3 MATERIALS AND METHODS
3.4 SOIL EROSION MODELING
3.4.2 MODEL PARAMETERIZATION
Model parameterization of the RUSLE was done for two first model runs on the soil erosion risk potential under natural conditions, and under former and current land use conditions. This enabled to analyze to what extent the Xiangxi catchment is prone to soil erosion without any human influence, and to analyze the effect of the land use change associated to the TGD on the soil erosion risk potential.
The soil erosion risk potential under natural conditions - or natural disposition to soil erosion - was estimated following (Eq. 1), however, without consideration of any soil conservation measure as the Xiangxi catchment was assumed to be undisturbed by human activities. Contrary to the approach used by DA SILVA ET AL. (2011) to predict the natural erosion potential for the Brazilian territory without taking into account potential natural vegetation, the Xiangxi catchment was assumed to be completely covered with woodland. Large woodland areas without any human impact resulting from agriculture or settlement can recently be observed for Shennongjia region in the northwestern headwater zone of the Xiangxi catchment (Figure 10). Moreover, based on experimental plot studies in a Chinese subtropical forest ecosystem, GEIßLER ET AL. (2010, 2013) proved the erosive power of throughfall drops and a distinct impact of forest vegetation on soil erosion. Thus, the C factor was set to 0.005 as proposed by LIU and LUO (2005) for woodland in the TGA.
The natural soil erosion risk potential in the Xiangxi catchment was calculated using (Eq. 2): Anat = R × K × LS × Cwood (Eq. 2)
where Anat is the average annual natural soil erosion risk potential (t ha -1
a-1) and Cwood refers to
the crop and management factor (dimensionless; -), however, is here dealt as vegetation cover.
The combined topographic, DEM-based factor LS was derived from pre-processed separate grids on S and L using the RUSLE approach according to RENARD ET AL. (1997) that was also applied by KONZ ET AL. (2009) for complex steep sloping areas using (Eq. 3) and (Eq. 4).
with
, and
for slopes < 5.14° (Eq. 4a) for slopes ≥ 5.14° (Eq. 4b) where θ is the slope gradient (°), λ is the real slope length referring to the flow length (m), m is the slope length exponent (-), and β is the susceptibility to rill erosion (-). Therefore, the S factor was derived from a slope grid based on the steepest slope algorithms introduced by TARBOTON (1997). Before computation of the L factor, the upstream flow length (m) was derived using Monte-Carlo- aggregation to simulate flow divergence reflecting a more natural spatial flow pattern in complex landscapes (BEHRENS ET AL., 2008).
Due to the facts that at the time of the first model runs precipitation data on a daily basis were only available for one climate station (Xingshan station near Gaoyang; Figure 4) and that the precise calculation of the R factor necessitates consistent long-term data with a high temporal resolution to assess single storm events (WISCHMEIER and SMITH, 1978; RENARD ET AL.,1997), a single R value was taken from literature. The used value on rainfall erosivity is R = 2,880 MJ mm ha-1 h-1 a-1 (SHI ET AL., 2004). It was derived for the mountainous Wangjiaqiao watershed in Zigui County 50 km northwest of the TGD. The value is based on the method by WISCHMEIER and SMITH (1978) and is assumed to correspond to the climate regime of the Xiangxi catchment.
As the classification of the soil texture throughout the SNSS (Table 1) refers to the Russian classification system according to KACHINSKY (1965) based on the ratio of particle sizes smaller than 0.01 mm, the K factor was calculated according to SHIRAZI andBOERSMA (1984) using (Eq. 5):
(Eq. 5)
where Dg is the geometric average particle size. The equation is based on the symmetric
Gaussian distribution of geometrically average particle sizes and was recommended by SONG ET AL. (2005) for Chinese soil textures and for short enough data. The K factor was calculated for the topsoil data of each of the 126 available soil profiles from the SNSS (Table 1) using the original soil particles data.
Aiming at the TGD-induced impact of the land use changes on the soil erosion, the soil erosion risk potential in the Xiangxi catchment was estimated using (Eq. 1) for agricultural land (c.f., Section 3.4.1). Therefore, the two land use classifications from 1987 and 2007(Table 1, c.f., Section
3.3) were used for model parameterization. Both classifications incorporate the same land use classes (Table 3; SEEBER ET AL., 2010).
For the first model set up, values on the C factor were taken from literature, as no previous adequate data were available and monitoring on specific vegetation and crop parameters in the study area was considered not practical and unfeasible. Those C factor values refer to the land use classes depicted for the Xiangxi catchment by SEEBER ET AL. (2010) as shown in Table 3 and have been assigned to the according land use pixel on the catchment scale. They represent common seasonal crop rotations within one year for subtropical agriculture in the Wangjiaqiao and Taipingxi watersheds in the TGA (SHI ET AL., 2004; LIU and LUO, 2005) close to the Xiangxi catchment. Thus, they are supposed to adequately represent the conditions in the study area. Contrary to the original purpose of the USLE to predict soil erosion solely on agricultural land (WISCHMEIER and SMITH, 1978), the potential soil loss in the Xiangxi catchment was modeled including all land use classes representing soil vegetative cover (i.e., woodland and grassland; Table 3) in order to assess the complete land cover-induced range of soil erosion. The classes 'bare ground' and 'reservoir' (Figure 10) were not considered. Assuming the settlements in the Xiangxi catchment (rural and urban) to be associated with small-scale home gardening - as often observed during the field campaigns from 2008 to 2010 - the class 'built-up' was also parameterized (Table 3). Generally, the higher the C factor values, the less the crop and vegetation cover and thus, protection against soil erosion.
Table 3 C factor values for model parameterization of the RUSLE for the land use in the Xiangxi catchment. The land uses refer to the classification by SEEBER ET AL. (2010).
Land use class C factor value Reference
Arable land 0.46 LIU and LUO (2005)
Orange orchard* 0.13 SHI ET AL. (2004)
Paddy field** 0.18 LIU andLUO (2005)
Steep garden plots** 0.1 LIU andLUO (2005)
Woodland 0.005 LIU and LUO (2005)
Grassland 0.2 ERENCIN (2000)
Built-up land 0.08 LIU and LUO (2005)
* The land use class 'orange orchard' refers to 'garden land' in the land use classification by SEEBER ET AL., (2010; c.f., Section 3.2.5). ** Only parameterized in the prediction of the soil erosion risk potential in Manuscript 2.
For both calculations, the further RUSLE factors K and R were kept constant. The LS factor was calculated with (Eq. 3) and (Eq. 4) using a slope grid based on the steepest slope algorithm introduced by TARBOTON (1997) as done in estimating Anat. The L factor also refers to the upstream
flow length (m) that was derived using Monte-Carlo-aggregation. However here, the flow lengths were calculated grid-based from the DEM individually for each agricultural class considered in the model parameterization according to the separate land use classification from 1987 and 2007. Doing so allows for simulating flow divergence reflecting the spatial agricultural pattern.
Since dry-stone walling bench terraces are common features in the Xiangxi catchment (c.f., Section 3.2.5), terraces were considered in the estimation of the soil erosion risk potential under former (1987) and current (2007) land use with the P factor assigned to the land use classes 'arable land' and 'orange orchard' (Table 3). As for the first model runs no data on the spatial distribution and design of bench terraces were available, data on the supporting conservation practices refer to rough estimates. Thus, P was set to 0.55 as suggested by SHI ET AL. (2004) and LIU and LUO (2005) for dry- stone walling level bed terraces on land sloping between 20° and 25°.
The grid-based estimates on the Anat, and A1987 and A2007 were then classified according to the
Chinese Soil Erosion Rate Standard (Technological Standard of Soil and Water Conservation SD238- 87). Thus, five numerical classes issued by the Ministry of Water Resources (LU and HIGGITT, 2000; XU ET AL., 2008; XU ET AL., 2009) give information about the soil erosion risk potential based on the calculated average annual soil loss.
The spatial resolution of all soil loss estimations in the Xiangxi catchment, the Backwater area, the sub-catchments Xiangjiaba and Quyuan is 45 m × 45 m as this is the spatial resolution of the database (c.f., Section 3.3).