Interactive comment on “ The amount and timing of precipitation control the magnitude , seasonality and sources ( 14 C ) of ecosystem respiration in a polar semi-desert , NW Greenland ” by M . Lupascu

Introduction Conclusions References

The amount and timing of precipitation control the magnitude, seasonality and sources ( 14 C) of ecosystem respiration in a polar semi-desert, NW Greenland Introduction

Conclusions References
Tables Figures

Back Close
Full

Abstract
This study investigates how warming and changes in precipitation may affect the cycling of carbon (C) in tundra soils, and between high arctic tundra and the atmosphere.We quantified ecosystem respiration (R eco ) and soil pore space CO 2 in a polar semidesert under current and future climate conditions simulated by long-term experimental warming (+2 • C, +4 • C), water addition (+50 % summer precipitation) and a combination of both (+4 • C ×+50 % summer precipitation).We also measured the 14 C content of R eco and soil CO 2 to distinguish young C cycling rapidly between the atmosphere and the ecosystem from older C stored in the soil for centuries to millennia.We identified changes in the amount and timing of precipitation as a key control of the magnitude, seasonality and sources of R eco in a polar semi-desert.Throughout each summer, small (< 4 mm) precipitation events during drier periods triggered the release of very old C pulses from the deep soil, while larger precipitation events (> 4 mm), more winter snow and experimental irrigation were associated with higher R eco fluxes and the release of recently-fixed (young) plant C. Warmer summers and experimental warming also resulted in higher R eco fluxes (+2 • C > +4 • C), but coincided with losses of older C. We conclude that in high arctic dry tundra systems, future magnitudes and patterns of old C emissions will be controlled as much by the summer precipitation regime and winter snowpack as by warming.The release of older soil C is of concern as it may lead to net C losses from the ecosystem.Therefore, reliable predictions of precipitation amounts, frequency, and timing are required to predict the changing C cycle in the High Arctic.

Introduction
Climatic changes and their effects on terrestrial ecosystems are amplified in the Arctic (Serreze and Barry, 2011).Globally, the Arctic is undergoing the largest temperature increase, with a predicted rise in surface air temperature of 3-8 • C by 2100 (Meehl et al., 2007), and permafrost degradation (Romanovsky et al., 2010).A deeper active layer associated with permafrost thaw is affecting the Arctic's surface hydrology: lakes are disappearing (Smith et al., 2005) and river run-off is increasing (Peterson et al., 2002).Other abiotic changes accompanying the warming include increasing cold season precipitation, declining duration of snow cover and regionally distinct changes in snow depth (Callaghan et al., 2011), wetting due to increased atmospheric transport of moisture into the Arctic (Zhang et al., 2013), and a decline in summer sea ice extent (Johannessen et al., 2004).The implications of these changes for the regional biogeochemistry are largely unknown, especially in the High Arctic (> 70 • N) where most non-alpine tundra ecosystems currently exist within 100 km of the coastline (Bhatt et al., 2010;Post et al., 2013).Its coastal proximity makes this ecosystem particularly vulnerable to changes in summer sea ice extend and associated warming and, or changes in precipitation.
Long-term multifactorial climate change experiments are a crucial tool to mimic the future and unravel how coupled abiotic and biotic changes are manifested in the biogeochemistry of terrestrial ecosystems, including high arctic tundra (Welker et al., 1997;Heimann and Reichstein, 2008).In particular, the cycling of C within soils and between the tundra and the atmosphere is only partially understood.Most of our current understanding of soil C dynamics is based on studies of organic soils in the Low Arctic (Post et al., 1982;Gorham, 1991;Tarnocai, 2000;Welker et al., 2000;Ping et al., 2008;Tarnocai et al., 2009).Fewer and only more recent studies have been conducted in the High Arctic (Lloyd, 2001;Welker et al., 2004;Illeris et al., 2003;Huemmerick et al., 2010;Lamb et al., 2011;Christiansen et al., 2012;Henry et al., 2012;Schaeffer et al., 2013).Introduction

Conclusions References
Tables Figures

Back Close
Full In the High Arctic, an estimated 12 Pg of organic C have accumulated in polar (semi-) deserts (Horwarth and Sletten, 2010), which cover approximately one-third of the ice-free Arctic (Wookey and Robinson, 1997;Jones et al., 2000).Changes in the C dynamics of the High Arctic and potential feedbacks to the climate system depend on the balance between enhanced plant fixation and microbial degradation of organic matter and C release (Welker et al., 2004).While no projections are available for the High Arctic only, models estimate emissions between 33 and 508 Pg C from circumpolar permafrost (equivalent to 0.04-1.69 • C warming) by 2100 (Koven et al., 2011;McDougall et al., 2012;Schneider von Deimling et al., 2012;Schuur et al., 2013).While there is evidence that the High Arctic is going to experience higher temperatures and precipitation levels (Vavrus et al., 2012), to date, only a few long-term field experiments have explored the interactions between a warmer and wetter High Arctic as it affects organism and ecosystem function (Welker et al., 1993;Wookey et al., 1993Wookey et al., , 1995;;Robinson et al., 1995;Sharp et al., 2013;Lupascu et al., 2013).Others have typically focused on the consequences of summer warming-only, with few studies capable of estimating how the level of warming in the near or longer term might influence tundra function and structure (Welker et al., 1997(Welker et al., , 2004;;Arft et al., 1999;Lamb et al., 2010;Elmendorf et al., 2012).However, results from these ∼ 2 • C passive warming experiments do not provide a means to forecast beyond the next decades (Welker et al., 1997), even as the Arctic could be warming 6 • C by 2050 (IPCC, 2007;Koenigk et al., 2012).Furthermore, some studies may suffer from short experimental monitoring periods (Lloyd, 2001;Illeris et al., 2003;Huemmerick et al., 2010).At high latitudes, terrestrial-atmospheric C exchange displays extreme interannual variability, which can exceed experimental treatment effects.Thus, short-term monitoring can bias the general understanding of the ecosystem response to environmental variations (Grøndhal et al., 2008).
In addition, (semi-) arid ecosystems, such as polar semi-deserts, are very sensitive to changes in precipitation regimes (Cable et al., 2011).Continuous precipitation or individual, but large events, may increase soil water availability due to deeper penetration of precipitation into soils, which in turn may stimulate net primary productivity, but hin-Introduction

Conclusions References
Tables Figures

Back Close
Full der microbial respiration and nitrogen mineralization if soils become saturated (Knapp et al., 2008;Pouliot et al., 2009).On the other hand, isolated and small precipitation events can trigger large pulses of R eco (Tang and Baldocchi, 2005;Sponseller, 2007;Vargas and Allen, 2008;Carbone et al., 2011).These pulses have been related to multiple factors, including the amount, seasonality, and intensity of precipitation, the timing between such events, and antecedent soil water content (SWC) (Huxman et al., 2004;Schwinning and Sala, 2004;Jarvis et al., 2007;Carbone et al., 2011).Currently, we do not know how C dynamics and the stability of permafrost C are affected by precipitation frequency.
Measurements of ecosystem respiration (R eco ) can be used to infer changes in soil C dynamics without compromising soil structure and biota (Heimann and Reichstein, 2008).And, understanding seasonal dynamics of R eco is key to understanding the interannual variability in ecosystem C budgets (Goulden et al., 1996).In addition, radiocarbon ( 14 C) analysis of R eco is a valuable tool to understand the decomposition of recent vs. older C. The rapid cycling of recently assimilated C between plants and soil microbes has almost no effect on atmospheric CO 2 levels, but decomposition of older C pools, formerly disconnected from the active C cycle, represents a net addition of C to the atmosphere.Thus, assessing soil C feedbacks to rising atmospheric CO 2 levels requires distinguishing R eco sources, i.e. plants and microorganism living in the rhizosphere from free-living microbes decomposing soil organic matter (Trumbore, 2006(Trumbore, , 2009)).
In this study we present measurements of R eco and belowground CO 2 and their sources conducted over three consecutive summers at a ∼ 10-year climate manipulation experiment in northwestern Greenland to address the following questions: 1. How does the natural, short-term variability in summer precipitation (amount and frequency) and previous winter snow affect the magnitude and sources of R eco ?
2. How does a simulated long-term increase in summer rainfall and, or temperature alter the seasonal patterns of soil CO 2 and R eco fluxes?Introduction

Conclusions References
Tables Figures

Back Close
Full 3. How do precipitation, irrigation and/or warming alter the sources of R eco ?
Addressing these temporally detailed questions complement our earlier report that focused on the annual C budgets of these systems (Lupascu et al., 2013).
2 Material and methods

Site description
This study is carried out in a polar semi-desert (aka prostrate dwarf shrub tundra) in the High Arctic of northwest Greenland near US Thule Air Force Base (76 • 32 N, 68 • 50 W; 200-350 m a.s.l.).Prostrate dwarf-shrub, herb tundra occupies approximately 8 % of the ice-free arctic land surface (Walker et al., 2005).At our site, vascular plant cover is approximately 50 % and the patterned ground is a mixture of nonsorted nets, weakly formed stripes and frost boils with patchy cryptogamic crust.The vascular plant community is dominated by the deciduous dwarf-shrub Salix arctica Pall., the graminoid Carex rupestris All. and the wintergreen dwarf-shrub Dryas integrifolia Vahl.The live biomass and litter of these three species account for approximately 70 % of vascular plant cover.
The soil is a Typic Haploturbel (USDA, 1999) with a maximum thaw depth of about 1 m.For the top 40 cm, soil bulk density is approximately 1.1 g cm −3 and the soil organic C content varies between 0.2-1.6 %-mass for vegetated areas and 0.1-0.2%-mass for bare areas.

Climate change experiment
Measurements are conducted from mid May to the end of August during the period 2010-2012 at a long-term climate change experiment established in 2003 to mimic climate scenarios of 2030 and 2050 (Sullivan et al., 2008;ACIA, 2005).The experiment consists of four treatments in a randomized complete block design: +2 • C soil warming Introduction

Conclusions References
Tables Figures

Back Close
Full   (relative to 1971-2000), while maintaining seasonal patterns (Sullivan et al., 2008).The plots (2.0 × 0.8 m 2 ) are oriented to span the transition between vascular plants and bare soil/cryptogamic crust, so that each comprised approximately 50 % of the plot area, to facilitate scaling from the plot to ecosystem level.

Climate trends 1952-2012
We used temperature and precipitation data from the Thule airport (THU) weather station for the period 1952-2012.Daily mean temperatures are calculated as the mean of the daily minimum and maximum temperatures.We calculate annual, summer (June, July, August), winter (December, January, February) temperature and total precipitation (rainfall plus snow) trends for the most recent climate normal period .In addition, to evaluate how temperature and precipitation change during the entire 61 yr period, we calculate the same trends for the last 20, 30, 40, 50 and 60 yr corresponding to 1992-2012, 1982-2012, 1972-2012, 1962-2012, 1952-2012 respectively, and we estimate how the trends over the shorter recent periods compare with the longer ones (Table 1).We used linear least square fit and regression line trend analysis to evaluate change for the various overlapping periods and on the latest climatological normal period.All reported trends are acknowledged to be significant if they exceed one standard deviation (1 − σ) of the respective dataset.Introduction

Conclusions References
Tables Figures

Back Close
Full We measure R eco and soil CO 2 concentrations approximately 2-3 times a week between 10:00 and 13:00 LT.The first one is measured using opaque, dynamic chambers (30 cm i.d., 8 L V = 8), a method that has been used for decades and its limitations have been widely discussed in the literature (e.g.Davidson et al., 2002).Chamber bases are inserted at the beginning of each measurement season to about 2 cm depth, sealed with soil material on the outside, and left in place for the entire sampling season.Vegetation is not clipped.We quantify CO 2 emissions by circulating the air in the chamber's headspace between the chamber and an infrared gas analyzer connected to a data logger (LI-840, LI-1400, LI-COR Biosciences, Lincoln, NE, USA) at a rate of 0.5 L min −1 .Emission rates are calculated from the slope of time vs. CO 2 concentration curves using a linear regression.
We use the daily R eco measurement as a proxy for mean daily R eco .When daily measurements are missing we estimate daily R eco values using the relationship between respiration and temperature (Lloyd and Taylor, 1994), described by an Arrhenius-type Eq. ( 1), where the effective activation energy for respiration varies inversely with temperature: with E 0 = 308.56K and T 0 = 227.13K and where A is a data-set dependent variable.Variable A was first obtained from the R eco data collected in situ.We hence calculate the daily R eco for the missing days using the average daily temperature (T ).We estimate cumulative summertime R eco as the sum of the daily values.
We monitor soil CO 2 concentrations using vertical, stainless steel gas wells (0.35 cm i.d., 0.6 cm o.d.) inserted into the soil to 20, 30, 60 or 90 cm depth and closed off with rubber septa (Blue septa, Grace, Deerfield, IL, USA).Wells were installed from the soil surface in 2010 as soon as the thaw of the active layer allowed and remained in the same location throughout the entire study period including the winters.Soil gases Introduction

Conclusions References
Tables Figures

Back Close
Full were obtained with 60 mL syringes (BD, Franklin Lakes, NJ, USA).We collect soil gas using 60 mL syringes (BD, Franklin Lakes, NJ, USA).We discard the first 60 mL sample and inject the second into an infrared gas analyzer connected to a data logger (LI-800, LI-1400, LI-COR), we then manually record the peak concentration.

Soil temperature and water content
With each R eco measurement we manually record soil temperature with a digital thermometer (15-077, Thermo Fisher Scientific, Waltham, MA, USA) at 5 and 10 cm depth and SWC at 10 cm (Hydrosense, Campbell Scientific, Logan, UT, USA).In the vegetated plots, we also continuously monitor SWC at 5 cm depth with Hydra II soil moisture and salinity sensors (SDI-12/RS485, Stevens, Portland, OR, USA) connected to a CR1000 datalogger (Campbell Scientific).Data are acquired every 15 min.

Sampling for isotope analyses
We collect gas samples for isotope analysis monthly.To sample R eco , we leave the chambers closed until the CO 2 concentration inside the chamber is ≥ 2× that in ambient air (up to 24-48 h).After each concentration measurement, the CO 2 is collected by circulating the air inside the chamber through drierite (W.A. Hammond Drierite Co. Ltd., Xenia, OH, USA) followed by a pre-conditioned, activated molecular sieve (13X powder-free 133 8/12 beads, Grace) trap at a rate of 0.5 L min −1 for 15 min (Gaudinski et al., 2000).Compared to other methods used to trap CO 2 in the field (e.g.Dörr and Münnich, 1986;Charman et al., 1999), small, light-weight molecular sieve traps are ideal for use in remote field locations as they do not require cryo-or caustic liquids.Small radiocarbon memory effects can be overcome by pre-conditioning traps (630 for 45 min under vacuum) and sampling enough CO 2 in the field (0.3-2 mg C).For each set of R eco samples, we collect two samples of CO 2 from ambient air on molecular sieve (15 min at 0.5 L min −1 ) in a well-ventilated area nearby the experimental site.Soil gas Introduction

Conclusions References
Tables Figures

Back Close
Full from the different depths was collected in pre-evacuated, stainless steel canisters via flow-restricting stainless steel capillaries (0.010 cm × 0.063 cm × 30 cm, Fisher).

Isotope analyses of CO 2
Radiocarbon ( 14 C) analysis is a valuable tool for understanding the sources of R eco (respiration of plants that fix CO 2 from the atmosphere vs. respiration of microbes that decompose soil C of various ages to CO 2 ).Plants respire CO 2 with a 14 C content that is similar or slightly enriched compared to atmospheric CO 2 .Soil microbes respire a range of C sources that vary in age from days to millennia.Due to radioactive decay, older soil organic matter is depleted in 14 C (t 1/2 = 5730 yr).In contrast, CO 2 derived from the decomposition of soil C made from photosynthetic products years to decades ago is enriched in 14 C.This is because during the late 1950s and early 1960s, testing of nuclear bombs aboveground almost doubled the naturally produced 14 C activity in atmospheric CO 2 .After test cessation, the amount of bomb-14 C (aka "modern" C) in the atmosphere has been declining as a consequence of mixing with terrestrial and ocean C pools and dilution of fossil fuel CO 2 emissions (i.e. 14 C-free).The mixing of this bomb-14 C tracer into terrestrial C pools over the last five decades can be used to infer C dynamics (Trumbore, 2006).
In order to analyze the 14 C content, CO 2 is released from molecular sieve traps by baking at 630 • C for 45 min or extracted from canisters using a vacuum line, purified cryogenically and converted to graphite using sealed tube Zn reduction (Xu et al., 2007).A split of the CO 2 is analyzed for its ∆ 13 C value (GasBench II, DeltaPlus, Thermo).The 14 C content of the graphite is measured with accelerator mass spectrometry (NEC 0.5MV 1.5SDH-2 AMS) at the KCCAMS laboratory at UC Irvine (Southon and Santos, 2007).Data are reported relative to NIST OX-I (SRM 4990a) and OX-II (SRM 4990c) following Stuiver and Polach (1977).The measurement uncertainty for

Conclusions References
Tables Figures

Back Close
Full The 14 C content of R eco is corrected for the amount of CO 2 from ambient air present in each chamber: with (∆ 14 C cor. ) being the actual 14 C content of R eco , (∆ 14 C obs. ) the measured 14 C content of R eco , (∆ 14 C air ) the measured 14 C content of ambient CO 2 , and (f air ) the fraction of CO 2 in R eco derived from ambient air, calculated from measurements of CO 2 concentrations in ambient air as well as in the chamber immediately before trapping.

Climate and weather
Data from the Thule operations site show a mean annual air temperature of −11.4 ±

1.3
• C and mean annual total precipitation of 122.6 ± 45.4 mm during the period 1952-2012.About half of the precipitation occurs during October-April as snow.
Table 1 shows annual, summer and winter temperature and total precipitations linear trends for the past 20, 30, 40, 50, 60 yr, all trends are significant at the 1 s level.Annual temperature and total precipitations trends for shorter recent periods are larger, indicating both accelerated warming and precipitation during the entire 60 yr period.Overall temperature increases more during the winter months.We observe a significant and consistent increase in total summer precipitation trends while in winter precipitation trends are very variable.
During the last climate normal period (1983 to 2012), the air temperature trend is The three years of this study show very variable weather conditions (Table 2).Cumulative summer precipitation varies by a factor of four during 2010-2012, with the largest values in 2012 and the lowest in 2010.The maximum values of cumulative snowfall are observed in winter 2011.Average summer air temperature is highest in 2011, with the maximum difference observed in July (+1-2 • C), similar variability is observed in 2010 and 2011.

Soil water content
We find that snowfall, summer precipitation patterns and irrigation (W and +4 • C ×W ) have a strong influence on the seasonal SWC pattern (Fig. 1a).increases average summer SWC by +3.8 ± 1.4 %-vol.relative to the control, but does not change the seasonal pattern.

Rates of ecosystem respiration
Ecosystem respiration is consistently affected by changes in seasonal, interannual and experimental changes in precipitation and air temperature.data not shown).
Point measurements of soil conditions indicate that in the control (Fig. 1c) and all treatments (data not shown), R eco fluxes are positively correlated to soil temperature (R 2 = 0.57 ± 0.13).Episodic cold snaps, commonly associated with summer rainfall events, dramatically reduced R eco fluxes within a few hours (Fig. 1b).Surprisingly, R eco fluxes are negatively correlated to SWC (R 2 = 0.37 ± 0.11) (Fig. 1d).Ecosystem respiration fluxes are highest at 10-15 % SWC (Fig. 2c and d).

Soil pore space CO 2 concentrations
Pore space CO 2 concentrations are strongly affected by changes in SWC, due to snowmelt, natural precipitation and experimental water addition and by air temperature changes, but not by experimental warming (Fig. 2).With depth, CO 2 concentrations increase to 60 cm followed by a decrease toward the permafrost table (Fig. 2  to the control (e.g.2011, control: 4488 ± 323, 3681 ± 266, 3248 ± 217 ppm), except at 90 cm in 2010 (Fig. 2 middle panel).On the other hand, experimental warming (+2, +4 • C) slightly increases CO 2 only in the upper mineral soil, while at greater depth concentrations are similar or lower than in the control.
Carbon dioxide concentrations at depths show little variation between years compared to the three-fold differences in R eco observed between 2010, 2011 and 2012.However, 2010 exhibits the lowest CO 2 concentrations compared to 2011 and 2012 (Fig. 2 middle and bottom panel).
In all years, a complex mixture of C sources contribute to R eco during snowmelt, including 1) older C, fixed before 1950 (up to −339 ‰), 2) modern C, fixed decades ago (up to +162 ‰) and 3) recently fixed C (+35, +32, +29 ‰ in 2010, 2011, and 2012, respectively).In July and August, recently fixed C dominates R eco , the period that corresponds to maximum above-and belowground plant growth.We find that episodically, very old C can, however, dominate R eco (Fig. 3).These pulses of ancient C efflux appear to be associated either with freeze-thaw cycles during spring or with rain pulses during summer.After about 7 days of no rain, small rainfall pulses (< 4 mm) results in emissions of old C from depth to the surface (Fig. 4a), with 14 C contents of up to −208 ‰.This old R eco coincides with an increase in both concentration and age of soil CO 2 near the permafrost table (up to −279 ‰; Fig. 4b and c).This phenomenon occurs rapidly: following experimental irrigation (∼ 3.2 mm; Fig. 4c insert) CO 2 concentrations within the soil profile increase up to 16 % within 6 h and then decreases within 24 h.Introduction

Conclusions References
Tables Figures

Back Close
Full Higher summer precipitation in 2010 and 2012 (Table 2) also significantly affect the mean age of R eco (F = 55.2, p < 0.001), but shows consistently younger R eco (summer average) than observed in 2011 (Fig. 3).Irrigation and/or warming do not statistically affect sources of R eco .

Climatic trends
Our study site, including the experimental treatments and control, is undergoing (winter-) warming and (summer-) wetting.Analysis of annual temperature and annual total precipitations trends show accelerated warming and precipitation during 1952-2012.Overall we find a larger increase in winter than in summer temperatures and larger increase in summer than in winter precipitation.During the past climate normal period we calculate a warming trend of 1.0 ± 0.2 • C decade −1 which is consistent with the strong recent warming detected on the west coast of Greenland since 1991 (Hanna et al., 2012).In the more recent period  the annual warming and precipitation trends are two times larger than in the last normal period , with significant contribution from winter temperature and summer precipitation increase.The observed climate trends show that our experiment is based on relevant scenarios of warming and wetting.

Magnitude and seasonality of ecosystem respiration
We show that the magnitude of R eco is strongly modulated by SWC and hence precipitation over the course of this three-year study.Early in the growing season R eco fluxes are positively correlated to the amount of snow accumulated over the previous winter, which is similar to the findings from more southern systems in a grassland and subalpine forest (Chimner and Welker, 2005;Monson et al., 2006).After snowmelt, the general seasonal pattern of R eco follows the rises in air and soil temperatures, with Introduction

Conclusions References
Tables Figures

Back Close
Full modulations in magnitude driven by changes in SWC.This is evident in the summer of 2010 with two clearly distinct periods, a dry and a wet one.Although R eco in both periods are correlated with temperature, the highest R eco fluxes occur during the wet period, due to the more favorable SWC conditions (10-15 %-vol, Fig. 1c and d).
While on short time scales R eco fluxes are negatively correlated to increasing SWC (Fig. 1d), over longer time periods R eco fluxes increase under wetter conditions.In previous work, Lupascu et al. (2013) showed that inter-annual differences in summer R eco budgets can be explained by difference in SWC.Ecosystem respiration is highest in 2012, the year with the maximum SWC, due to higher summer precipitation and higher snowfall levels in the previous winter.Further, irrigation with or without warming (W , +4 • C ×W ) strongly enhances R eco above control levels.The apparent incongruity between the short-and long-term response of R eco to water addition is likely due to a time-lag in the response of plant growth and respiration to water behind that of microbial respiration, which is commonly on the order of days (Carbone et al., 2011;Ogle and Reynolds, 2004), as well as a short-term decrease in soil temperature during precipitation events (data not shown).
Here, we present further evidence that polar semi-deserts are very sensitive to changes in SWC over multiple time scales.Warming by +4 • C stimulated R eco less than +2 • C warming -probably a result of reduced SWC and thus drought stress, which may decrease plant photosynthetic activity and respiration (Welker et al., 1993;Llorens et al., 2004;Sullivan and Welker, 2007;Pinhero and Chaves, 2011) and, or microbial activity (Schimel et al., 2006).Our results agree with previous studies showing that higher SWC stimulate R eco fluxes using experimental manipulations in the High Arctic (Illeris et al., 2003;Christiansen et al., 2012;Lupascu et al., 2013) by promoting leaf area, inferred by increases in the normalized difference vegetation index (NDVI; Sharp et al., 2013), and microbial biomass (Christiansen et al., 2012).
The seasonal trend of R eco fluxes is primarily controlled by temperature.Fluxes of R eco peaked in midsummer along with a maximum in air temperatures and active layer depth, and decrease dramatically during cold episodes, which typically coincide with Introduction

Conclusions References
Tables Figures

Back Close
Full rainfall events.Both experimental warming treatments (+2 and +4 • C) stimulate R eco fluxes above control levels (Fig. 1; Table 3).These findings cooberate earlier experimental warming studies in the High Arctic showing that higher soil temperatures stimulate R eco (Welker et al., 2004;Oberbauer et al., 2007;Strebel et al., 2010;Sharp et al., 2013) including those conducted previously under control condition-only at a similar polar semi-desert, 10 km from our experimental site (Czimczik and Welker, 2010).
Warming may be affecting soil respiration by directly stimulating microbial processes, or indirectly by stimulating plant growth above-and belowground; greater root exudation could in turn accelerate microbial litter and soil C decomposition and nutrient mineralization in all of these studies (Rustad et al., 2001;Robinson, 2002).Some of the observed warming effects in early summer may however, be related to co-occurring changes in plant phenology (Yuste et al., 2004), with root growth and exudation stimulating R eco (Sullivan and Welker, 2005;Sullivan et al., 2007).Collectively, these processes promote increased respiration in these nutrient-limited communities (Arens et al., 2008;Schaeffer et al., 2013).

Carbon dioxide concentrations within the soil profile
Measurements of CO 2 concentrations at different depths offer insights on CO 2 production along the soil profile (Davidson and Trumbore, 1995).Concentrations generally peaked at 60 cm depth, below the rooting zone (0-30 cm), with a minimum near the permafrost table at about 1 m depth, where temperatures are close to 0 • C. Similar to R eco , magnitudes of soil CO 2 concentrations are strongly affected by changes in SWC.In the topsoil, CO 2 concentrations are highest during the beginning of the growing season, following snowmelt (2011, 2012;Fig. 2 top panel).This probably reflects new microbial activity within the topsoil stimulated by water, C and nutrient inputs from melting snow leaching through the litter layer (Hirano et al., 2005;Scott-Denton et al., 2006) as well as the release of older CO 2 that was previously trapped in the frozen active layer with limited diffusivity during the winter (Albert and Perron, 2000;Schimel et al., 2006).Introduction

Conclusions References
Tables Figures

Back Close
Full In some cases CO 2 concentrations at depth also rise in response to snowmelt, but with a time delay of a few weeks that increases with depth.We find that CO 2 concentrations near the permafrost table are very sensitive to the occurrence and magnitude of precipitation events.Small precipitation events (< 4 mm) stimulate CO 2 concentrations in a matter of hours by either increasing CO 2 production or reducing the available gas-filled pore space.Larger precipitation events result in a sharp decline of CO 2 concentrations, as water pooling on the permafrost table likely restrict microbial activity by reducing oxygen availability or trapping existing CO 2 by disrupting the gas-phase connectivity (Stonestrom and Rubin, 1989).
Irrigation treatments further substantiate the importance of water on microbial activity.In general, long-term irrigation results in higher CO 2 concentrations, while longterm warming by 4 • C does not significantly alter CO 2 concentrations compared to control conditions.In a separate study focusing on inter-annual C budgets, Lupascu et al. (2013) show that long-term climate manipulations in these same experimental plots, dramatically change the 14 C content of CO 2 at depths.Furthermore, irrigation results in the presence of younger C respired compared to the control.In contrast, experimental warming (+4 • C) shows more depleted 14 C and hence a larger fraction of older C being released at all depths.In conclusion, these two studies show that water along with temperature is a crucial driver of microbial activity in these high arctic soils.

Sources of ecosystem respiration
Changes in plant density and community composition are additional manifestations of long-term changes in climate and permafrost regime.We find that R eco and soil CO 2 in vegetated areas are always younger compared to bare areas in this patchy landscape driven in large part by differential frost heave.Plant respiration is typically a large component of R eco with a 14 C content similar to, or occasionally slightly higher than, atmospheric CO 2 (Schuur and Trumbore, 2006;Czimczik et al., 2006), and that flux can mask the relatively smaller emissions of older CO 2 at depth.Microorganisms in the topsoil that decompose plant exudates or litter further contribute to the 14 C enrichment of Introduction

Conclusions References
Tables Figures

Back Close
Full R eco in vegetated areas.In addition, the bulk C pool in the vegetated areas is younger than in the bare areas due to the continuous input of fresh litter (data not shown).Thus, future assessments of high arctic R eco budgets need to carefully account for changes in vegetation cover and structure (Forbes et al., 2010;Bonfils et al., 2012) as greater plant biomass coupled with fresh C inputs to depth as well as associated changes in albedo are likely to affect the magnitude of old C emissions to the atmosphere.Seasonally, the contribution of old C to R eco should peak in mid to late summer when active layer depth is deepest, but its relative contribution should be highest after peak plant biomass.Plant respiration commonly masks contributions of older C in R eco , resulting in young emissions during the summer time as observed here and previously (Schuur et al., 2009;Czimczik and Welker, 2010;Nowinski et al., 2010). However Natali et al. (2011) show that the contribution of old C to R eco in a snowfence manipulation experiment in the Low Arctic of Alaska declines at the end of the growing season; they attribute this to water pooling on the permafrost table.
Here, we observe old R eco during two distinct time periods: (a) at the beginning of the growing season and (b) episodically after smaller precipitation events following a dry period.At the beginning of the growing season (Fig. 3), we identify a complex mixture of C sources with old C likely coming from residual CO 2 trapped over the winter (Schimel et al., 2006).This supports findings from our initial study in these NW Greenland polar semi-deserts (Czimczik and Welker, 2010).Ancient C can be a measurable part of R eco , especially before leaf emergence and senescence when fresh C inputs are minimal.Further work is needed to investigate potential emissions of old C during freeze-up as well as winter since these time periods is not covered in our study.We also observe episodic emissions of old CO 2 following smaller precipitation events (< 4 mm day −1 ) after about 7 days of no rain (Fig. 4).These small precipitation events enhance CO 2 concentrations at depth and triggered release of older C within 24 h of the rain event.Two processes may explain our observations.(1) Water rapidly percolating through the active layer down to the permafrost table, which can facilitate the diffusion of old CO 2 upwards into overlying soil horizons by displacing air space within Introduction

Conclusions References
Tables Figures

Back Close
Full the soil pores (Huxman et al., 2004).( 2) Decomposition of old C deeper in the active layer may be stimulated by cycles of drying and rewetting, the so-called "Birch effect" (Birch, 1964).Enhanced microbial activity and mineralization of soil C and nitrogen in response to drying and rewetting has been extensively described described in surface soils of temperate and semi-arid ecosystems in both field and laboratory studies (Fierer and Schimel, 2002;Jarvis et al., 2007).It is possible that the Birch effect also extends to permafrost soils, including the bottom of the active layer.The Birch effect has been attributed to rapidly increasing mineralization rates in response to changing moisture conditions (Inglima et al., 2009;Borken and Matzner, 2009;Unger et al., 2010) and, or C availability (Kieft et al., 1987;Fierer and Schimel, 2002;Jarvis et al., 2007).Different mechanism have been proposed to explain the change in C availability, including (1) drying and rewetting of soils shatters soil aggregates, exposing previously unavailable organic substrates to decomposition (e.g.Denef et al., 2001); (2) "priming" -increased decomposition of old, and potentially more recalcitrant C at depth via inputs of fresh labile C leached from the litter layer and, or rooting zone (Fontaine et al., 2007); (3) relocation of labile C produced by photo-degradation in the litter layer to depth (Ma et al., 2012); and (4) drying causing an increase in dead microbial biomass, which is rapidly recycled by new microorganisms and fungi after rewetting (e.g.Bottner, 1985).Recycling of microbial biomass at depth would result in the production of old CO 2 as microbes carry the same 14 C signature as their C source (Petsch et al., 2001).Here, we do not have enough data to identify which mechanism is responsible for this episodic old C release.Additional experiments, particularly at higher frequency, are needed to quantify the significance of water pulses on C cycling in permafrost soils and the mechanisms involved.
In (semi-) arid ecosystems, including polar semi-deserts, discrete precipitation events play a complex role in regulating the magnitude and sources of R eco and net ecosystem exchange (Huxman et al., 2004;Thomey et al., 2011).Small precipitation events cause immediate and strong increases in microbial respiration and net C loss from the ecosystem to the atmosphere.Larger rain events stimulate plant C uptake

BGD Introduction
Full and ecosystem C sequestration, but with a delay compared to microbial activity (Jarvis et al., 2007;Carbone et al., 2011).In our experiment we detect the CO 2 pulses but we are unable to detect the magnitude of these pulses.This might partially due to the fact that in order to collect sufficient CO 2 for 14 C analysis we left our chambers closed for at least 24 h which could have modified the natural concentration gradient in the active layer.In Mediterranean ecosystems pulses can account up to 10 % of the C lost over a year (Xu et al., 2004;Tang and Baldocchi, 2005;Jarvis et al., 2007;Carbone et al., 2011).
Our data demonstrate for the first time that losses of older C from high arctic permafrost soils can be episodic in nature and controlled by precipitation events -making them very difficult to quantify with discrete measurements.We find that contributions of older C to R eco are undetectable during wet periods with intense precipitation (> 4 mm day −1 ) during the summer.This is likely a consequence of water pooling on the permafrost table leading to a decrease of microbial decomposition of older C at depth.This data further supports earlier findings showing that interannually, wetter summers coincided with younger R eco being released due to inputs of recently-assimilated C from the litter layer and/or rooting zone (Lupascu et al., 2013).
While we find that precipitation events affect the short-term variability in the age of R eco , the irrigation treatments appear to not effect the age or seasonality of R eco fluxes.This does not necessarily indicate that episodic release of old C is minor component of the summertime R eco flux.It is likely a consequence of two factors.(1) Our sampling frequency was low (monthly), due to the high cost and effort required for

Conclusions
This study illustrates the complexity of temperature and water controls on the stability of old C in permafrost soils in semi-arid soils of the High Arctic.Accurately describing C cycling and feedbacks to the atmosphere requires not only an understanding of future temperature changes, but also of precipitation amounts, frequencies and form (rain or snow).Our analysis shows that soils of polar semi-deserts episodically release old C with light rain events.However, understanding the magnitude of these episodic contributions to the summertime and annual R eco budget and the overall loss of old C from high-arctic tundra requires further investigation.

BGD Introduction Conclusions References
Tables Figures

Back Close
Full  Full Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | by infrared radiation (+2 Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | decade −1 with the strongest warming during the winter months of 1.5 ± 0.5 • C decade −1 .During the same normal period we find a large positive trend in total annual precipitation with the strongest increase observed during summer of 10.1 ± 5.3 mm decade −1 Discussion Paper | Discussion Paper | Discussion Paper | In 2010, we started measurements (DOY 151) a couple of weeks after snowmelt commenced, and we divide the season into two distinct periods, dry and wet.From the beginning of June (DOY 151) to mid-July (DOY 195), R eco fluxes are low and stable (control: 0.28 ± 0.11 mmol C m −2 s −1 ), and become larger during the second half of the season (control: 0.60 ± 0.29 mmol C m −2 s −1 ; Fig. 1b).In the following years (2011, 2012), R eco fluxes follow a different seasonal pattern with a similar-sized maximum July (control: 2.36 ± 0.18 mmol C m −2 s −1 ; Fig. 1b).During all years, bare areas display similar sea-Discussion Paper | Discussion Paper | Discussion Paper | sonal patterns as vegetated areas, but with reduced R eco fluxes (< 1 mmol C m −2 s −1 ; middle panel).Throughout the season, CO 2 concentrations display a bi-modal pattern, with a first peak during the snowmelt and a second peak associated with maximum air temperatures and active layer depth (Fig. 2 top panel).The magnitude of the snowmelt peak between years is positively correlated with snowpack.Irrigation (W , +4 • C ×W ) has the strongest effect on pore space CO 2 concentrations.For instance, we observe higher concentrations at all depths (e.g.2011, +4 • C× W : 2900 ± 522, 8845 ± 419, 5577 ± 350 ppm for 30, 60 and 90 cm, respectively) compared Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | and in the laboratory.(2) Changes in contributions of older, 14 C-depleted C from depth are generally hard to detect in R eco as fluxes integrate C dynamics over the entire soil profile and are dominated by the 14 C-enriched respiration of plants and microorganisms in the topsoil.Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Fig. 1 .Fig. 3 .
Fig. 1.Seasonal patterns and correlations of ecosystem respiration (R eco ), soil water content (SWC) and soil temperature under control conditions (average ± SE, n = 1-3 plots) during the summers of 2010-2012: (a) continuous SWC at 5 cm depth, (b) daily R eco flux (c) correlation between daily R eco flux and soil temperature at 5 cm depth measured manually during the flux measurement, and (d) correlation between daily R eco flux and SWC at 5 cm depth measured manually during the flux measurement (in (c) and (d), black circles indicate the early, dry period -black squares indicate the late, wet period of the summer of 2010).
• C), +4 • C soil warming (+4 • C), irrigation (W ) and +4 • C soil warming × irrigation (+4 • C ×W ) alongside an ambient climate control.Each year we initiate warming during the first week of June, when the plots are about 50 % snow-free and we maintain it through the end of August.We use irrigation (+2 mm of deionized water every week in June and August, and +4 mm in July) to increase the magnitude of growing season precipitation by approximately 50 % At the beginning of the growing season SWC is strongly coupled to snowfall, with higher values in 2011 and 2012 than in 2010.In July and August, SWC is driven by rainfall regimes, with the highest average summer SWC observed in 2012 (31. 8 ± 3.7 %-vol.),compared to 2010 and 2011 (15.1 ± 4.4 and 17.2 ± 4.4 %-vol.).Weekly irrigation (W and +4• C ×W )

Table 2 .
Precipitation and air temperature (±SD) at Thule airport (THU) for the measurement period.
a Preceding winter; defined as snow pack height, not water equivalent.b Cumulative values for the snow-free period.