Soil was regarded as the largest organic carbon pool in terrestrial ecosystem. Estimates of SOC storage in the upper level of soils range from 1400 to 1500 Pg, which constitutes approximately two-thirds to three-fourths of the organic carbon in terrestrial ecosystems (Post et al., 1982; Schlesinger, 1990). Due to its huge storage capacity of SOC, even a slight change in it could have a significant impact on atmospheric CO₂ concentration (Xu et al., 2011). To be one of the most widespread vegetation types worldwide, grasslands cover nearly one-fifth of the world’s land surface (Lal, 2001). In grassland ecosystems, less than 10% of organic carbon is stored in the above biomass (Burke et al., 1997). The relatively large area of grasslands and huge SOC storage capacity make it a potentially important pool for carbon storage in terrestrial ecosystem. To detect carbon sequestration and its potential in grassland ecosystems would provide scientific basis for adaptation to global climate changes.

Research shows that grassland soil carbon storage occupies at least 10%-30% of the global soil carbon storage (Escaran et al., 1993; Anderson, 1991). Carbon storage in the soils of grasslands were previously studied based on different methods at a regional to global scale (Tao et al., 2006; Parton et al., 1995; Batjes, 1996; Jobbagy and Jackson, 2000; Ni, 2002; Zeng et al., 2004; Yang et al., 2010a; 2010b). Comparing those studies, we found that the results were not consistent, maybe because of different definitions of grassland and methods adopted. But all those works showed that grassland soils served as an important carbon pool (Scurlock and Hall, 1998). There is relatively little research on dynamics of SOC storage in grasslands of China and the existing estimates differ greatly. For the grasslands in northern China, Yang et al. (2010b) suggested that the SOC stock did not show a significant change by comparing current measurements with historical records derived from the Second National Soil Survey. In contrast, Xie et al. (2007) found SOC under grassland declined by 3.56 Pg over the last 20 years, which northern grasslands accounted for 94.7%.

The grasslands of Inner Mongolia are the main part of the temperate grasslands in northern China. The study of Inner Mongolia grasslands’ SOC stocks and their changes over time are important for China carbon cycling research. Scientists have already launched studies on SOC cycle in Inner Mongolia, including study on the carbon cycle of leymus chinensis steppe in the Xilin River Basin (Li et al., 1998), spatial heterogeneity of soil moisture, carbon and nitrogen in stipa communities (Bai et al., 2002), spatial characteristics of SOC and nitrogen in Inner Mongolia (Chen et al., 2003). From the perspective of human disturbance, there are studies such as variation of below-ground carbon sequestration under long term cultivation and grazing in typical steppe (Yan et al., 2008), effects of grazing on SOC and total nitrogen in typical steppe (Liu and Zhang, 2010), response of soil respiration to grazing intensity, water contents, and temperature of soil in different grasslands of Inner Mongolia (Yang et al., 2012). SOC changes along a 50-year grassland-to-cropland conversion chronosequence in an agro-pastoral ecotone of Inner Mongolia (Jiao et al., 2012). The current researches on SOC are mainly conducted in connection with specific regions, the estimation of SOC storage and its changes are still very limited in Inner Mongolia.

Since combination of remote sensing data and field survey data is a new trend in current SOC storage estimation, Yang et al. (2009) detected the positive correlation between remote sensing data normalized difference vegetation index (NDVI) and SOC in the Tibetan alpine grasslands, and estimated the SOC storage and its changes using this satellite-based methodology. In this paper, based on soil data from the second soil survey of China and field survey during 2011-2012, we first establish the regression relation between soil data and remote sensing data. We then use ArcGIS plus this linear equation to explore the changes and its spatial distribution of SOC storage in grasslands of Inner Mongolia over the past 30 years. Therefore, the objectives of this study are to (1) estimate SOC storage in grasslands of Inner Mongolia for a depth of 20 cm; (2) investigate the changes and its spatial distribution of SOC storage; (3) analyze the variation characteristics of SOC storage in grasslands of Inner Mongolia.

Based on the relationship between NDVI and SOCD, we estimated the sizes and the changes of topsoil organic carbon pool in grasslands of Inner Mongolia over the last three decades. The conclusions are as follows:

(1) The statistical test and analysis of equation results indicated that the predicted SOCD was well consistent with measured value, meaning this method can calculate SOC storage and its dynamic scientifically.

(2) SOC storage in grasslands of Inner Mongolia between the 1980s and 2010s was estimated to be 2.05 and 2.17 Pg C, with an average density of 3.48 and 3.69 kg C·m^-2, respectively. The spatial distribution showed a decreased trend from the meadow steppe, typical steppe to the desert steppe, corresponding to the temperature and precipitation gradient.

(3) SOC changes during 1982-2012 were estimated to be 0.12 Pg C, indicating that SOC storage in grasslands of Inner Mongolia remained relatively stable over this period. However, topsoil organic carbon of different grassland ecosystem showed different trends of carbon source/sink during the past three decades. Meadow steppe and typical steppe had sequestered 0.15 and 0.03 Pg C, respectively, served as a carbon sink; while desert steppe had lost 0.06 Pg C, served as a carbon source. It appears that SOC storage in grassland ecosystem may respond differently to climate change, related to vegetation type, regional climate type and grazing intensity.

Generally, the relationship-based SOC assessment can reduce the uncertainties derived from the large soil heterogeneity, but some uncertainties and application restricts still exist. Due to different data sources, regional differences and other factors, the estimates of grassland SOC storage are quite different. Our findings demonstrated that NDVI was positively correlated with SOCD in grassland ecosystems of Inner Mongolia. In the study of the major grassland types in Alex, Inner Mongolia, Fu et al. (2004) found that the most important factors affecting SOC in the 0-20 cm layer were plant cover, plant productivity and annual precipitation, and the partial correlation coefficients were 0.777, 0.697 and 0.429, respectively, reflecting the close relation between SOCD and NDVI. Our study calculate soil carbon storage through the relationship between SOCD and remote sensing data, the uncertainty of our findings include: (1) Remote sensing data used in this study are derived from two kinds of datasets: GIMMS and MODIS; the data conversion process will produce some errors. (2) Most soil profiles lacked the bulk density data in the Second National Soil Survey. In the previous studies, some use the relationship between bulk density and SOC concentration to estimate bulk density; some use the recent field survey data. In this paper, the lacked bulk density was obtained from related research, we use the soil bulk density of belonging soil subtypes to replace the missing data. (3) The 2010s’ sampling sites were mainly set in meadow steppe and typical steppe and little in desert steppe. Thus, compared with the 1980s, the correlation coefficient between predicted and measured values in the 2010s was relatively lower, which caused a potential uncertainty in estimating SOCD of desert steppe. However, considering the low proportion of desert grassland area and its low SOCD, it merely exerts little influence on the total SOC storage, especially for topsoil. These error sources suggest that more reliable field survey data and scientific model are required to improve our understanding on sizes and changes of SOC storage in the future study.

The dynamic process of grassland SOC contains two parts. One is carbon input—the humus forming process fixed carbon into soils, the other is carbon output acting as soil respiration. On the micro level, the micro-climates in soils affect organic matter’s decomposition rates, and then control the formation and decomposition of SOC to some extent (Yang et al., 2011). On the macro level, climate conditions and human activities are the main factors impacting grassland soil carbon storage. Climate change alters temperature, precipitation, CO₂ concentration and other environmental factors. Human activities (cultivation, grazing, mowing, etc.) also have an impact on grassland carbon storage, mostly exerting a negative influence. Our study found that the soil organic carbon storage in grasslands of Inner Mongolia were slightly increased over the past 30 years, but the main driving factors and their role in the process remain to be investigated.

The relationship between terrestrial ecosystems and the global carbon cycle is the front edge and hot spot in global change research, a lot of studies have been developed to estimate the grassland soil carbon pool. According to IPCC “B2 mid-low emission scenario”, the regional climate model predicts that the temperature in China in the 2020s, 2040s, and 2070s will rise by 1.16, 2.20, and 3.20℃, the corresponding precipitation increases are 3.7%, 7.0%, and 10.2%, respectively, as well as decrease in extreme cold damage, and increase in extreme heat (IPCC, 2007). Under such future climate change scenarios, how grassland soil carbon pool will change, how different grassland ecosystems will response, what kind of short-term and long-term effects will produce, all these questions need further study.