Soil Organic Carbon Storage and Influencing Factors in Southwest Mountainous Area

As the biggest "carbon sinks" and "carbon source" of terrestrial ecosystems, soil organic carbon plays a crucial role in global warming and agricultural production. Both of natural and human factors has a momentous influence on soil organic carbon. The cultivated soil of mountainous and hilly area located in Yunyang County was selected as the object area of research. Elevation, slope, parent materials, slope position and Topography Wetness (TWI) are considered as essential factors for soil organic carbon. In addition, category variables were introduced into the regression model through path analysis, the mechanism of factors on cultivated Soil Organic Carbon (SOC) density was discussed. The variability of SOC density was gained by border analysis and anisotropic analysis. The results show that the average, 0-20 cm, cultivated SOC density is 2.91 kg/m and the cultivated soil carbon storage is 1838.75×106 kg in the study area. The correlation between elevation and SOC density is significant (0.329**). Topography wetness index (TWI) also has great correlation with SOC density (0.256**). Areas covered by Graybrown purple mud (shale) efflorescence and Purple sand and mud (shale) efflorescence have lower SOC density. The sequence of SOCD in different slope position is: Valley>slope foot>ridge>slope shoulder>slope back. From spatial variation aspect, anisotropic analysis illuminates that, spatial variability of SOC density is more drastic in south-north orientation than in east-west orientation.


INTRODUCTION
Soil Organic Carbon (SOC) is an import ant soil component in farming systems.It is essential to improve soil and water quality and hence sustains food production (Singh et al., 2007;Longbottom et al., 2014).As the largest carbon reservoir in terrestrial ecosystems (Li et al., 2002), soil carbon library has an important effect on the greenhouse effect and global climate change4.
There is strong spatial variation in Soil Organic Carbon (SOC) (Xie et al., 2004;Grüneberg et al., 2010) and both of natural and human factors can bring a great influence on spatial variability of the SOC (Mou et al., 2005;Somaratne et al., 2005;Tan et al., 2004a).The significance of impact factors varies from scale to scale (Powers and Schlesinger, 2002).Likewise, the influence of different geographical regions on main controlling factors of the spatial distribution is different (Dai and Huang, 2006).In recent years, studies on the mechanism and spatial variation of Impact Factors affecting soil organic carbon have become a research hotspot.The traditional soil organic carbon analysis was often based on qualitative procedures (FAO, 1976).Recently, with the rapid development of computers and information technology (Soil Survey Staff, 1993;USDA, 2007), a more quantitative approach has been developed that may replace the traditional inventory techniques.
The method of agricultural management, cropping systems, vegetation, climate, terrain conditions and other factors are generally considered to have significant effects on soil organic carbon (Vallejo et al., 2008).To quantify the qualitative variables into regression model to analyze the influences of these factors on cultivated soil organic carbon density, using the above factors as auxiliary data has become possible.
As it has a significant effect on the predicted accuracy, that the soil parent material as the spatial variable, introduced into the soil organic carbon prediction model (Cresser et al., 2007).Alejandra function was used to quantify the qualitative factors and the soil parent material was introduced into the multiple regression models to realize the soil organic carbon density spatial prediction and mapping, where the results indicated that the prediction accuracy was improved (Cresser et al., 2007).
At present, researches on effects of soil parent material on soil organic carbon density and spatial prediction are of great importance but largely missing.
Meanwhile, effects of soil parent material on soil organic carbon density remain to be further explored.Hence, we have to quantify qualitative soil parent material to analyze inherent soil organic carbon variation caused by the interaction of different soil forming factors accurately.
Extensive literatures have reported the distribution patterns and spatial characteristics of soil organic carbon reserves from small scale to large scale (Tan et al., 2004b;Greenland and Szabolcs, 1994).However, due to the difficulties encountered in obtaining accurate information of soil organic carbon from the local scale, researches on the same region often draw different conclusions (Yu et al., 2005;Wang et al., 2000).
Although numerous studies have reported on field scale, the researches on regional soil organic carbon storage and distribution are still largely missing (Don et al., 2007).There is a need for further study of soil organic carbon content in small hilly areas, especially on regional scale, with limited data availability and considerable inherent soil variation caused by the interaction of different soil forming factors, in order to improve the accuracy of the carbon cycle from the regional, national and global scale.
Here, we focused on SOC content because these properties are important driving factors behind crop production and can be used in crop growth simulation models as indicators of soil fertility.We thoroughly analyzed influencing factors and the spatial variability of cultivated soil organic carbon at the regional scale, based on 5893 soil samples taken in the surface layer (0-20 cm) in the period from 2008 to 2010.This study may provide scientific guidance for farming and agricultural regional planning and have 4 main objectives: 1. to evaluate the soil organic carbon storage of Yunyang area; 2. to analyze quantitative relations of SOC and Impact Factors (including Soil parent material, soil type and topographical variables); 3.to analysis direct or indirect influence of Impact Factors on soil organic carbon and the influence path; and 4.to assess spatial distribution and variation characteristics of soil organic carbon.

MATERIALS AND METHODS
Study site: Our study located in southwestern China covers an area of 3649 km 2 with an elevation ranging from 139 to 1809 m above sea level  Following the farmland productivity evaluation survey and quality technical procedures, a total of 5893 soil samples were collected from a depth of 0-20 cm of cultivated horizon soil.The soil organic matter content and soil volume was determined using the dichromate oxidation (external heat applied) method and Ring knife method, respectively (Nelson and Sommer, 1975).The location of each soil sample was recorded with Geographical information systems (ArcGIS9.3)according to the latitude and longitude coordinates (Fig. 1).
Soil organic carbon density was calculated with the following equation (Zhang et al., 2005): where, A is gravel content (%), B is volume mass (g/cm 3 ), SOM is organic matter content (g/kg) and 0.58 is the conversion factor of Bemmelen.1.
Topographical variables: Elevation data was obtained by 30-m grid DEM (digital elevation model) (Fig. 2).Three topographical variables were derived from the DEM: (1) Topographic Wetness Index (TWI), ( 2) slope position and (3) slope.The topographic wetness index can accurately portray the terrain changes and their impact on soil runoff is an effective indicator to characterize soil moisture content (Zhang et al., 2005).
It can be written as Wilson and Gallant (2000) and Claessens et al. (2006): where, α = The specific catchment area (SCA, m 2 /m) β = The local gradient SCA is defined as the upstream catchment area of a unit contour.
Based on the similarity weighted fuzzy reasoning method (Qin et al., 2007(Qin et al., , 2009)), five slope positions, namely, ridge (Spr), shoulder (Sps), slopeback (Spb), footslope (Spf) and valley (Spv) were divided in this study.Among 5893 points, 937 locations are for ridges, 1330 locations are for shoulders, 1771 locations are for slopeback, 1410 locations are for footslope and 454 locations are for valleys.The basic statistics of sampling points are shown in Table 1.
All these variables were calculated using software SimDTA-V1.0.3 and ArcGIS9.3.Path analysis is the supplement and extension of regression analysis (Batjes, 1996).Multivariate linear regression equation was developed to achieve path analysis in SPSS18.0.MLR equation can be expressed as (Du and Chen, 2012):  [ ] where, N(h) is the number of pairs of sample points separated by h, h the lag distance and z(x i ) the value of the variable z at location of x i .As the most suitable model of semi-variogram to obtain the semivariogram in this study, Spherical model is described as: ( ) where, C 0 = The nugget value C 1 = The base value and a the distance parameter Trend and anisotropic analysis in ArcGIS 9.3 were used to compare SOC densities in South-North (SN), Southeast-Northwest (SE-NW), West-East (WE) and Northeast-Southwest (NE-SW) direction.The semivariogram value were calculated and exported in GS+9.0 and the theoretical variogram and test fitting effect were obtained by using matlab7.0 to fit semivariogram in each direction.

RESULTS
Descriptive statistics: The Kolmogorov-Smirnov test was used to analyze the level of the variables conformance to a normal distribution.The results (Table 2) showed that the values of skewness and kurtosis are close to 0 and Kolmogorov-Smirnov test (p = 2.9) at a significance level of higher than 0.05, implying that the data conforms to normal distribution.The standard deviation, basic statistical means are shown in Table 2. SOC concentration ranges from 0.01 to 7.60 kg/m 2 , with the arithmetic mean of 2.91 kg/m 2 .We integrated the standard deviation and arithmetic mean to obtain the coefficient of variation and the result shown SOC has a relatively moderate C.V. (38.49%).

Correlation analysis:
The results of the correlation analysis between SOCD and the variables are shown in Table 4.According to the results, there were highly significant (p<0.01)correlations of SOC with all the selected topographic and soil parent material properties, as well as between these properties.SOCD was positively correlated to elevation, slope, topographic wetness, footslope and valley, but negatively correlated to ridge, slope shoulder, slopeback and gray-brown purple mud (shale) efflorescence.Extremely significantly positive correlations were observed between slope and reddish brown thick mudstone (r = 0.036, p<0.01), elevation and ridge (r = 0.085, p<0.01).Among the negative correlations, topographic wetness index and elevation was highest (r = -0.99,p<0.01).
Stepwise regression: Relationships of SOCD with topographic and soil parent material properties were obtained by multiple linear regression analysis with the stepwise method.Taken seven factors as dummy variables into regressions formula 5.The results ( Path analysis: Path analysis (cf.Table 7) indicated that valley (b = 0.8630, R 2 = 0.1253) had the highest value of positively direct path coefficient and determination coefficient, followed by slope back (b = 0.4560, R 2 = 0.2079), however their indirect path coefficients and determination coefficients through other variables were lower.By contrast, elevation (b = 0.354, R 2 = 0.1253) and topography wetness index (b = 0.054, R 2 = 0.0.0029) had higher values of positive direct path coefficients, whereas others had negative direct path coefficients.
As shown in Table 4, elevation had extremely significantly positive correlations with SOCD, indicating SOCD increases as the increase of elevation, but an extremely significantly negative correlation with topographic wetness index, meaning that the soil moisture content in the high altitude areas is less.Path analysis (Table 7) showed elevation (b = -0.035)and SOCD had negative indirect path coefficients through topographic wetness index, as well as topographic wetness (b = -0.0053)index through elevation.
Purple sand and mud (shale) efflorescence (Pmp) and gray-brown purple mud (shale) efflorescence  ----------------------------------------------------------------------- (Pmm) had inhibitory effect on SOCD in correlation analysis, but the former did not reach a significant level.As far as both in the regression equation was concerned, inhibitory effect of purple sand and mud (shale) efflorescence (R 2 = -0.0017)was greater than gray-brown purple mud (shale) efflorescence (R 2 = -0.0006),this could be explained by the inhibitory effect strengthened in Pmp↔Spv→y path.8. Different sill value and the range of the theoretical semivariogram model (Table 8) suggested a zonal anisotropy of SOC density in the study area.The ratio of nugget value and base value in the four directions ranged from 25% to75%, indicating a strong spatial dependence of SOC density in the four directions.In the NE-SW (45°) direction (Sill is 1.174, Range is 8.7 km) the spatial variability of SOCD was the strongest and the spatial correlation function of the sample points range reached to the minimum.The spatial variability of SOCD in NE-SW (45°) and N-S (0°) directions were significant.In the W-E (90°) direction (Sill is 0.58 and Range is 10.25km) the spatial variability of SOCD was the weakest and the spatial correlation function of the sample points range is the maximum.Compared to the W-E (90°) direction, the spatial variability of SOCD in SE-NW (135°) was greater.

DISCUSSION
Effect of altitude and TWI on SOCD: SOC density increased obviously as the increase of elevation in this study is in agreement with the results of Xianfu Cheng and Tan (Chen and Xie, 2009;Chuai et al., 2012).There were highly positive significant correlations of SOC density with elevation, topography wetness index and valley; however topography wetness index and the distribution probability of valleys decreased with increasing elevation.A similar finding was made by Wang et al. (2012) in North China, nonetheless, such a result was surprising, elevation had a critical effect on SOC density and less susceptible to interference from other variables.This may be explained by indirect determination coefficient of SOC density through topography wetness index and valley was -0.0701 and -0.0496, respectively.Direct determination coefficient of SOC density, by contrast, was 0.1253.Other studies (Xu et al., 2010) showed that the increase of altitude lead to a decrease of temperature, reducing the decomposition of organic carbon, which was also one of the reasons in favor of organic carbon accumulation.Further study will take the temperature into account, which is ignored in this study.Topographic wetness index had highly significantly positive correlation with SOC density (0.256**) in this research, which consist with research results of Miao et al. (2010), who reported soil organic carbon content is high in the great soil moisture area lately.Meanwhile, probability distribution of the valley was large in the high topography wetness area, therefore, the accumulation of SOC may be explained by the repetition between topography wetness and valley.However, there was a contrasting result in our study; total decision coefficient of topographic wetness (0.023) on SOC density was not great.It was not difficult to find that the effect of topographic wetness index through TWI↔Spv→y path on soil organic carbon density (0.0321) was greater than the direct effect on organic carbon (0.0029).
Effect of soil parent material on SOCD: Purple sand and mud (shale) efflorescence had not shown significant correlation with SOC density (r = -0.013)and gray-brown purple mud (shale) efflorescence highly significant correlation (r = -0.032*).However, purple sand and mud (shale) efflorescence expressed a greatly negative impact on SOC density.A reasonable explanation was given by the multivariate statistical analysis.In the Pearson correlation, purple sand and mud (shale) efflorescence had significant positive correlation with elevation (positively correlated with SOCD), but negative with clouds weathered shale, limestone weathering and river alluvium (positively correlated with SOCD).Conversely, significantly negative correlation was found between gray-brown purple mud (shale) efflorescence and elevation and cannot be offset by the negative correlation with dolomite weathered material, limestone weathering and river alluvium.As the mainly parent material of purple (Fig. 4) soil in the study area, gray-brown purple mud Fig. 4: Maps of parents materials (shale) efflorescence developed into thick bony sandy soil may result in more soil erosion, leading to SOC decline also (Martin et al., 2010).
In all soil parent material, the maximum average SOC density (3.07kg/m 2 ) distributed in dolomite weathering area.As one of the main soil parent material, dolomite weathered material weathered into yellow soil by the impact between mountainously cold climate and coniferous forest.Level of Yellow Soil claying is not deep and organic matter accumulates a lot, therefore, soil organic carbon density is relatively high.
Effect of slope position on SOCD: Slope position contains different characteristics of soil properties in different topographic positions and becomes an important factor in geographical or ecological process model (Milne, 1934;Guo et al., 2010).Slope position was divided into three categories in the research of Guo et al. (2010) and Lu et al. (2013), who reported SOC content is highest on the bottom of the slope and lowest at the top of the slope lately.In our study, slope position was divided into five categories and results are in agreement with the consequences of previous studies.Moreover, the results suggest that the classification in our study can clarify the influence mechanism of SOC density and slope positions more accurately.
Correlations between SOC densities and slope position are found to be positive in valley and footslope, while negative in ridge, slopeback and shoulder.This is necessarily an indication that valley and footslope increased SOC accumulation, but others decreased.Slope and reddish brown thick mudstone shows extremely significant positive correlation with SOC density, implying palm red thick mudstone mainly distributed in relatively steeper slope.As the maximum in direct impact coefficients, valley on SOC density is 0.8630.It can be assumed that relatively soil erosion may result in redistribution of soil and water, leading to runoff and sediment rom the top of slope position accumulation in low-lying valley area.Consequently, more SOC concentrated in bottom of trench.

CONCLUSION
Overall, our study has revealed the average SCOD is 2.91 kg/m 2 , below the national average level of cultivated land (3.0 kg/m 2 ) and the total SOC storage is 1838.75×106kg on 0-20cm surface soil in the study area.High-density areas of SOC tend to distribute in limestone soil and yellow soil area.Clearly, high elevation, high topographic wetness, valley and dolomite weathered material have led to SOC density accumulation, in which elevation has the strongest relationship with SOC density.Reforestation, terracing, more organic fertilizer may necessary to increase SOC in low altitude, ridge or shoulder area, also no tillage and reduced tillage planting mode may necessary to increase agricultural soil carbon sequestration capacity.Likewise, the spatial anisotropic analysis have shown that spatial variability in NE-SW (45°) direction is strongest and the spatial correlation function of the sample points range reached to the minimum, followed by N-S (0°), however, weakest and maximum in the W-E (90°) direction.Further refinement of the quantitative qualitative variables is needed to express a certain variable on SOC density.

Fig. 2 :
Fig. 2: Maps of DEMPath and geo statistical analyses: Qualitative variables were converted to binary variables(Vallejo et al., 2008) by using Alejandra function before the correlation analyst.Path analysis is the supplement and extension of regression analysis(Batjes, 1996).Multivariate linear regression equation was developed to achieve path analysis in SPSS18.0.MLR equation can be expressed as(Du and Chen, 2012): organic carbon density x i = The Impact factor a = The constant b i = The regression coefficient ε = The error Geo statistics uses the semi-variogram to quantify the random and structured spatial variation of a regionalized variable and relevant statistical analysis methods to analyze the spatial distribution.The semivariogram is depicted as follows:

Geostatistical analysis :
There was a strong spatial anisotropy of soil organic carbon density in the study area from overall trend analysis.Four directional semivariograms which took range and semi-variogram as coordinates obtained by scatter diagram and spherical model were shown in Fig. 3. Nugget was the intercept of fitted curve and vertical axis given by Spherical model, as well as the base value.The variation as the step of the base value, Theoretical semivariogram model parameters for different directions were shown in Table

Table 1 :
Descriptive statistics of categorical auxiliary in the study area

Table 2 :
Classical statistical parameters of the SOCD in the study area

Table 3 :
Descriptive statistics of continuous auxiliary variables in the study area

Table 5
) of variance analysis (ANOVA) which sum of squares of regression and residua was 4127.73 and 3301.40,respectively, F statistic was 817.3 (p<0.05)indicated

Table 7 :
Results of path analysis -

Table 8 :
Theoretical semivariogram model parameters for different directions