Soil Respiration in an Old-Growth Subtropical Forest

JOURNAL OF GEOPHYSICAL RESEARCH: ATMOSPHERES, VOL. 118, 2981–2990, doi:10.1002/jgrd.50300, 2013

Soil respiration in an old-growth subtropical forest: Patterns, components, and controls Zheng-Hong Tan,1,2 Yi-Ping Zhang,1,3,4 Naishen Liang,2 Qing-Hai Song,1,3,5 Yu-Hong Liu,1,4 Guang-Yong You,1,3,5 Lin-Hui Li,1,3,5 Lei Yu,1,3,5 Chuan-Shen Wu,1,4 Zhi-Yun Lu,1,4 Han-Dong Wen,1,4 Jun-Fu Zhao,1,3,5 Fu Gao,6 Lian-Yan Yang,1,3 Liang Song,1 Yong-Jiang Zhang,1,7 Teramoto Munemasa,2 and Li-Qing Sha1 Received 6 December 2012; revised 20 January 2013; accepted 18 February 2013; published 4 April 2013.

[1] The patterns, components, and controls of soil respiration in an old-growth subtropical forest were investigated using an automatic chamber system. We measured soil respiration in three treatments (control, trenching, litter removal) over 15 months. The annual total soil respiration (1248 gC m–2 yr–1) showed considerable spatial variation (coefficient of variation = 27.8%) within the forest. Thirty samples were required to obtain results within 10% of the mean value at a 95% confidential level. A distinctive cosine-like diel pattern of soil respiration was observed; the time lag between gross primary production and soil respiration at this scale was calculated to be 4–5 h. Seasonality of soil respiration was strong (~1 mmol m–2 s–1 near the end of winter; ~6 mmol m–2 s–1 in midsummer). No time lag was discerned between gross primary production and soil respiration at the seasonal scale. Soil temperature at 5 cm below surface can explain most (>91%) of the observed annual variation in soil respiration. The apparent respiration temperature sensitivity index (Q10)was3.05.ThelowestQ10 value was observed in winter, when soil moisture was low. Soil respiration was overestimated by a Q10 function during both dry and wet periods. The relative contributions of soil organic matter (RSOM), litterfall decomposition (RL), and root respiration (RR) to total soil respiration are 65.25%, 18.73%, and 16.01%, respectively; the temperature sensitivity of these components differ: RL (Q10 = 7.22) > RSOM (2.73) > RR (1.65). This relationship between Q10 values for litter respiration, soil organic matter decomposition, and root respiration still holds after minimizing the confounding effect of moisture. A relatively constant substrate supply and/or thermal acclimation could account for the observed low-temperature sensitivity in root respiration. Given the high carbon stocks and fluxes, the old-growth subtropical forests of China seem important in the global carbon budget and climate change. Citation: Tan, Z.-H., et al. (2013), Soil respiration in an old-growth subtropical forest: Patterns, components, and controls, J. Geophys. Res. Atmos., 118, 2981–2990, doi:10.1002/jgrd.50300.

1. Introduction

[2] Human activity has signiﬁcantly altered global carbon Additional supporting information may be found in the online version of this article. cycling [Magnani et al., 2007; Vitousek et al., 1997]. Large 1Key Lab of Tropical Forest Ecology, Xishuangbanna Tropical Botanical amounts of carbon (~375 billion tons) have been released Garden, Chinese Academy of Sciences, 650223 Kunming, China. into the atmosphere through fossil fuel combustion, cement 2Global Carbon Cycle Research Section, Center for Global Environmental Research, National Institute for Environmental Studies, manufacturing, and tropical deforestation. Around half of Tsukuba, 305-8506, Japan. this carbon dioxide has been absorbed by the ocean and 3Global Change Research Group, Key Laboratory of Tropical Forest terrestrial biosphere, the remaining half remains in the atmo- Ecology, Chinese Academy of Sciences, 650223 Kunming, China. 4 sphere. Atmospheric carbon dioxide has increased from Ailaoshan Station for Subtropical Forest Studies, 676200, Jingdong, China. 280 ppm in the preindustrial era to 390 ppm in 2011 5University of Chinese Academy of Sciences, Beijing 100094 China. 6Lijiang Ecological Station, Kunming Institute of Botany, Chinese [WMO, 2012]. Carbon dioxide is one of many long-lived Academy of Sciences, Kunming, China. greenhouse gases. An increase in its atmospheric concentra- 7Department of Organismic and Evolutionary Biology, Harvard tion changes Earth’s radiation balance and climate system University, Cambridge, Massachusetts, 02138, USA. and leads to so-called global climate change. Corresponding author: L.-Q. Sha, Key Laboratory of Tropical Forest [3] In the study of climatic change and carbon cycling Ecology, Xishuangbanna Tropical Botanical Garden, Chinese Academy of interaction, there are two main research issues. One is the Sciences, 650223 Kunming, China. ([email protected]) “missing sink” in global carbon cycling [Pan et al., 2011; ©2013. American Geophysical Union. All Rights Reserved. Woodwell et al., 1978]. The other is the feedback between 2169-897X/13/10.1002/jgrd.50300 climatic change and carbon cycling [Cox et al., 2000]. As

2981 TAN ET AL.: SUBTROPICAL FOREST SOIL RESPIRATION mentioned earlier, the ocean and terrestrial biosphere are [7] We tried to investigate all aspects of the components, natural carbon sinks and have absorbed nearly half of all patterns, and controls of soil respiration at the study location artificial carbon dioxide emissions. In attempts to find the with field data collected from a soil automatic chamber, a size and location of terrestrial carbon sinks, the “missing nearby eddy flux system, and a nearby climate station. The sink” issue has been raised. It is still not well addressed main hypotheses of this study are: and remains full of uncertainties [Pan et al., 2011]. The [8] 1. Due to high soil-carbon densities and well-watered feedback issue is to understand whether global climatic conditions, the annual total soil respiration of the forest will change has a positive or negative impact on natural carbon be higher than the predicted value based on a global MAT- sinks [Denman et al., 2007]. If natural carbon sinks are Rs (MAT: mean annual temperature) relationship [Raich enhanced in climatic change, an increase in atmospheric and Schlesinger, 1992]. carbon dioxide will be mitigated and negative feedback will [9] 2. For an evergreen canopy and a year-long growing occur. The opposite holds true as well. The presence of feed- season, the seasonal variation of root respiration, both growth back is commonly accepted and plays a crucial role in and maintenance, will not be as strong as that of a temperate or predicting future climate trends. However, our understanding boreal forest [Bååth and Wallander,2003;Boone et al., 1998]. of them is still highly limited. Meanwhile, a low-temperature sensitivity of root respiration is [4] Soil is the largest carbon pool in terrestrial ecosystems, expected in the subtropical forest. containing 2700 Gt of carbon. This amount is more than the [10] 3. Soil water content covaried with temperature in the sum of carbon in atmosphere (780 Gt) and biomass (575 Gt) studied forest, which is influenced by a monsoon climate. [Lal, 2008]. Around 60–80% of photosynthetic production Water will have a strong confounding effect on the derived ap- is respired into the atmosphere through soil respiration parent temperature sensitivity of litterfall and soil organic mat- [Bond-Lamberty and Thomson, 2010]. The magnitude of ter decomposition. soil respiration (Rs) is about 13 times that of fossil fuel combustion. Therefore, a slight variation of soil respiration 2. Materials and Methods caused by biotic or abiotic factors can exert a strong impact on global carbon balance [Davidson and Janssens, 2006]. 2.1. Site Description Although numerous studies have been carried out to investigate [11] The study site is located in the Mt. Ailao Nature the soil respiration of different vegetation types, the regional Reserve (24320N, 101010E; 2476 m elevation) in Yunnan or global pattern of soil respiration, and its components and Province, SW China. In this area, an old-growth subtropical environmental controls, still needs to be addressed evergreen broadleaved forest is spread widely and well [Bond-Lamberty and Thomson, 2010]. protected. This forest has a stand age that exceeds 300 years [5] Region-wide field measurements in soil respiration and is free of management. The dominant vegetation species employ the same standards and, combined with manipula- in this forest are Lithocarpus chintungensis, Rhododendron tion treatments, could be used to address the “missing sink” leptothrium, Vaccinium ducluoxii, Lithocarpus xylocarpus, on the one hand and to quantify climate change carbon-cycle Castanopsis wattii, Schima noronhae, Hartia sinensis, and feedbacks on the other. Thus, we established a soil Manglietia insignsis. The tree density of the forest is respiration measurement network in Asia. An automatic 2728 ha–1; the median tree height is 9.0 m; median tree soil respiration chamber was applied to all the measurement diameter at breast height is 9.5 cm; and the median basal sites in the network. The network covers nearly all area in the forest is 91 m–2 ha–1. The leaf area index terrestrial ecosystems in Asia, from tropical rainforest in measured by the canopy analyzer (LAI-2000, Li-Cor Inc., Southeast Asia, to subtropical forest, temperate forest, Lincoln, NE, USA) is ~5.0. The estimated total stand boreal forest and tundra in west Siberia. At some of the biomass is 499 t ha–1. Mean annual air temperature is sites, soil was heated using infrared lamps to model a 11.3C, with monthly mean values ranging from 5.4 to warming effect. 23.5C. The site receives an annual average of 1840 mm [6] In this study, we report on the soil respiration mea- of precipitation, based on more than 20 years of data collected surements taken from one site of the network located in at a meteorological station. The region has two distinct an old-growth subtropical forest of China. Soil respiration seasons influenced by a monsoon climate. The wet season data collected in this area are very important for synthetic occurs from May through October, and the dry season studies; i.e., studies by Bond-Lamberty and Thomson occurs from November to April. The soils are loamy Alfisols. [2010] and Mahecha et al. [2010]. Furthermore, they are An organic carbon horizon is located 3–7 cm below ground important because these forests serve as regional carbon surface. It has a pH of 4.5 and organic carbon and total sinks. In Asia, mainly in China, there is the dry belt that nitrogen contents of 304 and 18 g kg–1,respectively is controlled by a subtropical high pressure resulting from [Tan et al., 2011]. the effect caused by the Tibetan Plateau [Kira, 1991]. Sub- tropical evergreen forests are potential vegetations in the 2.2. Experimental Design and Soil Respiration area, but not deserts or subtropical savannas. These sub- Measurement tropical forests were largely destroyed by humans in the [12] Twenty chambers were divided into four treatments past years. Only in the mountainous areas that are difficult (five chambers per treatment): control, trenching, above- to access are old-growth forests still found. In previous ground litter removal, and infrared light warming. Locations studies, these old-growth forests were reported to be strong of the chambers are shown in Figure S1 in the auxiliary carbon sinks [Zhou et al., 2006]. Whether these carbon material. A warming treatment was not included in this sinks persist or weaken under a warming climate is uncertain study. For the trenching treatments, we excavated trenches [Tan et al., 2012]. down to 50 cm in depth, lined them with plastic sheets,

2982 TAN ET AL.: SUBTROPICAL FOREST SOIL RESPIRATION and then refilled and packed them carefully with the original [17] The main components of the control box are an infrared soil. The trench depth of 50 cm was based on: gas analyzer (IRGA; Li-840, Li-Cor Inc.) and a data logger [13] i. The soil profile data showed that the main root sys- (CR10X, Campbell Scientific Inc.). During measurement, air tem of the studied forest trees is seldom deeper than 1.5 m. in the closed chamber is circulated through the IRGA by a [14] ii. Soil respiration was mainly contributed by fine root microdiaphragm pump (CM-50, Enomoto Ltd., Tokyo, but not coarse root. And the active fine root could also be Japan). The 20 chambers are closed sequentially by a home- reflected by the water uptake. A previous study in the same made relay board controlled by the data logger. The data forest showed that 82% water abstraction by roots 40 cm or logger acquires output signals from the IRGA every second shallower [Liu et al., 2003]. We could then believe a trench and records an average value every 10 s; the total sampling of 50 cm could at least exclude over 82% of the root period for each chamber was 180 s. For each chamber, 1 efflux respiration. value was obtained per hour. The efflux was calculated from [15] The aboveground litter was removed every two 18 records, as shown in equation (1): weeks in the aboveground litter removal treatment. ðÞ @ 16 fl VP 1 W c [ ]Soilefux was monitored by a multichannel automated Rs ¼ (1) measurement system developed by Liang at the Japan RST @t National Institute for Environmental Studies. The system measured soil efflux in a flow-through and non–steady-state where V is volume of the chamber (m3); S, the base area of manner, and was comprised of 20 automatic chambers and a the chamber (m2); R, a gas constant (8.314 Pa m3 K–1); T, control box. The chambers (90 cm 90 cm 50 cm) were the air temperature in the chamber (K); P, the air pressure made from clear PVC. The system incorporates several design (Pa); W, the water vapor mole fraction; and @C/@t, the rate features to prevent gas outlet [Liang et al., 2003]. Two lids at of increase in carbon dioxide mole fraction (mmol mol–1 s–1) the top of the chamber can be raised or closed; they are in the chamber calculated by the least squares method. operated by compressed air (MAX-E-12, Techno Fronto) [18] Soil temperature at 5 cm depth and air temperature regulated by a four-way valve (BK120, Techno Fronto). Two inside each chamber were measured with self-made thermo- fans (KMFH-12B, Kyoei, Tokyo, Japan) mounted in each couples. Soil moisture at 10 cm depth was monitored with chamber ensure sufficient mixing of air during measurement. time-domain reflectometers (CS-616, Campbell Scientific

Figure 1. (a) Diel pattern of soil respiration (Rs) and soil temperature (Ts) for 2011; (b) diel pattern of gross primary production derived from eddy covariance observations (P), photosynthetically active radiation (Qp), and atmospheric vapor pressure deﬁcit (de); and (c) diel pattern of Rs and modeled soil respiration (Rt)in autumn. In Figure 1c, the grey area shows the difference between observed and modeled soil respiration.

2983 TAN ET AL.: SUBTROPICAL FOREST SOIL RESPIRATION

Figure 2. (a) Annual variation of Ts and soil volumetric water content (Ws) measured at 5 cm in depth each hour, and (b) daily binned gross primary production (P) derived from eddy covariance observations. The grey line in Figure 2b shows results after smoothing. Daily binned respiration is shown for different treatments: control (open circles), no litter (grey circles), and no root (solid line).

Inc., Logan, UT, USA). Air pressure at 30 cm height in the mounted at the height of 34 m. Eddy flux data were first center of the plot was measured by a pressure transducer controlled and assessed; i.e., spike removal, coordinate rota- (PX2760, Omega Engineering, Inc., Stamford, CT, USA). tion, storage-flux correction, low friction velocity filtering, density correction [Tan et al., 2011]. Gap filling and flux partitioning were completed through an online procedure that 2.3. Complementary Measurements is maintained by Max Planck Institute (http://www.bgc-jena. [19] The eddy covariance system was comprised of an mpg.de/~MDIwork/eddyproc/index.php). Gross primary open-path IRGA (LI-7500, Li-Cor Inc., Lincoln, NE, USA) production (P) was the same as the gross primary production and a three-dimensional sonic anemometer (CSAT-3, Camp- that was derived from flux partitioning (GPP). bell Scientific Inc., Logan, UT, USA). The system was [20] Soil temperature was measured at a depth of 5 cm with a self-made thermometer and a time-domain reflectom- eter. Data were collected at hourly intervals, as controlled by a data logger (CR-1000, Campbell Scientific Inc., Logan, UT, USA). [21] Leaf area index was measured on a typical cloudy day each month with a canopy analyzer (LAI-2000, Li-Cor Inc., Lincoln, NE, USA). Aboveground litter was collected with 25 litter traps (1 m 1 m in area) every month. Litter was taken into the laboratory for analysis. Photosynthetic active radiation (Qp) was measured at a height of 34 m from the eddy flux tower. Air humidity and air temperature was measured by humidity sensors (HMP45C, Vaisala, Helsinki, Finland). The vapor pressure deficit (de) was calculated using the approach described by Campbell and Norman [1998].

Figure 3. Relative contribution of aboveground litter decom- 2.4. Calculations position (RL), root respiration (RR), and soil organic matter decomposition (RSOM)toRs in the year 2011 (annual) and in [22] We calculated soil respiration components from winter (December–February), spring (March–May), summer measurements obtained under different treatments. These (June–August), and autumn (September–November). components are related mathematically as

2984 TAN ET AL.: SUBTROPICAL FOREST SOIL RESPIRATION

RR ¼ Rc RNR (2) account during regression. After separating the confounding fi ¼ effects, Q10 values were calculated base on the tted parame- RL Rc RNL (3) ters. The model here is a typical multiplicative model of RSOM ¼ Rc RR RL (4) temperature and soil moisture and is given as ¼ ðÞðÞ= where RR, RL, RSOM, RC, RNR, and RNL are, respectively, Rs Rref exp ln Q10 Ts 10 (9) respiration of root, aboveground litter, soil organic matter, Q ¼ a a T þ a S þ a S2 (10) control, no-root, and no-aboveground litter treatments. 10 1 2 s 3 w 4 w [23]AQ function and Lloyd-Taylor equation [Lloyd and 10 where Rref and ai where i =1,2,3,4arefitted parameters. Sw Taylor, 1994] were used to obtain a temperature-respiration is volumetric soil water content at 5 cm depth. Initial param- relationship. These equations are given as eter values are 0.1, 3.0, 0.1, 10, and 0.1, respectively. ðÞ = [26] Nonlinear regressions, both univariate and multivari- ¼ Ts 10 10 Rs R10 Q10 (5) ate, were accomplished by using the nlinfit command in fi Rs ¼ R283 expðÞE0ðÞ1=ðÞ283:16 T0 1=ðÞTs T0 (6) Matlab 7.1. The con dence interval of regression parameters was estimated with the nlparci command. where Rs is soil respiration, Ts is soil temperature at 5 cm fi depth, R10 is the tted Rs at a soil temperature of 10 C, 3. Results Q10 is a temperature sensitivity index of Rs, R283 is the fitted Rs at a temperature of 10 C (283 K), and E0 and T0 are two 3.1. Diel Pattern of Soil Respiration and Related fitted parameters. The index Q10 is defined as the Rs at one Ecological Variables fl temperature over the ux at a temperature 10 C lower. [27] An obvious cosine-like diel pattern of soil respiration 24 [ ] A two-parameter exponential equation was also (Rs) was observed and is shown with hourly data in used to obtain a temperature-respiration relationship. This Figure 1a. Respiration peaks in late afternoon (between equation can be expressed as 17:00 and 19:00 h) and is lowest in early morning (between 08:00 and 10:00 h). The daily range of Rs is around Rs ¼ a expðÞbTs (7)

Q10 ¼ expðÞ 10b (8) where a and b are ﬁtted parameters. [25] Two approaches were taken to separate the confounding effects of temperature and water. One is called well-watered regression, which determines a temperature response after leaving out measurements when soil moisture is too low or too high (determined by temperature model residuals near to zero). The other is a multivariate regression that uses a multivariate mixed model. This approach takes soil moisture into

Figure 5. Explanation of the temporal variation of soil respiration with (a) soil temperature and (b) soil water content. Black circles represent raw hourly data in 2011. In Figure 5a, solid and dashed lines are ﬁtted by the Q10 function and Lloyd-Taylor equation. In Figure 5b, the residual values of Figure 4. Spatial pattern of soil respiration before treatments observed soil respiration and modeled values from the Q10 (from 8 October 2010 to 16 December 2010). There were a function were related to soil water content at 5 cm. The grey total of 20 permanent chambers with 5 rows and 4 columns. line indicates smoothing results.

2985 TAN ET AL.: SUBTROPICAL FOREST SOIL RESPIRATION

different diel pattern. Its peak value occurred in midday (between 13:00 and 14:00 h) when the highest photosynthetically active radiation (QP) occurred (Figure 1b). The diel pattern observed for P also occurred for QP but not for atmospheric water vapor deficit (de). The peak of Rs occurred 4 or 5 h after the peak of P. The relationship between Rs and P at the diel scale is comparable, roughly, to that between radiation intensity and temperature. The observed autumn soil respiration (Rs) is higher than the modeled soil respiration (Rt), especially during the latter afternoon (Figure 1c). The difference between Rs and Rt did not show a diel pattern that was similar to that of gross primary production (P)or radiation flux (QP). 3.2. Annual Pattern of Soil Respiration and Related Ecological Variables [28] Soil temperature varied markedly with season as shown in Figure 2a. The annual mean and range of soil temperature over the period of observation was 11.02C and ~10C. Maximum soil temperature coincided with the highest summertime soil water content. The minimum soil water content occurred when temperature increased rapidly in early spring. The annual mean and range of soil water content was 38.14% and ~11%. This feature indicates a mild seasonal variation of soil water content. Gross primary production (P), derived from eddy-covariance observations, –2 –1 –2 –1 ranged from 10 gCO2 m d in winter to 30 gCO2 m d in summer (Figure 2b). It increased gently from a moderate level in winter to a peak value during late summer and early autumn (August and September) and then declined smoothly. There is a similar and distinct seasonal pattern of soil respiration among all treatments: control, no litter, and no root (Figure 2c). Soil temperature increased after Figure 6. Relationship between soil respiration rate mid-January, but it did not induce a corresponding increase –2 –1 in soil respiration. A slight decline of soil respiration was (mmol m s )andTs ( C) for different seasons. Black circles represent raw hourly data. The solid line was fitted using observed despite an increase in temperature until mid-March. the two-parameter exponential equation (R = a*exp(bTs)). The decline of soil respiration was temporally correlated with s a decrease of soil water content in this period (from mid- January to mid-March). This correlation suggests that soil – – 0.25 mmol m 2 s 1 and accounts for ~7% of the mean respi- water content plays a leading role in regulating soil respiration ration rate. A similar diel pattern of soil temperature at 5 cm during the period (Figures 2a and 2c). Afterward, soil respira- depth (Ts) was observed, suggesting that Ts plays a strong tion increased rapidly under the well-watered condition and in role in controlling Rs at the diel scale. Gross primary produc- the warm climate of late spring and early summer (during tion (P), derived by the eddy-covariance method, showed a April and June) and peaked in midsummer.

–2 –1 bTs Table 1. (a) Parameters of the Relationship Between Soil Respiration Rate (mmol m s ) and Soil Temperature ( C) at 5 cm (Rs = a*e( )) for Different Seasons; (b) Parameters of the Relationship Between Soil Respiration Rate and Different Soil Respiration Components: Total Soil Respiration (Rs), Derived Root Respiration Calculated as the Difference Between Control and No Root Treatment (RR), Derived Above Ground Litter Decomposition Calculated as the Difference Between Control and No Litter Treatment (RL), and Derived Soil Organic Matter Decomposition Deﬁned as Control Minus the Sum of RR and RL (RSOM) 2 ab p rQ10(Conﬁdence Interval at 0.05) (a) Season WINTER (December–February) 1.3536 0.0479 <0.0001 0.1338 1.61(1.53–1.69) SPRING (March–May) 0.7712 0.1193 <0.0001 0.6488 3.29(3.12–3.46) SUMMER (June–August) 0.9054 0.1097 <0.0001 0.4795 2.99(2.84–3.14) AUTUMN (September–November) 1.0135 0.0975 <0.0001 0.9130 2.65(2.60–2.69) (b) Component RS 0.8626 0.1117 <0.0001 0.9168 3.05(3.02–3.08) RR 0.2924 0.0501 <0.0001 0.1220 1.65(1.59–1.70) RL 0.0591 0.1978 <0.0001 0.7160 7.22(6.91–7.54) RSOM 0.6460 0.1005 <0.0001 0.7545 2.73(2.68–2.77)

2986 TAN ET AL.: SUBTROPICAL FOREST SOIL RESPIRATION

3.3. Relative Contribution of Components to Total Soil 3.5. Environmental Controls on Soil Respiration Respiration [31] A strong dependence of soil respiration on tempera- 2 [29] The relative contribution of RR, RL, and RSOM to total ture was detected (r = 0.91, n = 7113, p < 0.0001), based soil respiration is shown in Figure 3. The main component on hourly data that were averaged over five chambers in contributing to total soil respiration is RSOM, which accounts the studied forest (Figure 5a). This finding indicates that soil for up to 65.25% annually. Component RL contributed temperature is the dominant factor in controlling soil respira- 18.73% to total soil respiration, and component RR contrib- tion both at diel (Figure 1a) and annual scales. The Q10 uted 16.01%. The relative contribution of RL is higher function and Llyod-Taylor equation describe the data set during summer and autumn, while that of RR is higher well (Figure 5a). Compared to the estimate of respiration during spring and winter. from the Llyod-Taylor equation, the estimate calculated from the Q10 function was slightly higher under high- and low-temperature conditions. The residuals between observed 3.4. Spatial Variation of Soil Respiration values of Rs and those from the Q10 function were related to [30] Respiration measurements were recorded before we soil water content (Figure 5b). The Q10 function overestimates –2 –1 started treatments to investigate the spatial variation of (residual ~1 mmol m s ) Rs under dry conditions, when soil soil respiration in the forest. A contour map of soil respi- water content Ws is near 30%, and works remarkably well ration illustrates that there was considerable spatial varia- under moderate water conditions (residual ~0 mmol m–2 s–1). tion in soil respiration (Figure 4). The mean soil There is a certain level of overestimation using the Q10 function respiration rate for all 20 chambers during this period is under wet conditions, when Ws is near 44%. The temperature –2 –1 3.74 1.04 mmol m s , where the second number sensitivity of Rs, indicated by Q10 values, ranges from 1.61 in denotes standard deviation. The dimensionless coefficient winter to 3.29 in spring (Figure 6 and Table 1a). Among the of variation was estimated to be 27.80%. The number of respiration components, aboveground litter decomposition samples needed to estimate soil respiration within 10% of (RL) was most sensitive (Q10 = 7.22) to temperature variation its mean value, at a 95% confidential level, was calcu- (Figure 7 and Table 1b). On the contrary, root respiration lated to be 30. (RR) showed the least sensitivity to temperature variation.

–2 –1 Figure 7. Relationship between soil respiration rate (mmol m s ) and Ts ( C) for different components: (a) RR, (b) RL, and (c) RSOM. Grey points represent raw hourly data, and black circles with error bars (standard deviation) are mean values of each 10 percentile. The solid line was ﬁtted using the two- (bTs) parameter exponential equation (Rs = a*exp ).

2987 TAN ET AL.: SUBTROPICAL FOREST SOIL RESPIRATION

4. Discussion Dinghushan. This idea, a high soil C stock induced high soil respiration, is supported by the high proportion of decom- 4.1. Annual Total of Soil Respiration in Primary position SOM (65.25%) in total soil respiration in the Subtropical Evergreen Broadleaved Forests Ailaoshan forest. 32 [ ] We compiled annual soil respiration measurements [34] The annual total of Rs estimated in this study from 16 old-growth subtropical forests in China (Table S1), (1248 gC m–2 yr–1)is also higher than that of a previous covering seven subtropical provinces of China. Mean annual study (1055 gC m–2 yr–1) in the same Ailaoshan forest soil respiration of old-growth subtropical forests derived from [Feng et al., 2008]. One possible explanation for this –2 –1 the data is 1279 gC m yr . This value is very close to that difference is that Feng et al. did not account for diel var- –2 –1 of a tropical forest (1260 gC m yr reported by Raich and iation [Rey et al., 2002; Saiz et al., 2006]. Measurements –2 –1 Schlesinger [1992]; 1286 gC m yr calculated from the taken from 09:00 to 11:00 will underestimate soil respiration Bond-Lamberty and Thomson [2010] data set updated in (Figure 1a). In addition, soil respiration measured with an 2012). Given the high carbon dioxide fluxes observed, much automaticsystemwillcoverthefulldieltimeperiod,even more attention should be paid to old-growth subtropical forests, under bad conditions (i.e., rainy days) when manual measure- both in merging global carbon cycle maps and studying forest ments are seldom made. and climate interaction. [35] The other main aspect of uncertainty in estimating an [33] There is a close relationship (Rs =614.91 annual total of soil respiration is spatial variation. Tradition- ((MAT 10)/10) 2 2.32 , r =0.46, p < 0.01) between MAT ally, the mean value of several fixed small soil chambers and soil respiration in a multisite plot for these forests, except (i.e., 10 cm diameter connected with Li-6400) is used to for the Ailaoshan subtropical forest (Figure 8). MAT of represent the respiration of a whole stand (scales in hectare). Ailaoshan is 11.3 C. Soil respiration predicted by the above We applied big square chambers (90 cm 90 cm) to reduce –2 –1 relationship is 686 gC m yr . The observed Rs is spatial uncertainty. Nevertheless, the minimum number of –2 –1 1248 489 gC m yr (mean SD) and is higher than samples needed to obtain results within 10% of its mean value, the expectation and also higher than the modeled value at a 95% confidential level, is still 30. Studies seldom take –2 –1 (589 gC m yr ) based on a global MAT-Rs relationship measurements from 30 chambers to obtain an annual total of [Raich and Schlesinger, 1992]. We hypothesized that well- soil respiration; usually, three to eight measurements are watered conditions would lead to a high respiration rate in taken, with five being the average. The 95% confidence inter- Ailaoshan forest. In fact, subtropical forests are occurred val based on five chambers in the control treatment is nearly all under well-watered condition (Table S1). Thus, 3.30 0.86 mmol m–2 s–1. More attention might be needed well-watered conditions appear to produce high levels of on spatial variation, rather than temporal, to obtain an annual respiration in these old-growth subtropical forests. However, total of soil respiration. well-watered conditions do not explain the exceptionally high soil respiration values in Ailaoshan (Figure 8). A high soil C 4.2. Temperature Sensitivity of Soil Respiration and its stock in Ailaoshan forest is probably the reason. Soil C stock Components –1 of the old-growth subtropical forest in Ailaoshan (286 tC ha ) [36] It is commonly accepted that soil respiration depends is largely higher than many other forests in China; i.e., on temperature [Lloyd and Taylor, 1994]. Subsequently, soil –1 rainforest in Xishuangbanna (95 tC ha ) and temperate forest carbon release is expected to increase with global warming, –1 in Changbaishan (118 tC ha )[Wang and Yang, 2010]. It is despite an acclimatization [Luo et al., 2001] and substrate also higher than the soil C stock in other old-growth subtrop- limit [Knorr et al., 2005]. Nonetheless, we found that tem- –1 ical forests; i.e., 97 tC ha in an old-growth forest of perature sensitivity of soil respiration is largely different in different seasons and under different component values. In 2000 a Danish beech forest, large seasonal changes in Q10 were Q =2.32 (CI: 1.07~3.57) observed [Janssens and Pilegaard, 2003]. Moreover, annual ) 10

-1 2 r =0.4639, p=0.0052 Q10 can be used as an indicator of canopy phenology [Yuste yr

-2 1600 et al., 2004]. The primary subtropical forest in this study is evergreen, yet seasonal variation of Q10 is apparent, ranging from 1.61 in winter to 3.29 in spring (Table 1a). A low soil 1200 water content restrained respiration and could account for the low Q10 in winter (Figure 2). A high Q10 in spring might be the result of leaf ﬂushing (Figure S2) [Zhao et al., 2012]. 800 [37] We observed differences in the temperature sensitiv- > Soil respiration (gC m ity of soil respiration components: RL (Q10 = 7.22) RSOM > 400 (2.73) RR (1.65) (Table 1). The apparent Q10, calculated 10 12 14 16 18 20 22 24 by a temperature-respiration regression, often was not o Mean annual temperature, MAT ( C) consistent with the deﬁnition of Q10 because, not only temperature, but other processes and conditions varied with Figure 8. Relationship between mean annual temperature time; i.e., phenophase, water condition, and photosynthesis and soil respiration for 16 old-growth subtropical forests in [DeForest et al., 2006; Kuzyakov and Gavrichkova, 2010; China (details of the data are shown in Table S2). Open Subke and Bahn, 2010]. Our study forest was evergreen circles represent outlier results and were obtained from the and has a year-round growing season [Tan et al., 2012]. Ailaoshan subtropical forest data. These results were not Compared to that of a temperate or boreal forest, the sea- included in the regression. sonal variation of phenophase and canopy photosynthesis

2988 TAN ET AL.: SUBTROPICAL FOREST SOIL RESPIRATION in a subtropical forest is small. Water condition, however, et al., 2005]. Nevertheless, the relationship between photosyn- could act as a strong confounding factor in deriving soil res- thesis and soil respiration (i.e., time lag) is still being piration temperature sensitivity. Dominated by the Indian quantified [Kuzyakov and Gavrichkova, 2010]. The relation- Monsoon, rainfall in the forest is mainly associated with ship between photosynthesis and soil respiration at diel high temperatures. We tried two methods to minimize the and annual scales was studied with a continuous data set confounding effect of water: well-watered regression and having a high temporal resolution by combining eddy covari- multivariate regression. After leaving out measurements at ance and soil chamber methods. After subtracting the times when moisture is too low or too high to limit respira- temperature-dependent soil respiration, the diel variation in tion, the conclusion that RL (Q10 = 3.93) > RSOM (3.12) > RR the residuals (Figure 1c, shaded area) did not show a pattern (1.61) still held (Table 2). The multivariate model combined similar to that of light intensity, which has been reported for the effects of temperature and moisture and also gave the a temperate deciduous forest in Oak Ridge, USA [Liu et al., same results (Table 2). Meanwhile, we found that water 2006]. In fact, the diel pattern between residuals and that of has little confounding effect on the temperature sensitivity light intensity were similar only when photosynthetic produc- of root respiration, but a strong effect on litter decomposition will immediately arrived soil and used as respiration tion. Temperature sensitivity of litterfall decomposition substrate with no time lag. Nevertheless, it is consistent with declined sharply from 7.22 to a range of 3.72 to 3.75 after observations which suggest a time lag from 7 to 12 h [Tang minimizing the confounding effect of water. et al., 2005] or, for a tree, from 4 to 5 days [Kuzyakov and [38] There is no unified conclusion on the temperature Gavrichkova, 2010]. It makes sense that temperature was not sensitivity of different components of soil respiration. A peaked at the time of the strongest radiation, but that its peak litter-manipulation experiment in Harvard forest suggested occurred usually several hours latter [Campbell and Norman, that temperature sensitivity of root is stronger than that of bulk 1998]. A very similar diel pattern was found in the relationship soil [Boone et al., 1998]. By contrast, no different Q10 values between soil respiration and soil temperature and in the were found between bulk soil and root respiration in a micro- relationship between canopy photosynthesis and light inten- cosm study [Bååth and Wallander, 2003]. A girdling experi- sity. This observation suggests that peaked soil respiration will ment in a Scott pine forest of Sweden suggested that root occur after the highest photosynthesis rate, like the radiation- respiration was less sensitive to that of bulk soil [Bhupinderpal temperature relationship in our studied forest. The time lag et al., 2003]. There was no significant correlation between root at the diel scale was calculated to be from 4 to 5 h. At an respiration and temperature in a secondary forest of Japan annual scale, soil respiration peaked earlier than that of gross [Lee et al., 2003]. The soil respiration component that was primary production (Figure 2). The relationship between least sensitive to temperature in the old-growth Ailaoshan photosynthesis can be fitted by a four-parameter sigmoidal subtropical forest is root respiration (Table 1). Moreover, logistic function (r2 =0.6502, p < 0.0001) (Figure S3). It is low temperature sensitivity was stable and did not change after not easy to obtain time lags at this scale because of the minimizing the confounding effect of moisture (Table 2). complex interaction of biotic and abiotic factors. There are two possible mechanisms for the low-temperature sensitivity of root respiration, compared to litter or soil organic matter decomposition. First, temperature sensitivity of root 5. Conclusions respiration is controlled by C availability and recent photosynthetic inputs to roots. The latter is rather constant over a year in [40] We have made five conclusions: the studied ecosystem (Figure 3b). Second, previous studies [41] 1. Annual soil respiration in Ailaoshan forest was have shown that respiration of plants, and thus roots and near the mean of 16 other old-growth subtropical Chinese mycorrhiza, acclimatizes to temperature, but there is less evi- forests, but high relative to its mean annual temperature. dence for compensatory thermal acclimation in free-living soil This may be due to high C stocks at this site. microbes [Atkin and Tjoelker, 2003; Bradford et al., 2008; [42] 2. Even though big size chambers have been intro- Hartley et al., 2008]. duced to make high spatial representative, the coefficient of variation of soil respiration is still 27.8%, which 4.3. Photosynthesis and Soil Respiration suggested 30 samples were required to obtain results within [39] The idea that photosynthesis drives or modulates soil 10% of the mean value at a 95% confidential level. More respiration has been a commonly accepted view in theoretical attention to spatial variation, rather than temporal, might be inferences and field campaigns [Hogberg et al., 2001; Tang needed to obtain an annual total of soil respiration. [43] 3. Photosynthesis-respiration coupling was detected at diel scale with time lag of 4 to 5 h. The seasonality of Table 2. Temperature Sensitivity (Q10) of Respiration Compo- soil respiration was strong and varied from a low value of –2 –1 nents RR, RL, and RSOM ~1 mmol m s in late winter to a peak value of ~6 mmol –2 –1 Method R R R m s in midsummer. A time lag between gross primary R L SOM production and soil respiration at the seasonal scale was not Well water regression 1.61(1.27–1.97) 3.93(2.80–5.07) 3.12(2.88–3.37) apparent. Multivariate regression 1.62(0.25) 3.75(0.29) 2.20(0.44) [44] 4. Soil temperature at 5 cm can explain more than We applied two methods to discard the confounding effect of water on 91% of the observed annual variation in soil respiration with soil temperature sensitivity: well-watered regression and multivariate re- a Q10 of 3.05. Soil respiration during winter was strongly af- gression. Details on these methods are given in the text. A value in parenthe- fected by low soil water content, and the lowest Q value ses for the well-watered regression method denotes a Confidence Interval 10 (CI), While one for the multivariate regression method denotes a Standard was detected in that time. Soil respiration was overestimated Deviation (SD). by the Q10 function during dry and wet periods; residuals

2989 TAN ET AL.: SUBTROPICAL FOREST SOIL RESPIRATION between observed and temperature-dependent soil respira- Knorr, W., I. C. Prentice, J. I. House, and E. A. Holland (2005), Long- term sensitivity of soil carbon turnover to warming, Nature, 433(7023), tion illustrate this feature. 298–301. [45] 5. Temperature sensitivity differed among several com- Kuzyakov, Y., and O. Gavrichkova (2010), REVIEW: Time lag between ponents of soil respiration. The highest sensitivity occurred in photosynthesis and carbon dioxide efflux from soil: a review of mechanisms – aboveground litter decomposition (R )(Q = 7.22), followed and controls, Glob. Chang. Biol., 16(12), 3386 3406. L 10 Lal, R. (2008), Sequestration of atmospheric CO2 in global carbon pools, by decomposition of SOM (RSOM)(Q10 = 2.73), and then root Energy & Environmental Science, 1(1), 86–100. respiration (RR)(Q10 = 1.65). The conclusion that Q10 for litter Lee, M.-s., K. Nakane, T. Nakatsubo, and H. Koizumi (2003), Seasonal respiration exceeds Q of soil organic matter decomposition, changes in the contribution of root respiration to total soil respiration in 10 a cool-temperate deciduous forest, Plant and Soil, 255(1), 311–318. which exceeds Q10 of root respiration holds even after mini- Liang, N., G. Inoue, and Y. Fujinuma (2003), A multichannel automated mizing the confounding effect of moisture. A rather constant chamber system for continuous measurement of forest soil CO2 efflux, substrate supply and thermal acclimation are possible mecha- Tree Physiol., 23(12), 825–832. Liu, Q., N. T. Edwards, W. M. Post, L. Gu, J. Ledford, and S. Lenhart nisms that can account for the low temperature sensitivity in (2006), Temperature-independent diel variation in soil respiration root respiration. observed from a temperate deciduous forest, Glob. Chang. Biol., 12(11), 2136–2145. [46] Acknowledgments. Funding for this study was provided by Liu, W., J. E. D. Fox, and Z. Xu (2003), Nutrient budget of a montane National Science Foundation of China (31200347, 41271056, 40571163, evergreen broad-leaved forest at Ailao Mountain National Nature Reserve, 41071071), Natural Science Foundation of Yunnan Province, China Yunnan, southwest China, Hydrological Processes, 17(6), 1119–1134. (2011FB110, 2011FA025), Development Program in Basic Science of Lloyd, J., and J. A. Taylor (1994), On the Temperature Dependence of Soil China (2010CB833501), and Knowledge Innovation Program of the Respiration, Functional Ecology, 8(3), 315–323. Chinese Academy of Sciences (KJCX2-YW-432-1, KZCX2-YW-Q1-05-04, Luo, Y., S. Wan, D. Hui, and L. L. Wallace (2001), Acclimatization of soil KZCX1-SW-01-01A). Thanks to reviewers for their excellent comments on respiration to warming in a tall grass prairie, Nature, 413(6856), 622–625. this work. Magnani, F., et al. (2007), The human footprint in the carbon cycle of temperate and boreal forests, Nature, 447(7146), 849–851. Mahecha, M. D., et al. (2010), Global Convergence in the Temperature References Sensitivity of Respiration at Ecosystem Level, Science, 329(5993), 838–840. Atkin, O. K., and M. G. Tjoelker (2003), Thermal acclimation and the dynamic Pan, Y., et al. (2011), A Large and Persistent Carbon Sink in the World’s response of plant respiration to temperature, Trends Plant Sci., 8(7), 343–351. Forests, Science, 333(6045), 988–993. Bååth, E., and H. Wallander (2003), Soil and rhizosphere microorganisms Raich, J. W., and W. H. Schlesinger (1992), The global carbon dioxide flux have the same Q10 for respiration in a model system, Glob. Chang. Biol., in soil respiration and its relationship to vegetation and climate, Tellus B, 9(12), 1788–1791. 44(2), 81–99. Bhupinderpal, S., A. Nordgren, M. Ottosson Löfvenius, M. N. Högberg, P. E. Rey, A., E. Pegoraro, V. Tedeschi, I. De Parri, P. G. Jarvis, and R. Valentini Mellander, and P. Högberg (2003), Tree root and soil heterotrophic (2002), Annual variation in soil respiration and its components in a respiration as revealed by girdling of boreal Scots pine forest: extending coppice oak forest in Central Italy, Glob. Chang. Biol., 8(9), 851–866. observations beyond the first year, Plant Cell Environ., 26(8), 1287–1296. Saiz, G., K. A. Byrne, K. Butterbach-Bahl, R. Kiese, V. Blujdea, and Bond-Lamberty, B., and A. Thomson (2010), Temperature-associated E. P. Farrell (2006), Stand age-related effects on soil respiration in increases in the global soil respiration record, Nature, 464(7288), 579–582. a first rotation Sitka spruce chronosequence in central Ireland, Glob. Boone, R. D., K. J. Nadelhoffer, J. D. Canary, and J. P. Kaye (1998), Roots Chang. Biol., 12(6), 1007–1020. exert a strong influence on the temperature sensitivityof soil respiration, Subke, J.-A., and M. Bahn (2010), On the ‘temperature sensitivity’ of soil Nature, 396(6711), 570–572. respiration: Can we use the immeasurable to predict the unknown?, Soil Bradford, M. A., C. A. Davies, S. D. Frey, T. R. Maddox, J. M. Melillo, Biology and Biochemistry, 42(9), 1653–1656. J. E. Mohan, J. F. Reynolds, K. K. Treseder, and M. D. Wallenstein Tan, Z.-H., et al. (2012), An observational study of the carbon-sink (2008), Thermal adaptation of soil microbial respiration to elevated strength of East Asian subtropical evergreen forests, Environ. Res. temperature, Ecol. Lett., 11(12), 1316–1327. Lett., 7(4), 044017. Campbell, G. S., and J. M. Norman (1998), An introduction to environmental Tan, Z., Y. Zhang, D. Schaefer, G. Yu, N. Liang, and Q. Song (2011), An biophysics, Springer-Verlag, New York. old-growth subtropical Asian evergreen forest as a large carbon sink, Cox, P. M., R. A. Betts, C. D. Jones, S. A. Spall, and I. J. Totterdell (2000), Atmos. Environ., 45(8), 1548–1554. Acceleration of global warming due to carbon-cycle feedbacks in a Tang, J., D. D. Baldocchi, and L. Xu (2005), Tree photosynthesis coupled climate model, Nature, 408, 184–187. modulates soil respiration on a diurnal time scale, Glob. Chang. Biol., Davidson, E. A., and I. A. Janssens (2006), Temperature sensitivity of 11(8), 1298–1304. soil carbon decomposition and feedbacks to climate change, Nature, Vitousek, P. M., H. A. Mooney, J. Lubchenco, and J. M. Melillo (1997), 440(7081), 165–173. Human Domination of Earth’s Ecosystems, Science, 277(5325), DeForest, J., A. Noormets, S. McNulty, G. Sun, G. Tenney, and J. Chen 494–499. (2006), Phenophases alter the soil respiration–temperature relationship Wang, B., and X.-S. Yang (2010), Comparison of carbon content and in an oak-dominated forest, Int. J. Biometeorol., 51(2), 135–144. carbon density of four typical zonal forest ecosystems, Journal of Hunan Denman, K. L., et al. (2007), Couplings between changes in the climate Agricultural University (Natural Sciences), 36, 464–469, doi:410.3724/ system and biogeochemistry, Cambridge University Press, Cambridge, SP.J.1238.2010.00464. New York, USA. WMO (2012), Greenhouse Gas Concentrations Reach New Record - WMO Feng, W., X. Zou, L. Sha, J. Chen, Z. Feng, and J. Li (2008), Comparisions Bulletin highlights pivotal role of carbon sinks, edited by N. 965, http:// between seasonal and diurnal patterns of soil respiration in a montane www.wmo.int/pages/mediacentre/press_releases/index_en.html. evergreen broad-leaved forest of Ailao Mountains, China, Journal of Woodwell, G. M., R. H. Whittaker, W. A. Reiners, G. E. Likens, C. C. Plant Ecology (Chinese Version), 32,31–39. Delwiche, and D. B. Botkin (1978), The biota and the world carbon budget, Hartley, I. P., D. W. Hopkins, M. H. Garnett, M. Sommerkorn, and P. A. Wookey Science, 199,141–146. (2008), Soil microbial respiration in arctic soil does not acclimate to Yuste, J. C., I. A. Janssens, A. Carrara, and R. Ceulemans (2004), Annual temperature, Ecol. Lett., 11(10), 1092–1100. Q10 of soil respiration reflects plant phenological patterns as well as Hogberg, P., A. Nordgren, N. Buchmann, A. F. S. Taylor, A. Ekblad, M. N. temperature sensitivity, Glob. Chang. Biol., 10(2), 161–169. Hogberg, G. Nyberg, M. Ottosson-Lofvenius, and D. J. Read (2001), Zhao, J., Y. Zhang, Z. Tan, Q. Song, N. Liang, L. Yu, and J. Zhao (2012), Large-scale forest girdling shows that current photosynthesis drives soil Using digital cameras for comparative phenological monitoring in an respiration, Nature, 411(6839), 789–792. evergreen broad-leaved forest and a seasonal rain forest, Ecological Janssens, I. A., and K. I. M. Pilegaard (2003), Large seasonal changes in Q10 Informatics, 10(0), 65–72. of soil respiration in a beech forest, Glob. Chang. Biol., 9(6), 911–918. Zhou, G., S. Liu, Z. Li, D. Zhang, X. Tang, C. Zhou, J. Yan, and J. Mo Kira, T. (1991), Forest ecosystems of east and southeast Asia in a global (2006), Old-Growth Forests Can Accumulate Carbon in Soils, Science, perspective, Ecological Research, 6(2), 185–200. 314(5804), 1417.

2990