Dynamics of the larch taiga–permafrost coupled system in Siberia under climate change

Larch taiga, also known as Siberian boreal forest, plays an important role in global and regional water–energy–carbon (WEC) cycles and in the climate system. Recent in situ observations have suggested that larch-dominated taiga and permafrost behave as a coupled eco-climate system across a broad boreal zone of Siberia. However, neither field-based observations nor modeling experiments have clarified the synthesized dynamics of this system. Here, using a new dynamic vegetation model coupled with a permafrost model, we reveal the processes of interaction between the taiga and permafrost. The model demonstrates that under the present climate conditions in eastern Siberia, larch trees maintain permafrost by controlling the seasonal thawing of permafrost, which in turn maintains the taiga by providing sufficient water to the larch trees. The experiment without permafrost processes showed that larch would decrease in biomass and be replaced by a dominance of pine and other species that suffer drier hydroclimatic conditions. In the coupled system, fire not only plays a destructive role in the forest, but also, in some cases, preserves larch domination in forests. Climate warming sensitivity experiments show that this coupled system cannot be maintained under warming of about 2 °C or more. Under such conditions, a forest with typical boreal tree species (dark conifer and deciduous species) would become dominant, decoupled from the permafrost processes. This study thus suggests that future global warming could drastically alter the larch-dominated taiga–permafrost coupled system in Siberia, with associated changes of WEC processes and feedback to climate.


Introduction
The world's forests influence planetary energetics, the hydrologic cycle, and atmospheric composition and climate through physical and biological processes (Bonan 2008). As one of the largest terrestrial biomes, Siberian boreal forests store huge amounts of carbon in their biomass and soil, and play an important role in climate feedback processes (e.g., lowers surface albedo, masking the high albedo of snow in winter, and large evapotranspiration in summer) (Bonan 2008, Lopez et al 2008. Therefore, taiga is thought to play an . In the schematic diagram: arrows show force directions between each factor. '+' ('−') means the following the forces direction, there is a positive (negative) correlation between two factors, and '±' means the correlation is still not clear. The thicker arrows indicate the processes which were better considered in the model simulation. Dashed arrow indicates the process which is not considered in this study. The words beside each arrow indicate the variables that control such process. important role in both the global carbon balance (Piao et al 2008, Euskirchen et al 2006 and regional or continental-scale hydroclimate (Betts 2000, Bonan et al 1992, Chapin et al 2005. Uniquely in eastern Siberia, the distribution of the nearly homogeneous larch-dominated taiga highly coincides with the zones of continuous and discontinuous permafrost (Osawa et al 2009, Stolbovoi andMcCallum 2002) (figure 1). Unlike other boreal ecosystems, recent in situ observations have shown that the larch taiga-permafrost system in Siberia has displayed distinct water-energy-carbon (WEC) exchange characteristics including in its temporal and spatial variations (Tanaka et al 2008, Sugimoto et al 2002, Ohta et al 2001. It suggests that larch has adapted better than other species to the permafrost environment, forming a larch-dominated taigapermafrost coupled system (Osawa et al 2009). Although soil freezing-thawing processes and permafrost have been introduced to a number of vegetation models to improve simulations of fire disturbance, soil respiration, and vegetation dynamics (Beer et al 2007, Sato et al 2010, Tchebakova et al 2009, Wania et al 2009, most of these studies focused on carbon exchanges and emphasized vegetation responses to climate variation. In other words, few models have treated 'larch taiga-permafrost' as a coupled system in terms of water and energy exchanges, which could be distinctly different from other boreal forest biomes.

Methodology
To address this issue, we need to introduce a new scheme that integrates vegetation and permafrost, two major components in the unique taiga-permafrost system, which emphasizes the hydrological feedbacks among these processes (figure 1).
Moreover, we also introduced a new biogeophysical process into DV-FSM in order to simulate the responses of vegetation physiological processes to soil hydrological seasonal and inter-annual variability, which in another sense link the simulation of long-term forest ecological dynamics and short-term soil hydrological processes (see supplementary, 'Model Description' available at stacks.iop.org/ERL/6/024003/mmedia). Using the model, this study aimed to prove the importance of the interactions of permafrost and Siberian larch taiga as a coupled system. Meanwhile, we conducted an additional experiment to evaluate the sensitivity of this coupled system to future warming climate. As a validation of the model experiment, we adopted the observed vegetation-permafrost processes at Spasskaya Pad near Yakutsk in eastern Siberia (Russia), where continuous observations of energy, water, and carbon fluxes of permafrostvegetation coupled system have been made since 1998 (Ohta et al , 2001. We conducted different simulations, all forced by the same 1 year length climate forcing dataset. The dataset is derived from averaged half-hourly observations from 2000-4. Time-integration of 1000 years for each plot-scale simulation started from bare ground condition was repeated eight times and then averaged for broader spatial scale. To clarify the importance of the existence of permafrost for maintaining the Siberian larch taiga-permafrost ecosystem, the following two sets of experiments were conducted. In the first set of experiments, we introduced four different runs: control run (CNTL, permafrost-vegetation-fire coupled run), no fire run (NF, without consideration of fire disturbance), no permafrost run (NP, without consideration of permafrost effect), and no permafrost and fire run (NPF). In the NP and NPF runs, we kept the soil water freezing-thawing process in the model, but not considering the effect of soil ice blocking the percolation of soil water. This blocking effect was considered to be a unique and essential feature of the permafrost system.
In the second set of experiments, four groups of simulations with different warming intensity were conducted to examine the sensitivity of the taiga-permafrost system to climate warming (see supplementary, 'Method' available at stacks.iop.org/ERL/6/024003/mmedia). We set warming intensity of study site to approximate 4.5 • C at the end of the 21st century projected by the Fourth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC-AR4). Similar to the previous experiment, each group included two runs in which we turned on/off the permafrost effect. In three groups, summer (JJA) surface air temperature were added by 1.12 • C, 2.25 • C, and 4.5 • C (abbreviated below as, for example, +1.12 • C), respectively, over the original forcing data. In another group, summer surface air temperature of +4.5 • C and precipitation +20% were added to examine the taiga-permafrost sensitivity to precipitation change under the strong warming condition. Hereafter, CNTL 1C represented the control run with temperature of +1.12 • C, and NP 1C was non-permafrost run with temperature of +1.12 • C. Accordingly, CNTL 2C & NP 2C and CNTL 4C & NP 4C represent the runs with temperature of +2.25 • C and +4.5 • C, respectively.

Results
The model control (CNTL) run was conducted to reproduce the taiga and permafrost interactions that were observed in long-term field experiments (Osawa et al 2009. The model successfully reconstructed the larchdominated forest in assembled simulation (figure 2) with birch and pine appearing occasionally as sub-dominant or scattered tree species. The simulation result coincides with the in situ observation (Osawa et al 2009 (also see supplementary 'Soil Water Content' available at stacks.iop.org/ERL/6/024003/mmedia) and in the large-scale  (Shvidenko et al 2007, Bonan andShugart 1989). The model also reproduced well the observed recovering process of forest biomass after fire disturbance (see supplementary, 'Fire' and figure 1 available at stacks.iop.org/ ERL/6/024003/mmedia). The annual maximum thawed layer thickness (i.e., the active layer is the surficial layer above the permafrost which thaws during the summer (Burn 1998); hereafter abbreviated as ALT) could reach 1.8 m in the model, which is close to the observed depth of 1.5 m. The model also well reflected changes of ALT in accordance with annual aboveground vegetation dynamics and disturbance (see supplementary figure 1 available at stacks.iop.org/ERL/6/ 024003/mmedia). In the time sequence, ground biomass higher (lower) above intercepts more (less) radiation and therefore shallows (deepens) the ALT depth (figure 2).
To clarify the effect of permafrost on the aboveground vegetation composition and its interaction with soil hydrology, we conducted a model run without permafrost, called the NP run, and compared it with the CNTL run. Although the amount of precipitation in the growing season is very limited (130 mm from May to June in the model forcing data), a wet spring to dry summer soil-moisture pattern appeared in the CNTL run, but in the NP run (see supplementary figures 2(a) and (b) available at stacks.iop.org/ERL/6/024003/ mmedia) an overall dry moisture condition appeared in spring through summer. Soil moisture in summer (JJA) of the CNTL run was 3-4% (absolute difference in soil volumetric water content) higher than that of the NP run (figure 4 leftmost square and diamond symbols). Accordingly, throughout the growing season, the CNTL run showed no clear depression in vegetation water uptake associated with the decrease of soil moisture ( figure 3(a)). On the other hand, when permafrost was removed (NP run), the model showed less vegetation water uptake and a resulting soil drought. The decrease in vegetation water uptake was found throughout the summer and peaked in mid-summer, despite the favorable temperature and radiation conditions for vegetation photosynthesis activity. This result suggests that vegetation photosynthesis activities of the taigapermafrost system might be more vulnerable to mid-summer drought than to the drought occurred in the early growing season.
The above results have proved that the active layer serves as an 'aquifer' and the permafrost serves as 'aquifuge'. To the soil water availability, this 'aquifer-aquifuge' system plays a 'buffer effect' through which intensive water input (by snow meltwater and summer precipitation) is evened out from early spring over most of the growing season until midsummer (see supplementary 'Soil water content' available at stacks.iop.org/ERL/6/024003/mmedia). By removing the permafrost, incoming water drained much faster because there was no frozen soil to hold the water. For soil moisture, the 'buffer effect' was stronger in spring than in summer due to the shallow ALT and intensive meltwater income (see supplementary figure 2(b) available at stacks.iop.org/ERL/6/ 024003/mmedia), as for vegetation water uptake. This effect was more obvious in mid-summer than in spring because of the low water supply and high potential vegetation water demand of summer ( figure 3(b)). Therefore, the absence of permafrost may lead to a larger summer water deficit.
The inter-seasonal hydrological effect of permafrost may further affect the long-term vegetation dynamics. Model output indicated that the existence of permafrost (CNTL run) helps larch survive the summer drought by providing a continuous supply of water to the top soil layer, whereas the absence of permafrost (NP run) might lead to fatally dry conditions for larch (see supplementary figure 3 available at stacks.iop.org/ ERL/6/024003/mmedia). Therefore, in the NP run, larch was eventually replaced by pine and birch as the dominant species, which have more roots in deeper soil and can tolerate extreme drought. Moreover, dry soil induced higher fire frequency in the NP run than in the CNTL run, resulting in a notable . Aboveground total biomass (with fractional components of major species) for CNTLrun (left bar) and NP run (right bar) simulated under five different climate conditions. Five bar groups represent those (from left to right) for the present climate condition; for temperature conditions of +1.12 • C (above the present); +2.25 • C; +4.5 • C, and +4.5 • C plus precipitation +20% (above the present). The name of each experiment was shown on the top of the bar. Two dashed lines in the upper part show changes of summer (JJA) mean soil water for CNTL run (square marks) and NP run (diamond marks), respectively. Note the changes of major species contributing to the total biomass under the different climate conditions. decrease in aboveground biomass in the NP run (figure 4, leftmost group of bars; see supplementary figure 4 and 'Fire' available at stacks.iop.org/ERL/6/024003/mmedia).
The vegetation dynamics and its phenological cycle (i.e., foliation and broadening and falling leaves etc) block radiation to prevent frozen soil from thawing in summer and also suppress the development of other permafrost processes (e.g., thermokarst, soil fluction, and thermo-erosion) . The limited ALT in summer, in turn, provides sufficient water for larch trees and lowers the fire frequency to preserve the aboveground taiga composition and its biomass. In this manner, under the present climate condition, the taiga ecosystem is stabilized by coupling with the hydrological process of permafrost. Thus, the taiga (larch)-permafrost system can be interpreted as a self-regulating coupled system.
Is this coupled taiga-permafrost system sustainable under future climate warming? The results of second set of experiments show a drastic change of dominant species and biomass depending on the degree of temperature increase. Under moderate warming (CNTL 1C, figure 4), larch trees were still the dominant species, and aboveground biomass increased. However, under intense warming (CNTL 2C and CNTL 4C), drastic larch extinction occurred, with decreased biomass and decreased soil moisture in mid-summer (figure 4, yellow squares). This drought condition was mainly attributed to the enhanced photosynthesis activity in early summer (figure 3(c)) and advanced thawing processes (see supplementary figure 2(c) available at stacks.iop.org/ ERL/6/024003/mmedia), both of which were caused by the temperature increase (see supplementary 'Warming' available at stacks.iop.org/ERL/6/024003/mmedia). In contrast, the sensitivity experiments without permafrost (NP 1C, NP 2C, NP 4C) showed no clear effect of temperature increase on changes in soil moisture and vegetation biomass (figure 4, red diamonds and right bar of each group). These experimental results indicate that the taiga-permafrost coupled system is more sensitive to the future warming compared to the ecosystem without permafrost. Another notable feature under intense warming (+2 • C or more) was the drastic change in biome (forest) composition as well as biomass, as shown in the experiments with the permafrost process (left bars of each experiment in figure 4). The difference between the CNTL 1C and NP 1C run is significant, but less difference between CNTL 4C and NP 4C in terms of forest composition, biomass (figure 4), and soil water uptake ( figure 3(d)). This suggests that under intense warming (+2 • C or more), the permafrost effect is drastically weakened, or more realistically, the taigapermafrost coupled system might no longer exist. However, in the additional experiments, by increasing the precipitation intensity by 20% over the NP 4C and CNTL 4C run, the permafrost effect again came into force (figure 4 rightmost bars), although this precipitation effect was not as remarkable as the temperature effect. That is, the vulnerability of the taigapermafrost system under future climate conditions depends greatly on the intensity of warming, but additionally on the intensity and temporal distribution of precipitation.

Conclusion and discussion
Climate has been considered as the most important driver to control the WEC activities in taiga-permafrost system on large scale. However, in the smaller scale, the lack of a definitive relationship with climate occurs because energy partitioning is strongly controlled by vegetation type and structure rather than directly by climate or latitude (McGuire et al 2002, Lopez et al 2008. The two sets of numerical experiments using the new dynamic vegetation-frozen soil coupled model have revealed that under the present hydroclimatic condition, the taiga (represented by larch forest) is tightly coupled with the permafrost, forming a taiga-permafrost coupled system. In this system, the permafrost maintains soil water for the taiga by controlling the ALT, while the taiga maintains the permafrost by controlling canopy radiation interception through vegetation long-term succession. Though the active layer seasonal thaw is mainly driven by climate seasonal rotate, its freezing-thawing process may notably controlled (changed) by vegetation dynamic.
The warmer climate runs have shown that this coupled system can be sustained under a temperature increase of about +1 to +2 • C or less. However, under intense warming of +2 to +4 • C or higher, a drastic change of vegetation (extinction of larch and its replacement by other boreal and sub-boreal tree species) is strongly suggested, where the coupling of forest and permafrost would be destroyed or weakened. Moreover, our results have predicted a decrease in biomass and the transpiration rate in mid-summer due to drought conditions in the soil layer under warmer climate ( figure 3(c)).
The feedback to climate from the taiga-permafrost system is still not clear. Some modeling studies showed strong continental-scale climate effects through vegetation processes (Betts 2000, Bonan et al 1992 and through soil freezing-thawing processes (Takata and Kimoto 2000), separately. The IPCC-AR4 suggested increasing precipitation in Siberia under the future warmer condition, but none of the models have considered the effects by the taiga-permafrost coupled system in their climate projections. The current study has suggested that decreases in transpiration and biomass might have negative feedback to the regional water cycle, including reduction of precipitation. Future climate prediction studies should, therefore, be conducted incorporating the vegetation-permafrost coupled system demonstrated here, in light of a potentially large impact of this coupled system in the global climate.