Biophysical controls on net ecosystem CO 2 exchange over a semiarid shrubland in northwest China

The carbon (C) cycling in semiarid and arid areas remains largely unexplored, despite the wide distribution of drylands globally. Rehabilitation practices have been carried out in many desertified areas, but information on the C sequestration capacity of recovering vegetation is still largely lacking. Using the eddy-covariance technique, we measured the net ecosystem CO 2 exchange (NEE) over a recovering shrub ecosystem in northwest China throughout 2012 in order to (1) quantify NEE and its components and to ( 2) examine the dependence of C fluxes on biophysical factors at multiple timescales. The annual budget showed a gross ecosystem productivity (GEP) of 456 g C m −2 yr−1 (with a 90 % prediction interval of 449–463 g C m −2 yr−1) and an ecosystem respiration ( Re) of 379 g C m−2 yr−1 (with a 90 % prediction interval of 370–389 g C m −2 yr−1), resulting in a net C sink of 77 g C m−2 yr−1 (with a 90 % prediction interval of 68–87 g C m−2 yr−1). The maximum daily NEE, GEP and Re were−4.7, 6.8 and 3.3 g C m −2 day−1, respectively. Both the maximum C assimilation rate (i.e., at the optimum light intensity) and the quantum yield varied over the growing season, being higher in summer and lower in spring and autumn. At the half-hourly scale, water deficit exerted a major control over daytime NEE, and interacted with other stresses (e.g., heat and photoinhibition) in constraining C fixation by the vegetation. Low soil moisture also reduced the temperature sensitivity ofRe (Q10). At the synoptic scale, rain events triggered immediate pulses of C release from the ecosystem, followed by peaks of CO2 uptake 1–2 days later. Over the entire growing season, leaf area index accounted for 45 and 65 % of the seasonal variation in NEE and GEP, respectively. There was a linear dependence of daily Re on GEP, with a slope of 0.34. These results highlight the role of abiotic stresses and their alleviation in regulating C cycling in the face of an increasing frequency and intensity of extreme climatic events.

2003).Although dryland ecosystems are characterized by low precipitation (PPT), soil fertility and productivity, they are important to the global carbon (C) budget as they account for approximately 20 % of total terrestrial net primary productivity (Whittaker, 1975) and 15 % of total soil organic carbon (Lal, 2004).The C cycling in desert ecosystems is particularly sensitive to climate and land-use changes, and may feed back to the climate system (Li et al., 2005).In order to accurately predict global C cycling under changing climate, it is necessary to understand how CO 2 exchange in dry areas responds to variations in climatic conditions (Gao et al., 2012).Currently, the C dynamics of desert shrub ecosystems and their responses to environmental factors are less-well-known compared to those of forests and grasslands (Gao et al., 2012).
Whether a dryland ecosystem is a net sink or source of CO 2 is affected by the way it responds to climatic variability (Liu et al., 2012).In semiarid and arid ecosystems, moisture related factors such as PPT, soil water content (SWC) and vapor pressure deficit (VPD) usually exert strong influences on diurnal, seasonal and interannual variations in the net ecosystem CO 2 exchange (NEE) (Fu et al., 2006;Gao et al., 2012).
Water deficit may depress gross ecosystem productivity (GEP) by limiting plant physiological processes (e.g., stomatal closure) and altering plant phenology (e.g., delayed leaf emergence) and canopy structure (e.g., reduced leaf area index, LAI) (Zhou et al., 2013).Low water availability may also limit ecosystem respiration (R e ) by reducing root activity, suppressing microbial decomposition of organic matters and restricting the diffusion of extra-cellular enzymes and C substrates in the soil (Wang et al., 2014).Moreover, the effects of water availability on GEP and R e depend not only on the sensitivity of related biotic processes and the magnitude of water stress, but also on the temporal pattern of water supply.For example, NEE in dryland ecosystems showed complex and inconsistent responses to rainfall events (Liu et al., 2011;Gao et al., 2012), indicating our lack of understanding on how dryland ecosystems respond to water stress and its relief.
Besides water availability, NEE in arid and semiarid ecosystems is also affected by other abiotic and biotic factors.Drought stress often accompanies thermal and irradia-Introduction

Conclusions References
Tables Figures

Back Close
Full tion stresses, as the cloudiness is usually low and the soil is readily heated up by solar radiation during dry periods.High leaf temperature can deactivate photosystem II, enhance the evaporative demand for plants and stimulate respiration (Fu et al., 2006).
Strong irradiation is common in arid and semiarid areas, and is likely to induce midday photosynthetic depression (Fu et al., 2006).In many ecosystems, canopy development (e.g., changes in LAI) is critical to the seasonal evolution of CO 2 fluxes (Xu and Baldocchi, 2004;Li et al., 2005).However, the large stochasticity of PPT and variability of soil moisture in arid and semiarid ecosystems can obscure the effects of LAI (Wang et al., 2008).Considering the inconsistent effects of these environmental stresses and biotic factors on CO 2 fluxes (e.g., Fu et al., 2006;Aires et al., 2008;Wang et al., 2008), it is needed to examine the relative importance of these biophysical controls and their interactions in desert shrub ecosystems.
China is one of the most threatened countries by desertification.Extensive revegetation and conservation practices have been carried out in northern China (Li et al., 2004).However, little has been done to quantify the C sequestration potential of the recovering vegetation (Gao et al., 2012).Shrubland ecosystems at the south edge of the Mu Us desert lie in a critical geographic transition zone between arid and semiarid climates, and between agricultural and pastoral land uses.Overgrazing on the natural shrublands and steppes caused severe desertification in this region (Chen and Duan, 2009).Grazing of natural vegetation has been prohibited since the late 1990s.Thus, the vegetation has been recovering for over ten years.The rehabilitation of desertified lands has been evidenced by the increasingly fine soil texture, increased nutrient contents and biodiversity, and reduced wind erosion (Chen and Duan, 2009).The ability of the recovering ecosystems to sequester CO 2 has not yet been assessed.This information, however, is essential to adaptive management under changing climate.
Using the eddy-covariance technique, we measured NEE over a shrub ecosystem at the south edge of the Mu Us desert throughout 2012.Our objectives were (1) to quantify NEE and its partitioning into GEP and R e at diurnal, seasonal and annual scales, (2) to examine the dependence of NEE and its components on abiotic and biotic factors Introduction

Conclusions References
Tables Figures

Back Close
Full   (Chen and Duan, 2009;Wang et al., 2014).The mean annual PPT is 287 mm, 62 % of which falls from July to September (Feng et al., 2013).The mean annual potential evapotranspiration is 2024 mm.The soil is sandy and has a bulk density of 1.54 ± 0.08 g cm −3 (mean ± standard deviation (SD), n = 16) in the upper 10 cm of the soil profile.The area is dominated by a mixture of deciduous shrub species including Artemisia ordosica, Hedysarum mongolicum and Hedysarum scoparium, and also has sparsely distributed patches of Salix psammophila and Agropyron cristatum.The canopy height is about 1.4 m.Water deficit is a limiting factor for plant photosynthesis and soil respiration (R s ) at the study site (Feng et al., 2013;Wang et al., 2014).

Eddy flux measurements
The eddy-covariance instrument was mounted at a height of 6.2 m on a scaffold tower and oriented in the prevailing wind direction (northwest).A 3-D ultrasonic anemometer (CSAT3, Campbell Scientific Inc., USA) was used to measure fluctuations in wind Introduction

Conclusions References
Tables Figures

Back Close
Full speed, direction and sonic temperature.A closed-path fast response infrared gas analyzer (LI-7200, LI-COR Inc., USA) was used to measure fluctuations in CO 2 and water vapor concentrations.The tube between the air inlet and the Li-7200 was 100 cm, the tube flow rate was 15.0 L min −1 .The tube inlet was situated about 14 cm south of, 16 cm east of, and 8 cm below the anemometer sampling volume.We calibrated the LI-7200 every three months, using 99.99 % nitrogen gas to calibrate zeros for both CO 2 and water vapor, and using a 650 ppm CO 2 standard and a dew point generator (LI-610, LI-COR Inc., USA) to calibrate the span for CO 2 and water vapor, respectively.A data logger (LI-7550, LI-COR Inc., USA) was used to store 10 Hz real-time data.The underlying surface of the shrubland was flat and extended over 250 m in all directions.
Footprint analysis using the flux source area model (FASM) (Schmid, 1997) showed that > 90 % of the fluxes originated from within 200 m of the tower.The CO 2 storage term was not added in estimating NEE because of the short canopy (1.4 m) that usually makes the term negligible (Zhang et al., 2007).In addition, CO 2 storage term tends to be close to zero when summed to daily and annual timescales (Baldocchi, 2003).

Meteorological measurements
Incident photosynthetically active radiation (PAR) was measured using a quantum sensor (PAR-LITE, Kipp & Zonen, the Netherlands).Net radiation (R n ) was measured using a four-component radiometer (CNR-4, Kipp & Zonen, the Netherlands).Air temperature (T a ) and relative humidity (RH) were measured with a thermohygrometer (HMP155A, Vaisala, Finland).All these meteorological sensors were mounted on the tower at 6 m aboveground.Soil temperature (T s ) and water content (SWC) profiles were monitored adjacent to the tower using ECH 2 O-5TE sensors (Decagon Devices, USA) at four depths (0.1, 0.3, 0.7 and 1.2 m).Five soil heat plates (HFP01, Hukseflux Thermal Sensors, the Netherlands) were placed at 10 cm below the soil surface, within about 5 m of the tower base.PPT measurements started from 15 May 2012 using a tipping bucket rain gauge (TE525WS, Campbell Scientific Inc., USA) at a distance of about 50 m from the tower.All micrometeorological variables were measured every 10 s and then aver-Introduction

Conclusions References
Tables Figures

Back Close
Full aged or summed to the 30 min resolution before being stored on data loggers (CR200X for PPT, CR3000 for all others, Campbell Scientific Inc., USA).

LAI measurements
For measuring LAI, we deployed a 4 × 4 grid of 16 quadrats (10 m × 10 m each) within a 100 m × 100 m plot centered on the flux tower in late March 2012.LAI was measured at roughly weekly intervals.The starting and ending dates for LAI measurements were specified for each species based on phenological observations.The plot-level LAI was calculated as the sum over all component species.The methods for LAI measurements were detailed in the Supplement.

Flux calculation
Raw data were processed using the EddyPro 4.0.0 software (LI-COR Inc., USA).Processing steps included spike removal, tilt correction (double axis rotation), correction for sensor separation, spectral correction, detrending (Reynolds averaging) and flux computation (Burba and Anderson, 2010).Correction for density fluctuations (WPL terms) was not used, however, as LI-7200 is capable of outputting CO 2 mixing ratios, i.e., thermal expansion and water dilution of the sampled air have already been accounted for (Burba and Anderson, 2010).Half-hourly fluxes were rejected if missing records, removed spikes and absolute limit violations together exceeded 10 % of the total records of any of the three components of wind velocity and/or CO 2 concentration.
CO 2 fluxes were also excluded from analyses when turbulent mixing was low during calm nights (friction velocity u Introduction

Conclusions References
Tables Figures

Back Close
Full data quality control procedure rejected 29 % of the annual dataset.Downward fluxes are counted as negative and upward fluxes as positive.The overall performance of the eddy flux measurement system was evaluated by the degree of energy balance closure (Li et al., 2005), which was 78 % in 2012 when taking into account the heat stored in the soil above soil heat plates.

Gap-filling and partitioning NEE into GEP and R e
Linear interpolation was used to fill small gaps (≤ 2 h).For larger gaps during daytime (i.e., PAR ≥ 5 µmol m −2 s −1 ), NEE-PAR relationships were used for gap-filling.A light response model (Eq. 1) which incorporates photoinhibition at high radiation (Ye, 2007) was used to estimate missing daytime data because net CO 2 uptake declined at high PAR, especially in summer (Fig. 2).
where NEE day is daytime NEE (µmol CO 2 m −2 s −1 ), Q is incident PAR in units of µmol m −2 s −1 , Q c is the light compensation point, and α, β, γ are fit values for the following calculations (Ye, 2007).

Conclusions References
Tables Figures

Back Close
Full In Eqs. ( 2)-( 6), Q m is the PAR intensity at the maximum rate of net CO 2 uptake (NEE max ), φ 0 and φ c are the quantum yield at Q = 0 and Q = Q c , respectively, R d is the model-derived bulk ecosystem respiration.For the estimation of missing NEE day , Eq. ( 1) was fit to consecutive windows of 500 non-missing daytime data points.The Q 10 model was used for filling nighttime gaps (Zha et al., 2004): where NEE night is nighttime NEE, T s the soil temperature at 10 cm depth, R e10 the R e at T s = 10, Q 10 the temperature sensitivity of ecosystem respiration.Equation ( 7) was only fit to the annual dataset because short-term data points were too scattered to establish any valid NEE night − T s relationships.T s at the 10 cm depth was selected because it produced a higher coefficient of determination (R 2 ) than T s at other depths and T a .In order to estimate annual CO 2 fluxes, missing T s values were gap-filled with the mean diurnal variation (MDV) method (Moffat et al., 2007), while missing PAR values were gap-filled using an empirical relationship to half-hourly PAR data from a meteorological tower about 3 km east.
Daytime R e during the growing season was extrapolated from the temperature response function for NEE night (Eq.7).Off-season R e was considered as 24 h NEE fluxes.GEP was estimated as: according to VPD (VPD ≤ 1 kPa, 1 kPa ).The NEE day values were then bin-averaged before parameters were fit for each group.To evaluate the relative importance of different abiotic factors, Eq. ( 1) was fit to all half-hourly NEE day values during June-August, and the residuals were then subjected to least-square regressions and a stepwise multiple linear regression against VPD, T a and SWC (Z-transformed data were used in the stepwise regression).In order to test NEE day -T a and NEE day -VPD relationships, as well as their dependence on SWC, NEE day was compiled with respect to SWC at 30 cm depth (SWC ≤ 0.1 m 3 m −3 , SWC > 0.1 m 3 m −3 ), and then bin-averaged into 1 • C T a and 0.2 kPa VPD intervals, respectively.NEE day -T a and NEE day -VPD relationships were fit with the quadratic model.
For examining the effects of SWC on the R e -T s (10 cm depth) relationship, we classified NEE night when T s > 0 • C into two groups with respect to SWC at 30 cm depth ), and then bin-averaged NEE night into 1 • C T s intervals.R e10 and Q 10 in Eq. ( 7) were estimated separately for each SWC group.
A minimum of ten data points were required for a valid bin for all abovementioned binaverages.The following surface fitting (T s -REW model) was then used to further examine the interaction between temperature and water availability in regulating half-hourly NEE night : on REW for both the Q 10 model (Eq.7) and the T s -REW model (Eq.10).SWC at 30 cm depth was chosen for both daytime and nighttime analyses because its effects were most pronounced among SWC at different layers.Linear regressions were used to examine the seasonal relationships between GEP and R e , and the responses of NEE and GEP to seasonal changes in LAI.For this purpose, daily-integrated values were calculated for C fluxes, and daily LAI was derived by linear interpolation between measurements.The uncertainty for annual fluxes were estimated by bootstrapping, in which the gap-filled annual dataset was sampled 2000 times and then SD calculated (applied to half-hourly data).

Seasonal variation in environmental conditions
Environmental variables showed clear seasonal patterns (Fig. 1).The daily mean T a ranged from −8.5 on 23 January to 23.4 • C on 11 July (Fig. 1a).T s had a minimum of −12.1 • C on 8 February and a maximum of 25.8 • C on 22 June (Fig. 1a).The daily PAR reached a maximum of 61.5 mol m −2 day −1 on 15 June, and was < 30 mol m −2 day −1 during winter (Fig. 1b).Daily mean VPD reached a maximum of 2.2 kPa on 9 June, and was lower than < 0.5 kPa during winter (Fig. 1c).PPT summed to 304.9 mm from mid-May to December, > 60 % of which fell between June and August.There were three rain events larger than 20 mm day −1 , among which the largest occurred on 27 June (49.8mm day −1 ) (Fig. 1d).Snowmelt and soil thaw in early spring resulted in a relatively wet soil (Fig. 1d).During the growing season, SWC (except for that at 120 cm depth) followed the pattern of precipitation, and SWC in deeper layers (30 and 70 cm) only responded to large rainfall events (Fig. 1d).Introduction

Conclusions References
Tables Figures

Back Close
Full

Seasonal variation in NEE and its biophysical controls
Daily NEE ranged from −4.71 g C m −2 day −1 (largest net CO 2 uptake) on 30 June to 1.63 g C m −2 day −1 on 30 July (Fig. 1f).GEP reached a maximum of 6.78 g C m −2 day −1 on 30 June.Maximum R e was 3.26 g C m −2 day −1 on 25 July.Annual net ecosystem productivity (NEP = −NEE) was 77±7 (SD) g C m to NEP, leading to an annual GEP of 456 ± 8 g C m −2 yr −1 .PAR was the direct control over NEE day during the growing season (Fig. 2), accounting for > 80 % of the variability in NEE day in most months (except for October) (Table 1).The absolute values of NEE max , φ 0 , φ c and R d were all highest in July, while lower in spring and autumn (Table 1).The direct effect of PAR was modified by other environmental factors (Table 2; Fig. 3).The magnitude of NEE max and φ c decreased, while Q m increased, with increasing VPD and T a .In addition, R d increased with T a .The magnitude of NEE max , Q m and R d were all lower under dry soil conditions.The NEE day residuals were positively correlated with VPD and T a , and negatively correlated with SWC (Fig. 3d-f).The stepwise regression produced the following relationship: residual = −0.30SWC+ 0.17T a + 0.11VPD (R 2 = 0.16, P < 0.01).
NEE day first decreased (towards higher CO 2 uptake), and then increased, with increasing T a and VPD (Fig. 4).Moreover, NEE day was more responsive to T a and VPD, and showed higher maximum CO 2 uptake rates under wetter soil conditions (SWC ≥ 0.1 m 3 m −3 ).For most of the VPD range, NEE day was more negative under higher soil water availability (Fig. 4b).NEE night showed close relationships with T s at 10 cm depth for both SWC groups (Fig. 5).However, Q 10 was much larger, and R e10 slightly smaller, under higher soil water availability.The surface fitting showed that Q 10 increased from 1.9 to 3.2, and R e10 increased with from 0.73 to 0.83 µmol m −2 s −1 , when REW increased from 0 to 1 (Fig. 6).The T s -REW model fit the measured half-hourly values better than the Q 10 model (Fig. 7).Half-hourly NEE night residuals showed a positive correlation with REW

BGD Introduction Conclusions References
Tables Figures
There was a linear dependence of daily R e on GEP (R 2 = 0.65), with a slope of 0.34 (Fig. 8a).Both daily GEP and NEE responded to the seasonal variation of LAI (R 2 = 0.65 and 0.45, respectively), with a slope of 4.12 for GEP, and −2.03 for NEE (Fig. 8b and c).

Synoptic variation in NEE as related to rain pulses
Pulses of NEE were observed during the growing season when precipitation occurred (Fig. 9a).In order to examine the effects of rain events and related environmental factors on NEE, half-hourly measurements around the largest rainfall event (DOY 179-180, 61 mm) were scrutinized (Fig. 9b and c).NEE during daytime was markedly depressed on DOY 179-180 compared to the day before rain (DOY 178), and NEE during nighttime was slightly higher on DOY 179-180 than on days without rain (DOY 181-184).As a result, a positive pulse in daily NEE was observed on DOY 179-180 (Fig. 9a).The positive NEE pulse was accompanied by a sharp increase in SWC, and decreases in T a , T s and PAR (Fig. 9b and c).Daytime NEE was dramatically stimulated by high PAR, temperatures and SWC on days immediately following the rain event (DOY 181-182), leading to a clear post-rain peak in CO 2 uptake.

Diurnal variations in NEE and meteorological factors
The monthly mean diurnal variations of non-rainy days showed that the diel amplitude of NEE varied dramatically over the growing season, being largest in July and smallest in October (Fig. 10a).NEE during nighttime was relatively low in magnitude in comparison to that during daytime in most months (except for October).As a result, the ecosystem was a C sink on non-rainy days from May to September.Net CO 2 uptake peaked before noon (at 09:30-10:00 LST, LST = GMT + 8) in summer months (June-Introduction

Conclusions References
Tables Figures
Both T a and VPD were lowest in early morning, and peaked in late afternoon (Fig. 10c  and d).

Magnitude of NEP
Dryland ecosystems can vary from a net sink to source of CO 2 on an annual basis, as mainly determined by water availability and the amount of primary producers (Liu et al., 2012).The total amount of C sequestered by the studied shrubland in 2012 (NEP = 77 ± 7 g C m −2 yr −1 ), with an annual PPT of at least 305 mm and a peak LAI of 1.2, was generally lower than that by forests and grasslands in humid and subhumid areas (e.g., Suyker and Verma, 2001;Zha et al., 2004;Zhou et al., 2013).However, it was higher than many reported values from semiarid and arid nonforest ecosystems (Wang et al., 2008;Gao et al., 2012).For example, a revegetated shrub ecosystem ∼ 200 km west of our site dominated by Caragana korshinskii and A. ordosica had a NEP of 14-23 g C m −2 yr −1 , with an annual PPT of < 150 mm (Gao et al., 2012).An semiarid steppe in central Mongolia showed a NEP of 41 g C m −2 yr −1 , an annual PPT of 260 mm and a peak LAI of 0.57 (Li et al., 2005).Liu et al. (2012) reported that a salt desert shrubland (with LAI < 0.4) in northwest China was near C neutral in a dry year (PPT = 17 mm), but a C sink of 40 g C m −2 yr −1 in a wet year (PPT = 48 mm).The shrub ecosystem we studied has a wide regional distribution, thus represents a huge C fixation potential.At our site, PPT in 2012 was at least 18 mm higher than the longterm mean, which calls for future studies that assess how semiarid shrub ecosystems respond to interannual variability in water availability.Introduction

Conclusions References
Tables Figures

Back Close
Full

Effects of abiotic stresses on NEE
The nature and relative importance of different biophysical factors in controlling NEE differ among ecosystem types (Fu et al., 2006).Water stress, which has the largest seasonal and interannual variation among all environmental variables, is the most common limitation to vegetation growth in dryland ecosystems (Fu et al., 2006).Our study also revealed the dominant role of soil water in limiting photosynthesis.Fu et al. (2006) found that water stress was responsible for photosynthetic depression in a semiarid steppe, while high temperature together with strong solar radiation caused a decrease of NEE day in an alpine shrub.Our findings that low SWC and high VPD depressed the maximum rate of CO 2 uptake (NEE max ) (Table 2; Fig. 3) were in agreement with previous studies in dryland ecosystems (Li et al., 2005;Wang et al., 2008;Yang et al., 2011).Both SWC and VPD affect plant hydraulic status, however, they reduce C assimilation through different mechanisms.Dry soil leads to reduced water supply for metabolism and cell expansion, while VPD affects CO 2 supply for photosynthesis by regulating stomatal conductance and evaporative demand (Zhou et al., 2013).Moreover, the two mechanisms did not act in isolation, but interacted to reduce CO 2 uptake under water-stressed conditions (Fig. 4).Li et al. (2005) concluded that the sensitivity of stomata to VPD becomes stronger once leaf water potential starts to drop because of deficiency of water supply from the soil.Low soil water availability may aggravate VPD-induced stomatal closure.
Water limitation of R e or R s has been found in various types of ecosystems (Gao et al., 2012).Our results showed a marked decrease in both the magnitude and temperature sensitivity (Q 10 ) of R e under low SWC (Figs. 5 and 6).Wang et al. (2014) measured R s in a nearby community dominated by A. ordosica (800 m north of the eddy-flux tower), and found that R s was closely correlated with T s when SWC at 10 cm depth was higher than 0.08 m 3 m −3 , while it was decoupled from temperature during dry periods.The reduction in Q 10 of R e under drought conditions was most likely associated with decreased C transportation to roots due to suppressed photosynthesis,

Conclusions References
Tables Figures

Back Close
Full inactivated rhizosphere and switched C pool being respired (e.g., from labile to recalcitrant) (Zhang et al., 2007;Wang et al., 2008;Gao et al., 2012).Air temperature is another factor affecting VPD besides humidity, which could explain the reduction in NEE max at high temperatures (Fig. 3b).Alternatively, high T a might have enhanced respiration and/or suppressed photosystem II, resulting in smaller NEE max .
The quadratic relationship between NEE day and T a was likely due to PAR-limited C assimilation on cloudy and rainy days at low temperatures, and to heat-and waterlimited NEE day at high temperatures (Fu et al., 2006).Similar to Li et al. (2005), we found that NEE day became less responsive to T a under drought conditions (Fig. 4a), reflecting drought limitations to both plant and microbial activities.A downward shift of optimum T a of NEE day under low SWC was attributable to different responses of GEP and R e to water deficit (Fu et al., 2006).The asymmetric distribution of net photosynthesis around noon (Fig. 10a) is commonly observed in arid areas (Zhang et al., 2007), and can be ascribed to either temperature-induced increases in R e (Fig. 10c) or VPD-induced stomatal closure (Fig. 10d) in the afternoon (Yang et al., 2011).The strong solar radiation at our site (Figs. 1 and 2) led to decreases in CO 2 uptake in summer months (Figs. 2 and 3).Insufficient thermal dissipation of leaves and consequent damage of photosynthetic apparatus are the likely causes for photoinhibition of CO 2 assimilation (Fu et al., 2006).

Effects of rain pulses on NEE
Rain events triggered pulsed dynamics of NEE in the shrub ecosystem during the growing period (Fig. 9a).Large positive daily NEE occurred immediately after a rainfall event, but the peak of CO 2 uptake lagged the pulse of CO 2 release by 1-2 days (Fig. 9a  and b).Similarly, Gao et al. (2012) found that a large precipitation event resulted in a rapid burst of positive C flux (CO 2 release) before negative values set in 1-2 days later in a revegetated shrubland in northwest China.Wang et al. (2014) also reported immediate pulses of R s following rain events in an A. ordosica dominated community at our site.In a semiarid steppe in central Mongolia, the respiration enhancement effect

Conclusions References
Tables Figures

Back Close
Full was even higher during the rain period itself (Li et al., 2005).In contrast, Liu et al. (2011) found in a saline desert that NEE took 4-5 days to reach its peak of CO 2 uptake after rain.The rapid stimulation of CO 2 release by precipitation may arise from the rapid activation of microorganisms and decomposition of soil organic matter due to the "Birch effect", which has been widely reported in regions with dry soils when a rainfall event occurs after a period of drought (Jarvis et al., 2007).Alternatively, soil water from rainfall may have degassed the CO 2 stored in soil pores, considering the sandy soils with high porosity in desert areas (Lee et al., 2004).The lagged responses of photosynthesis may be associated with physiological acclimation and recovery of plants from the preceding dry period (Liu et al., 2011), and related to the post-rainfall environmental conditions that are optimum for C assimilation (Fig. 9c).These results showed that fast and slow responses together contribute to the pulsed ecosystem behavior.

Biotic controls on CO 2 fluxes
Leaf area represents both the amount of photosynthetic tissues and the amount of intercepted light by the vegetation (Yang et al., 2011).Our results that photosynthetic parameters varied seasonally (Table 2; Fig. 2) with canopy development (e.g., changes in LAI) has been reported previously for different vegetations (Zha et al., 2004;Yang et al., 2011).Furthermore, we found that LAI accounted for 45 and 65 % of the seasonal variation in NEE and GEP, respectively (Fig. 8b and c), indicating the importance of canopy development in controlling C balance.Similar GEP-LAI and NEE-LAI relationships have been reported for steppe, grassland and pasture ecosystems (e.g., Tappeiner andCernusca, 1998, Flanagan et al., 2002;Yang et al., 2011).The slope of the GEP-LAI relationship reported here (4.1 g C (m 2 leaf area) −1 day −1 ) was comparable to that in a semiarid steppe (3.1 g C m −2 day −1 ) (Li et al., 2005) and two Mediterranean grasslands (3.9-4.1 g C m −2 day −1 , Xu and Baldocchi, 2004;Aires et al., 2008).
However, it was much smaller than that (7.5-8.7 g C m −2 day −1 ) found in a Canadian temperate grassland (Flanagan et al., 2002).A small GEP-LAI slope may be indicative of water and nutrient limitations (Li et al., 2005).Introduction

Conclusions References
Tables Figures

Back Close
Full Canopy productivity was shown to have a shadowing effect on R e as photosynthesis provides substrates to both autotrophic and heterotrophic respiration (Wan and Luo, 2003;Flanagan and Johnson, 2005).In our study, GEP accounted for 65 % of the seasonal variation in R e , with a slope of 0.34 (Fig. 8a).Li et al. (2005) reported a similar slope (0.33) for the R e -GEP relationship in a semiarid steppe.Xu and Baldocchi (2004) and Aires et al. (2008) showed slopes of 0.47 and 0.53 in Mediterranean grasslands, respectively.However, Liu et al. (2011) reported a larger slope (0.8) in a saline desert shrub ecosystem, which may reflect greater C allocation to respiratory tissues (stems, branches and coarse roots) or functions (e.g., maintenance respiration).

Conclusions
In line with our hypotheses, we found in the semiarid shrub ecosystem that (1) water stress exerted a major control over half-hourly changes in NEE during the peak growing season, and interacted with heat stress and photoinhibition in constraining C fixation; (2) rain pulses regulated NEE at the synoptic scale, highlighting the role of water supply in the alleviation of abiotic stresses; (3) canopy development largely determined NEE and GEP over the entire growing season.Climate modeling suggests a warmer and drier future climate in the semiarid and arid regions of Asia (McCarthy et al., 2001).Hence, more stressful environmental conditions in the future may lead to substantially lower carbon sequestration capacity in temperate semiarid areas.Also, the predicted higher variability in precipitation (Easterling et al., 2000), i.e., more extreme but less frequent rainfall events intervened by longer dry periods, accentuates the role of the temporal pattern of water availability in controlling NEE in the future., 10, 1492-1503, 2004.Zhang, W. L., Chen, S. P., Chen, J., Wei, L., Han, X. G., and Lin, G. H.: Biophysical regulations of carbon fluxes of a steppe and a cultivated cropland in semiarid Inner Mongolia, Agr. Forest Meteorol., 146, 216-229, 2007.

Conclusions References
Tables Figures

Back Close
Full Full and arid areas) cover over 40 % of the Earth's land surface, and are rapidly expanding as a result of climate change and human activities (Asner et al., Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | was fit monthly from May to October to evaluate the seasonal variation in light response of NEE.The regressions were conducted on bin-averaged data using 50 µmol m −2 s −1 PAR intervals.In order to test the dependency of the NEE day -PAR relationship on abiotic factors and exclude the confounding effects of plant phenology, we compiled NEE day during the peak growing season (June-August) into multiple groups Introduction Discussion Paper | Discussion Paper | Discussion Paper | 9) where a, b, c and d are fit parameters.REW means relative extractable water content (Zhou et al., 2013), which was calculated as: REW = SWC − SWC min SWC max − SWC min (10) where SWC max and SWC min are the minimum and maximum SWC during the period of T s > 0 • C, respectively.Linear regressions were used to compare measured vs. predicted half-hourly NEE night values and to examine the dependence of NEE night residuals Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | online at http://www.biogeosciences-discuss.net/11/5089/2014/ bgd-11-5089-2014-supplement.pdf.Discussion Paper | Discussion Paper | Discussion Paper | Zha, T., Kellomäki, S., Wang, K., and Rouvinen, I.: Carbon sequestration and ecosystem respiration for 4 years in a scots pine forest, Glob.Change Biol.
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Fig. 1 .
Fig. 1.Seasonal variations of air temperature (T a ) at 6 m aboveground and soil temperature (T s ) at 10 cm depth (a), incident photosynthetically active radiation (PAR) (b), vapor pressure deficit (VPD) (c), soil water content (SWC) and precipitation (PPT) (d), leaf area index (LAI) (e) and CO 2 fluxes (f) in 2012.NEE, net ecosystem CO 2 exchange; GEP, gross ecosystem productivity; R e , ecosystem respiration.Daily means are shown for T a , T s , VPD and SWC; Daily sums are shown for PAR, PPT and CO 2 fluxes; interpolated daily values are shown for LAI.The vertical dashed lines separate each month, and the horizontal dashed lines in (a) and (f) represent y = 0.

Fig. 4 .Fig. 7 .
Fig. 4. Responses of daytime net ecosystem CO 2 exchange (NEE day ) to air temperature (T a ) (a) and vapor pressure deficit (VPD) (b) under different soil water contents (SWC) measured at 30 cm depth.Only data from the peak growing season (June-August) were used to minimize the confounding effects of phenology.Half-hourly NEE day was bin-averaged into 1 • C T s (a) and 0.2 kPa VPD (b) intervals, respectively.Bars indicate standard errors.Response curves were fit using the quadratic model.The horizontal dashed lines represent y = 0.

Fig. 9 .
Fig. 9 1 at multiple timescales.We hypothesized that soil water shortage is dominant over other stresses in controlling NEE of dryland ecosystems, and could modify the responses of NEE to other environmental factors.We also proposed that the seasonal dynamics of LAI is an important determinant of productivity over the growing season, whereas at shorter timescales (e.g., hourly) abiotic stresses could impose critical constraints on CO 2 fluxes.
• 13.62 E, 1530 m a.s.l.), Ningxia, northwest China.The area is located in the midtemperate zone and characterized by a semiarid continental monsoon climate.The mean annual temperature) is 8.1• C and the frost-free season lasts for 165 days on average

Table 1 .
Parameter values describing the response of daytime NEE to incident PAR for each month and the entire growing season (GS).