Modeling Dynamics of Permafrost Degradation and their Impact on Ecosystems Across Entire Alaska: Arctic and Subarctic Engineering Design Tool (Part-1)

1. Introduction

Recent observations indicate a warming of permafrost in many northern regions with a resulting degradation of ice-rich and carbon-rich permafrost. In the last 30-40 years, warming in permafrost temperatures observed in the Northern Hemisphere has resulted in the thawing of permafrost in natural, undisturbed conditions in areas both close to the southern and north boundary of the permafrost zone. The impact of climate warming on permafrost and the potential of climate feedback resulting from permafrost thawing have recently received a great deal of attention. Most of the permafrost observatories in the Northern Hemisphere show substantial warming of permafrost since the 1980s [1,2,3,4]. The magnitude of permafrost warming has varied with location but was typically from 0.5 to 4°C. A more pronounced warming has been observed in cold permafrost in higher northern latitudes and a smaller increase in temperature was typical for the warmer permafrost (because latent heat). Permafrost temperatures were stable or even slightly cooling during the last several years in the lower northern latitudes.

A comprehensive system of permafrost observatories at Alaska was established in the late 1970s and early 1980s by the Geophysical Institute, University of Alaska Fairbanks along the Trans-Alaska Pipeline and at other locations in Alaska [5,6,7,8,9].

Permafrost extent and active layer depth are very important for the high-latitude temperature regimes and water cycle. Changes in the hydrologic cycle and soil temperature can have significant consequences for local people: degrading permafrost very often could lead to ecosystem changes and threatens to damage infrastructure. Understanding how the tightly coupled permafrost and hydrology will behave in a warming world is vitally important for local communities in Alaska and throughout the circumpolar North.

Alaska is warming very fast. Therefore, there is an urgent need to develop a robust modeling tool for projecting changes associated with thawing permafrost in this region. Many important questions about present day and future ecosystems change can be answered with such tools. Even with all the uncertainty of future climate change projections this model will be useful in creating estimations of the range of future permafrost change in Alaska. Many of the proposed developments will be experimental in nature, with a final goal of providing society with robust expert assessments of changes that are important for infrastructure for several decades ahead.

Climate warming is expected to have pronounced effects on high latitude ecosystems, especially regions underlain by warm permafrost such as the Interior of Alaska [10,11,12,13]. Large areas of permafrost in Alaska have been degrading for some time [14,15] and are expected to continue to do so as the global climate warms. These changes in permafrost distribution have dramatically affected the ecosystems across Alaska through widespread drying or wetland expansion [14,15,16].

In the discontinuous permafrost zone with ground temperatures near the freezing point， the presence of vegetation and organic soil layers strongly affect the ground thermal regime due to insulation effect. Shur and Jorgenson [17] classified permafrost according to patterns of formation and degradation in relation to climate and ecosystems. They propose five main permafrost zones:

1)

‘Climate-driven’.

2)

‘Climate-driven, ecosystem-modified’.

3)

‘Climate-driven, ecosystem protected’.

4)

‘Ecosystem-driven’.

5)

‘Ecosystem-protected’ permafrost.

Permafrost in each zone reacts differently to external disturbances. The permafrost types ‘climate-driven, ecosystem protected’, ‘ecosystem-driven’, and ‘ecosystem-protected’ are in a state of dis-equilibrium with the current climate; minimal disturbances have already caused widespread permafrost degradation [18].

During the second half of the 20^th century, permafrost has been thawing within the southern part of the northern Hemisphere permafrost domain. However, recent observations documented propagation of this process northward into the continuous permafrost zone. The close proximity of the exceptionally ice-rich soil horizons to the ground surface (Figure 1), which is typical for the arctic tundra biome, and loess sediments in interior Alaska makes tundra surfaces extremely sensitive to the natural and human-induced changes that resulted in development of processes such as thermokarst, thermal erosion, and retrogressive thaw slumps that strongly affect the stability of ecosystems and infrastructure. The most significant impact on ecosystems, infrastructure, carbon cycle and hydrology will be observed in areas where permafrost contains a considerable amount of ground ice in the upper few meters. Permafrost thawing is already causing serious damage to buildings and industrial facilities and is projected to continue. A major threshold is crossed when permafrost thaws after damage or removal of vegetation and upper soil organic layers, e.g., by wildfires as naturally or human activities artificially. Current permafrost temperature measurements indicate a relatively continuous warming of permafrost in most regions of Alaska over the last 30 years [2,3,9], increasing the sensitivity and vulnerability of permafrost in many regions to surface disturbances. Numerical models predict widespread warming and degradation of discontinuous permafrost until 2100, i.e., in interior Alaska. Permafrost degradation due to thermokarst processes is found in a broad variety of land surface features [15,19]. Future climate scenarios predict a roughly 2°C - 5°C increase in mean annual surface air temperature for the Alaskan Interior over the next 80 years [19]. This temperature increase will be enough to initiate permafrost degradation in interior Alaska [6,15,16,21,22], leading to widespread thermokarst development, groundwater-surface water interactions, and a deepening of seasonally thawed (active) layer [6,15,17,19].

Recently, several permafrost models were developed for various regions [21,22,23,24,25,26,27,28,29]. In many of the permafrost models, properties for the mineral and organic soil layers [21,22,24,30] or properties for the sediment and bedrock material [26], or values of n factors and thermal properties of the ground material [31] are producing a large number of distinct combinations of the input data set classes, the parameterization of which is a difficult task involving various assumptions and data sets. In some publications, the texture of mineral soils and other input parameters are used [28].

In this manuscript, we consider the ecosystem types and set of existing ground temperature observations across the individual ecosystem types. This combination allows estimation of the soil properties using the data assimilation technique for each input-data class to Alaska. The justification of this approach is based on the analysis of spatial and temporal variability of near-surface ground temperatures measured across entire Alaska [2,3,5,6,7,9,16].

To verify the developed model, we simulate the retrospective permafrost evolution and compare it with in situ observations of the active layer thickness. Consequently, we project the modeling results into the future according to the Intergovernmental Panel on Climate Change (IPCC) Representative Concentration Pathways (RCPs) 4.5 and 8.5 greenhouse gas emission scenarios [32,33,34]. The technology developed in this study is believed to be relevant and applicable to many Alaskan and other Circum-Arctic locations for the estimation of possible consequences of permafrost degradation in the 21st century.

The modes of permafrost degradation are highly variable, and its topographic and ecological consequences depend on the interactions of slope positions, soil texture, hydrology, ice content and others. However, it is still not clear how surface conditions could affect the hydrothermal dynamics of permafrost and active layer, how to determine the factors that are most important in estimating for surface condition. Knowledge of processes affecting the vulnerability of permafrost in respect to ecosystems stability, infrastructures, and socio-economic impacts, provides stakeholders with information for better understanding the range of possible future changes, to help guide land management and decision-making.

2. Permafrost Dynamics Modeling: Data and Methods

2.1. Simulation Domain

We defined the computational Alaskan domain (Figure 2a) with a spatial resolution of 1 by 1 km, which covered the entire Alaska except a small part of the Aleutian Islands. For computation the chosen area consisted of 1,497,645 grid cells. Each spatial grid cell contains a 100 m in depth multilayered soil column with 160 vertical computational grid points (layers). The total 239,623,200 computational nodes.

For the model validation and observed data incorporation into the model, we used measured data from more than fifty shallow (1.2-5.0 m in depth) boreholes across entire Alaska (Figure 2b). These high-quality ground temperature measurements (precision generally at 0.01°C) are available from the beginning of 1980s to 2023 across Alaska [2,3,5,6]. Soil water content and snow depth measurements were also available at most of these locations. In addition, more than twenty relatively deep boreholes from 15 m to 100 m in depth along the Dalton Highway were available for the model validation and calibration.

2.2. Climate Forcing

We prepared the SNAP (https://uaf-snap.org/) data set as a climate forcing to drive the GIPL2-MPI permafrost model. CMIP5/AR5 selected GCM's (https://uaf-snap.org/how-do-we-do-it) scaled down to 1 km spatial resolution were used to drive the GIPL2-MPI model. We used outputs for two operating separately GCMs GFDL-CM3 and NCAR-CCSM4 as well as for a 5-ModelAverage, and for two different greenhouse gas concentration trajectory RCP4.5 and RCP8.5.

The 5-ModelAverage is the average model output from the top Five Model Average that best replicate historical climate in Alaska and the Arctic regions (NCAR-CCSM4, GFDL-CM3, GISS-E2-R, IPSL-CM5A-LR, MRI-CGCM3) [35]. The 5-ModelAverage best for: looking at general climate trends over time, as it smooths over the year-to-year and decade-to-decade variability. NCAR-CCSM4 and GFDL-CM3 are the best for the Alaska region for exploring climatic extremes or annual variability [20]. Only individual models can provide that type of information.

RCP is a greenhouse gas concentration trajectory adopted by the IPCC [12,13]. Four pathways were used for climate modeling and research for the IPCC fifth Assessment Report (AR5) in 2013 and 2014 (Figure 3). The pathways describe different climate futures, all of which are considered possible depending on the volume of greenhouse gases (GHG) emitted in the years to come. The RCPs - originally RCP2.6, RCP4.5, RCP6.0, and RCP8.5 - are labelled after a possible range of radiative forcing values in the year 2100 2.6, 4.5, 6, and 8.5 W/m², respectively [32,33,34]. The RCP2.6 scenario was included in the RCP set to show what it would take to limit globally averaged warming to about 1°C by 2100. RCP2.6 is generally not used in presenting plausible futures because it requires such extreme mitigation that it is considered by many climate experts to be unrealistic. The RCP2.6 scenario required that after 2020, CO₂ emissions would decline and become negative by 2100 (i.e., we find ways to sequester more carbon than we are emitting in 2100). Given the present increase in the rate of emissions and CO₂ concentrations, a near-term stabilization and decrease of emissions is not plausible.

The RCP2.6 scenario was included in the RCP set to show what it would take to limit globally averaged warming to about 1°C by 2100. RCP2.6 is generally not used in presenting plausible futures because it requires such extreme mitigation that it is considered by many climate experts to be unrealistic. The RCP2.6 scenario required that after 2020, CO2 emissions would decline and become negative by 2100 (i.e., we find ways to sequester more carbon than we are emitting in 2100). Given the present increase in the rate of emissions and CO2 concentrations, a near-term stabilization and decrease of emissions is not plausible.

Table 1. Two Global Climate Models (GCM) and a composite of 5-ModelAverage GSMs scaled down to 1 by 1 km spatial resolution across Alaska used for permafrost dynamics simulation [35].

Project	Center	Model	Acronym
CMIP5/AR5	National Center for Atmospheric Research	Community Earth System Model 4	NCAR-CCSM4
CMIP5/AR5	NOAA Geophysical Fluid Dynamics Laboratory	Coupled Model 3.0	GFDL-CM3
CMIP5/AR5	5-Model Averaged	Calculated as the mean of the five GCMs*	5-ModelAverage

* Averaged of the five GCMs: NCAR-CCSM4, GFDL-CM3, GISS-E2-R, IPSL-CM5A-LR, MRI-CGCM3

Table 1. Two Global Climate Models (GCM) and a composite of 5-ModelAverage GSMs scaled down to 1 by 1 km spatial resolution across Alaska used for permafrost dynamics simulation [35].

Project	Center	Model	Acronym
CMIP5/AR5	National Center for Atmospheric Research	Community Earth System Model 4	NCAR-CCSM4
CMIP5/AR5	NOAA Geophysical Fluid Dynamics Laboratory	Coupled Model 3.0	GFDL-CM3
CMIP5/AR5	5-Model Averaged	Calculated as the mean of the five GCMs*	5-ModelAverage

* Averaged of the five GCMs: NCAR-CCSM4, GFDL-CM3, GISS-E2-R, IPSL-CM5A-LR, MRI-CGCM3

2.3. Permafrost Dynamics Model

To evaluate the vulnerability of permafrost under climate warming across the entire Alaskan domain in respect to ecosystems stability, infrastructure and socioeconomic we applied the physical process-based permafrost dynamics model developed in Geophysical Institute Permafrost Lab (GIPL) University of Alaska Fairbanks [21,22,24,25,37].

In this research, we employ the transient permafrost dynamics model GIPL2-MPI, which simulates soil temperature dynamics and the depth of seasonal freezing and thawing by solving 1D and 2D non-linear parabolic conductive heat equation with phase change using the finite difference method. The GIPL2-MPI model captures physical processes essential to robust and appropriate modeling of soil temperature dynamics, active layer thickness and talik development if so. Specifically, soil thermal properties are parameterized according to soil texture and organic matter. In this model, the process of soil freezing/thawing occurs in accordance with the unfrozen water content and soil thermal properties, which are specific for each soil layer and for each geographical location. Unfrozen water content was parameterized by power function θ(T) = ac*((Abs (Tfr-T))^bc), where ac > 0, bc < 0, T is soil temperature, and Tfr is a temperature of ice fusion or freezing point depression in soil. The absolutely stable implicit difference numerical method and enthalpy formulation of the energy conservation law [38,39,40] implemented in GIPL2-MPI makes it possible to use vertical resolution without loss of latent-heat effects in the phase transition zone, even under rapid or abrupt changes in the temperature fields [21,22,23]. Additionally, GIPL2-MPI includes thermal insulation of the snow cover, and geothermal heating at the appropriately selected depth. The upwards geothermal flux applied at the base of soil column (100 m in depth) was calculated as described in [41] for each spatially distributed grid cell. The GIPL2-MPI model also incorporates an efficient algorithm to estimate soil thermal properties using in-situ temperature measurements in the active layer and in permafrost. This simplifies model calibration for specific sites. The conceptual representation of permafrost dynamics in the GIPL2-MPI Model could be expressed as:

• Numerical finite difference solution (method) of heat diffusion;
• At least 100 m in depth vertical computational domain contains 128-256 vertical grid nodes;
• Moss insulation explicitly considered;
• Organic soil insulation explicitly considered;
• Snow insulation explicitly considered;
• Effect of unfrozen water on phase change explicitly considered;
• Thermal conductivity depends on soil moisture, temperature, and unfrozen water content.

The detailed description of the physical model and numerical solution are presented in the Appendix A.

2.4. The Ecosystem Approach of Permafrost Modeling

Definition: The ecosystem approach of permafrost modeling is based on field studies of the upper layers of soil section for various classes of ecosystems and incorporating this data into the permafrost model. For each ecosystem type the soil texture and thickness of the organic matter and mineral soil layers are parameterized according to field measurements and geographic location. In the frame of this method, the ecosystem types, and groups of existing ground temperature observations across the various ecosystem types are considered. This grouping allows estimating the soil properties using the data assimilation technique [37,50] and model validation for each input data class. An example to apply the ecosystem type approach to model permafrost across the North Slope of Alaska was presented in [24].

The 2018 Alaska Vegetation and Wetland Composite (AKVWC) represents the best-available data derived from 28 regional land cover maps that have been developed within the last 31 years (Figure 4b). The map is attributed with a uniform, two-tiered legend so that land cover classes that are similar in concept were used for determination of surface vegetation conditions, existence of organic matter and for prescribing of upper organic layers thermal properties (Figure 4a). The Alaska Land Carbon and Wetland Distribution Map (Figure 4b) was produced as part of the USGS Land Carbon Alaska assessment [42,43,44] and used for prescription of initial soil moisture content and thermal properties as well. The 2018 AKVWC data was used as one of the input datasets to the GIPL2-MPI model. The methods used to produce this wetland map as well their accuracy assessment is presented in He et al. [42]. This product was to provide regional estimates of specific wetland types (e.g. bog and fen) in Alaska, available wetland types mapped by the National Wetlands Inventory (NWI) program were reclassified into bog, fen, and other.

The map of ground ice distribution [36] was used for additional information to compile the initial soil ice/water content (Figure 6a). Ground ice volumes were estimated for the upper 5 m of permafrost using terrain relationships established by Kreig and Reger [48] and our field data. Variable ice (the legend component on the map) associated with buried glacial ice [36]. Ground ice volume near the surface is higher in colder regions due to active ice-wedge formation and ice segregation in fine-grained sediments. Buried glacial ice in old or stagnant young moraines is included but is irregularly distributed at this map scale. Depth to bedrock was used to compile the deep soil column structure in terms of the thickness of fine-grained soil layers or depth to bedrock (Figure 6b).

2.5. Data Assimilation and Model Validation

Different earth materials have varying thermal properties. The soil thermal conductivity and heat capacity vary within different soil layers as well as during thawing/freezing cycles and depend on the unfrozen water content that is a certain function of temperature. The method of obtaining these properties is based on numerical solution for a coefficient inverse problem and on minimization locally the misfit between measured and modeled temperatures by changing thermal properties along the direction of the steepest descent. The method used and its limitations are described in more detail elsewhere [21,22,37].

For the initial model calibration, we used data from more than fifty shallow boreholes (1-1.5 m in depth) across Alaska from north to south. These high-quality ground temperature measurements (precision generally at 0.01°C) are available from the beginning of 1980s to 2022 across Alaska [5,6,9]. Soil water content and snow depth measurements were also available at most of these boreholes. The temperature measurements in the shallow holes were performed with vertical spacing of 0.08-0.15 m. At most of these sites, soil water content and snow depth were also recorded. In addition, more than twenty relatively deep boreholes from 15 m to 100 m in depth along the same transect (Figure 7a) were available for the model validation in terms of permafrost temperature profiles, active layer depth and permafrost thickness. Figure 7 illustrates the results of the GIPL2-MPI model calibration for the West Dock (WD) permafrost observatory. The specific site WD is classified as an Arctic Tundra (70°22’28.08” N, 148°33’7.8” W). Reconstructed soil thermal properties for the WD site (Arctic Tundra) and other major ecotypes are shown in Table 2.

There are two basic approaches to calibration of modeled permafrost temperatures against the observed data, which can be distinguished by their use of temporal or spatial relationships. With the temporal approach, the quality of the modeling series is assessed by time series regression against measured data Figure 7b, Figure 8b, Figure 9b and Figure 10b. The quantitative relationship between simulated and measured data is then determined for a ‘‘calibration’’ period with some instrumental data withheld to assess the veracity of the relationship with independent data.

In the spatial approach, assemblages of the observed data from several different geographic locations with different landscape settings determine the quality of the modeling results. To achieve geographic correspondence between the scale of observation and modeling we utilized a regional scale permafrost characterization based on observations obtained from representative locations. Additional comparison of model produced ground temperatures, active layer thickness, and spatial permafrost distribution with measured ground temperatures at the Alaska sites shows a good agreement e.g. Figure 7b, Figure 8b, Figure 9b and Figure 10b.

Here we followed the idea of the variational approach and optimized a set of model parameters by minimizing difference between the observed and modeled soil temperature [37,50]. This task was associated with estimation of thermal properties of the ground material. Results of previous investigations [4,48] were used to determine the initial approximation to the model parameters. The soil characterization used in the GIPL2-MPI model was based on extensive empirical observations, conducted in representative locations that are characteristic for the major physiographic units across entire Alaska. Once the initial approximation was defined, we varied the set of parameters associated with the ground material in order to decrease the discrepancy between the simulated and observed temperature climatology at all other depths below the surface. Hence, at each step of the minimization procedure, if the quantity of parameters was updated, we recomputed the ground temperature dynamics for all relevant sites. After a few iterations, we obtained values for the ground thermal properties for the 26 ecotypes and soil classes for Alaska. Some year-to-year differences between the modeled and observed annual cycles exist. However, in general, the modeled and observed ground temperature climatology agree well with each other as shown in Figure 7b, Figure 8b, Figure 9b and Figure 10b. The error is typically less than 0.328°C, except when the snowmelt occurs. Some water may potentially penetrate into the borehole and may warm the sensors.

2.5.1. Model Validation

Model validation usually comes down to the process of comparing model simulations to an experimental or observed data set. Comparison with a real data set is the most reliable and preferred way to validate a simulation model.

The result of modeled and observed MAGT and ALT comparison confirmed that the model is predictive under the conditions of their intended use and satisfactorily reflects the real permafrost conditions across the entire Alaska (Figure 11). The averaged values of MAGT and ALT are simulated over the various periods of time between 1997 and 2017. There is some degree of inaccuracy, but generally, the modeling results demonstrate that the model is a reasonable representation of the actual permafrost environment and in very good agreement with observed data.

2.6. Sensitivity Analysis

Uncertainties in the presented model lie in specifying the thermal parameters such as heat capacity of frozen and thawed soils and thermal conductivity. But mostly uncertainties are affected by thermal conductivity and the volumetric soil moisture content. The employed values for the soil properties are recovered by assimilating in situ temperature observations at a series of the Geophysical Institute, UAF and USGS sites across Alaska. The model is verified by comparing with available active layer thickness at the Circumpolar Active Layer Monitoring (CALM) sites (https://www2.gwu.edu/~calm/), perma-frost temperature, and snow depth records from existing permafrost observatories in the Alaska North Slope region. These data sets may contain some biases and errors, and hence, to gain some knowledge about how they act together and influence the results. We conducted a sensitivity analysis for a typical permafrost observing station at the region of discontinuous permafrost distribution. For a few other locations within continuous and sporadic permafrost regions, we obtained similar results.

For the Smith Lake station, a station with a typical natural condition black spruce ecosystem at the interior Alaska region, we analyze sensitivity of the mean annual ground temperature (MAGT) at 1 m depth and active layer thickness (ALT) with respect to the volumetric soil moisture content, thermal conductivities of thawed and frozen soils between 0.25 and 1.0 m depth (Figure 12b–d) and surface conditions such as moss and upper organic layer (Appendix B). Our previous investigations showed that variations in the heat capacities frozen and thawed soils do not significantly affect the modeling results for ALT and MAGT [21,37,51,52]. The obtained sensitivity results for thermal conductivity, both thawed and frozen soils are shown in Figure 12c, d. As the MAGT increased, the depth of the ALT increased as well.

The most influence on the average values of ALT and MAGT is imposed by the values of thermal conductivity of thawed soils within the 0.25-1.0 m depth (Figure 12c). The most impact on the MAGT is imposed by the values of thermal conductivity of unfrozen and frozen soils for the soil layers between 0.5 m and 1 m in depth. The value of frozen conductivity affects the soil layer below 1 m, and soil water content for the layer between 0.5-1.0 m soil layers. Values of the thermal conductivity for the upper layers are the primary function of the organic matter content and soil moisture. Additional illustrations on the model sensitivity to surface conditions, for example presence or absence of moss and organic matter at the upper part of the soil column are provided in Appendix B.

3. Discussion

The modeling result shows the most significant changes in the upper layer (0.2-5.0 m) of ground temperature, active layer thickness as well as talik development, occurred under the climate scenario provided by the GFDL-CM3 climate model both 4.5 and 8.5 RCPs. As described by the IPCC the RCP-4.5 is an intermediate scenario [12,13]. Increase in soil temperature in RCP-4.5 peak around 2051-2070, then observed relative decline. RCP-4.5 is more likely than not to result in global temperature rise between 2.5°C and 3.5°C, by 2120 (Table 3). In RCP-8.5, ground temperature continues to rise throughout the 21st century.

The developed ecosystem approach of permafrost dynamics modeling attempts to address the problem of computing the ground temperature projections at the regional scale using the very high-resolution input dataset. It is sufficient to gain some understanding of how potential changes in permafrost could affect ecosystems and impact the infrastructure under various climate scenarios. The developed parametrization of the ground thermal properties and incorporation in-situ data into the model, which is obtained using a data assimilation technique [37,50]. Generally, the modeling results very well agree with observations.

We run the GIPL2-MPI permafrost dynamics model in order to project permafrost dynamics for the specific sites located along the Alaskan Permafrost-Ecological Transect from north to south across the entire Alaska. The modeling ground temperature dynamics at 2 m depth during 2021-2120 according to the Five-Model Average climate forcing illustrated Figure 13.

According to the modeling result the time in the calendar years of the 0°C threshold crossing of MAGT at 2 m depth generally decreases from north to south (Figure 13). It means at the south of Alaska the soil temperature crosses zero Centigrade threshold earlier in comparison with northern part of Alaska. An exception may be the Gulkana observing site, located near the southern border of the discontinuous permafrost distribution at the south of Alaska as well (Figure 7b). Here, a zero curtain at 2 meters depth could exist around between 2020 and 2090 according to Five-Model Average of the RCP-4.5 scenario (Figure 13a) and between 2020 and 2075 if climate change occurs in accordance with RCP-8.5 scenario (Figure 13b). These circumstances can be explained that the borehole located at the Copper River Basin within the area of the former Late Pleistocene ice-rich (up to 40-60%) glacial-lacustrine sediments [59,60,61,62]. Our permafrost observations showed that in this area the permafrost temperature varies between -0.5°C and -0.9°C. The thickness of the permafrost is 50-40 m and within the range of depths 10-30 m the permafrost is isothermal (there is no temperature gradient). This fact suggests that in some locations across this permafrost region the phase transition process is already occurring. It is very close to the research that was done in this region [59,60,61,62] and to our long-term permafrost field observations [2,3].

4. Conclusions

The developed transient model and ecosystem approach allows computation of ground temperature at various depths with daily resolution, ALT and talik development at the regional scale that are sufficient to gain understanding how probable changes in the permafrost might affect ecosystems and impact infrastructure under various climate scenarios. The developed parametrization of the ground thermal properties is obtained using a data assimilation technique [21,37,50]. The model validation and modeling results demonstrate that the model is a reasonable representation of the actual permafrost environment across the study area and in very good agreement with observed data.

In this research, we demonstrate the projections for two GCMs and Five Models Composite for RCP-4.5 and RCP-8.5 scenarios only and illustrate a significant difference in the future near-surface ground temperature regimes between 2070s and 2110s. For the RCP-8.5 scenario, we found that ALT or top of permafrost deepening, up to 1.5 m on average in 2021, increases by a factor of 2.5 by 2070 and by a factor 3.6 from 2070 to 2120, according to the GFDL-CM3 RCP 8.5 scenario. ALT continues to increase and widespread talik starts to develop in interior Alaska and even within the individual areas of the North Slope Alaska region. According to NCAR-CCSM4 RCP-8.5 scenario the factor of permafrost top deepening increased by 1.8 from 2021 to 2070 and by the factor 3.1 from 2070 to 2120. Development of the talik will have serious implications for ecosystems, human activities (infrastructure and subsistence lifestyle), and potential feedback to climate change. On the other hand, for the RCP-4.5 scenario, the current model predicts only a modest increase in the near-surface permafrost temperatures and a limited degradation of the near-surface permafrost in the Alaska North Slope region. Finally, we mention that according to our computer experiments, in order to keep permafrost from the substantial thawing, the RCP-4.5 scenario most likely needs to be a target for greenhouse gas emissions. Appendix C illustrates the GeoTIF Maps of permafrost dynamics, generated by GIPL2-MPI using GFDL-CM3 and NCAR-CCSM4 and Five Model Average RCP-4.5 and RCP-8.5 scenarios.

5. Glossary of Permafrost and Related Ground-Ice Terms

Glossary and Comments to the Terms used in the current Report: [53,54,55,56] and Marchenko’s comments.

• Active layer - the layer of ground that is subject to annual thawing and freezing in areas underlain by permafrost.

Comment: In the zone of continuous permafrost, the active layer (seasonally freezing layer) generally reaches the permafrost table; in the zone of discontinuous permafrost, sometimes it does not. The active layer includes the uppermost part of the permafrost wherever.

• Kurum - A general term for all types of coarse clastic formations on slopes of 2-3 to 40 degrees, moving downslope mainly due to creep.
• Permafrost Table or Top of Permafrost - the upper boundary of permafrost surface.

Comment: The depth of this boundary below the land surface, whether exposed or covered by a water body or glacier ice, is variable depending on such local fac-tors as topography, exposure to the sun, insulating cover of vegetation and snow, drainage, grain size and degree of sorting of the soil, and thermal properties of soil and rock.

Comment: We used the term active layer thickness (ALT) in case of freeze-up seasonal frost with permafrost table every year. However, we are using the term ‘top of permafrost position’ or ‘table of permafrost position’ in case of the existence of residual thawed layer (talik) between seasonal frost and top of permafrost. In other words, the freeze-up does not exist anymore.

• Permafrost base - the lower boundary surface of permafrost, above which temperatures are perennially below 0°C (cryotic) and below which temperatures are perennially above 0°C (noncryotic).
• Permafrost thickness - The vertical distance between the permafrost table and the permafrost base.
• Pereletok - is a layer of frozen ground which forms as part of the seasonally frozen ground (in areas free of permafrost or with a lowered permafrost table), re-mains frozen throughout one or several summers, and then thaws.
• Protalus - A pronival rampart, formerly referred to as a protalus rampart, is de-fined as a ridge, series of ridges or ramp of debris formed at the downslope margin of a perennial or semi-permanent snow bed, which is typically located near the base of a steep bedrock slope in a periglacial environment.
• Talik - a layer or body of unfrozen ground occurring in a permafrost area due to a local anomaly in thermal, hydrological, hydrogeological, or hadrochemical conditions.

Comment: Taliks may have temperatures above 0°C (noncryotic) or below °C (cryotic, forming part of the permafrost). Some taliks may be affected by seasonal freezing. Several types of taliks can be distinguished based on their relationship to the permafrost (closed, open, lateral, isolated, and transient taliks), and based on the mechanism responsible for their unfrozen condition (hadrochemical, hydro-thermal and thermal taliks).

• Talus. - An outward sloping and accumulated heap or mass of rock fragments of any size or shape (usually coarse and angular) derived from and lying at the base of a cliff or very steep, rocky slope, and formed chiefly by gravitational falling, rolling, or sliding.
• Isolated talik - a layer or body of unfrozen ground surrounded by perennially frozen ground.
• Open talik - a body of unfrozen ground that penetrates the permafrost completely, connecting suprapermafrost and subpermafrost water.

6. Acronyms and Abbreviations

5-ModelAvg	Five-Model Average. Composite of five GCMs that best replicate historical climate in Alaska and the Arctic regions (NCAR-CCSM4, GFDL-CM3, GISS-E2-R, IPSL-CM5A-LR, and MRI-CGCM3).
AKVWC	Alaska Vegetation and Wetland Composite.
ALT	Active Layer Thickness.
AR5	The IPCC Fifth Assessment Report.
ASL	Above sea level.
APET	Alaskan Permafrost-Ecological Transect.
CALM	Circumpolar Active Layer Monitoring.
CMIP5	Coupled Model Intercomparison Project Phase 5.
CRREL	Cold Regions Research and Engineering Laboratory.
DEM	Digital Elevation Model.
DoD	Department of Defense.
ERDC	Engineer Research and Development Center.
FMA	Five-Model Average.
GCM	General Circulation Model.
GFDL-CM3	Geophysical Fluid Dynamics Laboratory Climate System Model.
GHG	Greenhouse gases.
GI	Geophysical Institute.
GIPL	Geophysical Institute Permafrost Lab.
GIPL2-MPI	The Geophysical Institute Permafrost Lab Permafrost Dynamics Model of the Second Generation, Parallel-coded using the Message Passing Interface.
GIS	Geographic Information System.
GTN-P	Global Terrestrial Network for Permafrost.
HPC	The high-performance computing.
HPCC	High Performance Computing Center.
IPCC	Intergovernmental Panel on Climate Change.
LTER	Long Term Ecological Research Network.
MAAT	Mean annual air temperature.
MAGT	Mean annual ground temperature.
MMGT	Mean monthly ground temperature.
MPI	Message Passing Interface.
NWI	National Wetlands Inventory
NCAR-CCSM4	National Center for Atmospheric Research Community Climate System Model.
pfTop	Table of Permafrost (top of permafrost).
RCSG	Research Computing Systems Group.
RCS	Research Computing Systems.
RCP	A Representative Concentration Pathway is a greenhouse gas concentration trajectory adopted by the IPCC. Four pathways were used for climate modeling and research for the IPCC Fifth Assessment Report in 2014.
RGM	Regional Climate Model.
SERDP	Strategic Environmental Research and Development Program.
SNAP	Scenarios Network for Alaska and Arctic Planning.
UAF	University of Alaska Fairbanks.
USARAK	United States Army Alaska.
USGS	United States Geological Service.
VWC	Volumetric Vater Content.