of the predominant landscape FT status within each 25-km grid cell posted to an EASE-Grid projection
(Brodzik & Knowles, 2002) for the period 1979–2012 (Kim et al.,
2012; Kim, Kimball, Glassy and McDonald, 2014). The FT-ESDR is derived from 37 GHz brightness temperatures that are sensitive to land
surface FT conditions, but relatively insensitive to potential atmosphere
contamination effects (Holmes, De Jeu, Owe, & Dolman, 2009; Kim,
Kimball, Didan and Henebry, 2014). The FT-ESDR classifies the predominant frozen or non-frozen condition of the land surface on a daily basis
within each grid cell and does not distinguish among individual landscape elements within the sensor field-of-view (FOV), including vegetation, snow cover and soil components. Satellite microwave sensitivity to
these landscape elements is frequency dependent, whereby the 37 GHz
FT retrievals are expected to be more directly sensitive to land surface
conditions, rather than deeper vegetation, snow and soil active layer
properties. The estimated FT-ESDR mean annual spatial classification
accuracy is approximately 84–91% relative to in situ surface air temperature measurements from the global weather station network (Kim
et al., 2012). The seasonal FT-ESDR classification accuracy was generally
lower during spring and fall transitional periods, and higher during winter frozen and summer non-frozen periods (Kim et al., 2011); there was
also no significant difference in mean FT classification accuracy between
the spring (MAM) and fall (SON) periods.
PE has previously been determined manually using the ratio of annual degree days of freezing and thawing, mean annual air temperature,
and field observations (Gruber, 2012; Nelson & Outcalt, 1987). The
southern limit of permafrost corresponds roughly with the ± 1 °C
mean annual surface air temperature isotherm (Duguay et al., 2005;
Romanovsky et al., 2010). Recent studies showing favorable correspondence between satellite microwave derived surface FT state dynamics
and soil active layer thermal properties from in situ monitoring stations,
including Global Terrestrial Network of Permafrost (GTN-P) sites,
indicate potential utility for satellite-based permafrost monitoring
(Naeimi et al., 2012; Du et al., 2014). In this study, regional permafrost
extent and condition within the HNL domain was estimated from the
satellite microwave remote sensing based FT-ESDR, whereby land surface FT conditions captured by the sensor were assumed to be an effective indicator of underlying soil active layer thermal conditions affecting
permafrost. Based on this methodology, grid cells were classified as permafrost where the cumulative number of FT-ESDR defined frozen days
exceeded non-frozen days during a water year (September 1 to August
31) and over at least two consecutive years. This approach is consistent
with previous studies indicating that when the number of yearly frozen
days exceed non-frozen days, the top ground layer experiences longer
seasonal freezing, and frost penetration depth increases over time accordingly, leading to permafrost occurrence (Dobinski, 2011; Nelson &
Outcalt, 1987; Saito et al., 2013; Zhang et al., 2005). However, deviations
from this general premise can occur due to spatial heterogeneity in insulating surface properties, including snow cover, soil organic layer
thickness and vegetation cover, especially within southern permafrost
boundary regions.
The FT-ESDR PE results were evaluated against a static permafrost
map (Brown et al., 2014) derived from International Permafrost Association (IPA) inventory records. A linear trend analysis was applied to
quantify CHANGE and FT-ESDR based changes in PE condition from
1980 to 2009, where rate of PE change is used as a proxy for permafrost
stability. Tundra and boreal forest biomes within the PE domain were
categorized by a global terrestrial biome map (Olson et al., 2001), and
mean PE change rate was summarized for these individual biomes.
2.2. Estimate of active layer thickness based on MODIS LST
Active layer thickness (ALT) is predominately controlled by surface
FT seasonal regime, soil moisture content, and site topography; the
ALT is also influenced by thermal buffering of underlying soil by surface
organic layer thickness, snow cover, and vegetation structure (Duguay
et al., 2005; Zhang et al., 2014). ALT has been estimated indirectly
from in situ surface temperature measurements (Zhang et al., 2005)
and ground penetrating radar (Westermann, Wollschlager, & Boike,
2010), active LiDAR remote sensing (Hubbard et al., 2013), and dynamic
permafrost models (Burke et al., 2013; Lawrence et al., 2012; Park et al.,
2013; Riseborough, Shiklomanov, Etzelmuller, Gruber, & Marchenko,
2008). Zhang et al. (2005) developed an empirical method for estimating ALT using an annual thawing index (ATI) and an edaphic factor (EF)
that parameterizes the effect of land cover type on soil thermal state. A
similar method was applied to estimate ALT in this study using alternative surface temperature inputs from satellite thermal IR remote sensing
and global reanalysis data.
The MODIS Terra and Aqua land surface temperature (LST) product
(MOD11C1 and MYD11C1; Wan, 2008) for the 2003–2009 record was
used to derive ALT using an ATI defined as the cumulative number of
LST defined degree-days above 0 °C for each selected year; MODIS LST
H. Park et al. / Remote Sensing of Environment 175 (2016) 349–358
has also recently been used to quantify and monitor the occurrence and
thermal state of permafrost (Westermann, Østby, Gisnås, Schuler, &
Etzelmüller, 2015). Best quality (QC = 0) day/night LST retrievals,
screened for cloud and atmosphere aerosol contamination, were reprojected to the 25-km resolution EASE-Grid format using drop-inbucket averaging. Daily mean MODIS LST (LSTAVG) was calculated by
averaging MODIS LST day/night retrievals, and producing the ATI following Zhang et al. (2005) for the 2003–2009 record. The relationship
between ALT and ATI can be described as (Harlan & Nixon, 1978;
Zhang et al., 2005):
pffiffiffiffiffiffiffi
ALT ðmÞ ¼ EF ATI
ð1Þ
where, EF is a spatially variable edaphic factor (m2/°C-day)0.5.
Similar to Zhang et al. (2005), the EF is an empirical scaling parameter that was defined for individual HNL land cover classes using Eq. (1)
and MODIS LSTAVG defined ATI observations, and in situ ALT measurements obtained from regional Circumpolar Active Layer Monitoring
(CALM) sites and Siberian meteorological stations (Park, Scherstiukov,
Fedorov, Polyakov, & Walsh, 2014). Table S1 shows the resulting mean
EF values and spatial standard deviations within each HNL land cover
class defined from a MODIS 17-class IGBP global land cover classification (MCD12C1, Friedl et al., 2010). The mean EF values were
used for grid cell-wise estimation of ALT using MODIS LSTAVG derived
ATI estimates over the 2003–2009 record. Despite the relatively
short (7-year) MODIS LST record used in this study, the MODIS LST
based ALT estimates are derived solely from satellite observations
and provide for independent verification of other ALT estimates derived from global reanalysis data and land model simulations over
the HNL domain.
2.3. Estimate of active layer thickness based on ERA-Interim SAT
Average daily surface air temperatures (SATAVG) from 0.25° resolution ERA-Interim reanalysis data (Dee et al., 2011) were also used to estimate ATI and ALT over a longer 1980–2009 record. The resulting ATI
values were then used with the prescribed EF values (Table S1) in
Eq. (1) to estimate ALT annually on a per grid cell-wise basis. The ALT
maps were derived within the corresponding HNL PE area defined on
an annual basis using the FT-ESDR. The long-term (1979–2009) annual
mean SATAVG from ERA-Interim was also used to define the 0 °C isotherm over the domain, while all datasets were processed and analyzed
in a consistent 25-km EASE-Grid format.
2.4. Estimate of active layer thickness and permafrost extent
based on WATCH
A set of independent process model simulations were conducted to
estimate PE and ALT dynamics over the HNL domain for comparison
against alternative estimates of these parameters determined from the
remote sensing FT-ESDR and MODIS LST records, and sparse monitoring
sites. CHANGE (coupled hydrological and biogeochemical model, Park
et al., 2011) is a state-of-the-art process-based model that calculates
heat, water, and carbon fluxes in the atmosphere–land system, including soil thermal and hydrologic states with explicit treatment of soil
FT transitions extending up to 50.5 m depth, snow hydrology, plant stomatal physiology and photosynthesis. CHANGE numerically solves the
second law of heat conduction equation that requires soil heat conductivity and capacity, determined as functions of liquid water and ice contents, soil temperature (TSOIL), and vertically heterogeneous soil
textures. The model also includes effects of soil organic matter on soil
thermal and hydraulic properties. These effects vary depending on
prognostic changes in soil organic carbon, and water, carbon and energy
fluxes and conditions. The heat flux into the soil–snow layers, obtained
by solving the energy balance equation, defines the upper boundary for
351
estimating heat conduction, while the zero heat flux defines lower
boundary conditions.
A European Union project (Water and Global Change; WATCH) produced a global gridded meteorological dataset at 0.5° resolution
(Weedon et al., 2011) that allows global simulations of land surface processes at a three-hourly time step for the period 1979–2009. WATCH
data, including surface air temperature, solar radiation, precipitation,
specific humidity, surface air pressure and wind speed, were used as
meteorological forcings for CHANGE simulations over the HNL domain.
Land cover was assumed static under present-day conditions indicated
by the MODIS IGBP land cover map (Friedl et al., 2010), while the
CHANGE vegetation canopy leaf area phenology was prognostic depending on model estimated carbon and nitrogen fluxes. Model simulations were run at a three-hour time step and initialized by repeatedly
simulating the first 20 years of record until the soil carbon pool reached
a stable state over a minimum of 1800 years.
‘Permafrost’ was determined from CHANGE when a grid cell was frozen for two or more consecutive years in at least one soil layer within a
depth of 3.6 m from the surface. In a previous study (Park, Fedorov,
Zheleznyak, Konstantinov, & Walsh, 2015), CHANGE model derived PE
simulations were compared with independent PE assessments from
the IPA permafrost map and previous modeling studies (Burke et al.,
2013; Lawrence et al., 2012), where the other models defined lower
soil boundary layers of 3–4 m depth for their PE calculations; a prescribed soil boundary depth of 3.6 m was found to produce the best
CHANGE model PE performance relative to the other model based predictions. PE simulated from CHANGE represents the integrated permafrost area over the HNL domain, excluding permanent ice and snow
areas (e.g. Greenland). Here, permafrost degradation refers to an expansion of the area where estimated ALT exceeds 3.6 m over the simulation
period, rather than permafrost being completely absent. In individual
cells defined as permafrost, the simulated TSOIL in the eight discrete
model soil layers within the upper 3.6 m were linearly interpolated to
determine the depth at which TSOIL crossed the 0 °C isotherm; the maximum thaw depth over an annual cycle was defined as the ALT of that
year. A 3.6 m soil depth is assumed to be reasonable for comparing
near surface permafrost dynamics related to FT monitoring from satellite microwave remote sensing (Holmes et al., 2009; Kim et al., 2011;
Lawrence & Slater, 2005).
3. Results
Circum-Arctic permafrost and ground ice inventory data (Brown
et al., 2014) provides a static PE classification re-gridded from IPA inventory records (Fig. 1a). The PE spatial patterns indicated from the
FT-ESDR and CHANGE simulations are presented in Fig. 1(b and c), including the 0 °C isotherm of annual SATAVG defined from ERA-Interim
(in black). The domain for this study does not include PE areas below
45°N, which excluded some high elevation permafrost areas
(e.g., Tibetan Plateau). The FT-ESDR derived PE is largely consistent
with IPA permafrost inventory data. The PE boundary defined by the
FT-ESDR (Fig. 1b) generally coincides with the 0 °C air temperature isotherm defined from ERA-Interim SATAVG. However, FT-ESDR overestimation of PE relative to ERA-Interim is found in mountainous regions
of southwestern Canada and southern Siberia. There are also many FTESDR spatial gaps in northern Canada, resulting from masking cells
characterized as non-vegetated, or with large water bodies, or permanent snow/ice cover (Kim et al., 2012). PE defined from the CHANGE
simulations excludes western Siberia and southwestern and eastern
Canada, which generally consists of sporadic, isolated, and discontinuous permafrost, and where CHANGE and FT-ESDR PE differences are
evident.
Lower grid cell-wise PE temporal frequency (%) from the FT-ESDR
occurs along the southern permafrost boundary (Fig. 1b), implying
low probability of PE occurrence for the 30-year period (1980–2009);
here PE frequency is defined as the number of years of suitable
352
H. Park et al. / Remote Sensing of Environment 175 (2016) 349–358
H. Park et al. / Remote Sensing of Environment 175 (2016) 349–358
permafrost conditions over the long-term record. Regions where PE frequency is less than 100% of the 30-year record indicate unsuitable conditions for permafrost and potential vulnerability to degradation, with
lower frequencies indicating relatively greater degradation risk. Areas
with 100% PE frequency indicate stable permafrost and extend over
57% (12.7 million km2) of the PE domain. In contrast, the area of 25km grid cells showing moderate permafrost degradation risk (50–99%
frequency) covers 20% (4.4 million km2) of the domain, while severe
degradation risk (b50% frequency) areas cover 23% (5.2 million km2)
of the domain. Thus more than 40% (9.6 million km2) of permafrost
areas are vulnerable to degradation based on the 30-year satellite record. The results show that the mean PE frequency is greater in tundra
(95.7 ± 15.2%) relative to boreal (73 ± 35.2%) biomes, which implies
greater boreal forest vulnerability to PE degradation.
PE distributions from the CHANGE simulations and FT-ESDR are
binned by latitude and compared with IPA permafrost inventory data
(Fig. 2). The mean latitudinal distribution of FT-ESDR derived PE was
computed on an annual basis from cells classified as permafrost within
one-degree latitudinal bins. The PE spatial standard deviation within
each one-degree latitudinal bin was also calculated over the 30-year
(1980–2009) satellite record. The same spatial mask used for the FTESDR based PE processing was applied to produce the IPA and
CHANGE based PE maps. PE derived from the CHANGE simulations is
larger at higher latitudes (63°N b latitude b 70°N) and smaller than
IPA defined PE at lower latitudes (b 63°N). The smaller CHANGE derived
PE at lower latitude (49°N and 55°N) occurs from the coarse (0.5 degree)
resolution of the simulations, which constrain the model ability to delineate spatially heterogeneous permafrost subzones, particularly
along the southern permafrost boundary composed of discontinuous,
sporadic, and isolated types of permafrost (Fig. 1a). The FT-ESDR
defined PE shows large spatial variability with latitude and is overestimated near 66°N relative to the other PE estimates, due to the
large relatively greater FT-ESDR PE in northwestern Russia. Local heterogeneity in thermal and hydrologic conditions characterized by poor
drainage, dense vegetation, snow cover, and thick organic layers decouple surface and soil thermal conditions, and also lead to biases between
IPA, FT-ESDR and CHANGE based PE estimates.
The PE results (Fig. 3) show a decreasing stability trend
(− 0.033 million km2 yr−1; p b 0.05) indicated from the ensemble
mean of the FT-ESDR (0.02 million km2 yr− 1; p N 0.1) and CHANGE
(− 0.057 million km2 yr−1; p b 0.01) results. The ensemble mean is
used to reduce the effect of non-natural variations and to estimate the
trend uncertainty from the ensemble variance (Allan & Soden, 2007;
Mudryk, Kushner, & Derksen, 2014), which is represented as a two standard deviation range around the mean. The discrepancy in PE trends derived from the FT-ESDR and CHANGE simulations was attributed to
differences between the early and latter portions of the FT-ESDR time
series and the representation of insulating surface elements (i.e., snow
cover, vegetation, and soil organic carbon) in the CHANGE simulations.
Increasing Siberian snow depth trends in recent years have generated a
thicker insulating surface over permafrost (Park et al., 2015). Furthermore, soil organic carbon layers produce greater insulating efficiency
by enhancing soil-heat capacity during the winter season. These surface
insulating elements resulted in a significant declining PE trend in the
CHANGE simulations relative to a slight increasing trend in FT-ESDR derived PE. Decreasing PE stability coincides with an HNL SATAVG warming
trend indicated from both ERA-Interim (0.04 °C yr− 1; p b 0.05) and
WATCH (0.044 °C yr−1; p b 0.01) datasets; these results imply increasing permafrost degradation with recent climate warming. The CHANGE
simulated PE degradation is also significantly correlated with WATCH
353
based SATAVG changes (r = −0.70; p b 0.001). PE degradation is significant along HNL southern boundary regions (Fig. 1b), where the temporal frequency of suitable permafrost conditions is less than 100% of the
30-year satellite record. An apparent increase (0.25 million km2 yr−1;
p b 0.1) in FT-ESDR PE from 1980 to 1987 (Fig. 3) coincides with regional cooling indicated from the ERA-Interim (−0.08 °C yr−1; p N 0.1) and
WATCH temperature records (− 0.03 °C yr−1; p N 0.1), and indicates
significant FT-ESDR sensitivity to surface temperature. A decreasing
FT-ESDR PE trend after 1987 (−0.006 million km2 yr−1) is partially offset by a positive PE trend over the earlier (1980–1987) portion of record
(0.25 million km2 yr−1), resulting in an insignificant positive trend over
the longer (1980–2009) FT-ESDR PE period.
ALT results derived from ERA-Interim SATAVG and CHANGE were
compared with MODIS LSTAVG over CALM and Siberia sparse network
stations in Figs. S1 and S2. The MODIS LSTAVG and ERA-Interim SATAVG
based ALT results show generally strong correspondence (r N 0.764,
p b 0.05) with mean annual ALT observations from the regional ground
station network (Fig. S1). The MODIS LSTAVG derived ALT results show
similar interannual variations relative to ERA-Interim SATAVG derived
ALT (r = 0.993; p b 0.01) and CHANGE ALT (r = 0.939; p b 0.01) results.
The MODIS LSTAVG, and ERA-Interim SATAVG based ALT results are
overestimated by 0.01–0.19 m in DNF (deciduous needleleaf forest)
dominated land cover areas over CALM and Siberia sparse network stations (Fig. S2); the ALT results are similar to the ground observations in
MF, shrubland, savanna and cropland areas, but underestimated by 0.1–
0.5 m in grassland areas. Mean CHANGE ALT for MF (mixed forest),
shrubland, and savanna areas is deeper than other ALT results, but
shallower in DNF, grassland and cropland areas. The MODIS LSTAVG,
ERA-Interim SATAVG and CHANGE simulations show a general poleward
decrease in ALT, and strong sensitivity to the ATI (Fig. 4). ALT derived
from MODIS LSTAVG and ERA-Interim SATAVG are generally less than
CHANGE simulated ALT between 50 and 76°N.
Circum-Arctic permafrost is characterized by thick surface peat
layers and considerable soil organic matter, increasing soil insulation
and promoting warmer soil temperatures; the CHANGE ALT simulations
include these effects on soil thermal and hydraulic properties (Park
et al., 2011), while ALT from MODIS LSTAVG and ERA-Interim SATAVG
are derived only from surface temperatures. In contrast, CHANGE ALT
below 50°N is smaller relative to MODIS and ERA-Interim (Fig. 1c).
ALT derived from MODIS and ERA-Interim temperatures exhibit generally similar latitudinal patterns, which may be attributed to similar
LSTAVG and SATAVG behavior, and a consistent FT-ESDR defined PE domain used for estimating ALT. The long-term (1980–2009) ALT results
show increasing regional trends over the PE domain from ERA-Interim
SATAVG (0.03 m decade−1; p b 0.01) and CHANGE (0.06 m decade−1;
p b 0.0001), indicating widespread permafrost degradation.
Very few long-term and large-scale observations of ALT are available
for model validation over the Arctic. However, a consistent set of longterm TSOIL records from 31 Russian meteorological stations were used
to derive ALT trends over the 1980–2009 record, where ALT was derived
annually according to the seasonal position of the 0 °C isotherm in the
measured TSOIL profile at each site (Park et al., 2013). The Russian meteorological station records differ from the CALM sites used for EF estimation in Eq. (1) and ALT assessment in Figs. S1 and S2, in that the number
of Russian stations represented within the Siberian sub-region was
consistent over the 1980–2009 record, for a more robust ALT trend assessment. Soil temperatures were measured at nine vertical depths
under natural surface cover such as grass, periodically cut during the
warm season, and undisturbed snow cover during the cold season
(Gilichinsky et al., 1998). It is noted that these station maintenance
Fig. 1. Permafrost extent derived from IPA inventory records (a); grid cell-wise PE frequency (%) defined as the number of years of suitable permafrost conditions relative to the 30-year
(1980–2009) FT-ESDR (b); mean distribution of ALT (m) derived from CHANGE model simulations (c), and MODIS LSTAVG (d) for the period 2003–2009. The base maps in (c) and (d) have
been produced from CHANGE and FT-ESDR PE, respectively. The bold line within (a)–(d) indicates the 0 °C air temperature isotherm averaged during the period 1979–2009 using ERAInterim SATAVG. Open water bodies, barren ground, permanent ice and other masked areas excluded from the analysis are shown in white in (b) and (d). Areas excluded from IPA, FT-ESDR
and CHANGE PE are shown in white.
354
H. Park et al. / Remote Sensing of Environment 175 (2016) 349–358
Fig. 2. Latitudinal permafrost extent (PE) distribution projected from the CHANGE and FT-ESDR results in relation to the reference IPA permafrost map. Gray shading denotes two standard
deviation spatial variability around the mean FT-ESDR PE values. Areas of permafrost sub-zones are determined from the static IPA inventory map. Color plots display IPA permafrost subzone areas (continuous: blue, discontinuous: red, sporadic: light green, and isolated: purple).
Fig. 3. Inter-annual variation of permafrost extent (PE, million km2) derived from FT-ESDR and CHANGE, and their ensemble mean; the PE metric defines the spatial extent of suitable
permafrost conditions, while PE change is used as a proxy for permafrost stability. Gray shading denotes two standard deviation range around the ensemble PE mean. The dashed line
represents the linear regression trend of the ensemble PE mean.
H. Park et al. / Remote Sensing of Environment 175 (2016) 349–358
355
Fig. 4. Latitudinal variation of the 7-year (2003–2009) average of permafrost active layer thickness (ALT) derived from CHANGE, MODIS LSTAVG and ERA-Interim SATAVG. Gray shading
denotes two standard deviation range around the ensemble ALT mean.
activities may unavoidably cause some site disturbance despite consistent measurement depths at the stations; therefore any long-term
trends in active layer thickness potentially include this non-climatic
component. The annual ALT anomalies from the CHANGE, ERAInterim SATAVG, MODIS LSTAVG and station observations were compared
within the Siberia sub-region (60–70°N, 115–165°E) in Fig. 5, where the
annual anomalies are defined according to the long-term (1980–2009)
ALT means from individual stations and grid-cells. The ALT anomalies
for individual stations and grid-cells were then spatially averaged over
the sub-region. The ALT annual anomalies from the station observations
Fig. 5. Interannual anomaly variation (relative to long-term mean) and estimated regional trends of permafrost active layer thickness (ALT) derived from CHANGE, ERA-Interim SATAVG,
MODIS LSTAVG, and sparse network station observations (in red on inset map) within a Siberia sub-region (60–70 N, 115–165 E; in black on inset map) from 1979 to 2009.
356
H. Park et al. / Remote Sensing of Environment 175 (2016) 349–358
corresponded significantly with ALT variations derived from ERAInterim SATAVG and CHANGE over the 1980–2009 record. The resulting
ALT correlation was larger for CHANGE (r = 0.73, p b 0.01) than ERAInterim (r = 0.64, p b 0.01) results. Differences between the local station observations and simulations may be attributed to the low density
of available TSOIL stations, and spatial scale mismatch between the
sparse local observations and regional simulations (Park et al., 2013).
The SATAVG based ALT results from MODIS and ERA-Interim also show
strong correspondence (r = 0.97, p b 0.01) for the relatively shorter
2003–2009 period represented by the MODIS LST record; despite the
shorter MODIS observation record, the generally favorable results and
continuing MODIS operations provide a means for satellite based regional monitoring of ALT properties.
Both the observed and simulated ALT results show significant increasing regional trends over the 30 year record (Fig. 5). Relatively
stronger increases in ALT are apparent after 2000, consistent with
reported increases in regional SAT and snow depth (Bulygina,
Razuvaev, & Korshunova, 2009; Park et al., 2014). Recent increases
in winter snow depth (Bulygina et al., 2009; Park et al., 2014) and
summer soil moisture (Ohta et al., 2014) in eastern Siberia have resulted in TSOIL warming (Panda et al., 2012; Park et al., 2014) and
consequently deeper ALT (Hayes et al., 2014; Park et al., 2013). ALT
determined from ERA-Interim SAT does not consider the insulation
effect associated with increasing snow depth, leading to the lowest
ALT trend (0.02 m decade− 1).
4. Discussion and conclusions
Spatial and temporal patterns of PE from three independent sources
were compared over the HNL domain, including satellite FT-ESDR observations, CHANGE model simulations and IPA permafrost inventory
records. The dynamic PE metric defined from the CHANGE simulations
and FT-ESDR observations is used as a proxy for suitable permafrost
conditions and relative stability. Satellite microwave remote sensing observations indicated from the FT-ESDR show strong potential for regional monitoring of permafrost extent and condition, although finer scale
observations and model simulations are needed to improve permafrost
delineation over complex terrain and land cover conditions, and over
heterogeneous discontinuous, sporadic and isolated permafrost subzones. Major differences between FT-ESDR and CHANGE PE estimates
were associated with the representation and relative abundance of insulating surface elements, including snow cover, soil organic matter,
and vegetation cover; these elements can decrease thermal energy conductance and increase the complexity in relationships between surface
climate and sub-surface permafrost conditions. Previous studies necessarily required more intensive information and observation inputs to
identify permafrost presence/absence and distribution at regional
scales. The FT-ESDR used in this study primarily detects near surface
FT status and provides a means for observation based estimation and
regional monitoring of permafrost extent and condition in the HNL
domain. Additional ancillary data on vegetation structure, snow
properties, soil parameters, and terrain may enhance PE delineation
accuracy in future studies. The FT-ESDR defined PE metric also provides the possibility for including dynamic observation based estimates of permafrost conditions, rather than static permafrost data,
when defining regional land cover and land use changes over boreal
and Arctic zones. These results provide insights regarding potential
risk and variation of permafrost degradation from continued climate
warming, and potential critical inputs for higher order predictions of
regional water, energy and carbon cycle responses, and climate
impacts.
Regional HNL warming has resulted in permafrost temperature increases (Romanovsky et al., 2010; Park et al., 2014) and ALT deepening
(Park et al., 2013; Hayes et al., 2014), leading to permafrost degradation,
particularly along southern PE regions. Oberman and Liygin (2009) observed that the permafrost boundary shifted northward by 30–80 km in
the Russian European North during the past 50 years. Similarly, the permafrost boundary in eastern Canada shifted northward by 130 km during the past 50 years (Thibault & Payette, 2009). In western Canada,
permafrost degradation has also been observed during the last century
(Beilman & Robinson, 2003; Jorgenson, Racine, Walters, & Osterkamp,
2001). Our results also show recent widespread declines in suitable permafrost conditions and large variability of the southern PE boundary
that are generally consistent with observations from sparse regional
ground stations. The greater FT-ESDR PE interannual variability is a consequence of the 37 GHz FT retrieval uncertainties because the 37 GHz
frequency is more sensitive to near-surface FT changes, including
snow and vegetation, rather than CHANGE PE which is more sensitive
to sub-surface ground layers with larger thermal inertia (Kim et al.,
2012; Bateni, Huang, Margulis, Podest, & McDonald, 2013). Continuing
satellite observations, including FT-ESDR and MODIS LST records, provide potential for regional monitoring of permafrost and active layer dynamics spanning the entire pan-Arctic domain. The NASA SMAP (soil
moisture active passive) mission began operations in April 2015, inaugurating new global observations of landscape FT status and surface
soil moisture conditions with L-band (1.4 GHz) microwave sensitivity
to surface soil conditions, and near daily temporal fidelity of observations (Entekhabi et al., 2010); the enhanced resolution and precision
of these observations may enable improved delineation and monitoring
of permafrost and active layer conditions.
Surface temperatures obtained from satellite thermal IR remote
sensing and global reanalysis data were used within FT-ESDR defined
PE areas to estimate ALT, and compared against independent ALT estimates from CHANGE simulations and sparse network soil temperature
measurements. The ALT results derived from independent MODIS
LSTAVG and ERA-Interim SATAVG inputs show similar latitudinal distributions that differ from the CHANGE ALT results, due to the different PE
domains represented and because the process model simulations account for the important role of soil thermal and hydraulic components
affecting ALT. However, the observations and simulations all indicate a
deepening ALT trend linked to permafrost warming under recent climate
trends, and consistent with TSOIL warming (Romanovsky et al., 2010;
Park et al., 2014). Because ALT strongly depends on summer thermal
history and land cover properties (Osterkamp, 2007), heat propagation
is enhanced or reduced by soil insulation effects from snow cover, soil organic matter and vegetation composition (Park et al., 2013). These
coupled effects may increase uncertainties of temperature-based ALT
estimates from MODIS LSTAVG and ERA-Interim SATAVG.
Deepening ALT would lead to increasing permafrost degradation
over the HNL domain, although the ALT trend results in this study
are limited to the Siberian sub-region. Deepening ALT trends associated with permafrost degradation expose greater volumes of soil organic matter to mobilization, microbial decomposition and carbon
release to the atmosphere, potentially exacerbating global warming
(Hayes et al., 2014; Koven et al., 2011). Increase in ALT also affects
groundwater storage and soil drainage (Velicogna, Tong, Zhang, &
Kimball, 2012; Zhang et al., 2005), influencing soil organic carbon
stability, methane emissions (van Huissteden et al., 2011), soil nutrient availability (Hayes et al., 2014; Keuper et al., 2012) and vegetation productivity (Natali, Schuur, & Rubin, 2012). Further
integration of remote sensing and modeling of permafrost and active layer conditions developed from this study may facilitate regular and effective regional monitoring of these parameters, and
expand applications of remote sensing for examining permafrostrelated feedbacks and consequences for biogeochemical and hydrological cycling in the Arctic.
Acknowledgment
This study was supported in part by JAMSTEC and the Ministry of Education, Science, Sports and Culture, Grant-in-Aid for Scientists (C) No.
26340018.
H. Park et al. / Remote Sensing of Environment 175 (2016) 349–358
Appendix A. Supplementary data
Supplementary data to this article can be found online at http://dx.
doi.org/10.1016/j.rse.2015.12.046.
References
Allan, R.P., & Soden, B.J. (2007). Large discrepancy between observed and simulated precipitation trends in the ascending and descending branches of the tropical circulation.
Geophysical Research Letters, 34, L18705.
Bateni, S.M., Huang, C., Margulis, S.A., Podest, E., & McDonald, K. (2013). Feasibility of
characterizing snowpack and the freeze–thaw state of underlying soil using multifrequency active/passive microwave data. IEEE Transactions on Geoscience and Remote
Sensing, 51(7), 4085–4102.
Beilman, D.W., & Robinson, S.D. (2003). Peatland permafrost thaw and landform type
along a climatic gradient. In P. M, S. SM, & A. LU (Eds.), Proceedings of the Eighth
International Conference on Permafrost, Vol. 1. (pp. 61–65). Zurich, Switzerland: A.A.
Balkema.
Brodzik, M.J., & Knowles, K.W. (2002). Ease-Grid: A versatile set of equal-area projections
and grids. In M. Goodchild (Ed.), Discrete global grids. Santa Barbara, CA: National
Center for Geographic Information and Analysis.
Brown, J., Ferrians, O., Heginbottom, J.A., & Melnikov, E. (2014). Circum-Arctic map of permafrost and ground-ice conditions. Boulder, Colorado USA: National Snow and Ice Data
Center (http://nsidc.org/data/ggd318).
Bulygina, O.N., Razuvaev, V.N., & Korshunova, N.N. (2009). Changes in snow cover over
Northern Eurasia in the last few decades. Environmental Research Letters, 4, 045026.
Burke, E.J., Kankers, R., Jones, C.D., & Wiltshire, A.J. (2013). A retrospective analysis of pan
Arctic permafrost using the JULES land surface model. Climate Dynamics. http://dx.
doi.org/10.1007/s00382-012-1648-x.
Dean, K. G., & Morrissey, L. A. (1988). Detection and identification of Arctic landforms - an
assessment of remotely sensed data. Photogrammetric Engineering and Remote
Sensing, 54(3), 363–371.
Dee, D.P., Uppala, S.M., Simmons, A.J., Berrisford, P., Poli, P., Kobayashi, S., ... Vitart, F.
(2011). The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quarterly Journal of the Royal Meteorological Society, 137, 553–597.
Dobinski, W. (2011). Permafrost. Earth-Science Reviews, 108, 158–169.
Du, J., Kimball, J.S., Azarderakhsh, M., Dunbar, R.S., Moghaddam, M., & McDonald, K.C.
(2014). Classification of Alaska spring thaw characteristics using L-band radar remote
sensing. IEEE Transactions on Geoscience and Remote Sensing, 53(1), 542–556.
Duguay, C.R., Zhang, T., Leverington, D.W., & Romanovsky, V.E. (2005). Satellite remote sensing of permafrost and seasonally frozen ground. In D. CR, & P. A
(Eds.), Remote sensing in northern hydrology: Measuring environmental change.
Geophysical Monograph, Vol. 163. (pp. 91–118). Washington: American Geophysical Union.
Entekhabi, D., Njoku, E.G., ONeill, P.E., Kellogg, K.H., Crow, W.T., Edelstein, W.N., ... van Zyl,
J. (2010). The Soil Moisture Active and Passive (SMAP) mission. Proceedings of the
IEEE, 98(5), 704–716.
Friedl, M.A., Sulla-Menashe, D., Tan, B., Schneider, A., Ramankutty, N., Sibley, A., & Huang,
X. (2010). MODIS Collection 5 global land cover: Algorithm refinements and characterization of new datasets. Remote Sensing of Environment, 114, 168–182.
Gilichinsky, D.A., Barry, R.G., Bykhovets, S.S., Sorokovikov, V.A., Zhang, T., Zudin, S.L., &
Fedorov-Davydov, D.G. (1998). A century of temperature observations of soil climate:
Methods of analysis and long-term trends. In A.G. Lewkowicz, & M. Allard (Eds.), Proc.
of the Seventh Int. Conf. on Permafrost (pp. 313–317). Canada: Cent. d'Etud. Nord.,
Uiniv. Laval, Sainte-Foy, Oue.
Granberg, H.B. (1994). Mapping heat loss zones for permafrost prediction at the northern/alpine limit of the boreal forest using high-resolution C-band SAR. Remote
Sensing of Environment, 50, 280–286.
Gruber, S. (2012). Derivation and analysis of a high-resolution estimate of global permafrost zonation. The Cryosphere, 6, 221–233.
Harlan, R.L., & Nixon, J.F. (1978). Ground thermal regime. In O.B. Andersland, & D.M.
Anderson (Eds.), Geotechnical engineering for cold regions (pp. 103–163). New York:
McGraw-Hill.
Hayes, D.J., Kicklighter, D.W., McGuire, A.D., Chen, M., Zhang, Q., Yuan, F., ... Wullschleger,
S.D. (2014). The impacts of recent permafrost thaw on land–atmosphere greenhouse
gas exchange. Environmental Research Letters, 9, 045005.
Holmes, T.R., De Jeu, R.A.M., Owe, M., & Dolman, A.J. (2009). Land surface temperature
from Ka band (37 GHz) passive microwave observations. Journal of Geophysical
Research, 114, D04113.
Hubbard, S.S., Gangodagamage, C., Dafflon, B., Wainwright, H., Peterson, J., Gusmeroli, A.,
... Wullschleger, S.D. (2013). Quantifying and relating land-surface and subsurface
variability in permafrost environments using LiDAR and subsurface geophysical
datasets. Hydrology Journal, 21, 149–169.
Jorgenson, M.T., Racine, C.H., Walters, J.C., & Osterkamp, T.E. (2001). Permafrost degradation and ecological changes associated with a warming climate in central Alaska.
Climatic Change, 48, 551–579.
Keuper, F., van Bodegom, P.M., Dorrepaal, E., Weedon, J.T., van Hal, J., van Logtestijn, R.S.O.,
& Aerts, R. (2012). A frozen feast: Thawing permafrost increases plant-available nitrogen in subarctic peatlands. Global Change Biology, 18, 1998–2007.
Kim, Y., Kimball, J.S., McDonald, K.C., & Glassy, J. (2011). Developing a global data record of daily landscape freeze/thaw status using satellite passive microwave remote sensing. IEEE Transactions on Geoscience and Remote Sensing, 49(3),
949–960.
357
Kim, Y., Kimball, J.S., Zhang, K., & McDonald, K.C. (2012). Satellite detection of increasing
northern hemisphere non-frozen seasons from 1979 to 2008: Implications for regional vegetation growth. Remote Sensing of Environment, 121, 472–487.
Kim, Y., Kimball, J.S., Glassy, J., & McDonald, K.C. (2014a). MEaSUREs global record of daily
landscape freeze/thaw status. Version 3. [1979–2009]. Boulder, Colorado USA: NASA
DAAC at the National Snow and Ice Data Center (http://dx.doi.org/10.5067/
MEASURES/CRYOSPHERE/nsidc-0477.003).
Kim, Y., Kimball, J.S., Didan, K., & Henebry, G.M. (2014b). Response of vegetation growth
and productivity to spring climate indicators in the conterminous United States derived from satellite remote sensing data fusion. Agricultural and Forest Meteorology,
194, 132–143.
Kimball, J.S., McDonald, K.C., Running, S.W., & Frolking, S.E. (2004). Satellite radar remote
sensing of seasonal growing seasons for boreal and subalpine evergreen forests.
Remote Sensing of Environment, 90, 243–258.
Koven, C.D., Ringeval, B., Friedlingstein, P., Ciais, P., Cadule, P., Khvorostyanov, D., ...
Tarnocai, C. (2011). Permafrost carbon–climate feedbacks accelerate global warming.
PNAS, 108(36), 14769–14774.
Lawrence, D.M., & Slater, A.G. (2005). A projection of severe near-surface permafrost degradation during 21st century. Geophysical Research Letters, 32, L24401.
Lawrence, D.M., Slater, A.G., & Swenson, S.C. (2012). Simulation of present-day and future
permafrost and seasonally frozen ground conditions in CCSM4. Journal of Climate, 25,
2207–2225.
Leverington, D.W., & Duguay, C.R. (1997). A neural network method to determine the
presence or absence of permafrost near Mayo,Yukon Territory, Canada. Permafrost
and Periglacial Processes, 8, 205–215.
Minsley, B.J., Abraham, J.D., Smith, B.D., Cannia, J.C., Voss, C.I., Jorgenson, M.T., ... Ager, T.A.
(2012). Airborne electromagnetic imaging of discontinuous permafrost. Geophysical
Research Letters, 39.
Morrissey, L.A., Strong, L.L., & Card, D.H. (1986). Mapping permafrost in the boreal forest
with Thematic Mapper satellite data. Photogrammetric Engineering and Remote
Sensing, 52, 1513–1520.
Mudryk, L.R., Kushner, P.J., & Derksen, C. (2014). Interpreting observed northern hemisphere snow trends with large ensembles of climate simulations. Climate Dynamics,
43, 345–359.
Naeimi, V., Paulik, C., Bartsch, A., Wagner, W., Kidd, R., Park, S., ... Boike, J. (2012). ASCAT
surface flag (SSF): Extracting information on surface freeze/thaw conditions from
backscatter data using an empirical threshold-analysis algorithm. IEEE Transactions
on Geoscience and Remote Sensing, 50(7), 2566–2582.
Natali, S.M., Schuur, E.A.G., & Rubin, R.L. (2012). Increases plant productivity in Alaska
tundra as a result of experimental warming of soil and permafrost. Journal of
Ecology, 100, 488–498.
Nelson, F.E., & Outcalt, S.I. (1987). A computational method for prediction and regionalization of permafrost. Arctic and Alpine Research, 19(3), 279–288.
Nguyen, T.N., Burn, C.R., King, D.J., & Smith, S.L. (2009). Estimating the extent of nearsurface permafrost using remote sensing, Mackenzie Delta, Northwest Territories.
Permafrost and Periglacial Processes, 20, 141–153.
Oberman, N.G., & Liygin, A.M. (2009). Predicting permafrost degradation: Case study of
North European Russia. Exploration and Protection of Natural Resources, 7, 15–20.
Ohta, T., Kotani, A., Iijima, Y., Maximov, T. C., Ito, S., Hanamura, M., Kononov, A. V., &
Maximov, A. P. (2014). Effects of waterlogging on water and carbon dioxide fluxes
and environmental variables in a Siberian larch forest, 1998–2011. Agricultural and
Forest Meteorology, 188, 64–75.
Olson, D.M., Dinerstein, E., Wikramanayake, E.D., Burgess, N.D., Powell, G.V.N.,
Underwood, E.C., ... Kassem, K.R. (2001). Terrestrial ecoregions of the world: A new
map of life on Earth. Bioscience, 51(11), 933–938.
Osterkamp, T. E. (2007). Characteristics of the recent warming of permafrost in Alaska.
Journal of Geophysical Research: Earth Surface, 112, F02S02.
Panda, S.K., Prakash, A., Jorgenson, M.T., & Solie, D.N. (2012). Near-surface permafrost distribution mapping using logistic regression and remote sensing in interior Alaska.
GIScience and Remote Sensing, 49(3), 346–363.
Park, H., Iijima, Y., Yabuki, H., Ohta, T., Walsh, J., Kodama, Y., & Ohata, T. (2011). The application of a coupled hydrological and biogeochemical model (CHANGE) for modeling
of energy, water, and CO2 exchanges over a latch forest in eastern Siberia. Journal of
Geophysical Research, 116, D15102.
Park, H., Walsh, J., Fedorov, A.N., Sherstiukov, A.B., Iijima, Y., & Ohata, T. (2013). The
influence of climate and hydrological variables on opposite anomaly in activelayer thickness between Eurasian and North American watersheds. The Cryosphere,
7, 631–645.
Park, H., Scherstiukov, A.B., Fedorov, A.N., Polyakov, I.V., & Walsh, J.E. (2014). An
observation-based assessment of the influences of air temperature and snow depth
on soil temperature in Russia. Environmental Research Letters, 9, 064026.
Park, H., Fedorov, A.N., Zheleznyak, M.N., Konstantinov, P.Y., & Walsh, J.E. (2015). Effect of
snow cover on pan-Arctic permafrost thermal regimes. Climate Dynamics, 44(9),
2873–2895.
Payette, S., Delwaide, A., Caccianiga, M., & Beaucjemin, M. (2004). Accelerated thawing of
subarctic peatland permafrost over the last 50 years. Geophysical Research Letters, 31,
L18208.
Podest, E., McDonald, K.C., & Kimball, J.S. (2014). Multisensor microwave sensitivity to
freeze/thaw dynamics across a complex boreal landscape. IEEE Transactions on
Geoscience and Remote Sensing, 52(11), 6818–6828.
Riseborough, D., Shiklomanov, N., Etzelmuller, B., Gruber, S., & Marchenko, S. (2008). Recent advances in permafrost modelling. Permafrost and Periglacial Processes, 19(2),
137–156.
Romanovsky, V.E., Smith, S.L., & Christiansen, H.H. (2010). Permafrost thermal state in the
polar northern hemisphere during the international polar year 2007–2009: A synthesis. Permafrost and Periglacial Processes, 21(2), 106–116.
358
H. Park et al. / Remote Sensing of Environment 175 (2016) 349–358
Saito, K., Sueyoshi, T., Marchenko, S., Romanovsky, V., Otto-Bliesner, B., Walsh, J., ...
Yoshikawa, K. (2013). LGM permafrost distribution: How well can the latest PMIP
multi-model ensembles perform reconstruction? Climate of the Past, 9, 1697–1714.
Strozzi, T., Kaab, A., & Frauenfelder, R. (2004). Detecting and quantifying mountain permafrost creep from in situ inventory, space-borne radar interferometry and airborne
digital photogrammetry. International Journal of Remote Sensing, 25(15), 2919–2931.
Thibault, S., & Payette, S. (2009). Recent permafrost degradation in bogs of the James Bay
Area, northern Quebec, Canada. Permafrost and Periglacial Processes, 20, 383–389.
van Huissteden, J., Berrittella, C., Parmentier, F.J.W., Mi, Y., Maximov, T.C., & Dolman, A.J.
(2011). Methane emission by permafrost thaw lakes limited by lake drainage.
Nature Climate Change, 1, 119–123.
Vaughan, D.G., Comiso, J.C., Allison, I., Carrasco, J., Kaser, G., Kwok, R., ... Zhang, T. (2013).
Observations: Cryosphere. In T.F. Stocker, D. Qin, G.K. Plattner, M. Tignor, S.K. Allen, J.
Boschung, A. Nauels, Y. Xia, V. Bex, & P.M. Midgley (Eds.), Climate Change 2013: The
physical science basis. Contribution of working group I to the fifth assessment report of
the intergovernmental panel on climate change (pp. 317–382). Cambridge, United
Kingdom, and New York, NY, USA: Cambridge University Press.
Velicogna, I., Tong, J., Zhang, T., & Kimball, J.S. (2012). Increasing subsurface water storage
in discontinuous permafrost areas of the Lena River basin, Eurasia, detected from
GRACE. Geophysical Research Letters, 39, L09403.
Wan, Z. (2008). New refinements and validation of the MODIS land-surface temperature/
emissivity products. Remote Sensing of Environment, 112, 59–74.
Weedon, G. P., Gomes, S., Viterbo, P., Shuttleworth, W. J., Blyth, E., Österle, H., Adam, J. C.,
Bellouin, N., Boucher, O., & Best, M. (2011). Creation of the WATCH forcing data and
its use to assess global and regional reference crop evaporation over land during the
twentieth century. Journal of Hydrometeorology, 12, 823–848.
Westermann, S., Wollschlager, U., & Boike, J. (2010). Monitoring of active layer dynamics
at a permafrost site on Svalbard using multi-channel ground-penetrating radar. The
Cryosphere, 4, 475–487.
Westermann, S., Østby, T.I., Gisnås, K., Schuler, T.V., & Etzelmüller, B. (2015). A ground
temperature map of the North Atlantic permafrost region based on remote sensing
and reanalysis data. The Cryosphere, 9, 1303–1319. http://dx.doi.org/10.5194/tc-91303-2015.
Yoshikawa, K., & Hinzman, L.D. (2003). Shrinking thermokarst ponds and groundwater
dynamics in discontinuous permafrost near Council, Alaska. Permafrost and
Periglacial Processes, 14, 151–160.
Zhang, T., Frauenfeld, O.W., Serreze, M.C., Etringer, A., Oelke, C., McCreight, J., ...
Chudinova, S. (2005). Spatial and temporal variability in active layer thickness over
Russian Arctic drainage basin. Journal of Geophysical Research, 110, D16101.
Zhang, Y., Olthof, I., Fraser, R., & Wolfe, S.A. (2014). A new approach to mapping permafrost and climate incorporating uncertainties in ground conditions and climate projections. The Cryosphere, 8, 2177–2194.