Volume 123, Issue 4 p. 1366-1386
Research Article
Open Access

Depth-Resolved Physicochemical Characteristics of Active Layer and Permafrost Soils in an Arctic Polygonal Tundra Region

Yuxin Wu

Corresponding Author

Yuxin Wu

Climate and Ecosystem Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA, USA

Correspondence to: Y. Wu,

[email protected]

Search for more papers by this author
Craig Ulrich

Craig Ulrich

Climate and Ecosystem Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA, USA

Search for more papers by this author
Timothy Kneafsey

Timothy Kneafsey

Energy Geosciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA, USA

Search for more papers by this author
Robin Lopez

Robin Lopez

Climate and Ecosystem Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA, USA

Search for more papers by this author
Chunwei Chou

Chunwei Chou

Climate and Ecosystem Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA, USA

Department of Environmental Science, Policy and Management, University of California Berkeley, Berkeley, CA, USA

Search for more papers by this author
Jil Geller

Jil Geller

Energy Geosciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA, USA

Search for more papers by this author
Katie McKnight

Katie McKnight

College of Environmental Design, University of California, Berkeley, CA, USA

Now at San Francisco Estuary Institute, Richmond, CA, USA

Search for more papers by this author
Baptiste Dafflon

Baptiste Dafflon

Climate and Ecosystem Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA, USA

Search for more papers by this author
Florian Soom

Florian Soom

Climate and Ecosystem Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA, USA

Search for more papers by this author
John Peterson

John Peterson

Climate and Ecosystem Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA, USA

Search for more papers by this author
Susan Hubbard

Susan Hubbard

Climate and Ecosystem Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA, USA

Search for more papers by this author
First published: 30 March 2018
Citations: 5
This article has been contributed to by US Government employees and their work is in the public domain in the USA.


Permafrost physicochemical parameters play a key role in controlling the response of permafrost carbon to climate change. We studied the physicochemical parameters of permafrost in an Arctic tundra region to evaluate (1) how soil parameters vary with depth and whether and how they are interrelated, (2) whether and how permafrost soil differs from its overlaying active layer, and (3) whether soil property-depth relationships are different across geomorphic features (e.g., low, flat, and high centered polygons). We also explored the possible biogeochemical processes that led to these soil characteristics and how they may affect biogeochemical reactions upon permafrost thaw. We observed (1) consistent relationships between soil property and depth and between major parameters, (2) large contrasts of key soil parameters between active layer and permafrost, indicative of potentially different response of the permafrost carbon to warming when compared to the active layer, and (3) a correlation between soil hydraulic conductivity and topographic features that impacts soil hydrologic processes. Our analysis suggests that the permafrost has a marine-derived chemical signature that differs from the active layer and shapes the physicochemical fingerprints of the different geomorphic features. Specifically, we revealed the unique signatures of the high center polygons, indicative of possible microbial activity at depth (>1 m). Our study suggested consistent key soil parameter-depth correlations while demonstrating complex lateral and vertical variabilities. These results are valuable for identifying approaches to upscale point-based measurements and for improving model parameterization to predict permafrost carbon behavior and feedback under future climate.

Key Points

  • Key soil parameters show consistent changes with depth across polygon types with strong interparameter correlations
  • Soil parameters show large contrasts between active layer and permafrost, suggesting differential biogeochemical behaviors
  • Unique biogeochemical signature of high center polygons was revealed, indicative of microbial activity in the permafrost at depth (>1 m)

1 Introduction

The 22.8 million km2 of permafrost on Earth stores 1,307 Pg carbon and is vulnerable to degradation under a warmer climate (Hugelius et al., 2014; Zhang et al., 2008). The 2°C climate warming target set at the Paris climate conference (COP21) could result in ~40% loss of the permafrost worldwide (Chadburn et al., 2017). In the Arctic region, two thirds of the soil organic carbon (SOC) is stored in the perennially frozen permafrost as “old carbon” (Hugelius et al., 2014; Schuur et al., 2009; Tarnocai et al., 2009). There are significant uncertainties associated with predicting how permafrost thaw will alter the fate of this vast amount of stored carbon (Schuur et al., 2009). These uncertainties are propagated from the uncertainties of a large number of parameters and processes. These include, for example, the uncertainty of the stock of SOC and its decomposition rate and pathways; the uncertainty of vertical depth profile of SOC and its variability among different soil pedons; the complexity of geomorphological and hydrological properties of the tundra soils and its dynamics under a warmer climate; and our limited knowledge of the physicochemical and microbiological parameters of the permafrost soils, their effects on carbon degradation, and responses to future climate change.

A large amount of research has been conducted over the last few decades to survey and characterize permafrost physical, chemical, and biological properties and to understand its geomorphological, hydrological, and biogeochemical processes under a warmer climate (see Ping et al., 2015, for a recent review). A large fraction of these researches focused on mapping permafrost spatial distribution and thermal states (Brown et al., 1997; Olefeldt et al., 2016; Osterkamp & Romanovsky, 1999; Romanovsky, 2002), as well as the characteristics of the soil carbon in the permafrost, such as its inventory, depth distribution, and decomposability (Bockheim & Hinkel, 2007; Bockheim et al., 1999; Hugelius et al., 2010, 2013, 2014; Michaelson et al., 1996, 2008, 2013; Ping et al., 1998, 2008; Strauss et al., 2012, 2013; Tarnocai et al., 2009; Zhang et al., 2008). Specifically, through decades of research, the uncertainty associated with the total SOC stored in the permafrost region has been reduced due to increasingly available data across a wide range of soil pedons and improved models, and the total estimated volume has increased dramatically (Amundson, 2001; Eswaran et al., 1993; Hugelius et al., 2010, 2012, 2014; Jobbágy & Jackson, 2000; Ping et al., 2008; Rubey, 1951; Schlesinger, 1977; Tarnocai, 1998; Tarnocai et al., 2009). For example, the estimated total stored carbon at 0–3 m for the northern circumpolar permafrost region has increased from 294 Pg (Jobbágy & Jackson, 2000) to 1,024 Pg (Tarnocai et al., 2009) and recently to 1,307 Pg (Hugelius et al., 2014).

Depth-resolved characterization of permafrost could significantly improve climate modeling because climate modeling usually does not consider the vertical variability of soil properties. For example, by considering key permafrost parameters, such as depth distribution of soil organic matter (SOM) (Harden et al., 2012; Koven et al., 2009, 2011; Schaefer et al., 2011), cryoturbation, and thermal insulation effects (Koven et al., 2009), significant changes in the carbon feedback to climate in the Arctic region are predicted. In another example, accounting for the permafrost carbon release can change the Arctic region from a net carbon sink to a source by mid-2020s and offset 42–88% of the total global land sink (Schaefer et al., 2011). The depth distribution of permafrost soil carbon, its quality, and decomposition rate can account for ~50% of the uncertainty in future global temperature prediction (Burke et al., 2012). Continuing improvements in the estimation of the SOC stocks in the permafrost region and its fate under warmer climate demand continuing, fine-scale, and depth-resolved characterization of permafrost soil across different regions and soil types.

In addition to its quantity, the quality (or decomposability) of the permafrost carbon has been another important factor contributing to the climate prediction uncertainty. Over the millennial time scale, low-temperature and anoxic conditions in the permafrost region dramatically constrain SOC decomposition rate (Davidson & Janssens, 2006; Kaiser et al., 2007; Ping et al., 2010, 2015; Rodionov et al., 2007). In general, carbon stored in syngenetic permafrost (permafrost formed more or less concurrently with soil deposition) is relatively undecomposed, and rich in polysaccharides and proteins, thus more vulnerable to decomposition after thaw, when compared to those stored in epigenetic permafrost (permafrost formed after deposition of soil material; Ping et al., 2015). After burial, other key parameters that impact SOC decomposition include, for example, soil thermal and hydrological conditions, organic-mineral association (Davidson & Janssens, 2006; Diochon et al., 2013; Höfle et al., 2013), and pore water chemistry, such as pH (Grosse et al., 2011).

Permafrost ice, organic matter, and mineral contents significantly influence its thermal, hydrological, and mechanical responses to warming, thaw and deformation rate, and water infiltration and drainage. The abundance (or absence) of redox sensitive species, such as iron (Fe2+/3+), is tightly linked with the fate of permafrost carbon. For example, bonding or coagulation of Fe with organic carbon can strongly affect its degradation resilience (Herndon, Mann, et al., 2015; Herndon, Yang, et al., 2015; Herndon et al., 2017; Lalonde et al., 2012; Riedel et al., 2013). As a thermodynamically more favorable process, microbial iron reduction can also directly inhibit methanogenesis under anoxic conditions, a condition that is common in the northern permafrost of the Arctic tundra (Herndon, Mann, et al., 2015; Herndon, Yang, et al., 2015; Lipson et al., 2010; Miller et al., 2015). It is worth noting that despite its suppressive effects, concurrent iron reduction and methanogenesis have been observed on the permafrost coastal tundra as well (Herndon, Mann, et al., 2015; Herndon, Yang, et al., 2015; Roy Chowdhury et al., 2015).

The potentially significant contrast in the thermochemical, hydrochemical, and biogeochemical conditions between the active layer and permafrost can result in large variations of the decomposition rate and pathways of the SOC stored in these two different pools. While many recent studies of SOC stocks in permafrost have extended down to 3 m and have revealed a significant fraction of the stored SOC in the deep permafrost, characterization of other key soil parameters, such as permafrost geochemical and hydrological properties, has been largely focused on the shallow active layer (~ < 1 m) that experiences seasonable freeze-thaw cycles (Hinkel & Nelson, 2003; Mishra & Riley, 2014; Newman et al., 2015; Ping et al., 1998; Szymanski et al., 2016). Depth-resolved, fine-scale characterization of the Arctic soil geochemical and hydrological properties covering the deeper permafrost, and their implications for biogeochemical processes upon permafrost thaw, needs significant improvements. Specifically, comparison of the key soil parameters between the permafrost and the active layer can help understand how permafrost may respond differently to future warming when compared to the active layer, thus reducing the prediction uncertainty of the Earth system models.

In this research, we studied permafrost cores, some down to ~4 m below ground surface (bgs), collected at the Next Generation Ecosystem Experiments-Arctic (NGEE-Arctic) site within the Barrow Environmental Observatory (BEO, 71.29°N, 156.61°W), on the north slope of the Arctic coastal region. We investigated the depth-resolved tundra soil physicochemical characteristics at fine scales and explored the following questions: (1) what are the soil parameters and depth relationships and weather and how the key soil parameters are related to each other; (2) whether and how the permafrost soil differs from its overlaying active layer; (3) whether the geomorphic features (i.e., polygon types) have a dominant impact on the physicochemical parameters of the permafrost soils; and lastly, what are the possible biogeochemical processes that led to these characteristics and what are their biogeochemical implications under future climate conditions.

2 Study Site

This study used cores from the NGEE-Arctic site located on the Arctic Coastal Plain near Barrow, Alaska, and dominated by interlake polygonal ground, featuring low, flat, and high centered polygons (LCPs, FCPs, and HCPs), each at a different stage during the growth or decay of ice wedges (Black, 1976; Jorgenson et al., 2006; Lachenbruch, 1963; Ping et al., 1998, 2014; Wainwright et al., 2015; Zubrzycki et al., 2013; Figure 1). The marine deposit that underlies the tundra soil was formed by repeated ocean transgressions during the late Neogene (Pliocene) to Quaternary period (Brigham-Grette & Carter, 1992; Sellmann & Brown, 1973). The maturation of the tundra soil in this coastal, lowland region occurred under poor drainage conditions (Tedrow & Cantlon, 1958), resulting in the preservation of its marine characteristics at depth where hydrological exchange with freshwater precipitation at the surface is minimal.

Details are in the caption following the image
(a) The locations of the soil cores used in this study from HCPs (green), LCPs (black), and FCPs (red) and those used for hydraulic conductivity (K) measurements (pink), annotated on a map showing general microtopography. (b) Cartoon and example of polygon cross section and example aerial image (30-m side length) of HCP, FCP, and LCP (modified from Wainwright et al., 2015).

While recent studies have identified partially unfrozen soil below the permafrost table (Dafflon et al., 2016; Dou et al., 2016; Hubbard et al., 2013), the deeper permafrost that reaches an almost constant soil temperature around −9°C at 16 m bgs can be considered as continuous at the site (Jogenson et al., 2008). A shallow active layer with a thickness up to ~50 cm (Gangodagamage et al., 2014; Hubbard et al., 2013) rests on top of the permafrost and thaws seasonally. As the dominant topographic features, the different polygon types are the major organizing units in terms of their distinct hydrological features (Liljedahl et al., 2016) and soil physical and biogeochemical characteristics (Biasi et al., 2005; Gersper et al., 1980; Hubbard et al., 2013; Newman et al., 2015; Wainwright et al., 2015).

3 Materials and Methods

The frozen soil cores analyzed in this study were collected with a 7.62-cm inner diameter SIPRE soil corer during the late winter seasons (April–May) from 2013 to 2016. The corer was advanced with a hydraulic driven rotary coring platform (Big Beaver, Little Beaver Inc.) or manually with a gas-powered motor on a 4-m tripod auger. After retrieval, the cores were packed with dry ice in coolers and transported to the lab where they were kept in −25°C freezers for storage until analysis. Fifteen soil cores, five from each polygon type, with a maximal depth down to ~4 m bgs were analyzed for soil physicochemical parameters with additional six cores for hydraulic conductivity measurements (Table 1). The locations of the cores are shown in Figure 1a.

Table 1. Characteristics and Geomorphic Locations of the Cores Used in This Study for Physicochemical and Hydrological Measurements
Physicochemical analysis
Core name Polygon type Polygon features No. of depth intervals Core depth (cm)
AB434 LCP Center 8 53
AB446 LCP Center 5 180
S0-411 LCP Center 6 60
GR1 LCP Center 8 168
GR3 LCP Center 8 168
AB159B FCP Center 10 242
S0-305 FCP Center 8 148
S0-154B FCP Center 21 264
S0-146 FCP Center 10 262
ERT4 FCP Rim 4 26
AB136 HCP Center 10 384
L1 HCP Trough 4 255
GR6-DUP HCP Center 6 197
L2 HCP Center 7 256
ERT2 HCP Center 3 21
Hydraulic Conductivity Measurements
Core name Polygon type Polygon features No. of samples Sample depths (cm bgs)
AB117 HCP Trough 3 5.5–13; 13–24; 46–56
BD01 HCP Center 2 28–37; 37–42
DTLB40 LCP Center 2 34–46; 47–54
BD06 LCP Center 2 10–23; 23–27
DTLB19 LCP Rim 2 4–19; 20–37
DTLB38 LCP Rim 1 44–62.5
  • Note. The number of depth intervals indicates the total number of subsamples from each core.

X-ray Computed Tomography (CT) scans were performed on all frozen cores as the first step of analysis using a modified third-generation medical scanner (General Electric Lightspeed 16). CT scans can be used to visualize permafrost soil structures (Orsi et al., 1996; Taina et al., 2008). For example, the soil bulk density (BD) can be estimated using the CT data based on a calibration curve made from scanning known-density materials. The CT scans were used jointly with visual inspection as our primary means to identify the soil structure (e.g., active layer and permafrost boundary) and to select subsampling depths along the cores.

Following CT scanning and determination of the subsampling locations, each core was sliced into subsections at the selected depths with a rotary saw in a cold room (T < −18°C) to keep the soil frozen during sampling. Depending on core availability, a core subsection ranging from 2 to 10 cm in thickness was collected from each depth. In total, 118 subsamples were collected from these cores (Table 1). Each sample was subsequently measured for its weight and geometry (i.e., length and diameter) to calculate its wet bulk density (BDwet). The Archimedean immersion method was also used to determine the soil bulk density. Due to the imperfect shapes of many samples, the Archimedean immersion method was deemed more accurate and subsequently used as the primary method to determine BDwet. For this method, precalibrated thin and flexible vacuum bags were used to contain the core samples. The bags were vacuumed to remove trapped air before immersion into a large glass beaker filled with prechilled water (~1°C) on a balance to measure its water displacement (i.e., sample volume).

Following BDwet measurements, the frozen samples were transferred into Nitrogen (N2)-filled anaerobic bags to thaw for the next 3–4 hr. After thawing, macrorhizon samplers (pore size 0.15 μm, Rhizosphere Research Products Inc.) were used to extract soil pore water. Depending on the volume and water content of the soil sample, ~1–10 hr were required to extract 2–8 ml of the pore water from each sample. The extracted pore water was subsequently divided into multiple aliquots for analysis of its major ion composition with Ion Chromatography (IC, Dionex ICS-2100, Thermo Scientific) and Inductively Coupled Plasma-Mass Spectrometry (ICP-MS, Elan DRC II, Perkin Elmer). After pore water extraction, the thawed soil samples were oven dried at 60°C for multiple 24-hr periods until no further weight loss to acquire their dry bulk density (BDdry). The 60°C drying was adequate for our samples because the major composition of our samples were ice, OM, and sand with a minimal amount of clay. In addition, further drying at 105°C of selected samples did not yield further weight loss. After BDdry measurement, each soil sample was divided into multiple subsamples (using the coning-and-quartering method; IUPAC, 1997) for analysis of its particle size distribution, organic matter content, and carbon/nitrogen content, using laser diffraction (Mastersizer 3000, Malvern), loss on ignition (LOI) after 500°C ashing, and combustion-based elemental analysis (Flash 2000, Thermo Scientific) methods, respectively. The SOM volumetric content (referred later as SOM content) was calculated assuming an organic matrix density of 1.3–1.4 g/cm3 (Farouki, 1981). Similarly, the mineral and ice volumetric content was calculated using matrix densities of 2.64 g/cm3 and 0.92 g/cm3, respectively, based on the literature (e.g., Farouki, 1981).

In addition to the 118 subsamples analyzed following the procedure described above, another 12 samples from six additional HCP and LCP cores (Figure 1a and Table 1) were used for hydraulic conductivity measurements based on the falling head method. These samples include those that are organic rich, mineral rich, or are a mixture of mineral and ice based on their CT scan characteristics. In brief, the frozen soil core was inserted into a latex sleeve slightly stretched to provide a small (<1-psi) confining pressure on the core to minimize wall effects. The jacketed core was fitted with end caps, flow tubing, and positioned into a water-filled permeameter for subsequent thaw and hydraulic conductivity measurements. Due to the variable mechanical strength, structure, and ice content of the soils, core deformation occurred upon thawing, ranging from minor (<2%) to significant (>30% in length or diameter). After hydraulic conductivity measurements, these samples were dried and analyzed for their particle size distributions.

Correlation/regression analysis and Analysis of Variance (ANOVA) were the statistical tools used to explore the correlations between the key soil parameters, the significance of variances of these parameters between the different geomorphic features (low, flat, and high centered polygons), and between the active layer and permafrost. In addition, Principal Component Analysis was used to identify the primary components that represent the largest variances among all the parameters and visualize the key parameter variances represented in the primary components.

4 Results

4.1 Soil Structural Variability Across Polygon Types

Large variabilities of the soil structure within and across polygon types were revealed by the CT images (Figure 2). The relative distribution of the soil mineral (indicated by the red colors), ice, and SOM (white and blue) between the polygon types is not consistent. While the LCP cores generally display a layered structure in the active layer (<50 cm) with an organic-rich layer underlain by a mineral rich layer (Figure 2a), such a structure was generally not observed in permafrost or FCPs and HCPs (Figures 2b and 2c). For some LCP cores, a secondary organic-rich layer (up to 10-cm thick) was occasionally observed just above the permafrost table. A consistent pattern of this secondary organic-rich layer was not identified due to the limited number of cores analyzed. Across all the cores analyzed, the presence and structure of the ice vary significantly with, for example, reticulate ice veins observed in some cores, but absent in others. The significant variability of the core structures—with depth and as a function of polygon types—is challenging to represent in the models.

Details are in the caption following the image
Examples of the CT images (approximately the top 80 cm below ground surface) from (a) LCPs, (b) FCPs, and (c) HCPs. Higher densities in red represent mineral-rich zones, while lower densities in white and blue represent ice- and SOM-rich layers. Dark blue represents air-filled gaps in the cores.

4.2 Depth Profiles of Soil Physicochemical Parameters Across Polygon Types

The depth profiles of the key soil physicochemical parameters revealed their consistent correlations with depth with variabilities (Figure 3). Specifically, the soil wet bulk density and dry bulk density (BDwet and BDdry) show a generally increasing trend with depth across the polygon types with large variabilities at a given depth and polygon type. The BDwet of the active layer, roughly at <0.5 m bgs, ranges from 0.7 to 1.5 g/cm3, which increases to >1.5 g/cm3 at depths >2.5 m bgs. Similarly, BDdry varies between 0 and 1 g/cm3 in the top 0.5 m and increases to >1 g/cm3 at depths >2.5 m bgs.

Details are in the caption following the image
Depth profiles of the key soil physicochemical parameters. The units for porosity, SOM, mineral, and ice contents are percentage of the total volume of the soil. The unit for BDwet and BDdry is g/cm3. The ion concentrations are plotted on log scales, with the unit mg/L.

Unlike the bulk densities, the soil OM content decreases sharply with depth, roughly following an inversed power law correlation (R2 = 0.44), with the volumetric OM content ranges from 5 to 20% in the top 0.5 m, decreasing to <5% at depths below 2 m. Based on the regressed correlation between SOM and depth for the top 4 m of the soil, we calculated the accumulative, volumetric SOM content with depth, where ~35% of the total SOM is located in the top 0.5 m (approximately active layer), with another 35% contained in the next meter (0.5–1.5 m bgs), and ~30% from 1.5 to 4 m bgs.

The soil mineral content shows a clear increase with depth, similar to the trend of the soil wet and dry bulk densities, with 0–20% by volume at the shallowest 0.5 m to mostly >30% at 2.5 m and deeper. Unlike SOM and mineral contents that display a clear trend with depth, the ice content of the soils varied widely at a given depth; therefore, an obvious correlation between ice content and depth is not as clear. In general, the soil ice content varies between 40 and 90%, averaging ~65% along the whole depth profile from soil surface down to >3 m bgs. There are no distinguishable differences between the three polygon types in terms of these physical parameters. This is supported further by ANOVA analysis that will be discussed later.

Similar to the soil physical properties, the major ion concentrations in the extracted pore water also display consistent but different soil property-depth relationships. Sodium (Na+) concentration shows a sharp increase with depth for all three polygon types, with a larger than 2 orders of magnitude increase from the shallow active layer soils to the permafrost at 4 m bgs (Figure 3). Potassium (K+), magnesium (Mg2+), and calcium (Ca2+) depth profiles are similar to that of Na+, showing ~2 orders of magnitude increase with depth. The pore water strontium (Sr2+) concentration is generally small, showing an increasing trend with depth. Unlike the cations shown above, the pore water aluminum (Al3+) concentrations show a decreasing trend across all three polygon types with the highest concentration averaging at ~1 mg/L at the shallow depth. Except a few outliers, manganese (Mn2+) concentrations showed no significant changes with depth and average at 0.1 mg/L. Similar to the soil physical characteristics, the interpolygon-type variability of these cation species is negligible.

Unlike the cation species discussed above with uniform trends regardless of the polygon types, variations in total dissolved iron are observed as a function of polygon types. That is, while strong decreasing trends with depth are observed for LCP and FCP samples, the dissolved iron concentrations in HCP samples do not display an increasing or decreasing trend with depth, with an averaged concentration at ~5 mg/L at all depths. For LCP and FCP, the Fe concentrations average at >10 mg/L at shallow depths and decrease to <1 mg/L at >1 m bgs.

When comparing to the cations, the pore water major anions display similar variabilities among the different species, showing different trends of change with depth. Specifically, the changes of sulfate (SO42−) and chloride (Cl) concentrations are very similar to that of Na+ and K+, showing a significant increase with depth by roughly 2 orders of magnitude (Figure 3). The active layer sulfate concentrations average at <10 mg/L, which increase to >100 mg/L at >2 m depth. Chloride concentrations increase from ~30 mg/L in the active layer to ~2,000 mg/L at depth >2 m. The pore water nitrate (NO3) concentration behaves differently, showing different correlations with depth among the different polygon types (Figure 3). Namely, a sharp decreasing with depth is apparent for HCP, with no observable correlations with depth for LCP and FCP cores. It is interesting to note that when compared to LCPs and FCPs, HCPs display an anomalous behavior for both Fe2+/3+ and NO3, the two species with significant biogeochemical implications.

Unlike the major cations and anions discussed above, pore water nitrite (NO2), phosphate (PO43−), fluoride (F), and bromide (Br) concentrations do not display correlations with depth and display large variations at any given depth.

4.3 Interparameter Correlations Across Depths and Polygon Types

Results from correlation analysis of key soil physicochemical parameters revealed consistent patterns across depths and polygon types (Figure 4). For instance, BDwet is often used as a base parameter for correlation study. An excellent correlation exists between BDwet and BDdry (Figure 5), indicating an approach to estimate the soil dry mass based on wet bulk density measurements that can be collected nondestructively using CT scans. Note that the correlation between BDwet and BDdry will be likely weaker when larger variabilities of ice structure, segregation, and distribution are present.

Details are in the caption following the image
Correlation matrix for the key soil physical and chemical parameters across all depths and polygon types. The unit for SOM, mineral, and ice contents is percentage of the total volume of the soil. The unit for BDwet and BDdry is g/cm3. The ion concentrations are plotted on log scales, with the unit mg/L.
Details are in the caption following the image
The correlation between the soil wet bulk density (BDwet) and the dry bulk density (BDdry) across all the depths and polygon types. HCP: high centered polygon, FCP: flat centered polygon, LCP: low centered polygon.

One of the key soil parameters important for the understanding and prediction of climate change feedbacks from permafrost thaw is the SOM content. Our correlation analysis shows that while BDwet shows a fair correlation with the weight percentage of the SOM in the dried soil (R2 = 0.56) (Figure 6), similar to previous results (Bockheim et al., 2001), the volumetric content of SOM in the wet soil does not correlate well with BDwet (Figure 4). Contrary to SOM, the soil mineral content shows an excellent correlation (R2 = 0.94) with BDwet, indicating the dominant contribution of the soil mineral content to the bulk density. Figure 4 reveals that the correlation between BDwet and the soil ice content is also good.

Details are in the caption following the image
The correlation between the wet bulk density (BDwet) and the gravimetric organic matter content (percent weight) of the dried soil samples across all depths and polygon types.

In addition to the correlations between the soil physical parameters, our analysis also revealed correlations between the soil physical and chemical parameters, specifically between SOM content and some key ion species important for biogeochemical reactions (Figure 4). A positive correlation between SOM and the total dissolved iron revealed a significant increase of Fe concentration by more than 3 orders of magnitude when SOM content increases from 0% to 20%. Such a positive correlation is also observed for NO3 with 2 orders of magnitude increase with increasing SOM content from 0% to 20% (Figure 4). In addition to Fe2+/3+ and NO3, a negative correlation was observed between SOM content and SO42− and K+ concentrations in the pore water.

Correlations between the different ions are also observed across polygon types. For example, Na+, K+, Mg2+, Cl, and SO42− are all positively related to each other, suggesting the similar origin of the pore water across the site.

In order to better visualize the correlations between the multiple parameters discussed above, parameter correlation monoplot based on the results of the Principal Component Analysis was constructed (Figure 7). While the two most significant principle components (PC1 and PC2) represent only 72% of the total variances among all the parameters, interparameter correlations are clearly demonstrated. For example, the close approximation between SOM and Fe2+/3+ concentration (and between depth and the major ions) indicates their positive correlation, while the near-reversal approximation between depth and OM content indicates their negative correlation. The near-normal angle between depth and the ice content suggests that they are not well correlated with each other, as discussed above. The variances of NO3 and Mn2+ among the samples are not well captured by PC1 and PC2, indicated by their shorter vector length from the origin (Figure 7).

Details are in the caption following the image
The correlation monoplot between the key soil physicochemical parameters in relation to the first two principle components (PC1 and PC2) based on the Principal Component Analysis.

LOI has been used to estimate organic can carbonate content in sediments (Heiri et al., 2001). LOI-based SOM content measurements on our permafrost samples also show a significant correlation with total soil organic carbon (R2 = 0.96) and nitrogen (R2 = 0.94; Figure 8a). A roughly 2:1 ratio was observed between SOM and SOC and 40:1 ratio between SOM and SON (Figure 8a) regardless of depths or polygon types. These excellent correlations provide a model for estimating SOC and SON contents based on LOI measurements that are relatively straightforward to acquire. Further, an inspection of the soil carbon:nitrogen ratio (C/N) revealed a consistent decrease with depth from 22:1 at near surface to 11:1 at depth >2 m bgs (Figure 8b).

Details are in the caption following the image
(a) Correlation between loss on ignition (LOI)-based SOM measurements and soil organic carbon and nitrogen composition and (b) changes of soil carbon/nitrogen ratio (C/N) with depth across all measured samples.

4.4 Interpolygon-Type Variances and Active Layer-Permafrost Comparison

ANOVA analysis conducted on the samples between the different polygon types and cross the active layer-permafrost boundary revealed significant differences in most of the key parameters between the active layer and permafrost, indicated by the small p values (Table 2). On the other hand, the differences among the polygon types within the active layer or the permafrost are relatively small, indicated by large p values.

Table 2. ANOVA Results (p Values) Indicating the Magnitude of Differences for the Major Soil Physiochemical Parameters Between the Different Polygon Types in the Active Layer and the Permaforst and Between the Active Layer and Permafrost Across All the Polygon Types
Parameters Active layer Permafrost All polygon
LCP versus FCP versus HCP Significance LCP versus FCP versus HCP Significance AL VS PF Significance
Mineral content 3.25E−01 ns 6.16E−02 * 1.12E−04 ****
Ice content 3.76E−01 ns 6.25E−01 ns 4.91E−01 ns
OM content 1.59E−01 ns 5.18E−02 * 8.00E−12 ****
Porosity 2.24E−01 ns 3.46E−02 ** 7.96E−02 *
[Na] 8.64E−01 ns 2.23E−02 ** 2.59E−04 ****
[Fe] 2.50E−02 ** 9.22E−02 * 1.63E−04 ****
[K] 3.98E−02 ** 7.17E−02 * 3.73E−05 ****
[Al] 5.83E−01 ns 1.65E−03 *** 1.76E−08 ****
[Ca] 1.36E−01 ns 1.15E−01 ns 1.26E−09 ****
[Si] 1.40E−01 ns 3.58E−02 ** 7.65E−05 ****
[Mn] 1.85E−01 ns 6.68E−04 **** 1.30E−02 ***
[Mg] 9.46E−01 ns 3.01E−02 ** 1.75E−03 ***
[Sr] 1.26E−01 ns 4.88E−02 ** 1.35E−03 ***
[Zn] 2.27E−02 ** 1.60E−01 ns 2.27E−04 ****
[Ni] 5.80E−01 ns 3.00E−01 ns 4.01E−05 ****
[Cl] 1.03E−01 ns 1.11E−01 ns 6.16E−04 ****
[SO4] 2.03E−02 ** 5.76E−02 * 1.72E−03 ***
[NO3] 1.88E−02 ** 1.53E−02 ** 8.05E−02 *
[PO4] 4.31E−01 ns 7.59E−01 ns 6.27E−02 *
[NO2] 4.10E−01 ns 4.38E−01 ns 3.05E−01 ns
[F] 2.21E−01 ns 3.05E−01 ns 5.29E−01 ns
[Br] 2.53E−01 ns 1.90E−01 ns 1.73E−03 ***
  • Note. The asterisks indicate level of significance: p < 0.1 (*), p < 0.05 (**), p < 0.01(***), p < 0.001 (****), and not significant (ns).

Figure 9 shows the comparison of the key physicochemical parameters of the soil cores between the active layer and the permafrost regardless of the polygon types. Consistent with the observations from the depth profiles, on average, only a small difference in the soil volumetric ice content exists between the active layer and the permafrost samples regardless of the polygon types. Unlike ice content, large differences in SOM and mineral contents between the active layer and the permafrost are shown. Namely, the volumetric SOM content in the active layer averages at >10%, while it is only <4% for the permafrost, and the active layer volumetric mineral content averages ~10%, while it is ~20% on average for the permafrost. Geochemically, the pore water Na+ and K+ concentrations are, on average, more than 1 order of magnitude higher in the permafrost than in the active layer. This trend is reversed for Fe2+/3+ and Si4+. The pore water Ca2+ and Mg2+ concentration contrast between the active layer and permafrost is similar to Na+ and K+. The Ni2+ and Mn2+ concentrations in the pore water are generally low, with a higher concentration in the permafrost than the active layer.

Details are in the caption following the image
Comparison between the active layer (AL) and the permafrost (PF) for their key soil physicochemical parameters regardless of the polygon types. For each plot, the line inside the box indicates the median value and the cross represents the mean value. The bottom and top parts of the box indicate the second and third quartiles. The whiskers below and above the boxes represent the first and fourth quartiles of the data. The data outliers are shown as individual dots.

Among the major anions, the NO3 concentration is similar between the active layer and the permafrost, averaging at ~0.5 mg/L. This is similar for NO2 (1–2 mg/L for both active layer and permafrost). Consistent with the depth profile data, a much higher chloride and sulfate concentration is observed in the permafrost than in the active layer. In general, a higher averaged PO43+ concentration is observed in the active layer than the permafrost while the number of data points is limited.

Different from most of the key soil parameters, interpolygon-type comparison shows that Fe2+/3+ and NO3 concentrations are among a few parameters that displayed notable differences between the different polygon types in both the active layer and the permafrost (Table 2). As shown in Figure 3, these differences are mainly caused by the deviation of the HCP values from the other two polygon types. In addition to Fe2+/3+ and NO3 concentrations, our results also showed a significantly lower OM content in the active layer of the HCPs (averaging ~6% by volume) when comparing to the FCPs (~12%) and LCPs (14%), while their volumetric contents in the deeper permafrost are similar (averaging 2–5% across polygon types).

4.5 Soil Composition and Microtopographic Control on Hydraulic Conductivity

Hydraulic conductivity measurements on permafrost soils with large ice contents are challenging and often result in conductivity values ranging over a few orders of magnitudes in addition to large uncertainties. The significant ice content in the permafrost soils often results in deformation of the soil after thaw as observed in our study, making it challenging to interpret the results as well.

The saturated hydraulic conductivities of the selected permafrost cores show large differences between the different soil compositions and sensitivity to microtopographic features. In general, the organic-rich soils consistently have the highest hydraulic conductivity (averaging at 7.6E−4 cm/s) when compared to the mineral-rich soils (averaging at 8.29E−05 cm/s) or soils dominated by a mixture of mineral and ice (averaging at 3.54E−05 cm/s) by an order of magnitude (Table 3).

Table 3. The Saturated Hydraulic Conductivity of the Individual Core Samples and Their Averages for Each Soil Composition Group
Core composition Hydraulic conductivity (cm/s) SD (%)
Mineral rich 1.24E−06 19.5
Mineral rich 1.41E−05 10.2
Mineral rich 5.14E−05 31.2
Mineral rich 2.65E−04 19.6
Average 8.29E−05
Mineral + ice 3.07E−07 6.4
Mineral + ice 8.95E−05 37
Mineral + ice 9.30E−06 43.5
Mineral + ice 1.19E−05 47.3
Mineral + ice 6.60E−05 97.8
Average 3.54E−05
Organic rich 8.83E−04 9
Organic rich 6.69E−04 21.6
Organic rich 7.32E−04 6.4
Average 7.61E−04
  • Note. The standard deviation (SD) incuates the variation between the repreated measurements (at least 3 times) for each sample.

In addition to the different soil compositions, the microtopographic features, for example, center, rim, and trough, seem to have a large impact on the soil-saturated hydraulic conductivity. Figure 10 shows that the average hydraulic conductivity from trough to rim then center of the polygons increases consistently by roughly an order of magnitude in each step. The particle size distribution data revealed a higher percentage of sand (67%), and also a higher averaged sand particle size, in the samples from polygon center when compared to the trough and rim samples (60% and 62% sand contents, respectively).

Details are in the caption following the image
(a) The averaged particle size distribution and (b) the averaged hydraulic conductivities for soils from the three microtopographic features (trough, rim, and center). Standard deviations of the hydraulic conductivity are shown as whiskers.

5 Discussion

Despite the large variability of the soil structural and physicochemical characteristics across depth and polygon types, the results presented above highlight multiple consistent patterns and trends. These results are useful for upscaling fine-scale, point-based measurements to the larger, complex polygonal ground at our study site and beyond for model parameterization. We discuss the possible origins of these variabilities, patterns, and trends and their physical, hydrological, and biogeochemical implications under future climate.

5.1 Permafrost Origin and Structural Variability

The marine origin of the permafrost soil at our study site is supported by the depth profiles of the major ions, such as Na+, K+, SO42−, and Cl, which display a consistent and significant increase with depth (Figure 3). Specifically, at 3 m bgs or deeper, the major ion concentrations are comparable to those of the modern seawater. While studies have indicated the changes of the seawater composition over the last l00 million years (Coggon et al., 2010), no evidence suggested significant shifts of the major ion concentrations during the last 3 million years (Pliocene to Quaternary) when the sediments at our site were formed. Nevertheless, the marine depositional history and impact on pore water chemistry at depth are evident and have a profound impact on the physicochemical and mechanical characteristics of the permafrost soils at our study site. For example, Hubbard et al. (2013) and Dafflon et al. (2016) revealed a large subsurface zone characterized by extremely low electrical resistivity, suggesting partially unfrozen sediments. The presence of partially unfrozen sediments is supported by the high concentrations of the major ions shown in this study, resulting in freezing point depression that prevents the sediments from freezing completely at temperatures below 0°C (Banin & Anderson, 1974; Marion, 1995). This partially unfrozen, saline soil may be as shallow as 2 m bgs based on our core analysis, which has significant geomechanical and biogeochemical implications. Specifically, in addition to the significantly compromised mechanical strength of the saline permafrost shown in previous studies (Dou et al., 2016; Hivon & Sego, 1995; Nixon, 1987; Wu et al., 2017), slow yet continuous biogeochemical processes may occur in the unfrozen water, which are likely the cause of the significant shift of the geochemical characteristics of the deeper permafrost soils. We will discuss this later.

The CT scans revealed inconsistent structural patterns for FCP and HCP soils and a generally layered soil structure in the top active layer for LCPs (Figure 2), highlighting the significant lateral and vertical variability in soil composition with depth and polygon types. The lack of stratigraphy in the soil structure for FCPs and HCPs is due to cryoturbation during the evolution of the ice wedge polygons that lead to the mixing of the different soil horizons (Kaiser et al., 2007; Van Vliet-Lanoe, 2004). Cryoturbation can reach the depth of 80–120 cm on gentle to moderately slopped uplands and more than 3 m on exposed ridge tops, floodplains, and thaw lake basins (Michaelson et al., 1996; Ping et al., 2014, 2015; Shur & Jorgenson, 1998). In addition to homogenizing the different soil horizons in the active layer, cryoturbation could reduce SOM decomposition rate by moving it deeper, thus, experiencing a shortened thaw period during the growth season (Kaiser et al., 2007; Ping et al., 2008; Schirrmeister et al., 2002). Comparing to the FCP and HCP soils, the center of the LCPs is relatively undisturbed by cryoturbation; thus, a structured soil profile is generally preserved.

5.2 Ice and Mineral Content Contrasts Between Active Layer and Permafrost and Polygon Types

Comparison of the key soil physicochemical characteristics across the different polygon types and between the permafrost and active layer could provide important baseline data for understanding and predicting system response to permafrost thawing in terms of its biogeochemical reactions and rates. While there are apparent structural differences, the ice content of our cores does not show a significant change with depth and across polygon types, which is supported by direct comparison and statistical analysis (Figure 9 and Table 2). A consistently high, averaging 65%, ice (or water) content is observed down to ~4 m bgs (Figure 3) with minimal differences between the active layer and permafrost (Figure 9 and Table 2). Our volumetric ice content measurements are similar to previous investigations, such as those conducted along Alaska coastline where an averaged 77% of ice was founded in the upper few meters of soil (Kanevskiy et al., 2013; Shur & Zhestkova, 2003). Unlike the soil ice content, the averaged mineral content in the permafrost soil is twice that of the active layer (Figure 9; 20% versus 10%) and the reverse is true for the SOM. The significant ice content in the permafrost and ice wedge-dominated soil could result in dramatic shift of the landscape topography and hydrology during permafrost degradation (Liljedahl et al., 2016; Walvoord & Kurylyk, 2016). Subdecadal-scale ice wedge and permafrost degradation studies showed that significant surface topographic change resulted in dramatic alterations in the hydrology and water balance of lowland tundra, leading to reduced inundation and increased runoff (Liljedahl et al., 2016). Such a dramatic shift in surface topography and hydrology could significantly affect soil moisture conditions with dramatic impacts on the mobility, release rate, and form (i.e., CO2 or CH4) of the newly exposed permafrost carbon (Elberling et al., 2013; Schadel et al., 2016).

In addition to the comparison between the active layer and permafrost soil across polygon types, interpolygon-type comparison shows that the variability of the ice content is also nonsignificant between the different polygon types (Table 2). However, we note that the freeze/thaw state of this ice (or water) may be different between the different polygon types due to their different pore water salinities. This is supported by the large contrast of electrical resistivity between LCPs and HCPs (Dafflon et al., 2016; Hubbard et al., 2013) at the same permafrost depth (>0.5 m). Such a spatial difference could result in a variable thermal behavior across a site, with subsequent impact on soil deformation and hydrology under a warmer climate.

5.3 Permafrost SOM and C/N Ratio

Regardless of the variable soil structures and the effects of cryoturbation on the soil profiles, a consistent and rapid decrease of SOM with depth is observed for all cores and polygon types analyzed in this study (Figure 3). This is consistent with many previous observations (Bockheim et al., 1999; Harden et al., 2012; Hugelius et al., 2014; Schaefer et al., 2011; Tarnocai et al., 2009). Large SOM variability between different polygon types and microtopographic features exists, consistent with previous observations as well.

On average, the OM concentration in permafrost is much smaller than the active layer (Figure 9). While permafrost SOM concentration is smaller, its much larger thickness/volume when compared with the shallow active layer still results in two thirds of the SOM storage, particularly in the top meter of the permafrost (0.5–1.5 m bgs), which stores an equal amount of SOM as the active layer. With a 1 cm/yr permafrost degradation rate, the significant SOM concentration from 0.5 to 1.5 m bgs highlights the importance of the top meter of the permafrost in terms of its potential contribution to carbon degradation and release in the next 100 years. Studies have shown a significant loss of the permafrost carbon after thaw (Harden et al., 2012; Schuur et al., 2009), with the rate largely affected by the saturation conditions of this newly exposed permafrost carbon (Elberling et al., 2013; Schadel et al., 2016). In addition to the SOM distribution pattern and its implications for permafrost soil carbon cycling, our analysis also revealed strong positive correlations between SOM and Fe2+/3+, Al3+, and Si4+ concentrations in the pore water, supporting chelation and complexation effects as shown in previous studies (Chen et al., 2014; Wagai et al., 2013).

Beside the large differences in SOM content between the permafrost and the active layer (Figure 9), our interpolygon-type comparison also revealed a much lower active layer SOM content in HCPs when compared with FCPs and LCPs, while their differences in the permafrost is small. We postulate that this difference might result from (1) SOM migration/leaching driven by water drainage and runoff, which is more active in HCP-dominated areas; (2) a potentially higher carbon decomposition rate in HCPs due to a drier and predominantly aerobic condition in the active layer during the growth season when compared to the LCPs, where the soil stays largely anaerobic (Elberling et al., 2013); and (3) potential differences in plant productivity and organic carbon turnover rate between the different polygon types.

Similar to previous studies (Heiri et al., 2001), our results show that as a relatively easy-to-acquire parameter, LOI-based SOM content measurement is an excellent proxy for the estimation of total soil organic carbon and total nitrogen content (Figure 8), with a consistent correlation across topographic features, polygon types, and depths. While acquiring a depth profile of the SOM content is critical to predicting climate warming effects on permafrost thaw and carbon release, it is often time-consuming to perform carbon analysis on a large number of cores at large scales. Our results suggest a correlation (R2 = 0.56) between the soil wet bulk density and weight-based SOM content of the dried soil (Figure 6), which may provide a nondestructive method for quick SOM quantification based on CT scan or gravimetrically derived soil bulk density measurements. The volumetric SOM content does not correlate well with the BDwet directly (Figure 5), which is partly due to the insignificant contribution to the soil bulk density from SOM and large variations of ice and mineral contents in the upper transient permafrost layer (Michaelson et al., 2013). While limitations exist due to the small contribution of SOM to the bulk mass of the wet soil, estimation of weight-based SOM is possible based on the correlation (R2 = 0.93) between the soil wet and dry bulk densities. Unlike SOM, excellent correlations exist between BDwet and mineral content (R2 = 0.94) and between BDwet and ice content (R2 = 0.8) due to their major contributions to the total soil mass per unit volume. In brief, our results suggest that the wet soil bulk density is a relatively easy-to-acquire parameter that can be used as a proxy for estimating soil compositions.

In addition to SOM content, the soil C/N ratio depth profile (Figure 8b) can provide additional insights on the characteristics of the permafrost carbon when compared to the active layer and its potential impact on carbon decomposition after permafrost thaw. Comparing to soil carbon or nitrogen contents alone, C/N ratio has been suggested as a better indicator of the degree of SOM decomposition (Stevenson, 1994; Strauss et al., 2015). During the litter degradation process, loss of carbon and retention/recycling of nitrogen typically results in a decreasing C/N ratio with time (Chapin et al., 2002; Schädel et al., 2014). The decrease of C/N ratio with depth in our results is consistent with previous findings for the Arctic permafrost soils (Schädel et al., 2014) and likely indicates a higher degree of decomposition for the SOM stored in the permafrost at depth. This historic carbon decomposition occurred either at the early stage of soil development before permafrost formation in the epigenetic process or due to previous thawing of the permafrost during extreme warming periods, or both. The abundant lenticular or reticulated ice structures observed in the cores suggest that the upper 1–2 m of the permafrost are likely the transient layer that experienced previous thawing on the decadal scales (Shur, 1988; Shur et al., 2005), which could have resulted in the decomposition of the stored carbon. While this permafrost carbon with higher degradation levels is likely more recalcitrant to further decomposition, the low molecular weight organic molecules from previous decomposition can be mobilized, or trapped in pore water, which may result in rapid release of carbon to the atmosphere upon permafrost thaw, as has been observed recently (Drake et al., 2015).

5.4 Permafrost Geochemistry and Biogeochemical Implications

In addition to the soil physical characteristics discussed above, pore water geochemical analysis provides valuable information for understanding its potential impacts on soil biogeochemical processes during permafrost thaw. Previous studies on saline permafrost soils revealed a diverse microbial population that are active at subzero temperatures although the activity is low (Gilichinsky et al., 2005; Gilichinsky & Rivkina, 2011; Rivkina et al., 2007). This microbial activity occurs primarily in the unfrozen water films/inclusions in the otherwise frozen soil, and this slow yet continuous long-term microbial metabolism could result in sustained winter time carbon flux (Fahnestock et al., 1999), spring thaw burst (Raz-Yaseef et al., 2017), and significantly change the soil biogeochemical characteristics. Our pore water geochemical data support such a hypothesis. As discussed above, an examination of the dissolved iron concentrations shows that while a decreasing trend occurs in FCPs and LCPs with depth, the Fe concentration in HCPs stays fairly high even at large depth (e.g., ~10 mg/L at >3 m; Figure 3). The high dissolved iron concentrations in the shallow, active layer soil across all polygon types suggest active microbial iron reduction and possible translocation of Fe from the mineral layer to the organic layer (Herndon et al., 2017; Lipson et al., 2010). Similarly, the high dissolved iron concentration at depth for HCPs may suggest active microbial iron reduction in the permafrost as well. As previously discussed, the high salinity of the permafrost soil at our study site, particularly in the HCP-dominated area, renders the soils only partially frozen, which can provide suitable habitats for microbial metabolism that utilize soil iron (oxy)hydroxides as electron acceptors (Herndon et al., 2017; Lipson et al., 2010), thus producing dissolved ferrous iron. As discussed before, the association of SOM with iron through coprecipitation or chelation is an important stabilization mechanism for soil organic matter (Chen et al., 2014; Lalonde et al., 2012). In addition, the thermodynamically favorable iron reduction could inhibit the activity of methanogens (Miller et al., 2015). Therefore, the presence of a large amount of dissolved iron at depth suggests that iron cycling in the permafrost may play a critical role in shaping the biogeochemical characteristics of the permafrost soil and subsequently the fate of the carbon stored in permafrost.

Similar to Fe2+/3+, NO3 concentrations also display an anomalous behavior for HCPs when compared to FCPs and LCPs (Figure 3). Namely, instead of a constant concentration with depth for FCPs and LCPs, a significant decrease of NO3 concentrations with depth is observed for HCPs. This characteristic change in NO3 concentration suggests possible microbial denitrification at depth at >1 m, which further supports possible active microbial metabolism at depths in HCP dominated area. When averaged across all polygon types, NO3 concentration in the permafrost is comparable to that of the active layer (Figure 9). A recent subarctic peatland study suggests that the plant nitrogen uptake rate at the permafrost boundary is consistently higher than the shallow root zone (Keuper et al., 2012). Thus, the newly available NO3 from permafrost thaw may provide an additional resource that supports plant growth. However, competition between plant uptake and microbial denitrification will likely determine the fate of the newly available nitrogen from permafrost thaw. Based on nitrogen tracer experiments, a recent study shows that different Arctic plant species may adopt different strategies to compete with the microbes for nitrogen (Zhu et al., 2016); therefore, the fate of the permafrost NO3 is likely further impacted by the dominant plant functional types at the study site. To make predictions more complex, climate warming will increase not only plant nitrogen uptake and microbial denitrification (Xue et al., 2016) but also nitrogen fixation and mineralization (Natali et al., 2012); therefore, the dynamics of the total available nitrogen in the soil after permafrost thaw remains a process that likely has to be determined under specific site conditions.

Another critical nutrient for plant growth, PO43−, was detected at low concentrations down to 2 m bgs (Figure 3). This suggests the availability of additional yet limited phosphorus during permafrost thaw, which may promote plant growth to a certain degree. We note that the fate of the newly available nutrients, such as nitrogen and phosphorous, is strongly impacted by the changes in the hydrological conditions at the site. Because of its high mobility, NO3 is particularly vulnerable to loss via leaching as has been shown in previous studies on tundra soils (Frey et al., 2007; Harms & Ludwig, 2016). Whether the newly available NO3 from permafrost thaw will be utilized by plants/microbes or leached away remains an open question. The enhanced water drainage and runoff during the evolution of the polygonal ground could potentially promote the leaching of the newly available nutrients from the thawed permafrost before it can be utilized for biogeochemical activities by plants and microbes.

In addition to Fe2+/3+, NO3, and phosphorous, other ions of biogeochemical significance, such as Mn2+, also display a characteristic contrast between the active layer and permafrost (Figure 9). Microbial Mn oxidation has been shown to play an important role in the enzymatic decomposition of lignin (Keiluweit et al., 2015). Our data showed that both the mean and median Mn2+ concentrations in the permafrost are higher than the active layer (Figure 9), suggesting its potential contribution in promoting litter decomposition when release upon permafrost thaw.

5.5 Hydraulic Conductivity and Implications

The hydraulic conductivity measurements performed on selected permafrost cores revealed critical differences between the different soil texture, composition, and microtopographic features. Specifically, the organic-rich soil has an averaged hydraulic conductivity that is 1 order of magnitude higher than the mineral-rich soils, indicating increased water infiltration and drainage in the shallow soil dominated by organic matters (Table 3). More interestingly, a sequential decrease of hydraulic conductivity from center to rim then trough of the microtopographic positions by an order of magnitude each is observed (Figure 10b), which likely reflects the hydrological and associated sediment transport behavior during the evolution of the ice wedge dominated polygons. This is supported by the larger fraction of the fine sediments toward the troughs (Figure 10a). Studies by Liljedahl et al. (2016) suggest that the degradation process of ice wedge-dominated polygonal ground is accompanied by increasing drainage and shift of tundra hydrology. Fine sediment transport and accumulation in troughs likely accompany water drainage and runoff and have been observed in the field (Bring et al., 2016; Liljedahl et al., 2016). The accumulation of fine sediments in troughs is likely the reason for the lowest hydraulic conductivity measured in our study when compared with other microtopographic features, such as center and rim.

Together with the increasing dominance of the troughs as the main hydrological conduits on the tundra polygonal ground during landscape evolution, the reduced hydraulic conductivity in the troughs due to fine sediment accumulation could collectively result in promoted surface runoff over infiltration, which may affect the vertical mobility of dissolved oxygen and nutrients and subsequently soil biogeochemical processes in the subsurface.

6 Conclusions

Despite the high complexity of the geomorphological features and soil structures, consistent trends and patterns revealed in this study provide a means for upscaling point-based measurements to the much larger arctic coastal region dominated by polygonal grounds. Specifically, we show (1) significant changes of the major soil physicochemical parameters with depth, mostly showing a general depth trend across the different polygon types yet with large local variabilities; (2) strong correlations between the key soil physical and geochemical parameters that provide a means to infer important yet difficult to acquire parameters from other parameters, such as the estimate of SOM from bulk density measurements; (3) trends in soil hydraulic conductivity, with decreasing values from polygon centers to troughs, reflecting sediment transport process during the evolution of the polygonal ground; (4) unique biogeochemical signature of the high center polygons, suggesting slow, yet continuous microbial activity in the partially unfrozen permafrost at depth due to high soil salinity; and (5) large contrasts of the key soil geochemical characteristics between the active layer and permafrost, such as SOM content, NO3, and dissolved iron concentrations, suggesting that the fate of the permafrost carbon and its biogeochemical response during permafrost thaw may differ from the active layer.

This study highlights the complexity of soil property and depth relationships that climate models ignore due to a lack of spatially contiguous data. The data sets provided here represent, for future work, a crucial input to upscale point measurements and predict soil property and depth relationships across the much larger study area. While the representativeness of these data beyond the sampling locations needs to be further assessed, the consistent correlations between soil properties and depth regardless of the significant structural variability measured by the CT scans suggest their potential for upscaling. Particularly, because the soil wet bulk density can be measured relatively easily, and even nondestructively with CT scans, and that the bulk density correlates with other major parameters as shown by this study, it is a parameter that is the most feasible for upscaling. In addition to the physical properties, the relatively consistent depth trend for the major geochemical parameters across the polygon types can be upscaled to the larger region as well.

Our study provides important inputs for improving climate models. For example, the depth profiles of the major soil physicochemical properties, such as the SOM profile, can be used to establish a depth-resolved vertical soil biogeochemical profile to improve model parameterization. The significant correlations between SOM and soil carbon and nitrogen contents can be used to establish a baseline, depth-resolved soil carbon/nitrogen pool across the site and beyond. We also highlight the different biogeochemical baseline conditions between the different topographic features that the models may need to take into account. For example, the characteristic differences between the HCPs and LCPs/FCPs in terms of their key pore water chemistry, for example, Fe2+/3+ and NO3 concentrations, suggest their potentially different behaviors in terms of the rate of permafrost carbon degradation and nutrient availability. The large spatial extent of the partially unfrozen soil at our site and the arctic coastal region in general, which is likely responsible for the different biogeochemical fingerprints at depth of different polygon types, may need to be represented exclusively in the climate models.


The Next-Generation Ecosystem Experiments-Arctic (NGEE Arctic) project is supported by the Office of Biological and Environmental Research in the DOE Office of Science. We acknowledge David Graham (ORNL), Cathy Wilson (LANL), and Alexander Kholodov (UAF) for their help with field coring efforts. We thank the Editor, the Associate Editor, and the two anonymous reviewers for their constructive comments that significantly improved this manuscript. This NGEE-Arctic research is supported through contract number DE-AC0205CH11231 to Lawrence Berkeley National Laboratory. We gratefully acknowledge the project PI, Stan Wullschleger at ORNL. The data sets presented in this manuscript are available and can be accessed at and cited as Yuxin Wu, Craig Ulrich, and Timothy J. Kneafsey. 2018. Physical, Chemical, and Hydrologic Characteristics of Active Layer and Permafrost Soils of Arctic Polygonal Tundra, Barrow, Alaska, 2013–2016. Next-Generation Ecosystem Experiments Arctic Data Collection, Oak Ridge National Laboratory, U.S. Department of Energy, Oak Ridge, Tennessee, USA. Data set can be accessed on [date] at https://doi.org/10.5440/1358456.