A compiled soil respiration dataset at different time scales for forest ecosystems across China from 2000 to 2018

. China’s forests rank ﬁfth in the world by area, covering a broad climatic gradient from cold-temperate to tropical zones, and play a key role in the global carbon cycle. Studies of forest soil respiration ( R s ) have increased rapidly in China over the last two decades, but the resulting R s data need to be summarized. Here, we compile a comprehensive dataset of R s in China’s undisturbed forest ecosystems from the literature published up to 31 December 2018, including monthly R s and the concurrently measured soil temperature ( N = 8317), mean monthly R s ( N = 5003), and annual R s ( N = 634). Detailed plot information was also recorded, such as geographical location, climate factors, stand characteristics, and measurement description. We examine some aspects of the dataset – R s equations ﬁtted with soil temperature, temperature sensitivity ( Q 10 ), monthly variations, and annual efﬂuxes in cold-temperate, temperate, subtropical, and tropical zones. We hope the dataset will be used by the science community to provide a better understanding of the carbon cycle in China’s forest ecosystems and reduce uncertainty in evaluating of carbon budget at a large scale. The dataset is publicly available at https://doi.org/10.1594/PANGAEA.943617 (Sun et al., 2022).


Introduction
Soil respiration (R s ) refers to the total amount of CO 2 released by undisturbed soil, including autotrophic respiration and heterotrophic respiration, the former from plant roots and their microbial symbionts, and the latter from microorganisms decomposing litter and soil organic matter. As the second-largest terrestrial carbon flux, the recent estimations of global annual R s (80-98 Pg C yr −1 ) are over 10 % of the atmospheric carbon pool (750 Pg C) (Bond-Lamberty and Thomson, 2010b;Hashimoto et al., 2015;Raich et al., 2002;Warner et al., 2019); thus, accelerating soil respiration rates with climate warming have a strong potential to influence atmospheric CO 2 levels. It is therefore important to understand better soil respiration dynamics and their response to climate changes.
Forest area in China ranks fifth in the world (FAO, 2020) and covers a broad climatic gradient, including coldtemperate, temperate, subtropical, and tropical zones. In China, most R s measurements began only after 2001 , but have rapidly increased during the last 20 years . Several studies have summarized annual R s in China's forest ecosystems, but with small samples (e.g., N = 50 in Zheng et al., 2010; N = 62 in Chen et al., 2008; N = 120 in Zhan et al., 2012;N = 139 in Song et al., 2014). Yu et al. (2010) established a geostatistical model with a total of 390 monthly R s data from different ecosystems in China. With 1782 monthly R s in forest ecosystems across China, Jian et al. (2020) analyzed the spatial patterns and temporal trends from 1961 to 2014. However, numerous R s data are still unexploited, because they were only displayed in the form of monthly dynamics in the figures of the original papers. R s data at a subannual timescales are impor-H. Sun et al.: A compiled soil respiration dataset for forest ecosystems across China tant for upscaling global R s , which may yield different conclusions and deserve further exploration .
The lack of large-scale and observation-driven R s data is a main constraining factor in quantifying regional-to globalscale carbon budgets (Bond-Lamberty and Thomson, 2010a;Rayner et al., 2005). R s data and concurrently measured temperature thus provide not only a solid base for understanding the critical factors influencing R s , but also the opportunity for better simulation of R s at a large scale. We attempted to compile a complete forest R s dataset at different temporal scales in China, and to analyze temperature sensitivity (Q 10 ) as well as monthly and annual R s in cold-temperate, temperate, subtropical, and tropical zones.

Data sources
The terms "soil respiration", "soil carbon (or CO 2 ) efflux", or "soil carbon (or CO 2 ) emission" were searched in publications before 2018 in the China Knowledge Resource Integrated Database (http://www.cnki.net/, last access: 27 March 2021), China Science and Technology Journal Database (http://www.cqvip.com, last access: 27 March 2021), ScienceDirect (http://www.sciencedirect. com/, last access: 8 June 2020), ISI Web of Science (http: //isiknowledge.com/, last access: 8 June 2020), and Springer Link (http://link.springer.com/, last access: 8 June 2020). Means, minimums, and maximums of soil respiration during the observation periods were usually given in these published studies, and monthly patterns of soil respiration rates with the corresponding temperature were frequently shown in figures. WEBPLOTDIGITIZER, a graphic digitizing software, was used to acquire data from figures when values were not reported in the text (Burda et al., 2017).

Data collection criteria
The following criteria were used to ensure data consistency and accuracy: (i) R s was measured in the field without obvious disturbances or manipulation experiments, e.g., fire, cutting, nitrogen addition treatments, etc. (ii) Forested swamps and commercial plantations (e.g., orchard, rubber, etc.) were not examined. (iii) R s was measured either by static chamber and gas chromatography (GC) or by dynamic chamber and infrared gas analyzers (IRGA, model Li-6400, Li-8100, Li-8150 (LI-COR Inc., Lincoln, Nebraska, USA)), which are the most popular methods and provide methodological consistency Wang et al., 2011;Yang et al., 2018;Zheng et al., 2010).
Based on these criteria, a total of 10 288 monthly soil respiration data and 634 annual soil respiration data were assembled from 568 publications. Moreover, the related information was recorded, including geographical location (province, study site, latitude, longitude and elevation), climate (mean annual temperature and mean annual precipitation), stand description (forest type, origin, age, density, mean tree height and diameter at breast height), and measurement regime (method, time, frequency, collar area, height and numbers) ( Table 1). There were 155 study sites from 28 provinces in .08 • E, 7-4200 m) (Fig. 1). This forest region encompasses a large gradient of climate regimes, with mean annual temperature ranging from −5.4 to 23.8 • C and mean annual precipitation ranging from 105 to 3000 mm. The observation years were from 2000 to 2018.

Data verification
Soil temperature as a main influencing factor was usually concurrently measured with R s . Monthly dynamics of R s and soil temperature at 5 cm depth (T 5 ) and/or 10 cm depth (T 10 ) were shown with figures in many studies in the literature. In this study, most of the R s data (∼ 82 %) and the concurrent T 5 and/or T 10 were extracted with WEBPLOTDIGITIZER, while others (e.g., minimum, maximum) were given directly in the original papers. To verify the accuracy of the digital software, the means (R s , T 5 , T 10 ) averaged from the extracted data were compared with the corresponding means given in the original papers ( Fig. S1 in the Supplement). The root mean square error (RMSE) of R s , T 5 , and T 10 was 0.09 µmol m −2 s −1 , 0.35 • C, and 0.44 • C, respectively, and the coefficients of determination (R 2 ) were all greater than 0.99, indicating that the accuracy of WEBPLOTDIGITIZER is excellent. Moreover, the data from the same authors and different sources (e.g., master's thesis or PhD dissertation and journal article) were carefully cross-checked and supplemented.

Statistical analysis
Monthly and annual R s were averaged arithmetically in cold-temperate, temperate, subtropical, and tropical zones. Independent-samples t tests (two groups) and one-way ANOVA (≥ 3 groups) at the P = 0.05 significance level were used to test the differences among various forest types in the same climate zone and among the same forest type in different climate zones. Temperature sensitivity (Q 10 ) is defined as the factor by which R s is multiplied when temperature increases by 10 • C (Davidson and Janssens, 2006;Lloyd and Taylor, 1994), which is usually calculated with the van't Hoff equation (R s = ae β T and Q 10 = e 10 β ), where R s is soil respiration rate (µmol m −2 s −1 ) and T is temperature ( • C). All statistical analyses were performed with SPSS Statistics 21 (SPSS Inc., Chicago, IL, USA).

Relationship between soil respiration rate and soil temperature
Temperature is often the main factor determining soil respiration rates. The samples of the paired R s & T 5 and R s & T 10 were 6341 (69 %) and 2878 (31 %) in the dataset, respectively. There were statistically significant exponential relationships of R s with T 5 and T 10 in forest ecosystems across China, which could explain about 48 % and 52 % of the R s variations, respectively (Fig. S2). The exponential correlations were all significant in four climatic zones (R 2 = 0.23-0.93) (Fig. 2). RMSE values in cold-temperate and temperate zones (1.52-1.67 µmol m −2 s −1 ) were greater than those in subtropical and tropical zones (1.04-1.32 µmol m −2 s −1 ), except for the smallest RMSE from T 10 in the cold-temperate zone (0.42 µmol m −2 s −1 ).
Annual mean R s during January-December from low to high was cold-temperate (1.63 µmol m −2 s −1 ), temperate (1.93 µmol m −2 s −1 ), subtropical (2.47 µmol m −2 s −1 ), and tropical zones (2.57 µmol m −2 s −1 ). Annual soil carbon emissions were calculated with the annual mean R s : 621.91 g C m −2 yr −1 in the cold-temperate zone, 733.31 g C m −2 yr −1 in the temperate zone, 937.15 g C m −2 yr −1 in the subtropical zone, and 973.35 g C m −2 yr −1 in the tropical zone. Soil carbon emissions in the growing season (May-October) and in winter (November-April) accounted for 85 % and 15 % of emissions in the cold-temperate zone, 80 % and 20 % in the temperate zone, 69 % and 31 % in the subtropical zone, and 61 % and 39 % in the tropical zone. Subtropical and Figure 2. Exponential relationships of forest soil respiration rates with soil temperature at 5 and 10 cm depth in cold-temperate (a), temperate (b), subtropical (c), and tropical zones (d). * * P < 0.01. tropical zones maintained high soil respiration rates during November-April, which is the main source of their higher annual soil carbon emissions.

Annual soil carbon effluxes
There were 634 annual soil carbon effluxes, and most of the observations were conducted in the subtropical zone (61 %) and temperate zone (32 %) (Fig. 4) The annual soil carbon effluxes ranged from 260.10 to 2058.00 g C m −2 yr −1 in China's forest ecosystems, and the mean was 851.88 ± 12.75 g C m −2 yr −1 . The annual soil carbon effluxes increased with the increase in mean annual temperature and precipitation at the national scale (Fig. S3). The mean annual soil carbon emissions in the tropical, subtropical, temperate, and cold-temperate zones were 1042.01 ± 68.55, 928.91 ± 16.68, 697.85 ± 16.39, and 684.29±61.81 g C m −2 yr −1 , respectively. The first two were significantly higher than the last two, but the differences were not significant between the tropical and subtropical zones, and between the temperate and cold-temperate zones. The differences were not significant for evergreen broadleaf forest (EBF), evergreen needleleaf forest (ENF), and deciduous needleleaf forest (DNF) among different climate zones. The values for deciduous broadleaf forest (DBF) in the temperate (748.59 ± 25.18 g C m −2 yr −1 ) and subtropical zones (755.41 ± 58.26 g C m −2 yr −1 ) were similar, both of which were greater than that in the cold-temperate zone (284.20 ± 21.36 g C m −2 yr −1 ). Broadleaf and needleleaf mixed forest in the subtropical zone (977.35 ± 43.56 g C m −2 yr −1 ) had significantly higher emissions than in the temperate zone (733.44 ± 45.29 g C m −2 yr −1 ).

Discussion
4.1 Temperature sensitivity (Q 10 ) of soil respiration Q 10 is a key parameter in modeling the effects of climate warming on soil carbon release. The Q 10 values calculated with the exponential equations of T 5 and T 10 were 2.05 and 2.17, respectively, at the national scale (Fig. S2), which were lower than the averaged Q 10 from different studies in the syntheses of China's forest ecosystems (T 5 : 2.28-2.51 and T 10 : 2.74-3.00, Peng et al., 2009;Song et al., 2014;Xu et al., 2015;Zheng et al., 2009) and global forest ecosystems (T 5 : 2.55-2.70 and T 10 : 3.01-3.31, W. . Our results were close to the Q 10 of 2, which is commonly used in many biogeochemical models (e.g., Cox et al., 2000;Sampson et al., 2007), and the mean Q 10 of 2.11 estimated with inverse modeling in forest soils across China (Zhou et al., 2009).
Temperature was the most important limiting factor for soil microbial activity and root growth in cold regions; thus, R s was more sensitive to temperature changes (Lloyd and Taylor, 1994;Peng et al., 2009;Zheng et al., 2009Zheng et al., , 2020. The Q 10 increased from the tropical zone to the coldtemperate zone in this study, and varied from 1.63 to 3.74. Soil temperature at the depth of 5 and 10 cm could only ex-plain 29 % and 23 % of the R s variations and RMSEs were 1.09 and 1.13 µmol m −2 s −1 in the tropical zone, respectively (Fig. 2d). The difference of the mean R s between tropical moist forests (1260 g C m −2 yr −1 ) and tropical dry forests (673 g C m −2 yr −1 ) was about 2-fold (Raich and Schlesinger, 1992), indicating that soil moisture might play more important roles.

Comparisons of monthly and annual soil carbon effluxes
The lowest monthly R s occurred in January, and the highest values occurred in August in the cold-temperate and temperate zones and in July in the subtropical and tropical zones (Fig. 3). Similarly, monthly R s values of global terrestrial ecosystems reached their minima in February and peaked in July and August (Hashimoto et al., 2015;Raich et al., 2002). Due to the limitation of low temperature, winter observations of R s were relatively fewer in the cold-temperate and temperate zones. The R s in winter (November-April) was usually assumed to account for 20 % of the total annual R s (Geng et al., 2017;Yang and Wang, 2005), which was in agreement with the proportion in the temperate zone, but greater than 15 % in the cold-temperate zone.
Annual soil carbon emissions were synthesized for forest ecosystems across China, and the mean was 745.34 g C m −2 yr −1 , 764.11 g C m −2 yr −1 (Zhan et al., 2012), 917.73 g C m −2 yr −1 (Song et al., 2014) and 975.50 g C m −2 yr −1 (Chen et al., 2008); the mean of 851.88 g C m −2 yr −1 in the present study was in the mid-range. The mean annual R s in China's forest ecosystems was slightly lower than the mean R s of 990.00 g C m −2 yr −1 in global forest ecosystems . Warner et al. (2019) modeled global R s and found that the lowest and highest annual soil carbon emissions were in deciduous needleleaf forests (mean = 344.10 g C m −2 yr −1 ) and evergreen broadleaf forests (mean = 1310.47 g C m −2 yr −1 ), respectively. Compared with the predicted annual R s , deciduous needleleaf forests in the cold-temperate (mean = 734.56 g C m −2 yr −1 ) and temperate zones (mean = 555.15 g C m −2 yr −1 ) had higher values, but those of evergreen broadleaf forests in subtropical (mean = 1073.50 g C m −2 yr −1 ) and tropical zones (mean = 1065.09 g C m −2 yr −1 ) were lower (Fig. 4).
The mean annual soil carbon emissions from 634 annual R s and 5003 mean monthly R s were 684.29 and 621.91 g C m −2 yr −1 , respectively, in the cold-temperate zone, 697.85 and 733.31 g C m −2 yr −1 in the temperate zone, 928.91 and 937.15 g C m −2 yr −1 in the subtropical zone, and 1042.01 and 973.35 g C m −2 yr −1 in the tropical zone ( Figs. 4 and 3). The differences between the directly averaged annual R s and the accumulative mean monthly R s were smallest in the tropical zone (−8.24 g C m −2 yr −1 ), secondary in the temperate zone (−35.46 g C m −2 yr −1 ), and largest in the cold-temperate and tropical zones (62.38-68.66 g C m −2 yr −1 ). From Fig. 4 we can also see that the standard errors in the tropical and temperate zones (∼ 16 g C m −2 yr −1 ) were smaller than those in the coldtemperate and tropical zones (∼ 65 g C m −2 yr −1 ). Mean annual soil carbon emissions in the temperate, subtropical, and tropical ecosystems were 745, 776, and 1286 g C m −2 yr −1 at the global scale, respectively (Bond-Lamberty and Thomson, 2010a), which were comparable to our results.

Improvements of the dataset
R s measurements were mainly from Li-8100 (47 %) and Li-6400 (33 %), secondary from gas chromatography (18 %), and Li-8150 only accounted for 2 %. The differences of the four common measurement methods were proved to be small (∼ 10 %) Yang et al., 2018;Zheng et al., 2010). The sample sizes of annual R s were 50-139 (Chen et al., 2008;Song et al., 2014;Zhan et al., 2012;Zheng et al., 2010) and 634 in the current study, and increased more than 4-fold. The global soil respiration database (SRDB-V5) collected 523 undisturbed annual R s in China's forest ecosystems (Jian et al., 2021), but all methods were included, e.g., alkali absorption, gas chromatography, and various infrared gas analyzers. The alkali absorption method could underestimate R s (Chen et al., 2008;Jian et al., 2020). The total samples of mean monthly R s were 5003, which was much larger than the other dataset monthly samples of 1782 in China's forest ecosystems Steele and Jian, 2018). Additionally, we extended the dataset with the digital software (WEBPLOTDIGITIZER) from the monthly dynamics figures of the original papers, including the paired R s & T 5 (N = 6341) and R s & T 10 (N = 2878). Predicting soil respiration from soil temperature has gained extensive acceptance (Shi et al., 2014;Song et al., 2014;. These data could be used to establish a largescale soil respiration equation and acquire the key parameters of the carbon cycle. Compared with the aforementioned monthly or annual databases, this study collected all available R s data at different time scales. Figure S4 showed the length of the individual time series from the different sites; the high frequencies were 12 months (38 %), 6-7 months (20 %), and 13-24 months (15 %). Bamboo forests were seldom considered in the previous databases (Chen et al., 2008;Steele and Jian, 2018;Zhan et al., 2012;, which exhibited the highest soil carbon emissions (Mean = 1133.55 g C m −2 yr −1 , Fig. 4). With the area increasing at a high rate of 3.1 % per year (Song et al., 2017), bamboo forests would play an important role in the regional and even national carbon cycle. It is worth noting that the R s studies were fewer in the regions of latitude greater than 48 • (∼ 2 %) or elevation higher than 3000 m (∼ 4 %). The potentially under-represented forest types might affect the evaluation of the temperature sensitivity of soil respiration and of annual soil carbon emissions at the regional and national scales.

Data availability
The soil respiration dataset in China's forest ecosystems used to produce the results in this study is free to the public for scientific purposes and can be downloaded at https://doi.org/10.1594/PANGAEA.943617 (Sun et al., 2022).

Conclusions
In this study, we reviewed the R s -related literature and collected in situ R s measurements with common infrared gas analyzers (i.e., Li-6400, Li-8100, Li-8150) or gas chromatography to assemble a comprehensive and uniform dataset of China's forest ecosystems at different time scales. Besides the R s data reported directly in the original papers, the monthly patterns of R s and the concurrently measured soil temperature at 5 and/or 10 cm depth in the figures were digitized. Moreover, we made a preliminary analysis of the data. The results showed that soil temperature could explain 22.5 %-93.4 % of the R s variations. Tem-perature sensitivity (Q 10 ) was about 2.05-2.17 at the national scale, increasing from 1.63 in the tropical zone to 3.74 in the cold-temperate zone. Monthly R s showed a single-peak curve, and the highest values occurred in August (4.18-4.36 µmol m −2 s −1 ) in the cold-temperate and temperate zones, higher than the highest values in July (3.58-3.83 µmol m −2 s −1 ) in the subtropical and tropical zones. Mean annual soil carbon emissions decreased from tropical (1042.01 g C m −2 yr −1 ), subtropical (928.91 g C m −2 yr −1 ), temperate (697.85 g C m −2 yr −1 ) to cold-temperate zones (684.29 g C m −2 yr −1 ). This study provides basic data and a scientific basis for the quantitative evaluation of soil carbon emissions from forest ecosystems in China.
Author contributions. BJ designed the soil respiration dataset and searched the papers published until 2018. HS and BJ collected and digitized soil respiration data and compiled the associated information. HS and BJ prepared the manuscript. ZX provided many useful suggestions and reviewed the paper.
Competing interests. The contact author has declared that none of the authors has any competing interests.
Disclaimer. Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.