Quantitative characterization of soil organic carbon (OC) content is essential due to its significant impacts on surface–subsurface hydrological–thermal processes and microbial decomposition of OC, which both in turn are important for predicting carbon–climate feedbacks. While such quantification is particularly important in the vulnerable organic-rich Arctic region, it is challenging to achieve due to the general limitations of conventional core sampling and analysis methods, and to the extremely dynamic nature of hydrological–thermal processes associated with annual freeze–thaw events. In this study, we develop and test an inversion scheme that can flexibly use single or multiple datasets – including soil liquid water content, temperature and electrical resistivity tomography (ERT) data – to estimate the vertical distribution of OC content. Our approach relies on the fact that OC content strongly influences soil hydrological–thermal parameters and, therefore, indirectly controls the spatiotemporal dynamics of soil liquid water content, temperature and their correlated electrical resistivity. We employ the Community Land Model to simulate nonisothermal surface–subsurface hydrological dynamics from the bedrock to the top of canopy, with consideration of land surface processes (e.g., solar radiation balance, evapotranspiration, snow accumulation and melting) and ice–liquid water phase transitions. For inversion, we combine a deterministic and an adaptive Markov chain Monte Carlo (MCMC) optimization algorithm to estimate a posteriori distributions of desired model parameters. For hydrological–thermal-to-geophysical variable transformation, the simulated subsurface temperature, liquid water content and ice content are explicitly linked to soil electrical resistivity via petrophysical and geophysical models. We validate the developed scheme using different numerical experiments and evaluate the influence of measurement errors and benefit of joint inversion on the estimation of OC and other parameters. We also quantify the propagation of uncertainty from the estimated parameters to prediction of hydrological–thermal responses. We find that, compared to inversion of single dataset (temperature, liquid water content or apparent resistivity), joint inversion of these datasets significantly reduces parameter uncertainty. We find that the joint inversion approach is able to estimate OC and sand content within the shallow active layer (top 0.3 m of soil) with high reliability. Due to the small variations of temperature and moisture within the shallow permafrost (here at about 0.6 m depth), the approach is unable to estimate OC with confidence. However, if the soil porosity is functionally related to the OC and mineral content, which is often observed in organic-rich Arctic soil, the uncertainty of OC estimate at this depth remarkably decreases. Our study documents the value of the new surface–subsurface, deterministic–stochastic inversion approach, as well as the benefit of including multiple types of data to estimate OC and associated hydrological–thermal dynamics.
Soil organic carbon (OC) and its influence on terrestrial ecosystem feedbacks to global warming in permafrost regions are particularly important for the calculation of global carbon budget and prediction of future climate variation. Warmer air temperature leads to permafrost degradation, which is expected to enhance decomposition of huge pools of previously frozen OC, releasing carbon dioxide and methane to the atmosphere, and enhancing global warming (Koven et al., 2011; Schaphoff et al., 2013; Schuur et al., 2015). In that context, accurate estimation of OC content stored in both the active layer and permafrost is crucial for investigation of carbon stocks that expose microbial decomposition.
Predictive understanding of ecosystem feedbacks to climate in permafrost
regions requires quantitative knowledge of surface–subsurface
hydrological–thermal dynamics, which in turn are strongly governed by the
hydrological–thermal properties of soil OC (Jafarov and Schaefer, 2016). In
particular, there are dramatic differences between thermal and hydraulic
properties of OC and mineral soil, both of which typically co-exist in
shallow permafrost systems. OC's thermal conductivity (e.g.,
OC content is usually measured from core samples, which are collected from field sites and then analyzed in the laboratory (e.g., Kern, 1994). While this method is relatively accurate, it is labor intensive and typically limited in spatial coverage. Because OC and mineral content largely influence hydrological–thermal parameters (i.e., thermal conductivity, heat capacity, hydraulic conductivity and retention curve; see Appendix A), they are the main soil properties that control the subsurface hydrological–thermal dynamics. As a result, OC and mineral content can be potentially obtained by inverting observations of hydrological–thermal state variables (i.e., soil liquid water content and soil temperature) and their correlated observables (e.g., electrical resistivity). However, so far there has been no effort using this approach to indirectly estimate these soil properties.
Geophysical methods hold potential for characterizing the subsurface in permafrost regions as well as their associated physical, hydrological and thermal processes. Geophysical techniques offer an advantage over conventional point measurement techniques because they provide spatially extensive information in a minimally invasive manner (e.g., Hubbard and Rubin, 2005). For example, Arcone et al. (1998) and Chen et al. (2016) used ground-penetrating radar (GPR) to characterize the depth of the permafrost table. Hinkel et al. (2001) used GPR to estimate thaw depth, to recognize ice wedges and ice lenses, and to locate the organic–mineral soil interface. Schwamborn et al. (2002) combined seismic and GPR data to investigate the stratigraphy of both frozen and unfrozen parts of Lake Nikolay. Lewkowicz et al. (2011) and You et al. (2013) employed electrical resistivity tomography (ERT), ground temperature monitoring, frost table probing and coring to detect the permafrost depth. Hauck et al. (2011) developed a four-phase model of soil matrix, ice, liquid and air and used it to estimate soil liquid and ice content from combined ERT and seismic measurements in the Swiss Alps. Hubbard et al. (2013) combined lidar data with multiple geophysical (ERT, GPR, electromagnetic) and point measurements to characterize active-layer thickness and permafrost variability in a large area.
In spite of the potential benefits offered by geophysical data for characterizing permafrost systems, geophysical inversion approaches typically suffer from several challenges. First, inversion methods are often ill-posed due to the fact that geophysical observables are sensitive to different soil properties. Secondly, inversion approaches often typically require petrophysical models to link the geophysical observables with the property of interest. Finally, there are differences between the geophysical support scale and the scale of the imaging target (Hubbard and Linde, 2011). In order to take advantage of information inherent in geophysical signatures and minimize the non-uniqueness challenges described above, many recent studies have explored the value of coupled hydrogeophysical inversion frameworks for estimating soil properties (e.g., Johnson et al., 2009; Huisman et al., 2010; Irving and Singha, 2010; Kowalsky et al., 2011; Pollock and Cirpka, 2012; Busch et al., 2013; Herckenrath et al., 2013; Camporese et al., 2015; Tran et al., 2013, 2016). In these studies, the hydrological and geophysical models are coupled together so that geophysical data are used to estimate soil properties that control the subsurface hydrological–thermal dynamics. Of the geophysical techniques commonly used for monitoring the shallow subsurface, ERT is increasingly common because it can autonomously provide 2- or 3-D time-lapse measurements with a relatively high spatial resolution, is sensitive to properties influencing hydrological–thermal dynamics and is particularly suitable for field deployment over a long period of time. As a result, we use ERT data in this study.
Most coupled hydrogeophysical inversion approaches developed to date are not adequate for investigating permafrost systems due to several gaps. Developed methods have only been applied to terrestrial systems without consideration of the significant dynamics associated with the freeze–thaw transition. Developed coupled hydrogeophysical inversion approaches have also not yet incorporated surface–subsurface interactions (e.g., evapotranspiration, energy balance, plant water uptake). Finally, while a few studies have used soil–vegetation–atmospheric transfer (SVAT) models to qualitatively interpret geophysical data (e.g., McClymont et al., 2013), to date no study has coupled SVAT and geophysical models and data to improve property estimation.
Building on recent advances in the use of electrical methods in the permafrost (e.g., Minsley et al., 2015; Dafflon et al., 2017) as well as coupled hydrogeophysical inversion approaches described above, this study focuses on the development of an inverse approach that uses single or multiple datasets (soil liquid water content, soil temperature and electrical resistivity) to estimate OC content, which is a main factor that governs the subsurface hydrological–thermal dynamics. Our approach advances and couples several algorithms. We use a SVAT model known as the Community Land Model (CLM4.5; Oleson et al., 2013) to simulate water, heat and energy exchange from the bedrock to the top of the canopy. The model considers most of the land surface processes, ice–liquid phase change and surface–subsurface hydrological–thermal dynamics. For parameter estimation, we combined deterministic and stochastic optimization algorithms to concurrently obtain the best parameter estimates and their associated uncertainties. The deterministic optimization algorithm is employed to estimate the initial parameter set and covariance matrix of the proposal distribution. For the stochastic optimization, we used an advanced Markov chain Monte Carlo (MCMC) method known as delayed rejection adaptive Metropolis (DRAM; Haario et al., 2006). With this implementation of this adaptive MCMC algorithm, we expect to obtain the a posteriori probability density function (PDFs) of the desired model parameters more quickly than with the traditional MCMC technique. For hydrological–thermal-to-geophysical transformation, we explicitly consider the dependence of the soil electrical resistivity on the soil ice–liquid water content and soil temperature via petrophysical and forward geophysical models.
This study advances capabilities to estimate and understand the controls of OC on hydrological and thermal properties through developing a hydrological–thermal–geophysical inversion scheme and through exploring its potential to estimate the vertical distribution of OC and mineral content at several depths within a representative synthetic Arctic soil column. Herein, we use synthetic studies to (1) evaluate the relationship between the measurement error and uncertainties of parameter estimates; (2) examine the improvement in parameter estimation offered by including various datasets in the inversion, including apparent resistivity data; (3) investigate how OC estimation changes if the mineral and petrophysical parameters are unknown; (4) explore how parameter estimation changes when soil porosity functionally correlates with the OC and mineral content; and (5) investigate the uncertainty propagation from the OC and mineral content to the hydrological–thermal prediction.
The paper is organized as follows. Section 2 describes the development of the hydrological–thermal–geophysical inversion scheme. Section 3 analyzes and discusses the results of different synthetic experiments. Summary and concluding remarks are provided in Sect. 4.
Generally, the joint hydrological–thermal–geophysical inversion scheme
developed in this study (Fig. 1) includes two main components: (1) a forward
coupled hydrological–thermal–geophysical model that generates the subsurface
state variables (i.e., ice–liquid water content and temperature) and then
uses these variables to infer the apparent resistivity using a set of
petrophysical formulas and a forward electrical resistivity model (Fig. 1a),
and (2) a combined deterministic–stochastic optimization algorithm to estimate
the PDFs of desired model parameters (
In this study, we employed the CLM4.5 model (hereafter referred to as “CLM”), which can effectively simulate different land surface energy balance and surface–subsurface hydrological–thermal processes (Oleson et al., 2013). CLM represents horizontal heterogeneity using multiple parallel soil and snow columns having different land use and plant function types. The lateral flow between the soil columns is not accounted for in CLM. The model simulates the freeze–thaw dynamics by considering two phases of water: liquid and ice. The rate of phase change depends on the energy excess (for the ice-to-liquid transition) or deficit (for the liquid-to-ice transition) from the soil temperature to the freezing temperature. Given CLM's ability to simulate different hydrological–thermal processes in cold regions, we found it suitable for Arctic soil column simulations. The minimum requirements for the top boundary conditions in CLM include precipitation, incident solar, air temperature and wind speed. The land use and plant type information can be provided by users or extracted from the available model database.
Petrophysical parameters and soil properties information used for
synthetic simulation. Petrophysical parameters
CLM assumes that soil is a mixture of three soil types, namely, OC, sand and
clay. It calculates the soil hydrological–thermal parameters based on the
content (fraction) of these soil types and their corresponding
hydrological–thermal properties (see Appendix A for more detailed information
on these relationships). In CLM, the soil OC content (% OC) is defined as
For more detailed exploration of the vertical variability of subsurface
properties and associated hydrological–thermal dynamics, we increased the
default number of soil layers in the CLM from 15 to 32 layers and defined
the depth of layer
Information on parameter, “observation” data and inversion results
for nine scenarios. For scenarios 1–7, the soil OC content (three
parameters) at
Moreover, in order to explore how the soil porosity influences the estimation
of soil OC and sand content, we modified the CLM to consider two cases:
(1) the soil porosity profile was fixed and independent from the soil OC and
sand content (see scenarios 1–8 in Table 1), and (2) the soil
porosity was calculated from the OC and sand content as the default in the CLM
(see scenario 9 in Table 1) as below (Lawrence and Slater, 2008):
Soil thermal conductivity
Comparing the two cases shows that, when the soil porosity depends on OC and sand content, the soil thermal properties change in larger ranges with the variation of OC content, sand content and liquid water saturation. It is because, while the soil porosity is fixed at 0.7 in the first case (panels a–c), it varies from 0.36 (when soil is 100 % sand) to 0.9 (when soil is 100 % OC) in the second case (panels d–f). Because soil thermal properties strongly depend on soil porosity (see Eqs. A2 and A8 in Appendix A), together with the OC and sand content, the variation of porosity in the second case leads to rapid change of the soil thermal properties and, therefore, of the subsurface hydrological–thermal dynamics.
In our inverse scheme, we link the output of the hydrological–thermal
simulation described above (soil ice–liquid water saturation and
temperature) to soil electrical conductivity using Archie's law (Archie,
1942):
The water electrical conductivity (
The effect of soil temperature (
In this section, we present a combination approach of deterministic and
stochastic optimization algorithms to estimate the model parameters
In the stochastic parameter estimation, the objective is to find the
a posteriori probability distribution
Once the a posteriori density distribution Given the current parameter set where In delayed
rejection, once the candidate is rejected, instead of staying at the current
sample, a second ( If the second try is rejected, the third try can be generated and so on. The
number of tries is specified by users. One of the key
limitations of the MCMC technique is the selection of the proposal
distribution model. In adaptive Metropolis, the proposal distribution is
assumed to be Gaussian centered at the current sample In Eq. (14),
The assessment of convergence of the MCMC chain is analyzed by Geweke's
criterion, which compares the means and variances of the beginning and end
segments of the chain as below:
The speed of convergence of the MCMC optimization algorithm strongly depends
on the initial model parameter
To test the value of the developed joint inversion approach under a range of conditions and assumptions, we performed several synthetic case studies using the numerical soil column illustrated in Fig. 3. The synthetic column was developed to mimic typical soil and petrophysical properties associated with a high-centered polygon at an intensive study transect (NGEE Arctic, Barrow, Alaska) (Fig. 4). The transect is 35 m in length and covers three typical topography types in Barrow, namely, high-centered (HCP), flat-centered (FCP) and low-centered polygon (LCP). The thawing occurs during the growing season, lasting from the beginning of June to the end of September. In the growing season while the LCP is fully saturated, the HCP is relatively dry and unsaturated. The bottom of the thaw layer at the end of the growing season is located at about 0.3 and 0.5 m depth at the center of HCP and LCP, respectively. ERT measurements were performed along the transect daily using the Wenner–Schlumberger configuration with an electrode spacing of 0.5 m. Other measurements and conditions useful for our synthetic studies – including soil temperature, soil moisture, thaw depth, snow dynamics and climate conditions – were also measured (Dafflon et al., 2017). These data have been used here to develop conceptual models and synthetic columns, while they will be used for real application of the joint inversion scheme in a subsequent study.
Twenty-seven synthetic soil layers and soil properties (OC, sand content and
porosity) for simulating hydrological and thermal dynamics. The five bottom
bedrock layers are not shown in this figure. We assumed that the vertical
profiles of soil properties are constructed by interpolating their
corresponding values at
Soil properties and petrophysical information used for the synthetic studies are provided in Table 1. The “true” soil properties are based on the core sample analysis at the Barrow, AK site (Baptiste Dafflon, personal communication, 2016), and the “true” petrophysical parameters were obtained from Minsley et al. (2015). It is worth noting that soil is represented in the CLM as a mixture of OC, sand and clay. As such, in order to estimate the soil mixture, it was sufficient for us to consider OC and sand content (in sand–clay mineral mixture) only.
We assumed that the vertical profiles of soil properties (porosity, OC and
sand content) were constructed by interpolating their corresponding values at
four depths
We synthetically explored nine scenarios using the newly developed inversion
procedure (Table 2). The purposes of these scenarios are as follows: Evaluate the effect of measurement errors on uncertainties of soil OC
estimates (using electrical resistivity data as an example). Investigate the improvement in OC estimation gained by joint inversion of
multiple hydrological, thermal and geophysical datasets compared with
inversion of each single dataset. Study how the parameter estimates and their associated uncertainties
change if, in addition to OC content, sand content and petrophysical
parameters are unknown. Explore the effect of soil porosity on the parameter estimation by
comparing two cases: (1) soil porosity profile is fixed and independent from
the soil OC and sand content, and (2) soil porosity is defined as a function
of OC and sand content. Analyze the uncertainties, non-uniqueness, correlation and convergence of
the inverse problem, and evaluate the impact of parameter uncertainty
on prediction of hydrological–thermal dynamics.
General procedure used to perform the synthetic case studies.
For all scenarios, we used daily time step meteorological forcing data (including air temperature, wind speed, short-wave and long-wave radiation, and precipitation) collected at the Barrow site over a year period from 1 January to 31 December 2013, which includes a time period over which some of the soil and electrical datasets were also collected at the NGEE Arctic site. The plant functional type information was obtained from the CLM database for the Arctic region. The general approach that we followed to perform all synthetic scenarios is presented in Fig. 5.
In order to account for the measurement errors, we assumed that the error
distribution was Gaussian and added error to synthetic data to obtain
“noisy” synthetic data (hereafter referred to as observation data)
(Table 2). We set the standard deviation of ERT measurement error to 2 %
of synthetic data for scenario 1 (low measurement error) and to 5 % for
the other scenarios. We used a standard deviation of measurement errors of
0.5
For inversion, ranges were provided for unknown soil and petrophysical
parameters based on Hubbard et al. (2013) and Dafflon et al. (2017)
(Table 2). To minimize non-uniqueness in the inversion procedure, we ignored
the small OC content at 1 m and the small sand content at 0.1 m. For
scenarios 1–7, we estimated OC content at
In order to estimate the a posteriori PDF of OC and sand content as well as petrophysical parameters, we generate 8000 samples for scenarios 1–7 and 15 000 samples for scenarios 8 and 9. The number of samples in scenarios 8 and 9 is larger because there is a greater number of estimated parameters in these scenarios. We selected the last 5000 samples having a Geweke's score less than 0.4 to construct the PDFs of these parameters. Their best estimates and associated uncertainties are, respectively, represented by the means and standard deviations of the samples and summarized in Table 2. Discussion and comparison of the scenarios are presented below.
The influence of measurement error on the parameter uncertainties was
considered by comparing scenarios 1 and 2 using apparent resistivity as an
example. Scenario 1 assumed that the standard deviation of measurement error
is 2 % of synthetic apparent resistivity data (small measurement error),
while this value for scenario 2 is 5 % (large measurement error). For
these two scenarios, we estimated the OC content at
In order to investigate the non-uniqueness problem and the correlation
between parameters, we estimated the misfit (sum of square of absolute
differences) between the synthetic and sampled apparent resistivity data as a
function of the OC content at
The a posteriori probability of the soil OC content at
The effectiveness of the joint inversion of multiple datasets on the OC
content estimation (at
Misfit (sum of square of absolute difference) between synthetic
observations and MCMC sampling apparent resistivity data as a function of
soil OC content at
The a posteriori probability of the soil OC content at
In scenario 8, in addition to the OC content, we assumed that the sand
content and petrophysical parameters
The a posteriori probability of soil OC content at
The pairwise relationships between estimated parameters (Fig. 10) indicate
that the OC content at 0.1 m and petrophysical parameter
Pairwise relationships between estimated parameters. The calculation was based on 3000 MCMC samples of scenario 8.
The a posteriori probability of soil OC content at
Comparison of “observation” and predicted soil temperature at
Comparison between estimated and synthetic thaw depth over a year for scenario 8. The blue and red lines, respectively, represent the synthetic and estimated thaw depth. The grey region shows the confidence interval with a level of 95 %.
Comparison between “observation” and predicted apparent resistivity. The red line denotes the 1 : 1 line. The vertical error bar on the blue symbols represents the confidence interval of the predicted apparent resistivity with a confidence level of 95 %. The comparison was based on a posteriori samples of scenario 8.
In this section, we evaluate how the parameter uncertainties change when the porosity is determined as a function of the OC and mineral content by comparing scenarios 8 and 9. While the soil porosity in scenario 8 was fixed and independent from the OC and sand content, it was calculated from the OC and sand content in scenario 9 as shown in Eqs. (2) and (3).
Compared to scenario 8, all uncertainties of sand and OC content in
scenario 9 are smaller (Fig. 11). In particular, the uncertainties of these
parameters at
In this section, we evaluate the impact of parameter uncertainties on the
prediction of hydrological–thermal dynamics. A posteriori samples of the OC,
sand content, and petrophysical parameters
The synthetic and estimated thaw depth using results obtained from scenario 8 (Fig. 13) show that soil water thaws around the middle of June and freezes again around the middle of September. The thaw depth varies from 0.2 to 0.42 m. These results are compatible with our field survey data in Barrow (Dafflon et al., 2017), indicating that, although this is a synthetic study, its simulation is relatively compatible with the Arctic tundra field measurements. As for the influence of parameter uncertainties on the thaw depth estimation, we observed that the parameter uncertainties only cause thaw depth variations during the warmest period of the year (beginning of August to middle of September). During other times of the year, the thaw depths corresponding to different sets of parameters are quite similar.
The comparison between synthetic and predicted apparent resistivity data (Fig. 14) shows that there is a very good agreement between them with no bias, which implies that our inversion scheme converges to the lowest misfit region. The confidence ranges corresponding to a level of 95 % vary from 1.4 to 9.4 % of the “observation” resistivity, which is suitable with the relative measurement error of 5 %.
In this study, we developed and tested a surface–subsurface coupled hydrogeophysical inversion approach to estimate OC content and its influence on hydrological–thermal behavior under Arctic freeze–thaw conditions. In our inversion scheme, the CLM model serves as a forward model to simulate the land surface energy balance and surface–subsurface hydrological–thermal processes. The new scheme can jointly use different types of data for the inversion, including electrical resistivity data. The dependence of soil electrical resistivity on temperature and ice–liquid water content is explicitly accounted for within the inversion.
We developed an advanced optimization technique that combines the deterministic and stochastic optimization algorithms to obtain soil and petrophysical parameters and their associated uncertainties. The stochastic optimization estimated the a posteriori distribution of model parameters by using the Bayesian inference and adaptive MCMC algorithm DRAM. Meanwhile, the deterministic optimization algorithm was used to approximate the starting set of model parameters and the initial covariance matrix of the proposal distribution for the stochastic optimization, which helps to more quickly converge to the parameter a posteriori distribution.
We tested the inversion scheme using multiple synthetic experiments in a 1-D soil column representative of the Arctic tundra, where surface–subsurface hydrological and thermal regimes co-interact and are influenced by soil OC and mineral content. The obtained results show that the new inversion approach reproduced the synthetic data well in all experiments. The shallow (upper 0.3 m) active-layer OC and sand content and the petrophysical parameters can be reliably obtained using soil temperature, soil liquid/ice water content and ERT data. When the soil porosity is fixed, the uncertainties of OC and sand content are very high in the permafrost section (0.6 m), even when soil temperature, liquid water saturation and apparent resistivity data are jointly used in the estimation procedure. This suggests that, when the porosity is fixed, the inversion approach is unable to significantly improve the estimation of OC within the permafrost, due to the small magnitude of temporal variation of both temperature and soil moisture in that section. However, if the soil porosity is considered as a function of OC and sand content, the permafrost parameters can be reliably obtained because the variation of porosity with OC and sand content increases the sensitivity to ice–liquid water and temperature. Examining the relationship between measurement errors and parameter uncertainties, we found that the uncertainties of estimated parameters increase with increasing measurement error. We also explored the improvement in parameter estimation when jointly using multiple data for the inversion. Compared to single dataset inversion (temperature, soil moisture or electrical resistivity), joint inversion significantly reduces the uncertainties of estimated parameters, especially at 0.3 m depth. Finally, we quantified the influence of parameter uncertainties on the prediction of hydrological–thermal and thaw depth dynamics. The obtained results show that the soil liquid water content prediction is more uncertain than the soil temperature and apparent resistivity predictions, due to its large measurement error. The uncertainties in OC and sand content have an impact on the thaw depth estimation only during the warmest months of the year (August and September).
This study developed and tested a novel approach to estimating soil OC content using inverse modeling that can incorporate diverse hydrological, thermal and ERT datasets. In addition, the study also permitted exploration of surface–subsurface hydrological–thermal dynamics and spatiotemporal variations associated with freeze–thaw transitions. Given the importance of characterizing OC as part of ecosystem and climate studies, the typical challenges associated with collecting and analyzing “sufficient” core data to characterize the vertical and horizontal variability of OC associated with a field study site, and the increasing use of electrical resistivity data to characterize vertical, horizontal and temporal variability in shallow systems, the new inversion approach offers significant potential for improved characterization of OC over field-relevant conditions and scales. It also offers significant potential for improving our understanding of hydrological–thermal behavior of naturally heterogeneous permafrost systems. The successful validation of this approach using 1-D synthetic studies provides a foundation for extending it to 2-D and applying it to real field data, which is currently underway.
In this study, we concentrated on the indirect impact of the OC content on water electrical resistivity via soil water and temperature. Recent studies indicated that the soil OC content largely influences ionic mobility and, therefore, changes the polarization and relaxation time of soil response to the applied current, which can be measured by spectral induced polarization (SIP) (e.g., Schwartz and Furman, 2015). As a result, our future study will explore the possibility of integrating SIP measurements into our coupled hydrological–thermal–geophysical inversion scheme. In that case, the OC content is linked to SIP measurements both by its hydrological–thermal and by its electrical polarization properties. Hauck et al. (2011) indicated that combination of ERT and seismic measurements can improve the estimation of ice and liquid water. We will integrate this approach into coupled hydrogeophysical inversion to better constrain the inversion and reduce the non-uniqueness of parameter estimation.
With advancements in data acquisition, the surface–subsurface hydrological–thermal dynamics now can be monitored in real time at high temporal resolution using multiple above- and below-ground measurements including geophysical techniques. Our next step is to expand the inversion scheme so that it can assimilate these data into hydrological–thermal models to improve the model prediction in real time.
The meteorologial forcing data in Barrow that were used as boundary
conditions in this study are provided at
Soil thermal conductivity
The volumetric heat capacity (
As for soil hydrological characteristics, soil matric potential (
The exponent coefficient (
The authors declare that they have no conflict of interest.
The Next-Generation Ecosystem Experiments (NGEE Arctic) project is supported by the Office of Biological and Environmental Research in the DOE Office of Science. This NGEE Arctic research is supported through contract number DE-AC02-05CH11231 to Lawrence Berkeley National Laboratory. The authors would like to thank NGEE Arctic PI Stan Wullschleger (ORNL) for support and Thomas Günther for providing the BERT codes. Edited by: Julia Boike Reviewed by: two anonymous referees