| Soil organic carbon (SOC) comprises a major part of the terrestrial carbon reservoir. A slight change in SOC stock may influence global climate due to its large quantity stored in terrestrial ecosystems and its critical role in the global carbon cycle. Accurately estimating SOC stock has become a focus of present research on global change, and is considered as essential for assessing soil quality, sustainable development of agriculture, non-point source pollution, national food security, and global climate change, etc. Due to the strong spatial heterogeneity of soil properties, the result of SOC stock estimation at regional scale is helpful for improving the precision of SOC stock estimation at global scale, and can also serve as baseline data for regional policy-making, ecological and environmental conservation, and agricultural sustainable development. In this study, a soil survey database of Zhejiang Province, containing six scales of soil maps (1:50,000,1:200,000,1:500,000,1:1,000,000,1:4,000,000, and1:10,000,000) were used to estimate SOC stocks and to analyze their spatial patterns up to a maximum depth of100cm for the Province. Four methods, i.e., the soil profile statistics (SPS), mean, median, and pedological professional knowledge based (PKB) methods, were tested, which connect soil spatial data base with its attribute information. Moreover, the uncertainties in SOC stock estimation caused by up-scaling of soil properties from the county level to the provincial level, from lower classification category of soil species to up classification category of soil group, and from larger-scale (e.g.,1:50,000) to smaller-scale (e.g.,1:10,000,000) digital soil map were evaluated. Major results were summarized as follows:1. The SOC density values of the2,154soil profiles taken from county soil survey manual of the Second National Soil Survey of China presented a positively skewed distribution with skewness of14.17and a strong degree of variability with coefficient of variation of103.9%. The highest and lowest SOC density values of the2,154soil profiles were279.52kg/m2and0.10kg/m2, respectively. By excluding water and urban areas, the total soil area in Zhejiang Province was100,740.0km2, and the total SOC stock in the Province was831.49*106t, with a mean SOC density of8.25kg/m2based on the PKB method. Among the10soil groups, the highest and lowest SOC density values were found in Mountain meadow soils and Basic rock soils, respectively. Red soils occupied the largest areas and had the largest SOC stocks, accounting for39.4%of the total area and31.2%and of the total SOC stocks of soils in the Province, respectively.2. SOC density and stock decrease with the decrease of soil depth. Within the depth of0-20cm, its estimated SOC stocks account for41.1%of the total SOC stocks for the depth of0-100cm. For different land use types, the highest and lowest SOC density values were in grassland (14.42kg/m2) and unutilized land (5.75kg/m2), respectively. The SOC density under paddy land was higher than under both dry cropland and garden land. In Haining city of the province, the SOC density in the surface soil layer increased by11.6%from1983when the Second National Soil Survey was conducted to2005.3. With soil classification levels up-scaling from soil species to soil group, the estimated SOC stocks presented minor differences, and the estimated SOC stocks by using the median method were always the lowest. The difference in the estimated SOC stocks among the SPS, mean, median, and PKB methods was lowest at the soil species level. The estimated SOC stocks using the four methods at each of the five soil depths (0-20,20-40,40-60,60-80, and80-100cm) followed the order:PKB> SPS> mean> median. Due to the high variations in SOC density, Mountain meadow soils and Yellow soils presented significant spatial differences among the mean, median, and PKB methods. Compared with the other three methods, the PKB method had an obvious advantage whenpresenting the differences in spatial patterns of SOC distribution because spatial locations of all soil profiles and soil parent materials were taken into account during the procedure of linking soil profile properties to soil spatial database.4. The up-scaling of soil properties from the county scale to the provincial scale and the number of soil profiles had obvious influence on the SOC stock estimation. The estimated SOC stocks using161statistical profiles taken from provincial soil survey reports were12.4%and8.5%higher than that using1797typical profiles taken from county soil survey manual using the mean and PKB methods, respectively. Significant (P<0.05) differences were found among SOC stocks estimated by using different estimation methods. In terms of mapping scale, the estimated SOC stocks from1:50,000soil map presented significant (P<0.05) differences with that from1:500,000,1:1,000,000,1:4,000,000and1:10,000,000soil maps, while no significant difference of estimated SOC stocks was found between the1:50,000and1:250,000soil maps. As soil map up-scaling from1:50,000to1:10,000,000,(1) the estimated SOC density and stock presented an increasing trend;(2) the increased area of soils was mainly attributable to water and urban areas, the increased SOC stock at the1:10,000,000was mainly related to the conversion of Skel soils to other soils;(3) the disappearance of water and urban areas and to the increase of soil areas, the elimination of small soil map units, and change in number of soil profiles used at various scales directly influence SOC stock estimation;(4) the differences in the SOC densities among the different scales of soil maps presented a decreasing trend, the PKB method had an obvious advantage in showing the differences in spatial patterns of SOC distribution between different scale soil maps than both the mean and median methods.However, several limitations remain and further study should focus on the update of soil database, especially the update of soil properities, the effect of land-use and land-cover change on SOC stock estimation, the relationship among related factors, the application of new methods (e.g., geostatistical method, DNDC model, and decision tree model), and the scale matching and conversion for different data sources. |
