Spatial Variability Of Soil Detachment Capacity In A Small Watershed On The Hilly Loess Plateau

Soil erosion involves the processes of soil detachment, sediment transport, and deposition. Soil detachment is the removal of transportable materials from the soil mass by eroding agents, providing loose, non-cohesive sediment for subsequent transport and deposition. Soil detachment capacity is strongly influenced by soil properties since it occurs on the interface of flowing water and soil. Most soil properties exhibit significant spatial variations, which may lead to the spatial heterogeneity of soil detachment capacity by overland flow. Nevertheless, few studies have been conducted to detect the spatial variability in soil detachment capacity, and its potential influencing factors are still not clear.606 undisturbed soil detachment samples were collected from 202 sampling sites in ephemeral gully developed hillslope and 696 samples were taken from four different land uses on the red loess soil and six different land uses on the yellow loess soil in a small watershed of the Loess Plateau. To analyse the potential influencing factors that control rill erodibility, particle size distribution, bulk density, cohesion, water stable aggregate, organic matter content, root mass density, and litter mass density were also measured for each site. Soil detachment capacity was measured in a sand-glued hydraulic flume with adjustable bed gradients. Traditional statistics methods(ANOVA, paired T test, principal component analysis, correlation analysis, minimum data set method, linear and nonlinear regression), geostatistics methods(semivariogram and kriging interpolation), and state-space approach were used to detect the effect of ephemeral gully, soil type and land use on soil detachment capacity, and thus clarify the spatial variation in soil detachment capacity and rill erodibility and its potential influencing factors.(1) The study revealed the effect of ephemeral gully on soil detachment capacity, and showed the spatial variation in soil detachment capacity and its influencing factors. Soil detachment capacity varied within a wide range from 0.0004 to 1.25 kg m-2 s-1, with a mean value of 0.22 kg m-2 s-1. Coefficient of variation showed a strong variability in soil detachment capacity. Soil detachment capacity differed significantly between initial, upper, middle and lower slope, which was probably caused by the original soil distributions produced by erosion before ephemeral gully formed. Semivariogram of soil detachment capacity indicated a moderate spatial dependence. Sampling interval has a significant effect on the spatial patterns of soil detachment capacity. When the sampling interval decreased, nugget variance decreased, while structured variance, spatial dependence, and range increased. Distribution maps of soil detachment capacity, derived by Kriging interpolation, showed that samples in lower slope positions have statistically greater soil detachment capacity than the middle, upper, and initial slope positions. Significant correlations were found between soil detachment capacity and clay content, sand content, median soil grain size, bulk density, cohesion, water stable aggregate, and litter mass density. Principal component analysis(PCA) and a minimum data set(MDS) method identified that the median soil grain size, bulk density, and litter mass density were the principal factors to explain most of the spatial variability in soil detachment capacity.(2) Soil detachment capacity could be well predicted by state-space model at 2 m and 10 m intervals. Soil detachment capacity was positively correlated with sand content and soil organic matter(P<0.01) at 2 m interval, while it was only positively correlated with sand content at 10 m interval; soil detachment capacity both showed negative correlations with clay content, bulk density, cohesion, water stable aggregate and litter mass density at 2 m and 10 m intervals. At 2 m interval, all the variables were significantly autocorrelated, and soil detachment capacity was significantly cross-correlated with other 8 variables except silt content. However these spatial structures were changed at 10 m interval. Silt content, soil organic matter and litter mass density don’t manifest a significant spatial autocorrelation. Additionally, soil detachment capacity was merely significantly cross-correlated with clay, sand content, bulk density and cohesion. The state-space approach described the spatial pattern of soil detachment capacity much better than the equivalent classical regression methods both at 2 m and 10 m intervals and it predict soil detachment capacity better at 2 m than 10 m interval. State-space model was recommended as a useful tool for predicting soil detachment capacity at ephemeral gully developed hillslope on the Loess Plateau.(3) Both soil type and land use had significant effects on soil detachment capacity. For two tested soils, the mean soil detachment capacity of the yellow loess soil was 1.49 times greater than that of the red loess soil. For the red loess soil, soil detachment capacity of cropland was the maximum, which was 5.57, 5.85, and 34.08 times greater than those of shrub land, orchard, and grassland, respectively. For the yellow loess soil, cropland was much more erodible than other five land uses. On average, the ratios of the cropland soil detachment capacity to those of orchard, shrub land, woodland, grassland, and wasteland were 7.14, 12.29, 25.78, 28.45, and 46.43, respectively. The variability of soil detachment capacity under different land uses was closely related to soil properties, root systems, and tillage operations. Soil detachment capacity was positively related to shear stress, stream power and silt content, and inversely related to sand content, cohesion, water stable aggregate, aggregate median diameter, organic matter, and root density. The measured detachment capacity could be well estimated by measurable parameters of stream power, slope gradient, soil bulk density, median diameter, silt content, cohesion, and root density(R2=0.89, NSE=0.89).(4) Rill erodibility was affected by soil type and land use significantly. The average rill erodibility of yellow loess soil was 1.5 times greater than that of red loess soil. For the red loess soil, cropland had the maximum rill erodibility and followed by orchards, shrub land, and grassland. For the yellow loess soil, cropland also had the maximum rill erodibility, which was 9.31, 11.68, 26.32, 29.10 and 42.41 times greater than those of orchards, shrub land, woodland, grassland, and wasteland, respectively. Rill erodibility was considerably influenced by silt content, soil cohesion, water stable aggregate, soil organic matter, and root mass density. Rill erodibility increased with silt content, and decreased with soil cohesion, water stable aggregate, soil organic matter, and root mass density. In WEPP model, equations for rill erodibility underestimated the measured rill erodibility. A nonlinear regression showed that rill erodibility could be estimated well(R2=0.86, NSE=0.83) by soil cohesion, water stable aggregate, and root mass density in the Loess Plateau.(5) From 1938 to 2010, the average rill erodibility of Zhifanggou watershed increased rapidly first and then decreased gradually. Rill erodibility exhibited significant temporal variations. There was a second-degree parabola existed between rill erodibility and years. The rill erodibility increased significantly from 0.047 s m-1 in 1938 to 0.192 s m-1 in 1958, while it decreased to 0.039 s m-1 in 2010. Over the whole watershed, rill erodibility was very low, and the severe erosion area is only sporadic distribution in 1938. From 1958 to 1978, the rill erodibility was high, espectially in the upstream of watershed. In 1987 and 1990, the rill erodibility of the downstream changed very little compared to those in 1978, but rill erodibility of the upstream decreased to some extent; in 1999, due to the project of “Grain for Green” was implemented to reduce soil erosion on cropland, the eco-environment was gradually improved and the rill erodibility in general was low all over the watershed. Land use adjustment had strong effects on the temporal variation in rill erodibility. Rill erodibility increased as the area of cropland increased, while it decreased with shrubland and woodland increased. Thus, cropland is considered as a principal sediment source, and shrubland and woodland may be the excellent land management strategy for conserving soil source in the Loess Plateau.