Arctic land hydrothermal sensitivity under warming: Idealized off-line evaluation of a physical terrestrial scheme in a global climate model


  • Kazuyuki Saito

    1. International Arctic Research Center, University of Alaska Fairbanks, Fairbanks, Alaska, USA
    2. Frontier Research Center for Global Change, Japan Agency for Marine-Earth Science and Technology, Yokohama, Japan
    Search for more papers by this author


[1] A series of idealized one-dimensional off-line sensitivity experiments for the Arctic hydroclimate under a transitional warming environment were conducted to investigate the impact of different internal mechanisms and external forcing (excessive water input) in a physical terrestrial scheme, Minimal Advanced Treatments of Surface Interaction and Runoff (MATSIRO), used in a global climate model. The scheme has freeze/thaw processes and was run with the column depth greater than 50 m. The inclusion of the top organic layers and the physically based parameterization of soil hydrothermal properties led to a realistic seasonal amplitude of a subsurface thermal regime and mitigated the warming and deepening of the maximum active layer thickness (ALT). After a 6-K warming over 100 years, ALT increases by 67% (from 9.6 to 16 cm) with the top organic layers and increases by more than a factor of 2 for the default mineral layers (122% from 45 cm to 1.0 m). With the more realistic thermal property profile, the physically based parameterizations projected ALT after the 100-year warming to be about a half of what the original parameterization did (2.0 against 3.5 m). Finer soil layer thickness near the surface had impacts on the near-surface wetness and the energy and water exchange between the atmosphere and also showed greater tolerance to the misconfiguration of porosity profile, whose global-scale distribution is poorly known for global climate model applications. With much wetter forcing, water infiltration kept the soil close to the saturation, degradation of the frozen state proceeded faster, and the adaptation back to a drier condition occurred on a decadal time scale.

1. Introduction

[2] For the future ecoclimate projection under global warming, key areas include the polar regions, where the state of the frozen ground plays an important role in determining the direction and degree of induced change on regional and global scales. A number of studies using global climate models (GCMs) have shown that the subsurface freeze/thaw process in the physical terrestrial scheme (or land surface scheme) is vital for appropriate simulation of the high-latitude hydroclimate [Intergovernmental Panel on Climate Change (IPCC), 2007]. Most of the current terrestrial schemes in GCMs are too simplistic to properly simulate frozen ground dynamics. They tend to reproduce an active layer that is too deep, or the absence of permafrost under present climate conditions, and they allow frozen ground to degrade too fast under global warming [Saito et al., 2007; Lawrence and Slater, 2005].

[3] The importance of the top organic layers and unfrozen water under the freezing point has been suggested, either from observational or numerical results, or both [e.g., Waelbroeck, 1993; Romanovsky and Osterkamp, 2000]. Recent studies assured the importance of these processes, as well as the total soil column depth, for proper reproduction of the thermal regime under current climate conditions [Saito, 2008; Nicolsky et al., 2007; Alexeev et al., 2007].

[4] Performance of physical terrestrial schemes (or any components in the global climate models) used in climate change studies should be examined not only under an equilibrium environment, but also (at least the response and adaptation features) under a transitional environment. Yi et al. [2007] investigated the impact of the top organic layer on the evolution of the maximum thaw depth under transient warming conditions.

[5] However, the changes in the subsurface hydrothermal regime, and heat and water exchange between the ground surface and the atmosphere, remain to be examined. A recent observational study suggests that a wetter environment may accelerate warming of the upper soil layer under a continuingly warming condition [Iijima et al., 2007]. In the present study, sensitivity experiments are conducted to examine the impact of those internal mechanisms and the external forcing (precipitation amount), with respect to the direction and magnitude of the impacts, under a transient warming environment.

2. Model and Experiment

[6] A series of 100-year sensitivity experiments were conducted with the one-dimensional physical terrestrial scheme Minimal Advanced Treatments of Surface Interaction and Runoff (MATSIRO), a component in the CCSR/NIES/FRCGC MIROC3.2 coupled atmosphere-ocean global climate model (CGCM) suite. MATSIRO is a land surface model developed for climate studies, representing the exchange of energy and water between the land surface and atmosphere, including the effects of snow, canopy structure, and vegetation biophysical processes. It was originally developed by Takata et al. [2003] and has been used for the CGCM integrations, including those submitted to the IPCC AR4. Details of MATSIRO are provided by Takata et al. [2003]. The analysis and validation of the simulated frozen ground in the IPCC AR4 integrations by the CGCM are presented by Saito et al. [2007].

[7] The forcing data used in the study is derived from the observations at Barrow, Alaska, from the period 1993 to 1998 [Mölders and Romanovsky, 2006]. The control forcing was set to the 1990s climatology calculated from the observation. Barrow was selected because the permafrost in the area has been surveyed by a number of studies and projects for decades, and is well understood [e.g., Romanovsky and Osterkamp, 2000; Hinkel et al., 2001; Mölders and Romanovsky, 2006]. Continuous and coherent observational data for the near surface air and the ground taken at the site for more than 10 years in a continuous permafrost zone are available. Moreover, the soil profile and its physical parameters have been investigated and presented by Mölders and Romanovsky [2006].

[8] Examination and validation of the conventional MATSIRO for freeze/thaw processes under the current climate are presented by Saito et al. [2007] and Saito [2008]. Saito et al. [2007] compared the seasonal thermal regime in the upper (∼20 cm) and deeper (∼3.0 m) soil with the observed values. They showed faster and excessive cooling in the cold region. Saito [2008] gives the roles of (1) snow cover and the organic layer and (2) soil column depth in the subsurface thermal regime were examined by comparison with the observations at Barrow, Alaska, and Tiksi, Siberia. The results reaffirmed the importance of the snow cover and top organic layer for more realistic simulation, and suggested a necessary total depth of 15 to 20 m for physically consistent results in the regions where ground freeze/thaw cycle is present.

[9] Each run started after a spin-up with control forcing, of 100 years for the 50-m column and 600 years for the 100-m column. Two transient warming forcing data, at full and half warming rates, were produced by adding linear trends, respectively +6 K and +3 K over 100 years, to the control forcing. These rates of warming were obtained from the output of the 200-year IPCC AR4 integration under 20C3M and SRES A1B scenario at the grid point that includes Barrow [Saito et al., 2007]. The full warming case is shown in Figure 1a as a series of an 8-year climatology (an 8-year period was chosen arbitrarily for presentation purpose. The choice of other periods did not change the nature of the results shown in the paper). Similarly, downward shortwave and longwave radiation trends were taken from the 200-year integration outputs for the same grid point by scaling to the respective +6- and +3-K warming, and added to the climatology data. To determine the possible impact of external moistening on the hydroclimate state, a “big-stick” approach was chosen in that the system was forced with precipitation increased by a factor of 5 relative to the control (156 mm a−1). The purpose is a qualitative sensitivity examination of the direction and the magnitude of response under the strongly moist environment, rather than a quantitative assessment of the response to a realistic increase. Excess surface water, either due to snowmelt or rain, is set not stored at the surface but removed immediately from the system as surface runoff in this study, although the scheme has a capability to handle surface water storage. Excess soil water after the saturation is similarly removed as ground runoff. Other forcing terms were kept at the control values in all runs. The controlling factors and the conditions of the experiments are summarized in Table 1. The set of the experiments is summarized in Table 2. The values of the thermal and hydraulic soil parameters used in the study are given in Tables 3 and 4.

Figure 1.

(a) The series of surface air temperatures with the linear trend of 6 K in 100 years, shown by climatology series computed for each successive 8-year period (total of 12 periods in 100 years starting from year 3). Contour plot of soil temperature by depth for (b) experiment 7 (with top organic layers) and (c) experiment 8 (mineral layers only, GCM's default value). Contour interval is 2°C. Positive temperatures in degrees centigrade are drawn in red, 0°C are in black, and negative temperatures are in blue.

Table 1. Summary of the Controlling Factors in the Experiments and Explanation of Symbols for the Experiment Codes
Atmospheric Forcing
0constant 1990s climatology temperature and radiations
3(Ta, +3.0 K; L↓, +16.1 W m−2; S↓, −4.1 W m−2) per 100 years trend added
6(Ta: +6.0 K; L↓, +32.2 W m−2; S↓, −8.2 W m−2) per 100 years trend added
Total Soil Column Depth and Layer Thicknesses
C12 layers for 100 m at 0.01, 0.03, 0.08, 0.20, 0.45, 1.0, 2.0, 3.5, 6.0, 10, 20, 100 m
D13 layers for 50 m at 0.05, 0.25, 0.50, 1.0, 2.0, 3.5, 6.0, 9.0, 14, 21, 30, 40, 50 m
Hydrothermal Parameterization
Cconventional parameterization and effect of soil moisture only
Peffect of soil moisture and ground ice
Reffect of soil moisture and ground ice and presence of unfrozen water under 0°C
Porosity Profile
2values for Barrowa
1values reduced uniformly to 0.439
Physical Property of Soil Layers Factors
Mdefault values at 0.25 W m−1 K−1 (no-organic layers)
Oorganic layer (0.01 W m−1 K−1) at top 20 cm, otherwise default value
Bvalues for Barrowa
(null)1990s climatological values (pc = 156 mm)
_P5 times greater than pc for 100 years
_P05 pc for first 50 years, pc for second 50 years
Table 2. List of Experiments
Experiment NumberCode
Table 3. Thermal and Hydraulic Parameters of the Different Soil Layers Used in the Experiments for Barrowa
Soil LayerCs (106 J m−3 K−1)Ks (W m−1 K−1)θsat (m m−1)ψsat (m)Ksat (m s−1)b
  • a

    Here Cs and Ks are heat capacity and thermal conductivity of soil material, θsat is the saturated volumetric water content (porosity), ψsat is saturated matric potential, Ksat is the saturated hydraulic conductivity, and b is a pore size distribution index. Values are taken from Mölders and Romanovsky [2006].

Top organic (silt with organics)0.8511.0500.6−0.3891.1342.545
Mineral (silt)0.9710.7540.55−0.3962.4932.6875
Deeper than 20 m (silt and sand)1.160.6140.44−0.40591.622.85
Table 4. Thermal and Hydraulic Soil Parameters Used for Experiments 7 to 10a
Soil TypeCs (106 J m−3 K−1)Ks (W m−1 K−1)θsat (m m−1)ψsat (m)Ksat (m s−1)b
  • a

    Symbols are same as in Table 3. The values for the organic layer are taken from Beringer et al.'s [2001] Table 1 for peat except for the thermal conductivity.

Medium-size mineral (default)−0.35483.385.25

[10] The physically based improvement of the parameterization of the thermal properties (thermal conductivity and heat capacity) was done in the following manner. The thermal parameterization for heat capacity C and thermal conductivity K in the original scheme [Takata et al., 2003] were calculated as

equation image
equation image

where Cs and Ks are heat capacity and thermal conductivity of soil material (mineral, organic or the mixture of the both); ρw and Cpw are density and specific heat capacity of liquid water; w is total soil water content, A, B are empirically determined constants (respective default values are 0.25 and 6.0). This parameterization is referred to as the “conventional” or c case in the paper. It does not include the ground ice contributions to C and K.

[11] Since ice has a greater thermal conductivity but a smaller heat capacity than liquid water, the conventional parameterization underestimates the former and overestimates the latter when ground ice is present. A new thermal parameterization considers the effect of both liquid and solid water by the arithmetic average for heat capacity and the geometric average for thermal conductivity [cf. Lachenbruch et al., 1982]:

equation image
equation image

where ρi and Cpi are density and specific heat capacity of solid water; wl and wi are liquid and solid soil water content (w = wl + wi); and Kw and Ki are thermal conductivity of liquid and solid water. Heat capacity and thermal conductivity of air is neglected. The runs with this improvement but no unfrozen water are referred to as the “partially improved” or p runs hereafter.

[12] MATSIRO prognosticly calculates the vertical soil moisture flux in the unsaturated soil according to Richard's equation with the matric and gravitational potential. In the conventional scheme, no liquid water is allowed under the freezing temperature Tfreeze (set to 237.15 K). Soil temperature is kept at Tfreeze until all the soil moisture in the layer is frozen (or totally melted, in case of thawing). In the formulation used here, unfrozen water content at soil temperature Tg (in kelvins) under Tfreeze is parameterized after Flerchinger and Saxton [1989] as

equation image

where λmelt is the latent heat of melting (3.4 × 105 J kg−1), g is acceleration of gravity (9.8 m s−2), ψsat is saturated matric potential, and b is a pore size distribution index. The solute effect is neglected. The integrations with both the thermal and hydraulic improvement are referred to as the “refined” or r runs.

3. Results

3.1. Thermal Regime

3.1.1. Top Organic Layer

[13] In the full warming environment (6-K increase over 100 years), the maximum active layer thickness (ALT) increases by 67% (from 9.6 to 16 cm) with the top organic layers of 20 cm included, and more than doubles (from 0.45 to 1.0 m) with only the default mineral layers (Figures 1b and 1c) because of faster and deeper penetration of heat flux to the ground. The thermal and hydraulic parameters for the both soil textures are given in Table 4. The annual soil temperature amplitude was constantly greater without the top organic layers, almost twice the amplitude at 1 m depth during the last 8-year period (years 91–98). Ground ice melted but coexisted with liquid water at the bottom of the active layer for the mineral case during whole summer, forming a persistent zero curtain in the active layer. The freezing of the active layer followed a transition similar to the observed [cf. Hinkel et al., 2001, Figure 4], although upfreezing from the bottom of the active layer was not simulated in this study. By contrast, the top of permafrost remained within the organic layer after 100 years under the full warming in the organic case. A greater permeability of the organic layer (see Table 4) removed the liquid water effectively away from the melting front, leading to further ground ice melting in the layer until the ice completely melted, and soil temperature was allowed to rise above 0°C. This limited the period during which the active layer remained at 0°C to only the beginning and end of the thawing season.

3.1.2. Hydrothermal Parameterization

[14] The conventional parameterization increased the ALT from 1.6 to 3.5 m under 100-year full warming scenario (Figure 2c), where the values adapted to Barrow (taken from Mölders and Romanovsky [2006]) were used for the profile of porosity and soil thermal property and summarized in Table 3. This confirms the 200-year online results that the high-latitude soil temperatures increased too fast and ALT deepened too much, as shown by Saito et al. [2007]. ALT changed from 1.0 to 2.0 m in the partially improved case (Figure 2b), and from 0.85 to 2.0 m in the refined case (Figure 2a). The refined thermal parameterization (equations (3) and (4)) simulated a reasonable seasonal amplitude under the current climate condition. At the deepest layer of the observation available for the 1990s (1.0 m depth), the observations showed −1.9°C for maximum and −17.4°C for minimum (range of 15.5°C). The simulated seasonal soil temperature in the organic case was −1.4°C for maximum and −18.3°C for minimum (seasonal range is 16.9°C). In contrary, the maximum temperature was 1.5°C and the minimum was −12.3°C (annual range of 13.8°C) in the conventional case. The amplitude was smaller and shifted to warmer values. The refined parameterization also prevented excessive downward heat fluxes under the surface. Unfrozen water (equation (5)) further regulated the heat conduction, especially when the soil temperatures were so close to the freezing point that no prolonged zero curtain was produced.

Figure 2.

Same as Figure 1b except for (a) experiment 1 (physically based parameterization with unfrozen water; refined), (b) experiment 13 (physically based parameterization only; partial), and (c) experiment 14 (conventional parameterization). (d–f) Same as Figures 2a–2c except for excessive precipitation cases (experiments 17, 18, and 19).

3.1.3. Excessive Precipitation

[15] The climatological (Figures 2a2c and 3a) and the excessive precipitation cases (Figures 2d2f and 3b) showed a similar thermal regime and ALT under the current climate condition (first period), aside from a smaller seasonal amplitude and a longer 0°C period for the latter. However, the simulations show substantially different responses to warming: the subsurface thermal regime warmed at a faster pace with more dampened seasonal amplitudes in the excessive water input cases. Among those, the refined case produced the most conservative result, while the conventional and partial parameterizations resulted in layers permanently thawed (or at best at 0°C) below 1 m in some decades after the integration started. The increase of winter soil temperature was mitigated considerably when unfrozen water was considered (Figure 3a, lines at bottom).

Figure 3.

Soil temperature (lines at the bottom of each plot) and soil water content (lines at the top of each plot) at the depth of 0.45 m when prescribed precipitation is at the level of (a) the 1990s climatology for the whole period, (b) excessive precipitation for the whole period, and (c) excessive precipitation for the first 50 years and then the 1990s climatology for the second 50 years. Black lines represent the results from experiments with the refined parameterization. Blue and red lines show the results from partial and conventional parameterization, respectively.

3.2. Hydrological Response

[16] Implementation of unfrozen water changed the seasonal hydrological regime as well. With the excessive precipitation forcing, the upper soil layer remained saturated for all seasons for the first four decades in all cases (Figure 3b), while it was between 15 and 50% saturation for the climatology case (Figure 3a). Inspection of Figures 3a and 3b shows that, when soil was close to saturation, the presence of unfrozen water (difference between the r (black) and p (blue) runs) made a relatively small difference in both the thermal and hydrological regimes for the examined period: the r and p runs produced an almost identical results for soil temperature at 45 cm in the wet case (Figure 3b), while the p results for the dry case (Figure 3a) deviated more from the r results as the warming continued. As for the hydrological regime, the p results were closer to the r results for the first two cycles in the dry case (Figure 3a), then eventually became closer to the c results in the end. Under the wet conditions (Figure 3b) also, the hydrological response was almost identical between the p and r runs. The relative impact of unfrozen water is at minimum when saturated, since it is the absolute amounts of unfrozen water that is determined by the equation (5) as a function of soil temperature and soil porosity, not the percentage relative to the present soil water content. Water flow is also limited when it is frozen and close to the saturation. In contrast, because of the differences in parameterization schemes for thermal conductivity and heat capacity for a frozen state, the difference between the r and c (red) runs are already apparent regardless of the level of saturation (see equations (1)(4)).

[17] Another interesting result is the decrease in soil moisture (drying) during the winter period; with the excessive precipitation, the decrease first appears in the c case in year 28, starting from the topsoil layer, cascading down afterwards. This became visible in Figure 3b only after the fifth period in the 8-year climatology sequence. Figure 4 compares the hydroclimatic differences at the surface and at the depth of 45 cm in the consecutive 2 years in transition (years 27 and 28 for c runs). It illustrates the following physical mechanism of drying of the upper soil layer under the warming condition. In year 27, the bottom of the seasonal active layer was at 45 cm. Until this year, the permeability of the layer (the interface to the top of the permafrost table) was active only for the limited period (days 220–260). However, in year 28, water could still infiltrate into the layer even after the beginning of winter season (around day 260) because seasonal active layer became deeper in the year, and the thawed period for the ground above 45 cm became longer. Moreover, the delay in the start of the snow accumulation and freezing at the surface elongated the period of surface infiltration into the active layer. The end-of-season drainage of groundwater in the active layer to runoff, which occurred just before its shutting down by the freezing, decreased the amount of soil moisture at the bottom of the active layer. The layer remained drier until the next snowmelt season. Since the warming impact was strongest and appeared earliest in the c case, this phenomenon started to occur first in the c case in year 28. In the p case, exactly the same mechanism started later in year 66. In fact this drying phenomenon was more clearly seen in the drier (climatological) case (Figure 3a). The winter water content was drier by 0.1 m (1 m)−1 with the unfrozen water mechanism than that without it (Figure 3a, lines at top), because the above mentioned mechanism worked more efficiently when water was still allowed to flow in subsaturation situations.

Figure 4.

Seasonal change of subsurface hydrothermal regime for the conventional parameterization in (a) year 28 and (b) year 27 at the top layer (1.0-cm depth; scale at left; solid lines) for soil temperature (red lines; °C), soil moisture content (thick blue lines; dm m−1), ground ice content (thin blue lines; dm m−1), evaporation (thin black lines; ×10 W m−2), and snow accumulation (thick black lines; ×100 kg m−2). Dashed lines are for the 45-cm deep layer (scale at right) for soil temperature (red lines; °C), ground ice content (thin blue lines; dm m−1), and runoff (thin black lines; mg m−2 s−1).

[18] After a sudden forced decrease in precipitation to the control value in year 50, it took about three periods for the hydrothermal regime to adapt fully to the drier condition, irrespective of the level of the hydrothermal parameterization improvement (Figure 3c; see Figures 3a and 3b). With the improvement of thermal and hydrological parameterizations, early adaption in winter was much faster.

4. Discussion

[19] Figure 5 summarizes the results of the study with respect to surface heat and water flux to the atmosphere (Figure 5a), subsurface thermal (Figure 5b) and hydrological regime (Figure 5c). In the following we discuss the influences of the examined issues on the hydrothermal regime of the soil and the atmosphere.

Figure 5.

Summarizing plots of the sensitivity experiments for the last 8-year period (years 91 to 98), averaged for the snow-free (open diamonds and squares) and the snow-covered (solid diamonds and squares) periods for (a) sensible heat fluxes (diamonds) and latent heat fluxes (squares), (b) soil temperature at 0.5 m (diamonds) and 2.0 m (squares), and (c) water saturation (diamonds) and ratio of ice to water in the top 1-m layer (squares). The snow-covered period is defined as the continuous period when snow depth is greater than 10 mm, and the snow-free period is determined as 10 days after the complete snowmelt to 10 days before the continuous snow accumulation for each experiment.

4.1. Warming Versus Equilibrium Runs

[20] Comparison between experiments 1–3 gives the difference of the hydrothermal responses under different climatic forcing trends. The soil temperature increases were found at both the 0.5 and 2.0 m depths, with larger increases in summer at 0.5 m depth (8°C in summer and 5°C in winter), but smaller at 2.0 m (5°C in summer and 6°C in winter). In this study, seasonality of warming was not considered, although greater warming has been observed in winter than in summer [Osterkamp, 2005], and also projected for the future [Saito et al., 2007]. Winter warming, however, exerts less impact on the subsurface thermal regime because of the insulating effect of snow cover over the ground surface. Qualitative examination of the impact of the seasonality of warming will be a topic of the future research.

[21] Summer evaporation (from land to the atmosphere) decreased by about 15 W m−2 for the 6-K warming, corresponding to the drying tendency in the topsoil layers. It is worth mentioning that the amount of liquid water (i.e., water saturation times porosity times (unity minus ice ratio)) did not change substantially in the upper 1-m layer. In summer it ranged between 15.4 and 16.0%, while the ice content in the same season dropped from 60% of total water to 10%. Greater availability of moisture in the top 1 m of ground maintained more active and longer evaporation to the atmosphere in the season in the c run.

4.2. Soil Column Settings

[22] Columns C and D are different both in total depth (100 m for C and 50 m for D) and in near-surface layer thickness. Column C has finer layers near the top surface than D, e.g., the first layer is 1 cm for C and 5 cm for D. To ensure the thermal equilibrium in the initial conditions, all the experiments listed in Table 1 started after 600 years of spin-up for the column C cases, and after 100 years of spin-up for the column D cases, as previously noted.

[23] The total depth of the both columns was set greater than 15 m, which was found to be the minimum depth required for a simulation of the seasonal thermal regime in the polar regions to be physically sound and consistent with the boundary conditions [Saito, 2008]. In the control run, the zero amplitude was simulated in the layers below 20 m for both the columns. This is in a good agreement with the observed value of ∼20 m at the Barrow site (V. E. Romanovsky, personal communication, 2008). Under the full warming forcing (6.0°C warming in a 100 years), the mean annual amplitude at the 20 m depth in the last (twelfth) period was 0.07°C for both the case. Similarly, the warming at the depth after 100 years was similar in both cases: 3.77°C for column C, and 3.97°C for D. Those values are not unrealistic considering that warming in the range of 0.6°C to 1.9°C was observed at several sites on the Arctic Coastal Plain in Alaska between 1985 and 2004 [Osterkamp, 2005].

[24] Experiments 1–3 and 4–6 show the impact of the soil column depth with the porosity profile that was derived from model fitting to the observations [Mölders and Romanovsky, 2006]. Somewhat contrary to the expectation, no substantial sensitivity was found in any evaluated variables in this case. Next, interactions of the near-surface layer thickness and the porosity profile were investigated by specifying a value of 0.439 for porosity in all layers. A reduction in porosity means smaller capacity of water holding in the soil. The comparison of the results from experiments 11 and 12 with those from 1 and 4 delineate the interactions of the reduced porosity and soil layer thickness. Both the near-surface hydrological states and exchange of water and energy between soil and the atmosphere were affected; with the reduced porosity and coarser near-surface layers (experiment 12, compared to experiments 4 and 1), the top 1 m of soil became wetter and closer to saturation, and the surface available energy was prone to be partitioned more to evaporation (i.e., the Bowen ratio became less than one). Porosity configurations in GCMs are often very crude or even erroneous. However, this result suggests that, when the fine details of the near-surface layers are increased, the model results are less vulnerable to those inaccuracies.

4.3. Top Organic Layers

[25] Loss of the top organic layer, for example by forest fire, drastically changes the subsurface thermal regime. It has widely been found that burned forests, which lose their top organic layer by the fire, show warmer temperature in summer and deeper active layer than the adjacent unburned sites [e.g., Yoshikawa et al., 2002]. For the modeling of climate that includes ground freeze/thaw processes, the importance of the organic layer has also been noticed for more than a decade [e.g., Waelbroeck, 1993]. Yi et al. [2007] showed that inclusion of an organic layer buffered the unrealistic degradation of permafrost. The thermal effect of the top organic layer is primarily insulation between the atmosphere and the ground, especially in summer, when the ground is kept cooler because of the low thermal conductivity of the organic soil. In winter, however, the organic layer can be more conductive of heat because of the ice effect incorporated in equation (3). In the northern Alaska, the soils are usually wet in winter.

[26] Figure 5c (comparison among experiments 7–10) suggests that, with the top organic layer, water saturation was kept at a constant level under the warming environment (experiment 7 versus experiment 9), while the total water content was halved during the course in the mineral case (experiment 8 versus experiment 10). More rapid degradation of the thermal regime for the no-organic cases was previously shown in section 3. Formation of the deeper active layer is also shown for the summer soil temperature at 0.5 m (experiment 8; open diamond in Figure 5b). For the both cases, summer surface eddy heat fluxes decreased. However, sensible heat was more sensitive than latent heat to the inclusion of the top organic layers, possibly because of the slower soil warming relative to the air temperature warming, and constant soil moisture level. All these results reaffirm the importance of the organic layer for the water storage capability of soil under the warming environment [Waelbroeck, 1993], and also more correct accounting of changes in heat and energy exchange between land and the atmosphere.

4.4. Hydrothermal Parameterization

[27] Comparison of experiments 13 and 14 with 1 shows the effects of the thermal parameterization improvement, and the addition of the unfrozen waster mechanism, on the evaluated variables under the full warming condition. Similarly, comparison of experiments 15 and 16 with 3 contrasts the results under the climatological condition. The major deficiencies of the conventional thermal conductivity parameterization (equation (2)) are that it lacked the effect of solid ice, which has larger thermal conductivity than liquid water does. Also, the inadequate functional form of the conventional parameterization, leads to a substantial overestimation under the subsaturation conditions, regardless of the ratio of ground ice in a layer, as is easily seen from equations (2) and (4). Introduction of unfrozen water below the freezing point (equation (5)) enables a smoother and more realistic transition between the suprazero and subzero thermal regime in the soil. Figures 2 and 5c show respectively those thermal and hydrological impacts obtained by the improvement of the thermal parameterization, together with the presence of the unfrozen water. Both show the significant importance of retaining the summer ground ice in the upper soil layers under the current climatological conditions, and in mitigating the deepening of the maximum active layer thickness under the warming environment. To conclude, however, the presence of the unfrozen water may be more important in keeping ground ice from rapidly melting by providing larger thermal inertia and smoother phase changes between the frozen and thawed regime. The increase in soil temperature at the 20 m depth after the 100-year full warming was 3.77°C (see section 4.2) for the r case, 6.09°C for the p case, and 4.87°C for the c case. The increase was most conservative when both the thermal and the hydrological improvement were considered. This may still be instrumental to notice that the modeled maximum temperatures within the active layer were more or less the same (e.g., summer temperature at 0.5 m in the full warming cases, experiments 1, 13, and 14 in Figure 4b), because the both refinement makes a difference only for the frozen state.

4.5. Wet Environment

[28] There is an observational indication that a temporal wetting of soil under a warming environment in the Arctic may trigger large subsurface warming [Iijima et al., 2007]. We attempted to evaluate the impact of the wetter environment under warming conditions. The projected increase in annual precipitation in the Arctic region in IPCC AR4 was about 18% in median and no more than 40% [IPCC, 2007] (short of the 40% increase at the grid containing Barrow in the high-resolution MIROC3.2 A1B simulation). However, to highlight the possible different responses in the wetting environment, a “big-stick” approach was chosen in that the system was forced with precipitation of a fivefold excess. A distinctive response was obtained as displayed in Figures 2, 3, and 5. As Figures 3a and 3b contrastingly show, a warmer subsurface thermal regime (especially in winter) was obtained by the wetting of the soil under a warming environment. The thermal regime difference between the dry and wet environment is smallest in the r run. This implies that the terrestrial scheme that takes into account the more realistic thermal parameterization (i.e., equations (3) and (4)) and coexistence of solid and liquid water below the freezing point (i.e., equation (5)) will be more tolerant of deficiencies in projections of the future atmospheric states. Once the atmosphere changes to a dry condition, adaptation to the drier regime (as in Figure 3a) was slowest in the c run (Figure 3c).

[29] When water is drained and soil gets drier in the northern regions, evaporation will decrease, which is the major source of the region's summer precipitation. This will further dry ground, and decrease the soil heat capacity, possibly leading to an acceleration of subsurface warming under warming environment. Although the resolved scale is different from the one at which this study is aiming (i.e., usage for global or regional scale climate modeling: typically from 5 km to 100 km), the hydrological aspect of this transition from wetting to drying is relevant to the early stage of thaw lake development (i.e., from initial flooding to lake drainage) [e.g., Jorgenson and Shur, 2007]. Thaw lakes and drained thaw lake basins account for about half the surface areas on the North Slope of Alaska (median area of lakes and basins is 12.2 ha and 19.3 ha, respectively) [Frohn et al., 2005]. Thaw lake is one of the important factors in evaluating the impacts of their changes on the northern hydroecoclimate system under relatively rapid climate change. As mentioned in section 2, this physical terrestrial scheme has a capability to handle surface water storage. However, it was decided not to activate it in this study because the scheme still lacks subgrid-scale lateral transport of water and heat, and coupling of heat and water in vertical or lateral transfer (i.e., convective heat transfer by water or vapor is not implemented yet). We think these can be critical issues, and are at an attempting phase for implementation. The impact of hydrological environment change is worth the examination, but not by an offline simulation like in this study since land-atmosphere interactions (e.g., evaporation minus precipitation) are strongly concerned. This needs to be done with an atmospheric circulation model coupled with the terrestrial scheme that has the above processes feasible.

4.6. Future Implications

[30] This study demonstrated impacts and sensitivity of the internal mechanisms in the physical terrestrial scheme and excessive precipitation on the arctic subsurface and near-surface hydroclimate under a transient warming environment, although in an idealized framework. In the next generation CGCM experiments to investigate more comprehensive changes in global climate, ecosystem or dynamic vegetation will be included as important components in the CGCM suite. A result from a global ecosystem model SEIB-DGVM simulation for a Siberian larch forest pointed to a critical role of permafrost for the forest's existence (H. Sato et al., Modeling vegetation structure and function in an east Siberian larch forest using the dynamic vegetation model SEIB-DGVM, submitted to Global Biogeochemical Cycles, 2008). To provide the proper and realistic physical foundation for the simulation of cold region ecosystem simulations, adequate treatment of the freeze/thaw processes and accurate realization of the hydrothermal regime underground will be critical.

[31] The targeted area in this study was arctic tundra; specifically, forcing data was derived from observations at Barrow, Alaska. However, different responses may be obtained for another climate zone or area (e.g., Barrow and Tiksi shown by Saito [2008]; discontinuous, sporadic or isolated permafrost zones will likely respond with different sensitivity to produce different results [e.g., Delisle, 2007]. Moreover, with online experiments coupled to the atmospheric and other GCM components will further verify the results obtained in this study and establish their generality.

5. Conclusion

[32] In this study, a series of idealized one-dimensional off-line sensitivity experiments for the Arctic subsurface hydroclimate were conducted under climatological (equilibrium) and transitional warming environments, in an attempt to investigate the qualitative and quantitative impacts of different internal mechanisms and external forcing (excessive water input) in a physical terrestrial scheme MATSIRO used in a MIROC CGCM. We obtained the following major conclusion from the study.

[33] 1. The top organic layer showed an important insulating effect under warming environments. Under a 6-K increase of surface air temperature over 100 years, the ALT increased by 67% (from 9.6 to 16 cm) with the top organic layers, compared to a more than twofold increase for the default mineral layers (122% from 45 cm to 1.0 m).

[34] 2. The physically based parameterization of soil hydrothermal properties was also found to be important. With a more realistic thermal property profile, the improved thermal parameterization and presence of unfrozen water under the freezing point retained a realistic seasonal amplitude of a subsurface thermal regime, and mitigated the warming and deepening of the maximum active layer thickness: after the 100-year warming the active layer was about half as thick as with the original parameterization (2.0 m against 3.5 m).

[35] 3. Finer soil layer resolution near the surface had impacts on the near-surface wetness and on the energy and water exchange with the atmosphere. This also implies greater tolerance to the misconfiguration of porosity profile, whose global-scale distribution for use in GCMs has large uncertainties.

[36] 4. Under large increases of precipitation applied in a “big-stick” approach, water infiltration kept the soil close to the saturation, degradation of the frozen state proceeded faster, and the readaptation back to a drier condition occurred on a decadal time scale. The atmospheric hydroclimate therefore has a great potential to affect the course of subsurface hydrothermal regime change in future climate scenarios.


[37] All numerical integrations were run on the NEC SX-8 supercomputer at JAMSTEC. The author would like to thank the researchers at the permafrost laboratory, Geophysical Institute, University of Alaska Fairbanks, for providing observational data and John Walsh and two anonymous reviewers for their valuable comments to improve the manuscript. This work was supported by the following funds: (1) the Global Environment Research Fund (B-061) of the Ministry of the Environment, (2) Ministry of Education, Culture, Sports, Science and Technology (MEXT), Japan, and (3) the National Science Foundation under agreement ARC-0327664, USA.