spatiotemporal dynamics of soil erosion risk for anji county, china
TRANSCRIPT
ORIGINAL PAPER
Spatiotemporal dynamics of soil erosion risk for Anji County,China
Zhenlan Jiang • Shiliang Su • Changwei Jing •
Shengpan Lin • Xufeng Fei • Jiaping Wu
Published online: 24 April 2012
� Springer-Verlag 2012
Abstract Soil erosion, as a serious environmental prob-
lem worldwide, poses a great threat to human sustainabil-
ity. Spatiotemporal information on soil erosion is of vital
importance to finding a solution for this problem. A case
study was conducted to characterize the dynamics of soil
erosion risk in 1985, 1994, 2003 and 2008 for Anji County,
China, a region with seemingly high ecological quality.
Remote sensing and geographic information systems were
integrated to parameterize soil erosion-controlling factors.
By using the Revised Universal Soil Loss Equation, we
estimated annual soil loss, and generated categorical maps
of soil erosion risk in the County for the 4 years. Results
showed that, while appearing to improve in some areas,
soil erosion risk increased and eroded area expanded from
1985 to 2008. Spatial analysis revealed that the most vul-
nerable hotspots were erosion-free forests, where newly
eroded areas were most likely to occur. These results
implied that, similar to findings in many parts of the world,
soil erosion is an important issue in the study area, which
could be closely associated with local eutrophication and
algal blooms. Our research indicated that there should be
more focus on this issue. From a methodological point of
view, we believe that the approach used to estimate soil
loss in the study area has the potential to be applied in other
similar regions.
Keywords Soil erosion � RUSLE � Spatial variation �Temporal changes � Remote sensing � Geographic
information systems
1 Introduction
Soil erosion is one of the most serious environmental
problems in the world, given its negative impacts on agri-
cultural production, natural resources and the environment
(Onyando et al. 2005; Rahman et al. 2009). It is one of the
most significant forms of land degradation that is greatly
influenced by land use and management (Rey 2003; Jiang
et al. 2008). Soil erosion is closely associated with all three
major environmental challenges: water contamination, air
pollution and land degradation. Areas affected by soil ero-
sion account for 3.6 9 106 km2 in China, covering
approximately 37 % of the total national territory (Fu 2008).
The government has long been focused on erosion hotspots
suffering major soil-erosion problems, such as the middle
and upper reaches of the Yellow River (especially the Loess
Plateau) or the Yangtze River (Fan et al. 2004; Huang et al.
2007; Ouyang et al. 2010; Yang et al. 2011). Other areas,
however, such as the red soil region of Southeast China that
experiences severe soil and water losses due to high tem-
peratures and precipitation, mountainous and hilly land-
forms, and various large-scale construction activities (Deng
et al. 2010; Liang et al. 2010), usually attract scant attention
for management and conservation.
Selected from three nominated national eco-counties,
Anji, a typical red soil region, became the first, and thus far
the only, eco-county in East China. The term ‘‘eco-county’’,
highlighting harmony between economic development and
sustainable eco-environment, was proposed to facilitate the
implementation of China’s ‘‘new countryside’’ policy, one of
Z. Jiang � S. Su � C. Jing � S. Lin � X. Fei � J. Wu (&)
College of Environment and Natural Resources,
Zhejiang University, Hangzhou, China
e-mail: [email protected]
Z. Jiang
Department of Geographic Science, Minjiang University,
Fuzhou, China
123
Stoch Environ Res Risk Assess (2012) 26:751–763
DOI 10.1007/s00477-012-0590-0
the primary objectives of its 11th Five-Year (2006–2010)
Plan (Long et al. 2009). To meet this goal, many counties in
rural China have been adjusting their development plans to
the ‘‘Anji model’’. Therefore, characterizing spatiotemporal
dynamics of soil erosion in Anji can provide insight into
prospective soil and water erosion control in rural China, and
a case study of the County can serve as an example for the
scientific planning, management and development of new
countryside in China.
To draw public and governmental attention to soil ero-
sion, scientists should provide information about distribu-
tion patterns and dynamics of soil erosion, as well as its
ongoing processes (Tian et al. 2009; Wang et al. 2009). A
wide array of models can be applied for soil erosion esti-
mation, including the universal soil loss equation (USLE),
modified universal soil loss equation (MUSLE), revised
universal soil loss equation (RUSLE), European soil ero-
sion model (EUROSEM), etc. Among them, RUSLE, an
empirical soil erosion model designed on the basis of
USLE, has become the most frequently used model
worldwide, owing to its user-friendliness (Ismail and
Ravichandran 2008; Xu et al. 2008b). It can be used to
predict erosion rate of ungauged watersheds through
watershed characteristics and local hydro-climatic condi-
tions (Angima et al. 2003), and can also be applied to
obtain the spatial heterogeneity of soil erosion.
However, RUSLE is focused on estimating soil loss on
the basis of the water erosion ‘‘status quo’’. The prediction
of long-term soil erosion trend requires many years of
consistent and repeatable data (Gao and Liu 2010). This is
difficult in most cases, due to the dearth of such data from
the preceding decades (Masoudi et al. 2006). In these cases,
existing models cannot be directly applied in soil erosion
prediction (Van Rompaey and Govers 2002; Shrestha et al.
2004). In this study, soil erosion risk, which usually indi-
cates the relative incidence of erosion at a certain location
as compared to that of other locations in the mapped region
(Le Bissonnais et al. 2002), was used to provide information
about distribution patterns and dynamics of soil erosion and
its ongoing processes, because when planning land con-
servation practices, the spatial distribution of soil erosion
risk is more important than the absolute values of soil
erosion loss (Ma et al. 2003).
Soil erosion risk assessment involves comparing between
the present and past status to find out the incidence and rate
of degradation in the future. However, due to the lack of
quality geospatial data, an available model that fits locally,
and general public awareness, the task of soil erosion control
has long been delayed in the region. With the goal of
developing an approach for soil erosion risk assessment and
applying it in Anji County, we built up geospatial databases
and integrated RUSLE with remote sensing and geographic
information system (GIS) techniques. Specifically, remote
sensing was used to provide ground information because of
its regular revisiting capability, GIS was used to collect,
store, display, manipulate, and analyze various spatial data
and develop a spatial model, and RUSLE estimated soil
erosion risk. Thus, as a case study, Anji County, which has
great social and ecological significance in southeastern
China, was selected to characterize the spatiotemporal
dynamics of soil erosion risk.
2 Study area
With a total area of 1,886 km2 and a population of about
455,100, Anji County lies in the center of the Yangtze River
Delta, Zhejiang Province, China between 30�230 and
30�530N and 119�140–119�530E (Fig. 1). It covers more than
70 % of the West Tiaoxi watershed, the upper reach of the
Taihu Lake. Featuring a typical maritime monsoon climate
of the subtropical zone, the County has pleasant weather
with abundant sunshine and rainfall, as well as four distinct
seasons (Fig. 2). The yearly average temperature is 15.5�C
and the yearly average rainfall 1,400 mm. The dominant
soils are red soil, paddy soil, yellow soil, fluvo-aquic soil and
regosols (Zhejiang Office of Soil Survey 1985). The Coun-
ty’s topography is characterized by high elevation and steep
terrain in the southwest and relatively low and flat land in the
northeast. Its altitude ranges from 1,587 m to near 0 m
(above sea level). About 71 % of land surface has forest
coverage, with half being bamboo forests. Most of the for-
ested areas are located at the periphery of the county, while
the farmlands are primarily on a large plain in the middle.
During the 1950s and 1960s, the County fortunately
avoided the catastrophic impact of the ‘‘Steel Refinery
Movement’’, which prompted the felling of numerous trees
for firewood used in steel refining (Li et al. 2006). The
well-preserved ecological environment and tourist resour-
ces helped it become popularly known as ‘Bamboo Town’,
‘Land of White Tea’, ‘China’s First Ecological County’
and ‘Home of the Giant Panda’. It became more popular
when chosen as an important setting for the epic martial
arts movie ‘Crouching Tiger Hidden Dragon’, the 2001
Oscar winner for best foreign language film.
In 2000, Anji was designated to pilot a national program
on developing eco-friendly agriculture, industry and tour-
ism at the county level. On June 5, 2006, the 35th World
Environment Day, it was officially listed as a ‘‘National
Eco-County’’ by the Ministry of Environmental Protection,
China. Counties in rural China have followed its model of
planning and development, dubbed the ‘‘Anji model’’, to
facilitate their efforts in building ‘‘new countryside’’,
which were characterized with increased production, well-
off life, civilized rural customs, clean and tidy villages, and
democratic management (SCPRC 2006).
752 Stoch Environ Res Risk Assess (2012) 26:751–763
123
3 Methods
3.1 Data sets and preprocessing
Four satellite images were collected in the study, including
three Landsat Thematic Mapper (TM) images (30 m spatial
resolution) acquired in November 1985, May 1994 and
March 2003, and one Advanced Land Observation Satellite
(ALOS) image (10 m spatial resolution) acquired in
November 2008. They were used to derive vegetation frac-
tion cover (VFC) and prepare land use/land cover (LULC)
maps. Other data were digital topographical maps at a scale
of 1:10 000, a Digital Orthophoto Map (DOM) with 1 m
resolution, a soil survey geographic database originally from
soil field survey map at a scale of 1:50 000, and the digital
administration boundary of the study area. All the data were
geometrically rectified to the Universal Transverse Mercator
(UTM) projection system, zone 50 East.
Next, radiometric correction was performed on the
images in two steps. First, remotely sensed digital number
(DN) values were converted to at-sensor radiance (Lk) by
Eq. 1, which is based on the maximal and minimal spectral
radiance value for each band (Chander et al. 2009; Lim
et al. 2009).
Fig. 1 Location of Anji
County, Tiaoxi watershed,
China
Fig. 2 Monthly average rainfall
and temperature in Anji County
Stoch Environ Res Risk Assess (2012) 26:751–763 753
123
Lk ¼ Gaink � DNk þ Biask ð1Þ
where Lk is spectral radiance at the sensor’s aperture
(W m-2 sr-1 lm-1), Gaink is band-specific rescaling gain
factor (W m-2 sr-1 lm-1 DN-1), DNk is spectral digital
number, and Biask is band-specific rescaling bias factor
(W m-2 sr-1 lm-1). The gain and bias values are unique
for each spectral band acquired by a particular sensor.
Second, the at-sensor spectral radiance (Lk) was con-
verted to the top-of-atmosphere (TOA) reflectance, which
is required for quantitative analysis of multiple images
acquired on different dates. With regard to historical ima-
ges, in situ measurements of atmospheric parameters are
impossible to obtain for physically-based models. The
apparent reflectance model (Eq. 2) was selected to calcu-
late the TOA reflectance, because it does not call for any in
situ field measurements (Lu et al. 2002).
qk ¼pLkd2
ESUNk cos hSð2Þ
where qk is planetary TOA reflectance, p is a mathematical
constant equal to *3.141592, Lk is the same as Eq. 1, d is
Earth–Sun distance (astronomical units), ESUNk is the
mean exoatmospheric irradiance (W m-2 lm-1), and hS is
sun zenith angle (or 90-sun elevation angle) (degrees).
3.2 RUSLE
In the model of RUSLE, the mean annual soil loss is
expressed as a function of six erosion factors (Renard et al.
1997):
A ¼ R� K � LS� C � P ð3Þ
where A is the estimation of average annual soil loss
(t ha-1 year-1), R is the rainfall erosivity factor
(MJ mm ha-1 h-1 year-1), K is the soil erodibility factor
(t ha h ha-1 MJ-1 mm-1), LS is the combination of the
slope steepness and slope length measurements (unitless),
C is the cover-management factor (unitless), and P is the
support practices factor (unitless).
Of these factors, soil, topography and climate charac-
teristics are subject to natural conditions, while vegetation
and LULC are susceptible to human activities (Solaimani
et al. 2009); the latter are the most important factors,
because they have the greatest effect and can more easily
be changed to control soil loss and sediment yield (Ismail
and Ravichandran 2008; Meusburger et al. 2010).
RUSLE provides two options for erosion rate estimation
(Bahadur, 2009). The first, entitled ‘‘potential erosion’’, is
computed on the basis of only three factors, namely, R, K
and LS, while the second, entitled ‘‘actual erosion’’, is
computed with C and P factors taken into account in
addition to the three used for potential erosion. Therefore,
we treated the first three factors as the background factors,
i.e. RUSLE uses the same values of R, K, and LS for 1985,
1994, 2003 and 2008. The cover-management factor
(C) and the support practices factor (P) were taken as
dynamic factors, because they were driven by human
activities (Van Rompaey et al. 2001; Solaimani et al. 2009;
Di et al. 2010).
Average annual soil loss maps were generated using the
aforementioned RUSLE. Then, these maps were grouped
into different soil erosion categories followed the Stan-
dards of SL190-2007 (Table 1) (MWR 2007; Zhu et al.
2009). Figure 3 illustrates these steps.
3.3 Soil erosion risk assessment
3.3.1 Background factors
3.3.1.1 Rainfall erosivity factor (R) The R factor is an
index of rainfall erosivity, indicating the probability of
rainfall causing erosion. Renard and Freimund (1994)
developed a set of relationships for calculating the R factor
using mean annual rainfall data and the modified Fournier
index (MFI).
MFI ¼Xi¼12
i¼1
p2i
Pð4Þ
R ¼ 0:07397MFI1:847; when MFI\55 mm ð5Þ
R ¼ 95:77� 6:081MFI þ 0:4770MFI2;when MFI� 55 mm
ð6Þ
where MFI (mm) is the modified Fournier index value
(Arnoldus 1977), Pi (mm) is the mean monthly precipita-
tion of month i, P (mm) is the mean annual precipitation,
and R (MJ mm ha-1 h-1) is the RUSLE rainfall erosivity
factor. Equation 5 was applied to locations with modified
Fournier index values less than 55 mm, and Eq. 6 was used
for locations with modified Fournier index values above
55 mm.
Table 1 Standard for categorizing soil water erosion intensity in
China
Category Description Erosion modulusa Erosion depthb
1 Non-erosion \500 \0.74
2 Slight 500–2,500 0.74–1.9
3 Moderate 2,500–5,000 1.9–3.7
4 Severe 5,000–8,000 3.7–5.9
5 Very severe 8,000–15,000 5.9–11.1
6 Extreme [15,000 [11.1
SL190-2007 (Ministry of Water Resources, China, 2007)a Modulus in t km-2 year-1
b Depth in mm year-1
754 Stoch Environ Res Risk Assess (2012) 26:751–763
123
Practically most current efforts for mapping rainfall
erosivity at large-scale rely on spatial interpolation
approaches, such as kriging, which is based on erosivity
values derived from rain gauge data. However, for data-
poor regions this is not an option (Vrieling et al. 2010).
There is only one long-term meteorological station in Anji
County, so spatial interpolation approach is not applicable
in the study. Moreover, elevation, latitude and longitude
are found to be related to precipitation in the region, so
regression of R-factor with elevation and latitude/longitude
was used to map rainfall erosivity in the study area. At first,
R-factor values of the 11 meteorological stations sur-
rounding Anji County were calculated by using average
monthly precipitation data from 1985 to 2008. Then, a
regression relationship, with R as the dependent variable,
and elevation, product of longitude and latitude as inde-
pendent variables, was established:
R ¼ 58825:535þ 4:090h� 14:091uk ð7Þ
(Determination coefficient = 0.80, P \ 0.01) where h is
elevation in meters, u is latitude in degrees, and k is
longitude in degrees.
By using Eq. 7, a 30 m grid map of R-factor was gen-
erated from a 30 m elevation map, as well as longitude and
latitude information of the County.
3.3.1.2 Soil erodibility factor (K) The K-factor is the rate
of soil erosion per unit erosivity of rainfall, and represents
the effect of the inherent soil properties in soil erosion,
especially those related to the surface layer (Wu et al.
1997; Mati and Veihe 2001). In the Erosion Productivity
Impact Calculator (EPIC) (Williams et al. 1984; Zhang
et al. 2008; Rahman et al. 2009), soil organic carbon and
soil particle size distribution are used to calculate the K
values:
K ¼(
0:2þ 0:3 exp �0:0256SAN 1� SIL
100
� �� �)
� SIL
CLAþ SIL
� �0:3
1:0� 0:25C
C +exp(3:72�2:95CÞ
� �
� 1:0� 0:7SN1
SN1þ expð�5:51þ 22:9SN1Þ
� �ð8Þ
where SAN, SIL and CLA are sand fraction
(%, 0.05–2.00 mm diameter particles), silt fraction
(%, 0.002–0.05 mm diameter particles), and clay fraction
(%, \0.002 mm diameter particles), respectively, C is soil
organic carbon content (%), and SN1 equals 1-SAN/100.
The K value was calculated from the digital soil map.
Each soil type was associated with one K value, assuming
that the same soil type has the same soil properties
throughout the study area.
3.3.1.3 Slope length and steepness factor (LS) The LS
factor, accounting for the effect of topography on erosion
in RUSLE, is the expected ratio of soil loss per unit area on
a field slope to that from a 22.13 m length of uniform 9 %
slope. The slope length factor (L) represents the effect of
slope length on erosion, and the slope steepness factor (S)
reflects the influence of slope gradient on erosion. The
common equation used for calculating LS is an empirical
one by Wischmeier and Smith (1978):
LS ¼ k22:13
� �n
ð65:41 sin2 hþ 4:56 sin hþ 0:065Þ ð9Þ
where k is the slope length in meters, h is the angle of slope
in degrees, and n is a constant dependent on the value of
the slope gradient: n is 0.5 if the slope angle is greater than
2.86�, 0.4 on slopes of 1.72�–2.85�, 0.3 on slopes of 0.57�–
1.72�, and 0.2 on slopes less than 0.57�.
Fig. 3 Flow chart of soil
erosion risk assessment
Stoch Environ Res Risk Assess (2012) 26:751–763 755
123
The RUSLE-based ArcInfo Arc Macro Language
(AML) developed by Hickey et al. (1994) and Van Rem-
ortel et al. (2004) was downloaded from the website:
www.cwu.edu/*rhickey/slope/slope.html for computing
the LS factor. The basic input for generating an LS-factor
map in GIS is 30 m resolution DEM of Anji County, which
was derived from a 1:10 000 digital topographic map.
3.3.2 Dynamic factor
3.3.2.1 Cover-management factor (C) During recent
years, different approaches have been developed for C
factor mapping with the application of remotely sensed
data. Commonly used techniques in erosion assessment are
image classification, spectral indices and spectral unmixing
(Vezina et al. 2006; De Asis and Omasa 2007; Zhou et al.
2008). However, it is not possible to estimate C factor with
high accuracy through image classification alone, which
applies average C factor value to each vegetation type,
causing the smoothing of estimates and disappearance of
spatial heterogeneity and variability (Wang et al. 2002).
The most common spectral indices method is NDVI.
However, NDVI is only sensitive to fresh biomass, not
surface cover (Meusburger et al. 2010). Linear spectral
unmixing (LSU), which is applicable in deriving fractional
vegetation cover (FVC) (Vrieling 2006), has been proven
to be superior to NDVI in C factor mapping and was
successfully used to identify erosion features and areas
prone to soil erosion (Lu et al. 2004; De Asis and Omasa
2007; De Asis et al. 2008; Meusburger et al. 2010). Cau-
tion, however, should be taken in developing C factor.
While remotely sensed data capture the surface character-
istics only at the very moment of image acquisition, veg-
etation cover usually presents strong temporal dynamics
(Vezina et al. 2006; Vrieling 2006). The average C factor,
calculated from a multi-temporal evaluation of C factor,
can be used as input in the model (Vezina et al. 2006; De
Asis and Omasa 2007). In our study area, however, it is
even hard to find a cloud-free Landsat image in a year
because the region is cloudy most of the time (Jiang et al.
2012). There was even more difficulty in obtaining dif-
ferent seasonal images for calculating the average C.
To overcome this difficulty, we integrated image clas-
sification and LSU method in this study. For the LULC
types identified as built-up, water body, dryland, paddy
field and dense forest, the C-factor effects on soil erosion
often present high seasonal variability, but show similarity
within the type. These C-values were assigned following
the research of reported C values in the literature and local
conditions. For the remaining LULC types, i.e. sparse
forest, garden and bare soil, the C-factors show relatively
low seasonal variability, but have substantial annual
changes, and can vary by location. Therefore, they were
computed by LSU method. Detailed procedures for deter-
mining C-factor are as follows:
(1) LULC classification. Eight main LULC types, i.e.
dense forest, sparse forest, dryland, paddy field, built-
up, water body, garden, and bare soil, were classified
using remote sensing images in 1985, 1994, 2003 and
2008 by decision tree method (Quinlan 1993; Mahesh
and Mather 2003), with overall accuracy (OA) being
92.1, 93.3, 92.5 and 90.8 % in 1985, 1994, 2003 and
2008, respectively, with Kappa coefficients (Kappa)
of 0.89, 0.91, 0.91 and 0.88, respectively.
(2) The C-factor values of built-up, water body, dryland,
paddy field and dense forest were determined after a
study of the relevant reports and local conditions. The
C value for dense vegetation is 0, built-up 0.2, water
body 0, paddy field 0.05, and dryland 0.44 (Romken
1985; Shi et al. 2004; Fu et al. 2005; De Asis and
Omasa 2007; de Vente et al. 2009; Di et al. 2010).
(3) The C-factor value of sparse forest, garden and bare soil
were computed by LSU method. The spectral unmixing
approach assumes that the reflectance of each pixel is a
combination of spectral end-members, whose combi-
nations are usually green vegetation, non-photosyn-
thetic vegetation, bare soil, rock, and shadow (Adams
et al. 1995; Theseira et al. 2003). So the selection of
suitable end-members is the most important step in the
development of high quality fraction images.
In the study, a combination of automatic and supervised
end-member selections was performed on the Landsat/
ALOS images. At first, the minimum noise fraction (MNF)
algorithm was applied to the reflectance image. Then the
pixel purity index (PPI) obtained from MNF-transformed
data was used to find the spectrally purest pixels in the
images, i.e., the pixels with the highest PPI values were
selected as candidate end-members. At last, three end-
members, i.e., vegetation, water/shadow, and soil, were
determined by referring to the Landsat/ALOS images and
the results of field survey. The three end-members were
chosen because low correlation was found between the
fraction of the non-photosynthetic vegetation and the data
measured in the field, due to the confusion of reflectance
among different non-photosynthetic vegetation materials
(De Asis and Omasa 2007).
Then C-factor was obtained from Eq. 10 based on
vegetation, shadow, and soil fraction images derived from
LSU of remote sensing images (Lu et al. 2004; De Asis and
Omasa 2007).
C ¼ fsoil
1þ fgv þ fshadeð10Þ
where fsoil, fgv, and fshade are the three fraction values of
soil, green vegetation, and shadow. The values of fsoil, fgv,
756 Stoch Environ Res Risk Assess (2012) 26:751–763
123
and fshade parameters range from 0 to 1, and they sum to 1
(Lu et al. 2007).
3.3.2.2 Support practices factor (P) Factor P in RUSLE
is the ratio of soil with support practices like contouring,
strip-cropping or terracing to soil loss with straight-row
farming up and down the slope (Renard et al. 1997).
P value ranges from 0 to 1, where value 0 represents good
man-made erosion resistance facility and value 1 represents
lack of such facility. P value for paddy field was fixed at
0.1 (Vezina et al. 2006) because of its good conservation
practices. For other LULCs, the value was 1.0, assuming
no significant soil conservation practice (Irvem et al. 2007).
3.4 The average annual soil loss (A)
Average annual soil loss in 1985, 1994, 2003 and 2008 for
Anji County were computed with RUSLE and the afore-
mentioned parameters. Although the model yields quanti-
tative values, the model outcomes were usually suggested
to be applied in a qualitative way (Van Remortel et al.
2001; Van Remortel et al. 2004; Xu et al. 2008a). In the
study, soil erosion risk maps were generated by grouping
the annual soil loss into six categories according to SL190-
2007 (Table 1), i.e., non-erosion, slight, moderate, severe,
very severe and extreme erosion risk (Fig. 4) (MWR 2007).
The SL190-2007 is a national standard and has been used
to assess soil erosion at regional and national levels in
China (Zhu et al. 2009).
3.5 Soil erosion model validation
The most common method of validating an erosion model is
erosion survey, in which a visual estimation of erosion risk
is conducted on the basis of observed features (Metternicht
and Zinck 1998; Cohen et al. 2005). It was impossible,
however, to obtain in situ erosion information to validate
the results from historical periods. In this study, 240 random
points (40 validation points for each erosion level) repre-
senting non-erosion, slight, moderate, severe, very severe
and extreme erosion risk, were selected from the erosion
images of 1985, 1994, 2003 and 2008 respectively to create
a confusion matrix, which was used to compute overall
accuracy (OA) and Kappa coefficient (Kappa). Erosion risk
of the points were then validated with expert interpretation
following the criteria of soil erosion intensity in China
Fig. 4 Maps of soil erosion risk
in 1985, 1994, 2003 and 2008
for Anji County, China
Stoch Environ Res Risk Assess (2012) 26:751–763 757
123
(MWR 2007) (judged by slope gradient, vegetation cover
rate and land cover) based on field soil erosion surveys,
high-resolution DOM, Google Earth, topographic infor-
mation and LULC data. For uncertain areas of the historical
images, we consulted local agents about the erosion status
during the related periods. And the overall accuracy (OA)
and Kappa coefficient (Kappa) were then computed on the
basis of the error matrix (Congalton 1991; Foody 2002). OA
of the four classification maps was 90.4, 92.5, 91.3 and
90.4 %, and Kappa was 0.89, 0.91, 0.90 and 0.89 in 1985,
1994, 2003 and 2008, respectively.
3.6 Spatiotemporal analysis of erosion risk
A spatiotemporal analysis of erosion risk was conducted by
using GIS. At first, extent and intensity of erosion risk and
transition matrices for erosion risks in various periods were
calculated from erosion risk maps from the 4 years. Then,
soil erosion tendency was classified into three groups, i.e.
unchanged areas (areas with erosion risk unchanged during
the period), deteriorated areas (areas with erosion risk
increased to a higher level) and improved areas (areas with
erosion risk decreased to a lower level). In addition, the
rate of change and the percentage of different erosion risk
changes were calculated to characterize temporal dynamic
of soil erosion risk.
4 Results
Figure 4 shows the extent and the intensity of soil erosion
in the County for the 4 years. Generally, eroded areas
expanded and erosion risk increased continuously during
the study period. Table 2 summarizes the changes. Ero-
sion-free areas continually decreased from 87.6 % of the
total county area in 1985, to 85.2 % in 1994, 83.2 % in
2003, and 81.7 % in 2008, indicating a continuous
expansion of eroded areas. Erosion risk of eroded areas
kept increasing (Fig. 5). For the categories of Moderate,
Severe, and Very Severe erosion risk, the percentage of the
areas in the County increased gradually from 1985 to 1994,
2003, and 2008 in sequence. For the category of Extreme
erosion risk, its area increased from 2.5 % in 1985, to
3.9 % in 1994, and 5.9 % in both 2003 and 2008.
Table 2 Transitional matrices
of soil erosion risk during the
study periods (unit: %)
Non-erosion Slight Moderate Severe Very severe Extreme Total
1994
1985
Non-erosion 81.87 1.92 0.55 0.42 0.53 2.28 87.6
Slight 0.64 3.72 0.34 0.24 0.18 0.09 5.2
Moderate 0.42 0.07 1.05 0.22 0.19 0.14 2.1
Severe 0.37 0.01 0.06 0.51 0.22 0.15 1.3
Very severe 0.46 0.00 0.02 0.06 0.50 0.32 1.4
Extreme 1.43 0.00 0.00 0.01 0.08 0.93 2.5
Total 85.2 5.7 2.0 1.5 1.7 3.9 100
2003
1994
Non-erosion 77.92 1.64 0.63 0.52 0.64 3.86 85.2
Slight 1.74 3.30 0.22 0.20 0.18 0.09 5.7
Moderate 0.63 0.17 0.98 0.07 0.06 0.11 2.0
Severe 0.50 0.12 0.14 0.55 0.08 0.08 1.5
Very severe 0.64 0.12 0.06 0.14 0.58 0.15 1.7
Extreme 1.78 0.07 0.11 0.09 0.25 1.61 3.9
Total 83.2 5.4 2.1 1.6 1.8 5.9 100
2008
2003
Non-erosion 76.40 1.45 0.73 0.58 0.98 3.08 83.2
Slight 1.65 3.07 0.40 0.21 0.08 0.00 5.4
Moderate 0.50 0.38 0.90 0.07 0.20 0.09 2.1
Severe 0.39 0.24 0.20 0.49 0.06 0.19 1.6
Very severe 0.48 0.11 0.31 0.21 0.49 0.17 1.8
Extreme 2.32 0.04 0.13 0.30 0.71 2.40 5.9
Total 81.7 5.3 2.7 1.9 2.5 5.9 100
758 Stoch Environ Res Risk Assess (2012) 26:751–763
123
Figure 6 shows the spatial distribution of erosion risk
change in Anji County, and Fig. 7 quantifies the area and
percentage of erosion risk change for the three periods.
While the improved area of soil erosion increased steadily
from the period of 1985–1994, 1994–2003, and 2003–2008,
more area became deteriorated in the same period (Fig. 7a).
In terms of percentage, the improved area of soil erosion
increased from 3.63 % in 1985–1994, 6.57 % in
1994–2003, to 7.98 % in 2003–2008, but the deteriorated
area remained about 8 % in the three periods (Fig. 7b).
Non-erosion risk and extreme risk were two main categories
leading to dynamics of eroded areas. These two erosion risk
categories were most likely to transform into each other’s
type over the period (Table 2).
5 Discussion and conclusions
5.1 Method for soil erosion estimation
By subdividing background factors (representative of spa-
tial variations) and dynamic factors (representative of both
spatial and temporal variations), the method developed in
the study took advantage of both remote sensing and GIS
techniques and reasonably parameterized all controlling
factors of soil erosion. In particular, a combination of
LULC classification and LSU of the remote sensing image
generated the C factors. In this way, we determined the C
factors for those LULCs with strong temporal dynamics
and high seasonal variability, and computed the C factors
for other LULCs with high spatial heterogeneity and
variability.
Fig. 5 Areal summary (%) of soil erosion risk in 1985, 1994, 2003,
and 2008 for Anji County, China
Fig. 6 Maps of erosion risk
change in Anji County
Stoch Environ Res Risk Assess (2012) 26:751–763 759
123
In this region, RUSLE had been used for estimating soil
erosion for two decades. However, limited by the avail-
ability of digital soil survey geospatial database, surface
geological data were used for estimating K factor. Given
the data availability, the method we reported clearly
showed some advantages, and could be applied to other
regions nearby. However, this reported method has limi-
tations: (1) accurate prediction of soil erosion depended on
the accuracy of C-factor, which was closely associated with
the accuracy of LULC classification and with end-member
selection in LSU. Classification errors of LULC will be
introduced to the C factor map. Incorrect selection of end-
members in a physical sense will cause errors in fractional
abundances and consequently the C factor; (2) the method
is suitable for predicting spatial distribution of soil erosion
risk, as opposed to absolute values of soil erosion loss,
since it takes R, K and LS as the background factors.
Nevertheless, new algorithms and techniques aimed at
increasing classification accuracy (Lu and Weng 2007) or
obtaining appropriate end-members (De Asis and Omasa
2007) have become available in recent years. They helped
in producing accurate classification maps and appropriate
LSU fractions for soil erosion analysis. In addition, soil
conservationists and environmental managers intended
more to identify areas that need immediate soil conserva-
tion supports, rather than absolute values of soil loss.
Therefore, the outcomes produced by using this method
could be helpful in soil and water conservation planning
and practices: (1) the generated erosion risk maps show the
relative incidence of erosion in a certain location as com-
pared to other locations in the region; (2) erosion change
maps delineate areas that are more likely to change into the
worst degradation, such as a change from moderate to
severe state of erosion, and measure risk of this adverse
change. Such soil erosion maps and analysis results are
certainly more helpful for regional planners and policy
makers to initiate controlling measures and prioritize
conservation areas (Ma et al. 2003; Masoudi et al. 2006;
Rahman et al. 2009).
5.2 Soil erosion in Anji County
Soil erosion risk in Anji has had a trend of deterioration,
with a significant increase in both extent and intensity. Soil
erosion in the study area featured two simultaneous pro-
cesses: increasing area where originally extreme risk areas
were controlled and reduced, and expanding new eroded
areas at an accelerating rate. This indicated that the real
threat of soil erosion in the study area came from new
eroded areas that expanded in the erosion-free forests. The
existing eroded areas, on the other hand, gradually recov-
ered with the implementation of conservation practices.
Therefore, top priority should be given to natural vegeta-
tion preservation, land-use management and other conser-
vation practices aimed at soil erosion control. The erosion-
free forests are the most important areas for taking con-
trolling measures due to their dominant deteriorating pro-
cess and high probability of erosion. It is crucial to prevent
further erosion and to restore destroyed vegetation in these
areas.
The case study of soil erosion risk in Anji gives insight
into the prospective of soil erosion risk in China. The
results of this study from a county with seemingly high
ecological quality were initially surprising to us. We sus-
pected that public and government did not give adequate
attention to the issue of soil erosion. The likely reasons
could be: (1) eroded areas in such regions were widely
scattered and almost imperceptible in the national soil
erosion map at a scale of 1:4000000 (MWR et al. 2010);
(2) the subtle process of soil erosion in these regions often
Fig. 7 Area (a) and percentage
(b) of erosion risk change for
the three periods
760 Stoch Environ Res Risk Assess (2012) 26:751–763
123
went on unnoticed, at least during the initial stage (Lu et al.
2007); (3) these regions boast a high forest coverage of
more than 52 % and are mainly covered with pure forests
of Masson pine, bamboo and plantations, while large areas
of ground surface under the forests were exposed, without
shrubs and grass. The high forest coverage, however, often
causes a neglect of the ‘‘erosion under forests’’ (MWR
et al. 2010; Liang et al. 2010); (4) policy makers and local
farmers failed to give due consideration to the impact of
soil erosion in achieving the short-term goal of economic
development.
5.3 Soil erosion and water environmental issue
in the region
Eutrophication and algal blooms have been considered the
biggest environmental issue in this region over last decade.
To address the issue, the government at local and national
levels plan to invest over one billion Yuan on a pilot
research and engineering project in the Tiaoxi watershed in
the coming decade. Besides, a huge investment will be
made on pollution control in other fresh water systems of
the region. But soil erosion has not been blamed and
connected specifically to water pollution.
It is commonly accepted that non-point sources are the
main contributors of nitrogen and phosphorous, the two
dominant nutrients causing eutrophication in this region
(Su et al. 2011, Liang et al. 2011). Immediate measures for
soil conservation must be taken to control the increased soil
erosion risk discovered over the study period. Otherwise
the discharge of nitrogen and phosphorus from farming
land will cause eutrophication and algal blooms on surface
water as well as reduce soil fertility (Eroglu et al. 2010;
Zhu 2011).
5.4 Conclusions
By selecting Anji County (China), an area with seemingly
high ecological quality as the study area, we integrated
remote sensing and GIS techniques and quantified soil
erosion controlling factors. Then, by using RSULE, we
estimated annual soil loss, and generated categorical maps
of soil erosion risk for the years of 1985, 1994, 2003 and
2008. Soil erosion risk increased and eroded area expanded
for the period. The most vulnerable hotspots were erosion-
free forests. Our results indicated that soil erosion was a
critical issue in the region, which was closely associated
with eutrophication and algal blooms of fresh water. Since
it is applicable to similar regions, the method developed in
this study can serve as an operational basis for environ-
mental managers and planners.
Acknowledgements We were very grateful to the Editor-in-Chief
and the two reviewers for their providing constructive comments and
suggestions. This work was partly supported by the Fundamental
Research Funds for the Central Universities, Fujian Education
Department Project Fund (No. JB11150), and Minjiang University
Project Fund (No. YKY1106).
References
Adams JB, Sabol DE, Kapos V, Almeida R, Roberts DA, Smith MO,
Gillespie AR (1995) Classification of multispectral images based
on fractions of endmembers-application to land-cover change in
the Brazilian Amazon. Remote Sens Environ 52:137–154
Angima SD, Stott DE, O’Neill MK, Ong CK, Weesies GA (2003)
Soil erosion prediction using RUSLE for central Kenyan
highland conditions. Agric Ecosyst Environ 97:295–308
Arnoldus JMJ (1977) Methodology used to determine the maximum
potential average annual soil loss due to sheet and rill erosion in
Morocco. Food Agric Org Soils Bull 34:39–51
Bahadur KCK (2009) Mapping soil erosion susceptibility using
remote sensing and GIS: a case of the Upper Nam Wa
Watershed, Nan Province, Thailand. Environ Geol 57:695–705
Chander G, Markham BL, Helder DL (2009) Summary of current
radiometric calibration coefficients for Landsat MSS, TM,
ETM?, and EO-1 ALI sensors. Remote Sens Environ
113:893–903
Cohen MJ, Shepherd KD, Walsh MG (2005) Empirical reformulation
of the Universal Soil Loss Equation for erosion risk assessment
in a tropical watershed. Geoderma 124:235–252
Congalton RG (1991) A review of assessing the accuracy of
classification of remotely sensed data. Remote Sens Environ
37:35–46
De Asis AM, Omasa K (2007) Estimation of vegetation parameter for
modeling soil erosion using linear Spectral Mixture Analysis of
Landsat ETM data. Isprs J Photogramm 62:309–324
De Asis AM, Omasa K, Oki K, Shimizu Y (2008) Accuracy and
applicability of linear spectral unmixing in delineating potential
erosion areas in tropical watersheds. Int J Remote Sens
29:4151–4171
de Vente J, Poesen J, Govers G, Boix-Fayos C (2009) The
implications of data selection for regional erosion and sediment
yield modelling. Earth Surf Proc Land 34:1994–2007
Deng H, Zhang B, Yin R, Wang H, Mitchell SM, Griffiths BS, Daniell
TJ (2010) Long-term effect of re-vegetation on the microbial
community of a severely eroded soil in sub-tropical China. Plant
Soil 328:447–458
Di BF, Zeng HJ, Zhang MH, Ustin SL, Tang Y, Wang ZY, Chen NS,
Zhang B (2010) Quantifying the spatial distribution of soil mass
wasting processes after the 2008 earthquake in Wenchuan, China
A case study of the Longmenshan area. Remote Sens Environ
114:761–771
Eroglu H, Cakir G, Sivrikaya F, Akay AE (2010) Using high
resolution images and elevation data in classifying erosion risks
of bare soil areas in the Hatila Valley Natural Protected Area,
Turkey. Stoch Environ Res Risk Assess 24:699–704
Fan JR, Zhang JH, Zhong XH, Liu SZ, Tao HP (2004) Monitoring of
soil erosion and assessment for contribution of sediments to
rivers in a typical watershed of the Upper Yangtze River Basin.
Land Degrad Dev 15:411–421
Foody GM (2002) Status of land cover classification accuracy
assessment. Remote Sens Environ 80:185–201
Fu BJ (2008) Blue skies for China. Science 321:611–611
Fu BJ, Zhao WW, Chen LD, Zhang QJ, Lu YH, Gulinck H, Poesen J
(2005) Assessment of soil erosion at large watershed scale using
Stoch Environ Res Risk Assess (2012) 26:751–763 761
123
RUSLE and GIS: a case study in the Loess Plateau of China.
Land Degrad Dev 16:73–85
Gao J, Liu YS (2010) Determination of land degradation causes in
Tongyu County, Northeast China via land cover change detec-
tion. Int J Appl Earth Obs 12:9–16
Hickey R, Smith A, Jankowski P (1994) Slope length calculations
from DEM within ARC/INFO grid. Comput Environ Urban
18:365–380
Huang CC, Pang J, Su H, Yang Q, Ha Y (2007) Climatic and
anthropogenic impacts on soil formation in the semiarid loess
tablelands in the middle reaches of the Yellow River, China.
J Arid Environ 71:280–298
Irvem A, Topaloglu F, Uygur V (2007) Estimating spatial distribution
of soil loss over Seyhan River Basin in Turkey. J Hydrol
336:30–37
Ismail J, Ravichandran S (2008) RUSLE2 model application for soil
erosion assessment using remote sensing and GIS. Water Resour
Manag 22:83–102
Jiang X, Huang C-h, Fushui R (2008) Impacts of land cover changes
on runoff and sediment in the Cedar Creek Watershed, St. Joseph
River, Indiana. United States. J Mt Sci-Engl 5:113–121
Jiang Z, Qi J, Su S, Zhang Z, Wu J (2012) Water body delineation
using index composition and HIS transformation. Int J Remote
Sens 33:3402–3421
Le Bissonnais Y, Montier C, Jamagne M, Daroussin J, King D (2002)
Mapping erosion risk for cultivated soil in France. Catena
46:207–220
Li YK, Ni J, Yang QK, Li R (2006) Human impacts on soil erosion
identified using land-use changes: a case study from the Loess
Plateau, China. Phys Geog 27:109–126
Liang Y, Li D, Lu X, Yang X, Pan X, Mu H, Shi D, Zhang B (2010)
Soil erosion changes over the past five decades in the red soil
region of southern China. J Mt Sci-Engl 7:92–99
Liang XQ, Xu L, Li H, He MM, Qian YC, Liu J, Nie ZY, Ye YS,
Chen YX (2011) Influence of N fertilization rates, rainfall, and
temperature on nitrate leaching from a rainfed winter wheat field
in Taihu watershed. Phys Chem Earth 36:395–400
Lim HS, MatJafri MZ, Abdullah K, IEEE (2009) Turbidity measure-
ment from ALOS satellite imagery. Oceans 2009—Europe, vols
1 and 2. IEEE, New York, pp 1155–1159
Long HL, Liu YS, Wu XQ, Dong GH (2009) Spatio-temporal
dynamic patterns of farmland and rural settlements in Su–Xi–
Chang region: implications for building a new countryside in
coastal China. Land Use Policy 26:322–333
Lu D, Weng Q (2007) A survey of image classification methods and
techniques for improving classification performance. Int J
Remote Sens 28(5):823–870
Lu D, Mausel P, Brondizio E, Moran E (2002) Assessment of
atmospheric correction methods for Landsat TM data applicable
to Amazon basin LBA research. Int J Remote Sens
23:2651–2671
Lu D, Li G, Valladares GS, Batistella M (2004) Mapping soil erosion
risk in Rondonia, Brazilian Amazonia: using RULSE, remote
sensing and GIS. Land Degrad Dev 15:499–512
Lu D, Batistella A, Mausel P, Moran E (2007) Mapping and
monitoring land degradation risks in the Western Brazilian
Amazon using multitemporal landsat TM/ETM plus images.
Land Degrad Dev 18:41–54
Ma JW, Xue Y, Ma CF, Wang ZG (2003) A data fusion approach for
soil erosion monitoring in the Upper Yangtze River Basin of
China based on Universal Soil Loss Equation (USLE) model. Int
J Remote Sens 24:4777–4789
Mahesh P, Mather PM (2003) An assessment of the effectiveness of
the decision tree method for land cover classification. Remote
Sens Environ 86:554–565
Masoudi M, Patwardhan AM, Gore SD (2006) Risk assessment of
water erosion for the Qareh Aghaj subbasin, southern Iran. Stoch
Environ Res Risk Assess 21:15–24
Mati BM, Veihe A (2001) Application of the USLE in a Savannah
environment: comparative experiences from East and West
Africa. Singap J Trop Geogr 22:138–155
Metternicht GI, Zinck JA (1998) Evaluating the information content
of JERS-1 SAR and Landsat TM data for discrimination of soil
erosion features. Isprs J Photogramm 53:143–153
Meusburger K, Banninger D, Alewell C (2010) Estimating vegetation
parameter for soil erosion assessment in an alpine catchment by
means of QuickBird imagery. Int J Appl Earth Obs 12:201–207
MWR (Ministry of Water Resources, PRC) (2007) National profes-
sional standards for classification and gradation of soil erosion
(SL190-2007). China Hydraulic and Hydropower Press, Beijing
(in Chinese)
MWR (Ministry of Water Resources, PRC), Chinese Academy of
Sciences (CAS), Chinese Academy of Engineering (CAE)
(2010) Soil erosion control and eco-security in China-A volume
on red soil area of South China. Science Press, Beijing (in
Chinese)
Onyando JO, Kisoyan P, Chemelil MC (2005) Estimation of potential
soil erosion for River Perkerra catchment in Kenya. Water
Resour Manag 19:133–143
Ouyang W, Skidmore AK, Hao FH, Wang TJ (2010) Soil erosion
dynamics response to landscape pattern. Sci Total Environ
408:1358–1366
Quinlan R (1993) Programs for machine learning. Morgan Kaufman,
San Mateo
Rahman MR, Shi ZH, Chongfa C (2009) Soil erosion hazard
evaluation—an integrated use of remote sensing, GIS and
statistical approaches with biophysical parameters towards
management strategies. Ecol Model 220:1724–1734
Renard KG, Freimund JR (1994) Using monthly precipitation data to
estimate the R-factor in the revised USLE. J Hydrol
157:287–306
Renard KG, Foster GR, Weesies GA, McCool DK, Yoder DC (1997)
Predicting soil erosion by water: a guide to conservation
planning with the revised universal soil loss equation (RUSLE).
Handbook #703. US Department of Agriculture, Washington,
DC
Rey F (2003) Influence of vegetation distribution on sediment yield in
forested marly gullies. Catena 50:549–562
Romken MJM (1985) The soil erodibility factor: a perspective. Soil
Erosion and Conservation, Soil Conservation Society of Amer-
ica, Ankeny, lowa
SCPRC (The State Council of the People’s Republic of China) (2006)
Guidelines of PRC’s 11th Five-Year Plan for national economic
and social development. People’s Press, Beijing (in Chinese)
Shi ZH, Cai CF, Ding SW, Wang TW, Chow TL (2004) Soil
conservation planning at the small watershed level using RUSLE
with GIS: a case study in the Three Gorge Area of China. Catena
55:33–48
Shrestha DP, Zinck JA, Van Ranst E (2004) Modelling land
degradation in the Nepalese Himalaya. Catena 57:135–156
Solaimani K, Modallaldoust S, Lotfi S (2009) Investigation of land
use changes on soil erosion process using geographical infor-
mation system. Int J Environ Sci Technol 6:415–424
Su SL, Li D, Zhang Q, Xiao R, Huang F, Wu JP (2011) Temporal
trend and source apportionment of water pollution in different
functional zones of Qiantang River, China. Water Res
45:1781–1795
Theseira MA, Thomas G, Taylor JC, Gemmell F, Varjo J (2003)
Sensitivity of mixture modelling to end-member selection. Int J
Remote Sens 24:1559–1575
762 Stoch Environ Res Risk Assess (2012) 26:751–763
123
Tian YC, Zhou YM, Wu BF, Zhou WF (2009) Risk assessment of
water soil erosion in upper basin of Miyun Reservoir, Beijing,
China. Environ Geol 57:937–942
Van Remortel RD, Hamilton ME, Hickey RJ (2001) Estimating the
LS factor for RUSLE through iterative slope length processing of
digital elevation data. Cartography 30:27–35
Van Remortel RD, Maichle RW, Hickey RJ (2004) Computing the LS
factor for the Revised Universal Soil Loss Equation through
array-based slope processing of digital elevation data using a
C?? executable. Comput Geosci-UK 30:1043–1053
Van Rompaey AJJ, Govers G (2002) Data quality and model
complexity for regional scale soil erosion prediction. Int J Geogr
Inf Sci 16:663–680
Van Rompaey AJJ, Govers G, Van Hecke E, Jacobs K (2001) The
impacts of land use policy on the soil erosion risk: a case study in
central Belgium. Agric Ecosyst Environ 83:83–94
Vezina K, Bonn F, Van CP (2006) Agricultural land-use patterns and
soil erosion vulnerability of watershed units in Vietnam’s
northern highlands. Landscape Ecol 21:1311–1325
Vrieling A (2006) Satellite remote sensing for water erosion
assessment: a review. Catena 65:2–18
Vrieling A, Sterk G, de Jong SM (2010) Satellite-based estimation of
rainfall erosivity for Africa. J Hydrol 395:235–241
Wang G, Wente S, Gertner GZ, Anderson A (2002) Improvement in
mapping vegetation cover factor for the universal soil loss
equation by geostatistical methods with Landsat Thematic
Mapper images. Int J Remote Sens 23:3649–3667
Wang K, Wang HJ, Shi XZ, Weindorf DC, Yu DS, Liang Y, Shi DM
(2009) Landscape analysis of dynamic soil erosion in Subtrop-
ical China: a case study in Xingguo County, Jiangxi Province.
Soil Till Res 105:313–321
Williams JR, Jones CA, Dyke PT (1984) A Modelling approach to
determining the relationship between erosion and soil produc-
tivity. Trans ASABE 27:129–144
Wischmeier WH, Smith DD (1978) Predicting rainfall erosion losses-
a guide to conservation. Agricultural Handbook 537. US
Department of Agriculture, Washington, DC
Wu J, Nellis MD, Ransom MD, Price KP, Egbert SL (1997)
Evaluating soil properties of CRP land using remote sensing and
GIS in Finney County, Kansas. J Soil Water Conserv
52:352–358
Xu YQ, Peng J, Shao XM (2008a) Assessment of soil erosion using
RUSLE and GIS: a case study of the Maotiao River watershed,
Guizhou Province, China. Environ Geol 56:1643–1652
Xu YQ, Shao XM, Kong XB, Peng J, Cai YL (2008b) Adapting the
RUSLE and GIS to model soil erosion risk in a mountains karst
watershed, Guizhou Province, China. Environ Monit Assess
141:275–286
Yang T, Xu CY, Zhang Q (2011) DEM-based numerical modelling of
runoff and soil erosion processes in the hilly-gully loess regions.
Stoch Environ Res Risk Assess. doi:10.1007/s00477-011-0515-3
Zhang KL, Shu AP, Xu XL, Yang QK, Yu B (2008) Soil erodibility
and its estimation for agricultural soils in China. J Arid Environ
72:1002–1011
Zhejiang Office of Soil Survey, China (1985) Zhejiang soil. Zhejiang
Science Technology Press, Hangzhou (in Chinese)
Zhou P, Luukkanen O, Tokola T, Nieminen J (2008) Effect of
vegetation cover on soil erosion in a mountainous watershed.
Catena 75:319–325
Zhu, MY (2011) Soil erosion risk assessment with CORINE
model:case study in the Danjiangkou Reservoir region, China.
Stoch Environ Res Risk Assess. doi:10.1007/s00477-011-0511-7
Zhu D, Wang TW, Cai CF, Li L, Shi ZH (2009) Large-scale
assessment of soil erosion using a neuro-fuzzy model combined
with GIS: a case study of Hubei province, China. Land Degrad
Dev 20:654–666
Stoch Environ Res Risk Assess (2012) 26:751–763 763
123