Introduction

Soils contain approximately 1,500 Pg organic carbon (C) in the global upper 100 cm, which is about three times the amount stored in terrestrial vegetation (550 Pg) and twice that stored in the atmosphere (750 Pg)1, 2. The soil organic carbon (SOC) pool plays important roles in the cycling and balance of global C3. The global storage of SOC in cropland is about 128–165 Pg C4, which is approximately 8% to 10% of the terrestrial SOC pool5, 6. In addition to being an important part of global SOC storage, SOC in cropland is the most active SOC pool among terrestrial ecosystems7.

Global warming has already had observable effects on the environment8. Although warming is expected to accelerate SOC decomposition, the responses of SOC to warming still exhibit many uncertainties3. One of the key factors leading to these uncertainties is the temperature sensitivity (Q 10) of SOC decomposition3. SOC storage in cropland is not only influenced by climate change, but is also regulated by human activities over a short period of time. Therefore, understanding the Q 10 of SOC decomposition in cropland is important for understanding global C cycling. Unlike natural soils, cropland soils have relatively low organic C concentrations, and thus a higher C sequestration potential4. Sustainable land use and management practices in agroecosystems could have the potential to sequester approximately 55–78 Pg SOC9.

Many studies have focused on soil respiration in croplands (e.g. Lohila et al.10, Fiener et al.11, and Campos12). Field-measured soil respiration is the CO2 flux emitted from the soil surface, which commonly includes both autotrophic and heterotrophic respiration13, with autotrophic respiration occupying 30% to 70% of the total14, 15. Therefore, field measurements can not clearly distinguish between SOC decomposition from topsoil (TL) as opposed to subsoil layers (SL). In laboratory soil incubation experiments, live roots are commonly removed from soil samples16, 17. The measured CO2 flux in this case is attributed to heterotrophic respiration, i.e., SOC decomposition18, reflecting the SOC decomposition potential under the specified conditions. For example, Arevalo et al.19 reported that the amount of mineralized C from root-free soil over a 370-d incubation period ranged between 2% and 9% of the initial total SOC in four land use system soils. However, the cropland soil samples used in most studies have been collected from the TL (e.g. Iqbal et al.16, Thiessen et al.17, and Guntinas et al.20). Little is known about organic C decomposition in the SL in cropland21.

Because of differences in the intensity of human disturbances, cropland soils have distinct soil layers: TL and SL. The TL is often disturbed by ploughing and other agricultural activities22 and is characterized by a porous soil structure with high permeability23, rapid changes in moisture and temperature24 and abundant nutrients and organic materials from crop root turnover and exudation25. TL compaction by mechanical farm operations causes the subsoil to have a dense and hard pan (commonly referred to as plough sole) of high bulk density26, restricting the exchange of gas, water, nutrients and other substances between soil layers27, 28. Moreover, variations in temperature and water conditions are smaller in the SL than in the TL29. Thus, SOC pools, decomposition, and dynamics may differ between TL and SL30, 31.

SOC stored in the SL as well as in the TL plays an important role in SOC storage within the entire soil profile32, 33. For example, Xie et al.34 reported that SOC in the surface and subsurface layers constituted about 30% and 70%, respectively, of the total SOC pool up to 1 m depth in Chinese croplands. Jenkinson and Coleman35 found that treating the entire soil profile (1 m depth) as a homogenous unit, instead of dividing it into different soil horizons, overestimates the impact of global warming on C decomposition. It is therefore essential to consider the differences in SOC decomposition in different soil layers in predictive models.

In this study, we aimed to estimate differences in SOC decomposition and its temperature sensitivity between topsoil and subsoil layers in paddy-upland rotation and upland farming systems in China. We hypothesized that SOC quality in the TL would be higher than in the SL. Because low-quality soil C is more sensitive to temperature change than high-quality soil C3, we also expected that SOC in the SL would be more sensitive to temperature increases than SOC in the TL in both paddy-upland rotation and upland farming systems.

Results

Soil properties and soil respiration

SOC content was significantly higher in the TL than in the SL across sites (Table 1; P < 0.05). On average, the SOC in the TL was 14.2 g kg−1 for paddy field and 13.0 g kg−1 for upland soil, while the values in the SL were 12.1 g kg−1 and 9.9 g kg−1 for paddy field and upland soil, respectively. Furthermore, permanganate oxidizable C (POXC) as an index of soil labile C was significantly higher in the TL than in the SL for both paddy field and upland soil (Table 1; P < 0.05), suggesting that SOC had a higher quality in the TL than in the SL. The POXC in the TL was 3.2 g kg−1 for paddy field soil and 1.5 g kg−1 for upland soil, which was about 154% and 122% higher than in the paddy field and upland soil in the SL, respectively. Soils in the Meishan (MS) area had significantly higher N content than those in the Huaping (HP) area (Table 1), possibly because of different agricultural practices in the two areas. The soil physicochemical properties of each soil sample are listed in the Supplementary Tables.

Table 1 Soil physicochemical properties of the topsoil layer (TL) and subsoil layer (SL).
Full size table

Soil respiration and its temperature sensitivity (Q 10)

The relationship between soil respiration and temperature fitted well to an exponential model (Eq. (2)), with R 2 varying from 0.95 to 0.99 for all the samples (Fig. 1). For paddy field soil, the SOC decomposition rate at 20 °C (R 20) in the TL (0.0492~0.2431 μg CO2-C g−1 soil h−1) was about 84% higher than in the SL (0.0259~0.1415 μg CO2-C g−1 soil h−1); for upland soil, the value in the TL (0.0759~0.1873 μg CO2-C g−1 soil h−1) was about 149% higher than in the SL (0.0360~0.0833 μg CO2-C g−1 soil h−1) (Fig. 2). Two-factor analysis of variance showed that the SOC decomposition rate at 20 °C in the TL was significantly higher than in the SL (P < 0.01).

Figure 1
figure 1

Variation of respiration rate with incubation temperature. Error bars indicate standard deviation (n = 4). TL, SL, and MS represent topsoil layer, subsoil layer, and Meishan, respectively.

Full size image
Figure 2
figure 2

SOC decomposition rate (at 20 °C) in topsoil layer (TL, black bars) and subsoil layer (SL, gray bars). Error bars indicate standard deviation (n = 4). Means are the average decomposition rates for six sites. HP and MS represent Huaping and Meishan, respectively.

Full size image

For each sampling site, Q 10 values for SOC decomposition in the TL were significantly lower than those in the SL regardless of different land use types (Fig. 3). On average, Q 10 values in the TL were 55 ± 20% and 40 ± 13% lower than those in the SL for paddy field and upland soil, respectively. In the paddy field, the mean Q 10 values in the TL were 4.38 ± 0.31 and 3.52 ± 0.36 for HP and MS, respectively, and those in the SL were 6.85 ± 0.36 and 5.31 ± 0.64, respectively. In the upland soils, the average Q 10 values in the TL were 6.37 ± 0.64 and 6.88 ± 0.26, respectively, and in the SL they were 9.41 ± 1.11 and 9.05 ± 0.64, respectively. SOC decomposition in the paddy field soil was significantly less sensitive to temperature changes than in the upland soil. The Q 10 values averaged over sites and soil layers were 5.14 ± 1.33 and 8.05 ± 1.54 for the paddy field and upland soils, respectively.

Figure 3
figure 3

The Q 10 value for topsoil layer (TL, black bars) and subsoil layer (SL, gray bars) of paddy field and upland soils. Error bars indicate standard deviation (n = 4). Means are average Q 10 values for six sites. HP and MS represent Huaping and Meishan, respectively.

Full size image

Correlations

In order to eliminate the influences of soil type (paddy field soil and upland soil) and sampling area (HP and MS), we assessed the correlations between R 20 and POXC, and between Q 10 and POXC, for each soil type in each location. The results showed that POXC was positively correlated with R 20 for each soil type at each location (Table 2). However, the correlations between Q 10 and POXC were negative and significant for the two soil types at both HP and MS (Table 3).

Table 2 Pearson correlation coefficients (r) between R 20 values and POXC.
Full size table
Table 3 Pearson correlation coefficients (r) between Q 10 values and POXC.
Full size table

Discussion

On average, R 20 in the TL was 0.1728 ± 0.0416 (mean ± SD) μg CO2-C g−1 soil h−1 for paddy field soil and 0.1030 ± 0.0389 μg CO2-C g−1 soil h−1 for upland soil; this was significantly higher than in the SL where the mean values were 0.0484 ± 0.0182 μg CO2-C g−1 soil h−1 and 0.0534 ± 0.0065 μg CO2-C g−1 soil h−1 for paddy field and upland soil, respectively. Firstly, fresh organic matter derived from residues, root secretions, and/or organic fertilizer is often concentrated in the surface layer36, leading to a higher POXC in the TL than in the SL (Table 1). Secondly, the differences bwtween R 20 in the TL and the SL may be caused by differences in soil microbes. Babujia et al.37 found significantly higher microbial biomass in the TL than in the SL of soybean-wheat rotation soils. Though experimental conditions, especially oxygen conditions, were the same between the TL and the SL in our study, there might be more anaerobic and inactive microbes in the SL relative to the TL38. A study by van Leeuwen et al.39 also showed that microbial biomass and activity decreased with soil depth in cropland.

In addition, R 20 was significantly higher in paddy field soils than in upland soils, except in the TL of the MS sites where R 20 was not significantly different between paddy field and upland soils. This pattern of R 20 may be due to higher SOC content in paddy fields relative to uplands (Table 1) and it is consistent with previous studies of cropland soils40. The SOC content in the upland soils of the MS sites was the highest among all the sites.

Q 10 is a critical factor in predicting future changes in soil C pools41. Although field studies more closely replicate natural conditions than do laboratory studies, Q 10 values for soil respiration from field studies are usually confounded by variation in root respiration and other environmental factors (e.g., substrate and water conditions). Although laboratory incubation is commonly criticized for being unnatural42, it is an effective complementary way to study the Q 10 of SOC decomposition43. In this study, the Q 10 values of cropland soil at depths of 0–20 cm and 20–40 cm ranged from 3.1 to 11.6, with an average of 6.5, which was higher than the commonly reported Q 10 values (1.73~4.67)19, 44, 45. These differences in Q 10 values may be the result of differences in the incubation methods used in this study and in others. Zhu and Cheng found that Q 10 values were higher under constant temperatures than under diurnally-varying temperatures in farm and grassland soils through a 122-day incubation46. In most reported studies, soil samples were incubated at two or three constant temperatures to estimate the Q 10 value (e.g., Arevalo et al.19, Conant et al.45, and Haddix et al.47). However, this approach has shortcomings. Firstly, it may underestimate the Q 10 value48, because the depletion rates of C substrate incubated at different temperatures are different, with samples incubated at relatively high temperatures using up the more labile fractions of SOC sooner than those incubated at lower temperatures49. Secondly, microbies may adapt to different constant temperatures during the long-term incubation, which may lead to contradictory measurements of Q 10 values3. The uncertainties in the estimated temperature sensitivity of SOC decomposition caused by the methodology used have not yet been examined, and the underlying mechanisms are not yet known. Further studies are warranted to reliably predict the feedback between soil C storage and global climate change.

The decomposition of SOC in upland soil had higher Q 10 values than the decomposition in paddy field soil, and this pattern was consistent across sampling sites in both the HP and MS regions. Iqbal et al.16 reported similar results from incubation experiments, with a Q 10 of 2.3 in upland soil that was significantly higher than the 1.9 value in paddy field soil. In general, this pattern of Q 10 values could be explained by the fact that SOC in paddy field soil is more labile and the decomposition of labile C components is not as temperature sensitive. The difference in Q 10 values between paddy field and upland soil may also be partly attributed to variation in soil microbial community composition and microbial activity17. Chen et al.50 reported that soil microbial communities and activity levels were significantly influenced by cropland land type, showing that the relative amount of bacterial phospholipid-derived fatty acids (PLFAs), fungal PLFAs, and total PLFAs were greater in paddy fields than in uplands, while the bacterial/fungal PLFA ratio was greater in uplands than in paddy fields.

Under the same incubation conditions and model fitting, even in this study, Q 10 values in the SL were found to be significantly higher than in the TL. These results seem to be consistent with the current prevailing opinion that the decomposition of resistant C is more sensitive to temperature changes47, 51, 52. Because significantly higher POXC was found in the TL than in the SL, however, significant negative relationships were observed between Q 10 values and POXC in our study. Fierer et al.53 and Karhu et al.52 found that decreasing C quality (degree of resistance to microbial decomposition) with soil depth was responsible for the increase in Q 10 values in grassland and boreal forest soil profiles. A larger proportion of recalcitrant substances in deeper soil have also been shown to significantly increase Q 10 values with increasing depth in a peat ecosystem18. Another possible explanation for higher Q 10 values in the SL than in the TL may be the differences in microbial community composition and metabolic activity in the two soil horizons, but this needs to be experimentally verified.

Our results suggest that the differences in SOC decomposition between paddy fields and uplands and between the TL and SL should be considered in soil C models for predicting future C cycling. Furthermore, the results of this study have implications for agricultural management. On one hand, Q 10 values in upland soils were higher than in paddy field soils, suggesting that the upland soils may lose C more easily than the paddy field soil under future global warming. On the other hand, because of the higher Q 10 values in the SL relative to the TL, we should try to ensure shallow tillage because deep plowing could increase air circulation and increase soil temperature in the SL, leading to soil C loss in the SL in a warmer world.

Conclusions

SOC decomposition and its temperature sensitivity are two important indicators of the character of soil C processes in the context of global climate change. In arable soils, the TL relative to the SL has a significantly higher POXC content and SOC decomposition rate. However, SOC decomposition in the TL is significantly less temperature sensitive than decomposition in the SL. These differences between TL and SL need to be incorporated into soil C models in order to more reliably predict the response of arable soils to global climate change.

Materials and methods

Soil sampling and analysis

Cropland soils were sampled from two areas: HP and MS. HP (26°21′–26°58′N, 100°59′–101°31′E) is located in the northwestern part of Yunnan Province, China. The area has a mean annual temperature (MAT) of 19.8 °C and a mean annual precipitation (MAP) of 870 mm. MS (29°34′–30°21′N, 102°49′–104°49′E) is located in Sichuan Province, China, with a MAT of 17.0 °C and MAP of 1236 mm.

Paddy-upland rotation and upland farming are two typical cropping systems in Yunnan and Sichuan because these cropping systems are the major food production methods used in the southwest China. These systems are sensitive to climate change as are other Chinese agricultural ecosystems54. Rice, beets, and wheat are the main crops used for paddy-upland rotation, whereas wheat, corn, and soybean are used for upland farming. Six paddy-upland rotation sites (hereafter referred to as paddy field sites) and six upland sites were selected from both HP and MS. Distances between any two sites were about 5 to 10 km of paddy fields or uplands. All sites had been used for conventional farming for over 15 years.

Soil samples were collected in December 2013. Three sampling points (10 m apart) were selected randomly from each site. After surface litter and aboveground plants were removed, an auger with an inner diameter of 8 cm was used to collect soil samples from the 0–20 cm (TL) and 20–40 cm (SL) layers. Xie et al.34 showed that the average depths of the surface soil layers in Chinese paddy fields and uplands are 15.2 cm and 19.4 cm, respectively. Thus, we established sampling depths of 0–20 cm and 20–40 cm for the TL and the SL, respectively. After the roots were removed, the individual samples from each soil layer at each site were passed through a 2-mm sieve and individually mixed, yielding six TL and six SL samples for paddy field and upland soils in each area, respectively. The soil samples were transferred to the laboratory and stored at 4 °C for analysis and incubation.

Soil water holding capacity (WHC) was gravimetrically determined. Samples were saturated with water, allowed to drain through Whatman #1 filter paper, placed on a glass funnel for 24 h, and then dried at 105 °C for 48 h to a constant weight. Soil pH was measured in a water extract (water to soil ratio of 2.5:1) by using a glass electrode. Soil total carbon (TC) and total nitrogen (TN) were measured using an NC analyzer (FlashEA 1112 Series; Italy). SOC was determined using a TOC analyzer (Multi N/C 3100; Jena, Germany) after removal of carbonates with 0.1 M HCl. Permanganate oxidizable C (POXC) was measured using the method of Culman et al.55.

Soil incubation and measurement of respiration

In laboratory incubation, soils have often been incubated at three to five constant temperatures for several months or years (e.g., Conant et al.45 and Karhu et al.52). Incubating soils at constant temperatures for a relatively long term in this manner may have biased the estimated Q 10 values of soil respiration, because the substrate, SOC composition, and other soil properties very likely changed with time42. We therefore used a short term method of incubating soils (often several days) with changing temperatures56. Four replicates of each soil sample, equivalent to a dry weight of about 50 g each, were placed in 250 mL incubation jars. Soil samples were adjusted to 60% WHC and pre-incubated at 20 °C for 72 h to stabilize soil respiration57. The incubation jars were then placed in a cryogenic thermostatic bath (DC0530; Bilang Instrument Corp. Ltd., Shanghai). The incubation temperature was initially set to 20 °C, with a step length of 4 °C, increased to 28 °C, and then decreased to 8 °C and back to 20 °C. A similar method using vaying temperature has been reported by Fang et al.56 and Chen et al.57. Water loss during the incubation period was periodically checked gravimetrically and adjusted accordingly.

Soil incubation and respiration measurement followed Chen et al.57. Fresh air via a gas distribution system was continuously passed through the headspace of each incubation jar at a rate of 0.75 L min−1. The jars were allowed to remain at each incubation temperature for 2.5 h in order to stabilize soil respiration. Soil respiration was measured by closing the incubation jars and immediately removing a 5-mL gas sample from the headspace of the jar. The same volume of CO2-free air was injected into the jar to balance the air pressure. After some time, a second gas sample of 5 ml was obtained, and the incubation jar was opened to allow fresh air circulation. The CO2 concentration in the gas samples was measured using a gas chromatograph (Agilent 6890; Agilent Corp., USA) equipped with a flame ionization detector. Soil respiration rate was calculated as the difference in headspace volume, time interval, and CO2 concentration between the first and second gas samples57, according to the following equation:

$${R}_{s}=\,\frac{M}{22.4}\frac{P}{{P}_{0}}\frac{{T}_{0}}{T}\frac{{\rm{\Delta }}C}{{\rm{\Delta }}t}\frac{V}{m}$$
(1)

where R s is the soil respiration rate (μg CO2-C g−1 soil h−1), M is the molar mass of CO2-C (g mol−1), 22.4 is the molar mass of the gas under standard conditions (273 K, 1013 hPa) (1 mol−1), T 0 and P 0 are the temperature (K) and pressure (hPa) of the air under standard conditions, respectively, T and P are the air temperature (K) and pressure (hPa) at the time of gas sampling, respectively, ∆C/∆t is the change in CO2 concentration (ppm) in the jar by time (h), V is the headspace volume of jar (l), and m is the dry weight of incubated soil (g).

Data analysis

The variation in soil respiration rate in response to temperature change was described by an exponential model57:

$${R}_{s}=a{e}^{bT}$$
(2)

where R s is the soil respiration rate in μg CO2-C g−1 dry soil h−1; T is temperature (°C); and a and b are fitting parameters. Parameter a can be referred to as respiration rate at 0 °C, and b defines the temperature dependence of soil respiration.

The Q 10 value (the factor of respiration rate increase related to a temperature increase of 10 °C) of soil respiration was then calculated as follows:

$${Q}_{10}={e}^{10b}.$$
(3)

The MAT was about 17.0 °C and 19.8 °C in MS and HP, respectively. To compare SOC decomposition across soil samples, we chose the decomposition rate at 20 °C because the MAT in the two sampling areas was approximate 20 °C. SOC decomposition at 20 °C, R 20, was defined as

$${R}_{20}=a{e}^{20b}.$$
(4)

Data were analyzed using a two-factor analysis of variance method to determine the effects of soil horizon and sampling site on SOC, POXC, R 20, and Q 10 in each cropping system (paddy field and upland). All statistical analyses were performed using SPSS 19.0 (IBM/SPSS Inc., Chicago, IL, USA).