The soil system represents the dominant terrestrial reservoir of carbon in the biosphere. Deforestation, poor land management, and excessive cropping lead to a decrease in soil carbon stocks, but intensive cropping can reverse this trend. We discuss long-term soil organic carbon data from two major rice-growing areas: Java (Indonesia) and South Korea. Soil organic carbon content in the top 15 cm for both countries has increased in recent decades. In South Korea, the top 15 cm of soils store about 31 Tg (1012 g) of carbon (C) with a sequestration rate of 0.3 Tg C per year. In Java, the agricultural topsoils accumulated more than 1.7 Tg C per year over the period 1990–2010. We attribute the increase in measured SOC mainly to increases in above- and below- ground biomass due to fertilization. Good agronomic practices can maintain and increase soil carbon, which ensures soil security to produce food and fiber.
 Climate change including change in temperature, elevated CO2concentrations, increased rainfall variability, and altered land-use will have a great impact on soils. The influence of these factors will create a dynamic feedback between soil and the environment [van Noordwijk et al., 1997]. Soil organic carbon (SOC) and its potential to become a managed sink for atmospheric CO2 has been discussed widely [Jobbágy and Jackson, 2000; Smith et al., 2008]. The role of SOC as a primary control on soil quality and soil function has been acknowledged for over 70 years [Hardon, 1935; Jenny, 1941]. Soil structure, water holding capacity, pH buffering capacity and nutrient supply are just a few of the soil properties that are affected by SOC. Because of food-security requirements [Ash et al., 2010], it is important to attempt to sequester soil carbon under increased yields. The amount of carbon stored in the soil is highly variable and depends on inputs, soil type and the decomposition rate of SOC. There are numerous studies reporting that SOC storage is influenced by land use, with certain emphasis placed on the role agricultural activity on the depletion of SOC stocks [van Noordwijk et al., 1997]. Studies of Australian soil systems have shown that conversion of forested and grassland areas into cultivated agriculture has led to an overall decline in SOC stock in those soils [Dalal and Mayer, 1986; Dalal and Chan, 2001]. A study in the UK suggested climate change is the primary cause of soil carbon decrease in England and Wales [Bellamy et al., 2005]. The decline in SOC can be broadly attributed to changes in soil structure, diminished microbial activity, altered aeration and moisture regimes, reduced inputs of organic matter and more intense soil erosion [Dalal and Mayer, 1986]. However, better land management such as minimum tillage has shown some tentative increases. This applies largely to cereal production in Europe, North and South America [Post and Kwon, 2000; West and Post, 2002]. There have been several long-term studies on rice in Asia [Jung et al., 2007; Pampolino et al., 2008; Zhang and He, 2004], however the impact has not been analyzed fully at regional scales.
 With a warming climate, the rate of carbon flux to the atmosphere through soil respiration is expected to increase [Bond-Lamberty and Thomson, 2010]. But this does not necessarily mean that soils lose a greater proportion of their large carbon stores to the atmosphere. Because of the inherently long-term nature of soil carbon change, especially accumulation, addressing the questions of the influence of climate change requires empirical data over decades.Trumbore  recommended that more work needs to be done to evaluate the responses of soil to global environmental change at large scales that occur along natural environmental gradients over decadal timescales. The accelerating pace of human intervention is believed to cause the decrease in soil carbon stocks [Bellamy et al., 2005]. However, in some parts of the world, agricultural activities can result in increasing soil carbon sequestration. For example, recent SOC measurements across China showed that in some areas there has been an increase of SOC content when compared with values collected from a national soil survey conducted in 1979–1982 [Yan et al., 2011]. The increase in SOC was mainly attributed to the large increase in crop yields since the 1980s. Meanwhile areas that have been recently established for crop production show a SOC decline.
 While soil carbon turnover models have given us a tool with which we can study the response of SOC to climate change [Deryng et al., 2011; Jenkinson and Coleman, 2008], stronger evidence arises from empirical studies over large areas to assess the actual impact of climate change and land management on soil carbon sequestration. In this paper we will illustrate the dynamics of SOC in two areas in the world that have been cultivated continuously since the 1960s, in South Korea and Java, Indonesia. Both areas have experienced increases in mean maximum and mean minimum temperature over the past four decades [Griffiths et al., 2005].
 A national soil monitoring program was established in 1999 for major agricultural lands in South Korea, including paddy fields, upland agriculture systems, orchards, and plastic-film houses. Here we only focus on the monitoring data for paddy fields. The sampling design is based on land use, soil geography and agro-climatic zonal areas. Specifically, soil sampling sites were determined based on 19 agro-climatic zones, by soil series and by land use (paddy fields in this case). Topography, soil texture, drainage class, effective soil depth, gravel content, plough pan, slope were considered as soil geographical factors. Sampling sites were distributed to Province, City/County, and Myeon administrative levels. The locations of the sampling points were located by pedologists who know the representative soil series. Seeauxiliary material Figure S1 in Text S1 for the spatial distribution of the data. The program was repeated every 4 years (data from 1999, 2003 and 2007 were used in this study).
 Soil sampling was conducted in the spring before fertilization and irrigation in paddy fields. Soil samples were taken from a depth of 0–15 cm. The organic matter content was determined based on the Tyurin method [Tyurin, 1931]. SOC was calculated using the van Bemmelen conversion factor, assuming SOC is 58% of organic matter. Carbon stocks for the top 15 cm of the soil (in kg m−2) were calculated as follows:
where C is SOC content in mass percentage (g 100 g−1), ρb is the soil bulk density (g cm−3) and t is the thickness of the soil (m).
 Since bulk density was not measured in the monitoring program, we used the Korean soil database to derive an estimate of the bulk density. The national soil survey data, which was collected in the early 1970s, were also used to represent the average SOC content in 1970 [Jung et al., 2000]. The survey data is represented by a soil map for the whole country at a scale of 1:25,000 describing 380 soil series. Each soil series is represented by a typical (modal) profile where its physical and chemical properties were fully analyzed (Table 1). There is only a limited number of bulk density data in the database (n = 108), therefore we derived a pedotransfer function (PTF) for the mineral bulk density (ρm):
where depth is the mean depth of the sample (cm), and Sand is the percentage by mass of sand (50–2000 μm) (g 100 g−1). Each soil series was represented with an average mineral bulk density. The soil bulk density was calculated to include the influence of organic matter based on the model of Adams :
where OM% = mass percentage of organic matter (g 100 g−1), and ρOM = average organic matter bulk density = 0.224 g cm−3, and ρm is the mineral bulk density (g cm−3). Applying equations (2) and (3) to the Korean soil database provided a goodness of fit R2 = 0.49 (n = 108). Table 1 shows the distribution of the raw soil monitoring data.
Table 1. Soil Organic Carbon (0–15 cm) From Paddy Soils in South Koreaa
Number of Observations
Median SOC Content (g kg−1)
Median C Stock (kg m−2)
Numbers in parentheses represent the 2.5 and 97.5 centiles of the data.
 While there are 152 soil series in the South Korean soil taxonomy system that are paddy soils, only 130 soil series were sampled for all 4 periods (1970, 1999, 2003 and 2007). We grouped the soil monitoring data based on the 130 soil series.
 To calculate the total carbon storage in the paddy field area, the data from the soil monitoring sites were interpolated to the whole area of paddy fields in South Korea using ordinary kriging [Webster and Oliver, 2007] with a grid spacing of 90 m × 90 m. The delineated area of paddy fields was derived from a remotely sensed image classification.
2.2. Java, Indonesia
 A data set of soil properties from Java was compiled from soil survey data conducted from 1960 to 2010 by the Indonesian Center for Agricultural Land Resources Research and Development (ICALRD) in Bogor, Indonesia. For the period of 1960 to 1990, we used the database that was compiled by Lindert . We updated the database to include new data from surveys in Java conducted from 1990 to 2010. We calculated the SOC content for 0–15 cm depth, resulting in a database of 1032 soil profiles. Since the samples were collected at various depths based on soil horizon designation, the average SOC content for 0–15 cm was calculated as a depth-weighted average. The exact geographical coordinates of the samples were not usually recorded, only the name of the closest village was given. SOC was mainly analyzed using the Walkley-Black method [Walkley and Black, 1934]. Bulk density was not measured, and predicted using a PTF. A PTF for mineral bulk density for soils in tropical regions was used [Minasny et al., 2011]:
where ρm is the mineral bulk density (g cm−3), d is the mean depth of the sample (cm), and Sand is the percentage by mass of sand (50–2000 μm) (g 100 g−1). Soil bulk density (ρb) was then predicted from mineral bulk density adjusted for organic matter content, calculated using the model of Adams (equation (3)) [Adams, 1973]. Equations (3) and (4) had a goodness of fit R2 of 0.46 when calibrated on tropical soils [Minasny and Hartemink, 2011].
 Since the data covered time unevenly (see auxiliary material Figure S14 in Text S1), the soil carbon stock data were grouped per decade from 1960 to 2010.
3.1. South Korea
 Rice is mostly grown in alluvial soils in South Korea. No significant trend was observed between SOC content and elevation (see auxiliary material Figures S2 and S3 in Text S1). A general increase in annual mean temperature over the Korean Peninsula was observed from 1974 to 2002, with an average of 0.2C° per decade for the rural areas [Chung et al., 2004]. We only found a trend of increased SOC content with increasing rainfall (Figure 1). No significant trend was observed between SOC content and mean annual temperature. This suggests that moisture conditions are the main environmental factor affecting SOC in paddy fields. Human activities also override the influence of topography.
 We analyzed the SOC content based on data from the 130 soil series that were sampled at the aforementioned discrete times (Table 2). The data distribution (Figure 2, see also auxiliary materialFigures S4–S7 in Text S1) shows that there has been an increase in the median topsoil organic carbon content and carbon stock from 1999 to 2007. We used the median rather than the mean, as we do not need to assume a normal distribution for the data. The non-parametric Van der Waerden rank scores test showed that the medians are significantly different. We did not have enough data to show the trend from 1970 to 1999, however SOC content consistently increased from 1999 to 2007. Seeauxiliary material Figures S8 and S9 in Text S1 for carbon stock comparison between each monitoring period. The median SOC content in 1999 was 12.8 g kg−1, and increased at an average rate of 1.2% per year or 0.25 Mg C ha−1 year−1.
Table 2. Median of Soil Organic Carbon in the Top 15 cm in South Korea Based on 130 Soil Seriesa
Median SOC Content (g kg−1)
Median SOC Stock (kg m−2)
Numbers in parentheses represent the 2.5 and 97.5 percentiles of the data.
 We estimated the SOC storage for the paddy field areas in South Korea in 1999, 2003, and 2007 using ordinary kriging, a spatial interpolation method, assuming a constant area of 14 221 km2 (Figure 3). The areas where SOC has increased have been mainly on the plains in the east to southeastern parts of the country. Using a grid spacing of 90 m × 90 m, we calculated that on average, the top 15 cm of soils store about 31 Tg (1012 g) C with a sequestration rate of 0.3 Tg C per year (from 1999 to 2007) (Table 3).
Table 3. The Total SOC Storage (0–15 cm) in Paddy Fields in South Korea (Assumed Area 14,221 km2)
C Storage (1012 g or Tg)
Rate of C Accumulation (Tg C year−1)
3.2. Java, Indonesia
 For SOC in Java, we examined historical soil survey data collected by the Indonesian Center for Agricultural Land Resources Research and Development (ICALRD) in Bogor, Indonesia [Minasny et al., 2011]. We selected data with agricultural landuse from 1960 to 2007. SOC content was calculated for the top 15 cm (Table 4) using a depth-weighted average. SOC has its lowest level of 7.0 g kg−1 in 1960–1980 (Figure 4). This is below the critical level for tropical soils of 11 g kg−1 [Aune and Lal, 1997]. This low value indicates that the soil had become impoverished because of continuous food production to meet the demand for a national staple food supply. However, SOC content has been steadily increasing since the 1970s. The increase is mainly attributed to the green revolution which introduced high-yielding varieties, and chemical fertilizer application. The rate of accumulation in the top 15 cm of soil was on average 0.12 Mg C ha−1 year−1 in 1980–1990, and 0.32 Mg C ha−1 year−1 since then. With an estimated average area for rice of 53 000 km2, we estimated that in each year, the soil in Java can accumulate more than 1.7 Tg C per year. Although the overall SOC content is less than South Korea, the area of rice cultivation in Java is about 4 times larger and the carbon sequestration rate is six times higher. Increased biomass production and the return of crop residues, green compost and animal manure application have been mostly responsible for the increase of SOC content in Java.
Table 4. Median of SOC Content for 0–15 cm in Java, Indonesia Grouped by Decadesa
Median SOC Content (g kg−1)
Median SOC Stock (kg m−2)
Numbers in parentheses represent the 2.5 and 97.5 percentiles of the data.
 The data highlighted the changes in topsoil organic carbon content for rice-growing areas in South Korea and Indonesia. We realize there are limitations in the data, namely:
 1. The change in sample locations over time.
 In Korea, the monitoring program started in 1999 and has been repeated every 4 years, but not all locations were revisited at every round. There were only around 50% of the locations that were revisited from 1999 to 2007. The sampling locations were allocated based on the known soil series. Nevertheless at each period, samples were collected evenly over the rice-growing area with a sufficient sample size of between 2000 to 3000 observations.
 In Java, the data are more problematic as they came from legacy soil survey data with no statistical criteria for sampling. The surveys can be selective and change with space and time. This may lead to biases in the areas being sampled. The geographic coverage in Java at different times is uneven, thus we grouped them by decades to ensure that each decade covered the same areas (administrative district) evenly [Minasny et al., 2011].
 2. The accuracy of laboratory methods used is unknown, in addition to the lack of cross-time calibration of laboratory methods.
 Although the methods used for measuring SOC in both countries are consistent throughout the period, we do not know the accuracy and the consistency of the measurement. SOC content in Korea was measured using the Tyurin method, while in Java the Walkley and Black method was used.
 3. Only soil properties in the top 0–15 cm were considered.
 The sampling in Korea only focused on the top 15 cm soil, as it is where the roots are mostly concentrated and most relevant for soil fertility management. SOC can vary with depth and we do not account for SOC at depths greater than 15 cm.
 4. The change in soil bulk density.
 We use a standard depth of 0–15 cm for comparisons between different times, however bulk density can change due to tillage or compaction. The change in bulk density means the soil mass over certain depths will differ at various times, thus the calculation of carbon density may be affected. While the influence of SOC was accounted for when calculating bulk density (equation (3)), comparison of soil carbon density in differing soil masses may not be appropriate.
 These limitations are assumed to have a minor influence at a national scale; we argue that because we have a large number of observations, we can represent the average soil C level for each period.
 SOC dynamics depend on various factors, such as soil type, environmental conditions, climate, and landuse management. Assuming that the soil type is constant over time, the main driver for SOC change regionally therefore is climate and landuse management. Climate change has been reported for both countries, with a trend of increasing temperature. An increase in temperature would suggest a more rapid SOC decomposition as hypothesized by [Bellamy et al., 2005], however we are reporting empirical evidence that SOC is increasing in the two countries. Based on historical evidence and the literature, we suggest some hypotheses of possible causes.
 1. Increased yield and root biomass.
 In South Korea, rice production has been almost uniformly managed with irrigation, flooding, deep tillage, and puddling throughout the whole country. Rice yields increased rapidly until 1980 because of the introduction of the high-yielding rice varietyTongil and the adoption of modern farming practices. However, severe cold weather caused a yield decline in 1980. New Japonicahigh-yielding rice varieties were developed and replaced theTongil variety in 1990. Since then, rice yield in Korea has maintained its maximum with variations due to climate and fertilization schemes. However, because of land pressure, the rice cultivation area has declined at a rate of 150 km2 per year since 2000. Meanwhile, Indonesian rice yields have increased steadily from 1960 to 1990 because of government support including expansion of the irrigated area, adoption of improved varieties, and the use of fertilizers accompanied by extension efforts [Cassman, 1999]. The increase after 1990 has been slower because of the removal of fertilizer subsidies. The production figures for both countries show a similar pattern of increase, particularly after the economic crisis in 1998 (Figure 5). Yan et al.  also observed increases in SOC for agricultural soils in China. They observed not only an increase at 0–20 cm but also at 20–100 cm. This suggests that increased productivity, especially root biomass plays an important role in SOC increase [Jobbágy and Jackson, 2000; Rasse et al., 2005]. In Java, most of the crop residues are incorporated back into the soil. In South Korea, most of the residues are also returned, however in the past 4 years they have been removed and used for animal feed. The root exudates and root biomass represent an important component of SOC build up [Lu et al., 2003; Rasse et al., 2005]. Root C has a longer residence time in soil than shoot C, and root activities also enhance soil organic matter protection via physical and chemical mechanisms [Rasse et al., 2005]. In Korea and Indonesia, where the soil is not ploughed after harvest, the decomposition of the intact root system was found to be much slower and the root conserved more organic C in soils compared with the incorporation of fresh residues [Lu et al., 2003]. Additionally, more SOC was gained from the below-ground biomass (root C) compared to the incorporation of fresh biomass from plant residues.
 2. Increase in fertilizer use.
 The increase in fertilizer use and efficiency is the main driver for increased crop production. In South Korea, chemical fertilizer consumption (N, P, and K) increased rapidly from the 1960s to 1990s, however fertilizer use has been decreasing since 1997 because of better nutrient management, which is due to the availability of a national soil fertility analysis and decision-support system. The rate of fertilizer application is around 200–300 kg ha−1 (Figure 6). However, the use of organic fertilizers in the form of green manure has been increasing dramatically since 2005 with a rate of 100–300 kg ha−1 (Figure 6). A long-term fertilizer experiment on rice in Korea, which was established in 1954, found that green-manure application in addition to NPK fertilizers continually increased SOC content [Jung et al., 2007]. SOC content was increased by 0.26 g kg−1 yr−l when 7.5 Mg ha−1 yr−l of rice straw was applied [Jung et al., 2007]. Fertilization produces higher above ground and root biomass which is incorporated in the soil. The use of nitrogen (N) fertilizer was found to increase SOC levels [Paustian et al., 1997; Li et al., 2006]. Nevertheless, the influence of this effect depends on management [Alvarez, 2005]. The average ratio of soil carbon to nitrogen is about 10:1, therefore an increase in SOC requires an increase in N. If N is supplied mainly from fertilizer, this may not be beneficial for greenhouse gas mitigation because of the fossil fuel energy required for the production of inorganic fertilizer Figure 6.
 3. Rice cultivation.
 Rice cultivation is known to enhance the SOC stock. It is suggested as one of the recommended management practices that may enhance SOC sequestration at the rate of 0.2–0.3 Mg ha−1 yr−1 [Cai, 1996; Lal, 2002]. The incorporation of rice residues and continuous flooding has become common in tropical areas through intensification of rice-cropping practices. Flooding can reduce C mineralization, furthermore there are also substantial inputs of C and N from aquatic biomass (mainly algae) [Sahrawat, 2003]. Long-term experiments in the Philippines [Pampolino et al., 2008] and China [Zhang and He, 2004] showed that continuous cultivation of irrigated rice with balanced fertilization on submerged soils maintained or increased SOC. In a flooded rice system, decomposition of the plant-derived materials not only plays an important role in nutrient availability and crop yield, but also increases methane production and emissions [Lu et al., 2003]. The balance between the magnitude of C storage in the soil, CH4emissions into the atmosphere, and long-term soil fertility need to be fully explored.
 These empirical data presented here suggest that agricultural practices can increase SOC storage in the soil despite a warming climate. Studies from China [Yan et al., 2011] and Italy [Fantappiè et al., 2011] have also indicated that the management effect overrides the increase in soil temperature. The increase of SOC content is mainly attributed to the large increase in crop yields and root biomass, increases in fertilizer use and the nature of the rice cultivation system. Therefore food security and soil security are highly related. Soil security is the maintenance or improvement of the world's soil resource so it can supply the world on an ongoing basis with sufficient food and fiber and a variety of ecosystem services. Furthermore, soil security contributes to energy sustainability and climate stability. A principal mechanism for achieving soil security is the management and sequestration of soil carbon through active land management systems and technologies. Sequestering soil carbon is positive for mitigating greenhouse gas emissions and for food and soil security.
 The soil data presented in this study are not public domain; they belong to the Rural Development Administration (RDA) South Korea and the Indonesian Center for Agricultural Land Resources Research and Development (ICALRD) respectively. The South Korean national soil database (1970 data) can be obtained from the publication by Jung et al. . The South Korean monitoring data is managed by RDA. The Java soil data from 1960–1990 can be obtained from the publication of Lindert . Data post 1990 is maintained by ICALRD. The authors thank Meine Van Noordwijk from World Agroforestry Center, Bogor for his constructive comments. Budiman Minasny is supported by the Australian Research Council Discovery project Methodologies for global soil mapping. Alex McBratney acknowledges Australian Research Council support through the Discovery project Global space-time soil carbon assessment. The Korean study was supported by PJ008546012012, Rural Development Administration, Republic of Korea.