Water Air Soil Pollut (2014) 225:2103
DOI 10.1007/s11270-014-2103-x
Establishment and Validation of an Amended Phosphorus
Index: Refined Phosphorus Loss Assessment of an Agriculture
Watershed in Northern China
Bin Zhou & Rolf D. Vogt & Chongyu Xu & Xueqiang Lu &
Hongliang Xu & Joshi P. Bishnu & Liang Zhu
Received: 14 May 2014 / Accepted: 24 July 2014 / Published online: 6 August 2014
# Springer International Publishing Switzerland 2014
Abstract Phosphorus (P) loss from non-point sources
is a main cause of freshwater eutrophication in agricultural regions. Knowledge-based watershed management
plans, aimed at reducing the diffuse flux of phosphorus
from specific land-use and site characteristics to freshwater resources, are needed in order to curb eutrophication in agriculture regions. In this context, the use of a
phosphorus index provides a simple and practical method for identifying hot-spot source areas and to estimate
their potential for contributing a flux of P to the surface
waters. However, as a semi-quantitative tool, the P index
is usually difficult to validate due to inadequate data
representation relative to large spatial and temporal variation in P fluxes. An amended P index scheme is
therefore developed and validated, based on comprehensive synoptic soil study and stream water monitoring
as well as a previous study that had applied the former P
index in the studied watershed in northern China (Zhang
et al. 2003). The amendments include the use of data
from the individual village units (mean area, ca.
30.6 ha), use of the degree of P saturation (DPS) in the
B. Zhou (*) : R. D. Vogt : J. P. Bishnu : L. Zhu
Department of Chemistry, University of Oslo,
Oslo 0315, Norway
e-mail: bin.zhou@kjemi.uio.no
C. Xu : H. Xu
Department of Geosciences, University of Oslo,
Oslo 0315, Norway
B. Zhou : X. Lu
Tianjin Academy of Environmental Sciences,
Tianjin 300191, China
source factor scheme, adoption of flow length factor and
modified water course erosion factor into the P transportation scheme, and an adjustment of the organization
structure of the P index scheme. The validation of the
amended P schemes was performed by comparing the
modeled average P index values with the corresponding
measured P fluxes for 12 different sub-catchments. The
results indicate an improved precision in the simulated
potential for P loss using the refined P index scheme.
Measured fluxes of total P (r=0.825), particulate P (r=
0.867), and less-studied yet more relevant dissolved P
(r=0.627) all showed significant correlations with the
modeled P index values in the amended P scheme. The
primary direct finding of the current research is that the
areas with close proximity to rivers and the reservoir, as
well agricultural land around villages, are found to be
the main hot-spot sources for P loss to the reservoir.
Keywords Amended P index . Degree of P saturation .
P fluxes . Validation
1 Introduction
In modern intensive agriculture, the addition of phosphorous (P) is a key factor to increase crop production
(Cordell et al. 2009). Globally, 500 million tonnes of P
has been applied as fertilizer during the second half of
last centenary (Brown et al. 1997; FAO 1950–1995).
This has led to an excess of P in the top Ap horizon of
the agricultural soil, which increases the P loss from
terrestrial to aquatic ecosystem. Numerous studies have
2103, Page 2 of 16
subsequently identified diffusive phosphorus loss from
agricultural land as the main cause of water eutrophication in main developed countries (Agnew et al. 2006;
Elliott et al. 2006). The increased flux of P from diffuse
agricultural sources is thus posing a potential threat to
the quality of freshwater systems (Correll 1998;
Sharpley et al. 2001a). Moreover, as the contribution
of P from point sources have been significantly and
efficiently limited, especially in Europe and USA, the
diffusive loading of P from agriculture has become the
remaining cause for P flux to surface water (Orderud
and Vogt 2013; Sharpley et al. 1999). Identifying and
assessing potential for P loss from agricultural land is
thereby the key prerequisite in order to select the most
cost-efficient abatement strategies in the struggle against
eutrophication. Compared to point source pollution, the
sources of diffusive phosphorus loss from agricultural
land are inherently more difficult to identify and quantify. The continuous development of water quality management models, especially those based on physical
mechanisms, such as the SWAT and AGNPS models
(Bingner et al. 2003; Neitsch et al. 2011), have greatly
improved our understanding of hydro-geochemical processes governing the mobilization and transport of phosphorus in the watershed (Tominaga 2013). However, the
heavy data requirements of these models greatly limit
their practical application as routine management tools.
Instead, the simplified P loss indicator was introduced
by Lemunyon and Gilbert (1993) as a proxy for assessed
potential for P flux. As a semi-quantitative tool, the P
index is based on arithmetic computations of source and
transportation factors, most of which are readily available data (Bechmann et al. 2005; Buczko and
Kuchenbuch 2007). This indicator is therefore an applicable environmental management tool for local water
quality managers: In the USA, the original P index has
been routinely applied for assessments of P loss from
agricultural fields and identification of critical areas with
high susceptibility for P loss (Sharpley 1995; Stevens
et al. 1993). In Germany, the P index combined with a
GIS framework has been applied to identify areas most
susceptible for phosphorus leaching (Behrendt et al.
1996).
Since the P index was introduced in 1993, the system
has grown more comprehensively by incorporating
more governing factors: Elliott et al. (2006) introduced
the phosphorus source coefficients (PSCs) into the
source factor module, enabling the identification of
different load rates of dissolved P in runoff from various
Water Air Soil Pollut (2014) 225:2103
types of organic fertilizer. Jokela et al. (1998) introduced
the role of reactive aluminum as an important P suppression factor in acid soils (Jokela 1999; Jokela et al.
1998). In the transport factor module, McFarland et al.
(1998) incorporated the distance to water body factor
and Sharpley et al. (2001b) introduced the leaching
potential factor. More recently, Li et al. (2007) introduced the quality of receiving water factor into the
modified phosphorus ranking scheme and Zhou and
Gao (2011) adapted the P index in order to use it also
in large-scale agricultural catchments. The development
and implementation of the phosphorus index is still in its
early stages in China, where a broad application of the
phosphorus index system by agricultural managers is
yet to be realized (Li and Guo 2010).
Assessment of P loss using P index at various scales
(e.g., plot, field, and watershed) have been thoroughly
described in the literatures. Most watershed studies tend
to use data with a relatively low spatial resolution. For
example, Andersen and Kronvang (2006) introduced
the use of an averaged catchment (1,000 km2) value
for the measured soil test phosphorus parameter (STP,
a measure of the pool of bioavailable P (BAP) in the
soil), instead of applying the spatially distributed measured data. Zhou and Gao (2008) estimated the distribution of degree of phosphorus saturation (DPS %),
based on measurements of only 80 soil samples, in a
large agricultural watershed (13,349 km2) in southern
China. By doing so, the modeling efficiency was significantly improved, and an overall estimate of phosphorus
loss at a large scale was achieved. However, the use of
such relatively rough input data is inherently associated
with limited capability for interpretation, due to compromised spatial accuracy. The use of high-resolution
spatial data, such as maps of soil texture, land use, and
vegetation cover, greatly improves the possibility for
interpretation at a relatively small scale by inducing
their conceptual link to the potential for P loss. Most
of the phosphorus mobilized during rainfall periods is
found to be immobilized again by the soil. For example,
Tuo et al. (2009) found that the concentration of especially particle-bound phosphorus in surface runoff was
reduced (more than 80 %) upon subsequently passing
through soils as interflow prior to entering the stream.
Watersheds in near proximity to stream channels or the
target water therefore commonly possess relatively
higher risk relative to regions located at some distance
away (Johnes and Heathwaite 1997). An erosion factorbased distance (water course factor) was therefore
Water Air Soil Pollut (2014) 225:2103
introduced into the P index assessment system
(McFarland et al. 1998). However, this relationship
was initially generalized as a simple straight line distance function, with equal distance weights. This made
it impossible to represent the actual transportation mechanisms, which are generally based on non-linear functions. The actual migration distance is generally longer,
especially in regions with complex terrain structures.
The transport length was instead considered as the actual
flow distance within the given basin, meaning the distance traveled by the water from any point along the
flow path (Akhtar et al. 2009).
The soil parameters total P and BAP are usually
selected as primary assessment indicators due to their
strong correlation with the loss of P as well as their
ready availability from environmental monitoring departments (Bechmann et al. 2009; Drewry et al. 2011;
Sharpley 1995; Shen et al. 2011). However, the use of
only BAP might lead to erroneous estimations of the
potential for soil P loss due to differences in the P
sorption capacity of different soil types (Wang 2010).
For instance, Sharpley (1995) showed that different soil
types although having the same BAP (200 mg kg−1
measured by Mehlich-3 P method) supported runoff
water which differed markedly in the concentration of
dissolved reactive P (0.28 and 1.36 mg L−1). The degree
of phosphorus saturation (DPS) has instead been introduced as a promising indicator to improve the estimate
for risk of P loss from agricultural soil. This proxy is
defined as the amount of BAP in the soil relative to the P
sorption capacity (PSC) of the soil.
A credible validation method for the P index is a
major challenge, especially on a watershed scale, as it
is a semi-quantitative tool. In most of the P index studies
found in the literature, the main attention has been on the
comparison of P index value with the flux of total P in
the stream draining the watershed, e.g., Sharpley (1995)
found a strong relationship between the P index value
for 30 different watersheds in the USA and the measured
total P fluxes (0.1–5 kg P ha−1 year−1).
Yuqiao Reservoir is the drinking water supply for the
6 million urban population of Tianjin municipality by
the diversion channel (Fig. 1). The reservoir has received considerable attention due to annual algae
blooms over the past decade. Several studies have concluded that increasing concentrations of phosphorus is
most likely the main factor governing the algae blooming in the reservoir (Chen and Zhu 1991). Monthly
mean and maximum concentrations of total P in the
Page 3 of 16, 2103
reservoir reached 0.043 and 0.055 mg P L−1, respectively, in 2012.1 The reservoir water authorities have found
that the major (64 %) source of P flux to the reservoir is
from the local watershed (TMWA 2010; Zhang et al.
2003). Identification of hot-spot areas and assessment of
risk for P loss in the local watershed of the Yuqiao
reservoir is therefore a necessary prerequisite for curbing the eutrophic trend through the selection of optimum
and cost-efficient abatement actions.
The objective of this study has been to (1) improve
the P index assessment system by modeling a set of
small-scale watersheds in which a detailed database was
established, compiling existing and generated high spatial resolution soil data (maps) and temporal
(monitoring) data of P fractions in runoff; (2) introduce
the concept of degree of P saturation (DPS%) and flow
length factor to the P index assessment system to acquire
an improved estimation of P loss; (3) modify the organization structure of the P index scheme in order to
consider more correctly the internal relationships between indicators and thereby better capture the conceptual mechanism of P loss; (4) introduce a non-linear
watercourse erosion factor in order to better assess the
spatial distribution of erosion in the watercourse; and
finally (5) compare and evaluate the former and
amended P index through correlations with measured
flux of P fractions.
2 Materials and Methods
2.1 Study Area
The study area is the local watershed of the Yuqiao
Reservoir (117° 25′–117° 43′ E, 39° 56′–40° 18′ N)
located in the north of Tianjin municipality, in northern China (Fig. 1). The targeted watershed covers an
area of 540 km2. The region has a sub-humid continental monsoon climate, with an annual mean temperature of 14 °C and an average annual precipitation of 653 mm, of which nearly three-fifths rains
between July and September (JCBS 2011). Mean
elevation of the local watershed is less than 200 m.
The region is characterized by both flat plains and
hilly morphology. The lowlands, which have an
average gradient of less than 2°, account for 22 %
of study area. In addition, there is an area with
1
Data source: Yuqiao Reservoir Administration Department.
2103, Page 4 of 16
Water Air Soil Pollut (2014) 225:2103
Fig. 1 Location and digital elevation map (DEM) of studied watershed
relatively flat landscape, with gradients of only 2–6°
(plain), comprising an additional 25 %. Together,
these two landforms constitute almost half of the
watershed. Hilly land with gradients of 6–15° shared
23 %, low mountain region with gradients of 15–25°
account for 19 %, while gradients above 25 %
(mountain region) comprise only 11 %. On a macro
scale, the terrain declines from the north to the south
(Fig. 1) where the reservoir is situated. Most of the
soil material, especially in the lowland area, is deltaic alluvial sediments. On the hills and mountains,
the soils are developed through weathering of the
parent sedimentary bedrock consisting of sandstone
and limestone. There have a mix of Gleysols in the
lowland and Lithosols in the mountain. The majority
of the ca. 130,000 local residents are living of agricultural crops and livestock breeding. Agricultural
land is the dominating land use, accounting for 36 %
of the targeted area. Crop rotation with “summer
maize-winter wheat” is common in the lowland
and plains, while agriculture in the hilly and mountainous area is predominantly orchard. The local
environmental protection bureau (JCEPB 2012) reports that applying excessive amounts of inorganic P
fertilizer is a general practice in this region. In
addition, due to excessive livestock breeding and
lack of sewage treatment system, the large amount
of animal’s dung and domestic sewage were
disposed of either in heaps on wasteland or directly
into the drainage channels.
2.2 Background Data
Background data used for the study of the P index
are listed in Table 1. The soil sampling strategy
aimed for an even spatial distribution as well as
capturing the span in soil types and land use. Soil
sampling and analysis were conducted in
cooperation with Joshi (2014) as a part of their
master studies. Stream water samples, used for the
validation of the P indexes, were collected and
a na l yz ed i n c oo pe r a t i o n w i t h t he Tia n j i n
Academy of Environmental Sciences (TAES).
More detailed information and further parameters
on the soil and water samples may be found in
their thesis.
2.3 Source Factors
A total of 126 surface soil samples (0–15 cm)
were collected and analyzed for soil test P (STP)
and P sorption index (PSI) as part of the Master
studies by (Joshi (2014)). DPS in the soils was
determined using Eq. 1 (Wang 2010; Wang et al.
2010b).
Water Air Soil Pollut (2014) 225:2103
Page 5 of 16, 2103
Table 1 Background data types used for the P index scheme
Data type
Scale
Resolution
Data description
Source
Meteorological data
Four stations
Daily
Precipitation amount
Ji County Meteorological
Bureau; Yuqiao Reservoir
Administration Department
Hydrological data
1:25,000
River distribution
Field investigation and local
Water Bureau
Topographic data
1:10,000
Digital elevation model
The United States National
Aeronautics and Space
Administration (NASA)
global digital elevation
mapa
Soil data
126 Soil samples
Land use
1:10,000
30 m×30 m
Soil physical and chemical
Field sampling and laboratory
properties (pH, STP, PSI
analysis; Ji County soil
(phosphors sorption index),
survey report (1982);
DPS, soil texture)
Harmonized World Soil
Database (2012)b
2.5 m×2.5 m
Land-use classification
Interpretation of satellite
remote sensing data
(Spot 5); Ji County Land
Resource Bureau
Vegetation cover
30 m×30 m
Normalized difference
vegetation index (NDVI)
Landsat 8 Operational Land
Imager (OLI) data
Agricultural management data
Village scale
P fertilizer application;
(mean area 30.6 ha)
livestock breeding
Field investigation; local
bureau of agriculture; local
statistics yearbook
Environmental protection data
Village scale
(mean area 30.6 ha)
Local environmental bureau
Environmental protection
situation/livestock manure
and domestic sewage
treatment situation
a
http://asterweb.jpl.nasa.gov/gdem.asp
b
http://webarchive.iiasa.ac.at/Research/LUC/External-World-soil-database/HTML/
DPSð%Þ ¼
STP
STP þ PSI
100
ð1Þ
The denomination for STP and PSI is mg P kg−1.
Olsen P (Olsen et al. 1954), Bray-1 P test (Bray and
Kurtz 1945), and Mehlich-3 P (Mehlich 1984) are commonly used methods for measuring STP: Olsen P method is suited for calcareous soil, while Bray-1P method is
more reliable for acid and neutral soils. The Mehlich-3
method has been reported as well suited for a wide range
of soils (Pierzynski, 2000), yet it has been argued that
the Mehlich-3 method tends to overestimate the STP in
calcareous soil (Basta et al. 2000; Watson and Mullen
2007). The Olsen and Bray-1 methods are therefore
used, based on whether the soil pH was above or below
7.4, respectively, following the recommendation from
the US Department of Agriculture (Elrashidi 2010). The
final grid values of the degree of P saturation (DPS)
were calculated by the inverse distance weighted interpolation method (Childs 2004). The phosphorous sorption index (PSI) is measured as an estimate of soil P
sorption capacity, which was calculated by Eq. 2 (Bache
and Williams 1971):
X
PSI Lkg−1 ¼
logC
ð2Þ
where X is the amount of sorbed P (mg P kg−1) on the
soil, and C is the remaining P concentration in solution
at equilibrium (mg P L−1).
The loading of inorganic P fertilizer (P2O5) applied to
the fields was calculated from the amount used by each
village divided by their agricultural land area. This was
subsequently recorded for each village vector and finally converted to raster format. Non-agricultural lands
were considered as not fertilized. Livestock manure
2103, Page 6 of 16
Water Air Soil Pollut (2014) 225:2103
production (total P) was calculated based on various
breeding species and stocks in each village. The P
content in the feces of each species was based on recommended reference values issued by the Chinese
Ministry of Environmental Protection (CPSC 2008)
and previous research results (Fan et al. 2011). The use
of this manure as an organic fertilizer product was based
on village values supplied by the local agricultural management department. The final results were divided by
agricultural land area of each village. The detailed calculation method is given in Eq. 3:
n
X
Qi P i
Lj ¼
ð3Þ
1 λj
Sj
S
A ¼ R K LS C P
where Lj is the average total phosphorus load applied
to the fields from livestock breeding for village j
(kg ha−1a−1), Qi is the breeding amount of farmed specie
i, and Pi is total P emission coefficient of farmed specie i
(kg unit−1a−1), determined by related emission coefficient manual from China’s Ministry of Agriculture
(2009). Sj is agricultural area of village j, λj is manure
utilization rate of village j.
Only the P load from sanitary sewage was considered
for this scheme, as there is a good mechanism of household garbage pickup though the region lacks treatment
of sewage water. The P emission coefficient was determined from the recommended value (0.07 kg
capital −1 a −1 ) by the People’s Republic of China
Ministry of Housing and Urban (2010). The final results
were divided by agricultural land area of each village, as
shown in Eq. 4:
where P is mean annual precipitation (mm), and Pi is
the mean monthly precipitation (mm). The R values
were calculated using precipitation data monitored from
2006 to 2012 at four stations distributed within and in
the vicinity of the watershed. R values for the total study
area were obtained through interpolating with inverse
distance weight (IDW) method using Arcgis 10 spatial
analysis module (Hillier 2011).
The soil erodibility factor (K) is computed according
to Eq. 7 (Williams et al. 1983):
where A is the computed soil loss per unit area
(t ha−1 year−1), R is the rainfall and runoff factor
(MJ mm ha−1 h−1 year−1), K is the soil erodibility factor,
LS is the topographic factor, C is the cover and management factor, and P is the support practice factor.
The procedures for the determination of each factor
are described below.
The amount of rainfall (R) is calculated by Eq. 6
(Wischmeier and Smith 1978):
R¼
Q j Ps
Si
ð4Þ
where Dj is the average total phosphorous load from
sewage of village j, Qj is the total population of village j,
and Ps is the total P emission coefficient of sanitary
sewage per capital, Si is agricultural area of village i.
2.4 Transportation Factors
Potential for soil erosion was calculated using the
Revised Universal Soil Loss Equation (RUSLE)
(Eq. 5) (Renard et al. 1997). This equation is a commonly used empirically derived model that provides an
improved manner of computing soil erosion based on
the original Universal Soil Loss Equation (USLE)
(Wischmeier and Smith 1965).
12 h
X
1:1735 101:5lgðPi
2
=PÞ−0:8188
i
ð6Þ
i¼1
j
Dj ¼
ð5Þ
K
Si
¼ 0:2 þ 0:3exp −0:0256S d 1
100
0:3 Si
0:25C
1
Þ
C i þ S i
C þ expð3:72−2:95C
Sd
Sd
1−
þ exp −5:51 þ 22:9 1
100
100
ð7Þ
where Sd, Si, and Cl are the soil sand, silt, and clay
content (%), respectively. C denotes the organic carbon
content (%), which was obtained from local soil survey
reports and the Harmonized World Soil Database.
The topographic value (LS) was calculated by Eq. 8
(Wischmeier and Smith 1978):
LS ¼
λ
22:1
m
8
m ¼ 0:2
>
>
<
m ¼ 0:3
m
¼ 0:4
>
>
:
m ¼ 0:5
65:41 sin2 θ þ 4:56 sinθ þ 0:065
slope < 1%
1%≤ slope ≤ 3%
3% < slope < 5%
slope ≥ 5%
ð8Þ
ð9Þ
where L is slope-length factor (m), θ is slope gradient
(in radians), and m is gradient exponent as specified
relative to the percent slope. Related parameters were
Water Air Soil Pollut (2014) 225:2103
Page 7 of 16, 2103
derived from Global Digital Elevation Map (GDEM)
using the ArcGIS spatial analyst module.
Crop management factor (C) was computed using
Eqs. 10 and 11 (Chai et al. 2000; Ma and Ma 2001):
watercourse erosion model by introducing the watercourse erosion attenuation trend weight (E) (Eq. 13),
which was used in this study.
8
<C ¼ 1
C ¼ 0:6805−0:3436lgI c
:
C¼0
F ðxÞ ¼
Ic ¼ 0
0 < I c < 78:3
I c >78:3
ð10Þ
I c ¼ 108:49NDVI þ 0:717
ð11Þ
where Ic is the vegetation coverage (%), NDVI is
normalized difference vegetation index, which was derived from Landsat satellite remote sensing data with
Erdas software image interpreter module (Imagine
2006).
The support practice factor (P) reflects the effects of
best management practices (BMP) that will reduce the
amount and rate of surface runoff and thus reduce the
amount of soil erosion. Due to the lack of related plot
experiments in this study, this factor was simply adopted
from the recommended value by USLE and previous
studies based on different land-use types (Huang et al.
2004; Sivertun and Prange 2003; Stone et al. 2000).
The surface runoff factor (R) was calculated by
Eq. 12 (Zhou and Gao 2011):
Rrunoff ¼ P α
ð12Þ
−1
where Rrunoff is total annual runoff (mm year ), P is
mean annual rainfall (mm), and α is annual runoff
coefficient. The mean annual rainfall of each station
was calculated using daily monitored data from 2006
to 2012. Annual runoff coefficients for different land use
were adopted from previous studies in this study area
(Chen and Zhu 1991).
Soils with poor drainage commonly have higher soil
erosion due to more surface runoff with a greater potential for P loss (Marjerison et al. 2011; Wang et al. 2006).
The soil drainage class factor was therefore directly
generated from the soil database (the Harmonized
World Soil Database) and introduced into P index
scheme.
The watercourse erosion factor was determined as the
actual distance from each cell to a river network. In most
previous studies, a simple straight line distance function
is used to assess the water course erosion factor
(Bechmann et al. 2005; Wang et al. 2010a). However,
the mechanism of watercourse erosion is in most cases
non-linear. Sivertun and Prange (2003) improved the
0:6
e0:002x
0:4
ð13Þ
where x is distance from each cell to its nearest river
network (m), F(x) is the value of water course factor.
The flow length factor is the actual migration distance of surface runoff. Thus was calculated using the
digital elevation model data by means of the Arcgis 10
hydrological analyst module.
The irrigation amount factor is the transportation
factor potentially influenced by human activities.
Flood irrigation is still the dominant irrigating method
in the study zone due to the lack of modern irrigation
(Chen and Zhu 1991; Kendy et al. 2004). The irrigation
erosion factor was generated by extracting the actual
irrigation water volume of each village to the corresponding agricultural land unit.
2.5 P Index Organization and Calculation
A weighting value (0.75, 0.85, or 1.00) for each factor
was used in order to adjust the influence of each factor
on the total potential for P loss (Table 2). These values
were assigned based on previous studies (Drewry et al.
2011; Sharpley 1995; Zhang et al. 2003). The weighted
value for each factor was finally converted to a five-step
unified risk rating value (2, 4, 6, 8, 10) relative to
ascribed ranges given in Table 2. All risk rating values
were determined for each 50 m×50 m grid based on the
coordinate system (WGS 84/UTM zone 50N).
The structure of P index scheme was re-organized in
order to better account for the interactions among the
factors. In previous studies, the mutual restriction between source and transportation factor group was handled through a multiplicative approach. The parallel
relation, which is implied within each factor group, is
accounted for through a weighted summation approach.
The problem with this is that an additive approach is not
sound in regards to the transportation factors since the
migration distance factors have restriction to other
erosion-based transportation factors, i.e., the effect of
erosion-induced transport is restricted by the migration
length into relevant water course. In order to account for
this in the amended P index scheme, the transportation
factor was sub-grouped into erosion process factor (I)
and immigration distance factor (II) (Eq. 14).
2103, Page 8 of 16
Water Air Soil Pollut (2014) 225:2103
Table 2 Factor weightings and loss ratings for the amended P index scheme
Factors
Unit/range
Weight
Risk rating value
Very low
2
Low
4
Medium
6
High
8
Very High
10
1.0
<5
5–10
10–15
15–25
>25
Source factors
DPS
%
−1
−1
P fertilizer
kg P ha
0.75
0
1–30
30–90
90–150
>150
Livestock manure
kg P ha−1 year−1
0.75
0
1–30
30–90
90–180
>180
Daily life sewage
kg P ha−1 year−1
0.75
0
0–8
8–16
16–24
>24
1.0
<5
5–10
10–15
15–25
>25
>240
year
Transportation factors I: erosion process
Soil erosion
ton ha−1 year−1
−1
Runoff amount
mm year
1.0
<195
195–210
210–225
225–240
Irrigation amount
mm ha−1 year−1
0.75
0
0–380
380–420
420–460
>460
Soil drainage class
0–5
0.85
5
4
3
2
0
Transportation factors II: migration distance
Water course erosion
0–1
0.85
0–0.2
0.2–0.4
0.4–0.6
0.6–0.8
>0.8
Flow length
km
0.85
>50
38–50
28–38
18–28
<18
PI ¼
hX
i hX
i
Sαwα TDβwβ
hX
i
TEγWγ
ð14Þ
where Sα and wα are the source factor α rating and its
weighting value, TDβ and wβ are the transportation
factor β rating and its weighting value, which are based
on migration distance, TEγ and Wγ are the transportation factor γ rating and its weighting value, which are
based on erosion process, PI is P index value.
calculation of the water course factor, was replaced by
an exponential function. In addition, the flow length
factor was not included in the former PI scheme. As
for the organization structure, the weighted summation
approach was applied within each of the transport factor
groups in the amended P index, while all the transportation factors were summarized in the former model, i.e.,
migration distance factors were not distinguished from
the other transportation factors in the former module. An
overview of the changes made in the amended P index is
given in Table 3.
2.6 Comparison of the Amended and Former P Index
2.7 Validation with the Measured P Losses
To validate the amended P index, a previous P index
(Zhang et al. 2003) applied in Yuqiao Reservoir watershed was rebuilt as a reference for the former P index
scheme. In the amended P scheme, the spatial resolution
was based on village scale data while in the previous
study the coarser town scale was used. The finer scale
was selected in order to incorporate the calculation of
relevant source and transport factors (sewage emission,
livestock manure emission, P fertilizer, and irrigation
factor). Data on DPS were used as source factor instead
of STP as DPS is considered to better reflect the potential for loss of dissolved P, which is more relevant than
total P for the eutrophication issue. In the transportation
module, the linear distance function, used for the
How well the former and amended P index schemes
manage to index the measured flux of different P fractions was evaluated using 12 different sub-catchments
within the local watershed. Discharge data was collected
six times a day using a portable river flow meter
(LS300). Water samples were collected three times a
day at the outlet of each sub-catchment during the rainy
season (June 25, 2012 to July 7, 2012) as this is the only
period with significant flow in the streams. During the
monitoring period, the maximum daily rainfall reached
62.8 mm, while the daily average was 10.4 mm. P
concentrations in all samples were determined using
the colorimetric molybdenum blue method according
Water Air Soil Pollut (2014) 225:2103
Page 9 of 16, 2103
Table 3 Comparison of the former and amended P index
Improved content
Former
Source factors
STP (kg ha−1)
DPS (%)
Transportation factors
Water course erosion based on
linear function
Water course erosion based
on exponential function
–
Flow length (km)
Town scale (mean area 477.4 ha)
Village scale (mean area 30.6 ha)
PI=∑(SiWi)×∑(TjWj)*
PI=[∑Sαwα]×[∑TDβwβ]×[∑TEγwγ]
Data resolution:
(P fertilizer application, manure and sewage
P production and irrigation factor)
Organization structure
Amended
*Si and Wi are the source factor i rating, and its weighting value, Tj and Wj, are the transportation factor j rating and its weighting value
to the NS-EN-ISO 6878 (ISO 2003). Prior to analysis,
the phosphorous was fractionated according to the
scheme shown in Fig. 2.
P load (mg s−1) from each sub-catchment was calculated by the flow (L s−1) at the time of sampling multiplied with the concentration of different P fractions
(mg L−1), and the final P load of each sub-catchment
is an average P load.
The flux of measured P fractions were correlated with
the mean P index value of each sub-catchment, which was
separately clipped and extracted from the final P index
grid by Arcgis 10 spatial analysis module (Hillier 2011).
3 Results and Discussion
3.1 Validation by Correlation with Fluxes of P Fractions
A comparison of how well the phosphorous index (PI),
generated by the amended and former P index schemes,
is correlated to the measured flux of different P fractions
from the 12 sub-catchments is shown in Fig. 3. The
Fig. 2 P fractionation scheme
applied on the water samples
former scheme achieved poor correlations with fluxes
of all the P fractions, having correlation coefficients (r)
less than 0.48 for total P and particulate P and only 0.38
for dissolved P (sum of dissolved inorganic and organic
P). With the amended P index scheme, significant correlations were achieved with the flux of total P (r=
0.825, p<0.001) and the dominating particulate P fraction (r=0.867, p<0.0002), implying that the overall
phosphorus flux of P is better ranked by the amended
P index scheme. Soil erosion could be the main cause of
the increasing of particle P in water body, and the
potential for soil erosion was relatively accurately captured by the RUSLE equation in both P index models.
The main cause for the significantly better merits of the
amended P index scheme regarding particulate P loss is
thus the change in the model structure and data resolution. The significance level of the amended P index
factors in regard to the dissolved total phosphorus (r=
0.627, p<0.029) and dissolved inorganic phosphorous
flux (r=0.586, p<0.005) remained moderate. However,
in comparison to the former scheme, the explanatory
capacity for dissolved P fractions has improved
2103, Page 10 of 16
Water Air Soil Pollut (2014) 225:2103
Fig. 3 Comparison of the
average correlation coefficients
between the PI modeled by the
two P index schemes and fluxes
of P fractions from 12 different
sub-catchments. a Former P index
scheme and b amended P index
scheme
significantly (Fig. 3). The main cause for this improvement may be the higher resolution of the main P source
factors, such as manure and sewage production factors,
and the use of DPS instead of STP as the main soil
source factor. The latter is likely the principal cause for
the improved correlation with the dissolved P fractions
as the DPS is prone to better reflect the potential for
mobilization of dissolved P than the STP. These parameter sensitivity assessments are based on conceptual
reasoning. A thorough statistical parameter sensitivity
analysis of the model is in process and will be published
in a forthcoming paper. A relatively weak correlation
and significance level (r = 0.492, p < 0.105) was
achieved between the new P index and the dissolved
organic P. This is likely due to low concentrations of
dissolved organic P (Avg. conc., 0.017 mg/L), which
thus is greatly influenced by “unknown unknowns”
such as direct disposal of sewage into stream channels
or flushing of manure piles during storm flow.
3.2 Spatial Characteristics of Factors Governing the P
Index
allowing for elevated phosphorus mobilization and
thereby high dissolved P concentrations in drainage
water (Pautler and Sims 2000). Approximately 24.6 %
of soil samples exceed the critical limit value (25 %),
above which the potential for P loss through runoff is
found to significantly increase (Breeuwsma et al. 1995).
DPS of the soils exhibited some clear regional distribution characteristics. Areas with relatively high levels of
DPS are mainly situated close to the Yuqiao Reservoir,
especially in the lowlands on the northern shore of
reservoir (Fig. 4a). The lowest overall DPS levels are
found in the mountain region. Relatively high amounts
of inorganic P fertilizers (>150 kg ha−1 year−1) were
applied in the eastern part of the local catchment
(Fig. 4b). This is the region containing most of the
farmland, along with a high population density. A similar distribution can therefore be found for sewage loading (Fig. 4c). The area surrounding the reservoir is also
the region with the majority of livestock, producing a
high loading of livestock manure in this zone (Fig. 4d).
Based solely on the source factors, the highest relative
risk levels for loss of P appears to be mainly distributed in
the region surrounding the Yuqiao Reservoir.
3.2.1 Source Factors
3.2.2 Transportation Factors
The degree of phosphorous saturation (DPS) in the soils
of the local watershed ranged between 0.3 and 50 %. At
high DPS, the phosphorus is poorly retained by the soil,
In contrast to the characteristics of the spatial distribution of source factors, the relatively high level of
Water Air Soil Pollut (2014) 225:2103
Page 11 of 16, 2103
Fig. 4 Spatial distribution maps of source and transportation
factors. Warm colors represent the zone with relatively high risk
levels, while cold colors imply relatively low risk levels. Source
factors: a degree of phosphorus saturation (DPS, %), b fertilizer
phosphorus application rate (kg P ha−1), c daily life sewage
emission (total P) (kg ha−1 year−1), d livestock manure emission
(total P) (kg ha−1 year−1). Transportation factors: e potential soil
erosion (ton ha−1 year−1), f surface runoff based on calculation
(mm), g soil drainage class (0–5), h irrigation
(kiloton ha−1 year−1), i water course (0–1), j flow length (km)
potential soil erosion is mainly found in the mountain
and hilly area, located generally further away from the
Yuqiao Reservoir (Fig. 4e). Concurrently, the region that
receives the highest amount of rain, and thus the highest
potential for surface runoff, lies mainly in the northwest
part of the study area (Fig. 4f).
Soil drainage class mainly depends on soil types and
texture. Soils with good drainage are mainly distributed in
the mountain area with sandy soil, while soils with relative
poor drainage class are found in the plain lowland area
with clay soil (Fig. 4g). A fairly homogeneous distribution
of irrigation on agricultural land was assumed, with the
farmland near the reservoir being slightly more irrigated
(Fig. 4h). The water course factor, reflecting the distance
water travels before it reaches a water channel, depends on
the distribution of rivers and agricultural channel. Fewer
river courses are distributed in the northern part of Yuqiao
Reservoir (Fig. 4i), suggesting a greater possibility for the
water of losing its P load during the catchment transportation in this region. As for the flow length in the river
channels, the northern mountain areas commonly had a
relatively longer transportation distance, due to the greater
distance to the reservoir and stronger morphology, while
the transportation distance is shorter in the plain area,
located close to the reservoir and with a simpler topographic situation (Fig. 4j).
3.2.3 Spatial Characteristics of PI Using the Amended
and Former P Index
The PI results from the two P index schemes were
categorized based on Jenks natural breaks classification
method (McMaster 1997). The PI was subdivided into
five categories as follows: very low, low, medium, high,
and very high (Table 2). Detailed comparison of the
spatial distribution of PI categories from the two P
Index schemes (Fig. 5) reveals that the amended P index
scheme (Fig. 5b) exhibited a higher discrimination,
especially in the plain lowland area with significant
anthropogenic activities. This can be rationalized by that
the source data from the individual village unit, rather
than the town scale, provides a higher resolution in this
area. In addition, a more apparent risk of P loss was
found around the water courses. Furthermore, a high
2103, Page 12 of 16
Water Air Soil Pollut (2014) 225:2103
Fig. 5 P index rating maps: a
results from the former P index
scheme, b results from the
amended P index scheme
risk for P loss around the reservoir was further emphasized in the new scheme, mainly due to the relatively
higher DPS of soil in this region (Fig. 4a).
3.2.4 Spatial Characteristics of the Amended PI Values
In order to further explore the distribution relationship between the amended PI values and corresponding geographic factors, a total of 6,983 grid
points (250 m × 250 m) were extracted from the
study area. In Fig. 6, the PI for each point is plotted
against the distance to each individual village and
watercourses as well as its elevation. Most grid
points are distributed in the regions with less than
“medium” risk level categories. This implies that the
critical source areas with a major potential for P loss
are limited. The three geographic factors showed a
general decreasing trend with increasing PI level.
Most of the grids with high and very high risk for
P loss are found in places that are located closer than
1 km from a village or less than 500 m from a river.
A more significant negative trend was found for the
elevation, indicating that most of the high and very
high risk points are located at an elevation level less
than 100 m above the lake level.
In general, the plain lowland area, riverine zones, and
village regions are the sensitive source areas for P flux to
the reservoir.
3.3 Land-Use Structure Within Different Risk Levels
for P Loss
The areas with high and very high risk for P loss cover
29 % of the local watershed (Fig. 7). With increasing
potential for P loss, there are clear trends of decreasing
proportion of natural land use (forest and shrub) and
increasing share of farmland, with the latter contributing
46 and 86 % of the high and very high risk zones,
respectively. This reflects that the anthropogenic factors
are the primary drivers for P loss within the high risk
zones.
4 Conclusions
The precision of the P index scheme was greatly improved by refining the spatial resolution of input data to
the scale of village unit, introducing more specific indicators in both source and transport system and readjusting the organization structure of the index. The
amended P index scheme is demonstrated to have the
capability of capturing the spatial differences in loss of P
fractions from watersheds with different land use.
Despite that the correlation with the flux of dissolve P
fractions remain relatively moderate, a significant improvement compared to the former scheme was
Water Air Soil Pollut (2014) 225:2103
Page 13 of 16, 2103
Fig. 6 Spatial distributions of PI
values relative to main
explanatory geographic factors. a
distance to the each village, b
distance to each watercourse, c
elevation above reservoir level
(Jenson and Domingue 1988)
achieved, thereby increasing the credibility of amended
P index scheme for the estimation of dissolved P fluxes.
Areas in the vicinity of the reservoir and rivers, the
plain lowland area, and land surrounding the villages
generally constitute a relatively high risk potential for P
Fig. 7 Land-use distribution
within each category of risk for P
loss
loss. In regard to different land use, the study demonstrates that farmlands have the highest potential risk for
loss of P. The areas with “very high” and “high” potential of P loss occupy only a small proportion (29 %) of
the total study area. This emphasizes the need to have
2103, Page 14 of 16
specifically targeted and differentiated abatement actions, rather than inducing a general “flat” target goal.
With the amended P scheme, it is possible to significantly improve both the identification of important hotspot source areas for P flux to surface waters as well as
assess their relative potential for P fluxes in streams and
rivers. This refined risk discrimination and P loss assessment by means of an applicable index is a required
tool for policy makers, enabling them to conduct
knowledge-based decisions regarding environmental
management, ensuring a more sustainable use of the
ecosystem services providing our freshwater resources.
Acknowledgments The authors are grateful to the Research
Council of Norway for funding the SinoTropia Project (Project
no. 209687/E40). The Ji County Environmental Protection Bureau, Ji County Meteorological Bureau, Yuqiao Reservoir Administration Department, Ji County Statistics Bureau, and Ji County
Land and Resources Bureau are all highly acknowledged for their
valuable assistance in providing the essential background data.
The authors would also like to thank the numerous local farmers
and Ms. Ellen Pettersen and Mr. Wycliffe Omondi Ojwando for
their kind assistance during the fieldwork.
References
Agnew, L. J., Lyon, S., Gérard-Marchant, P., Collins, V. B.,
Lembo, A. J., Steenhuis, T. S., & Walter, M. T. (2006).
Identifying hydrologically sensitive areas: bridging the gap
between science and application. Journal of Environmental
Management, 78, 63–76.
Akhtar, M., Corzo, G., Andel, S. v., & Jonoski, A. (2009). River
flow forecasting with artificial neural networks using satellite
observed precipitation pre-processed with flow length and
travel time information: case study of the Ganges river basin.
Hydrology and Earth System Sciences, 13, 1607–1618.
Andersen, H. E., & Kronvang, B. (2006). Modifying and evaluating a P index for Denmark. Water, air, and soil pollution, 174,
341–353.
Bache, B., & Williams, E. (1971). A phosphate sorption index for
soils. Journal of Soil Science, 22, 289–301.
Basta, N., Zupancic, R., & Dayton, E. (2000). Evaluating soil tests
to predict bermudagrass growth in drinking water treatment
residuals with phosphorus fertilizer. Journal of
Environmental Quality, 29, 2007–2012.
Bechmann, M., Krogstad, T., & Sharpley, A. (2005). A phosphorus index for Norway. Acta Agriculturae Scandinavica
Section B Soil and Plant, 55, 205–213.
Bechmann, M., Stålnacke, P., Kværnø, S., Eggestad, H. O., &
Øygarden, L. (2009). Integrated tool for risk assessment in
agricultural management of soil erosion and losses of phosphorus and nitrogen. Science of the Total Environment, 407,
749–759.
Water Air Soil Pollut (2014) 225:2103
Behrendt, H., Lademann, L., Pagenkopf, W.-G., & Pothig, R.
(1996). Vulnerable areas of phosphorus leaching-detection
by GIS-analysis and measurements of phosphorus sorption
capacity. Water Science and Technology, 33, 175–181.
Bingner, R. L., Theurer, F. D., & Yuan, Y. (2003). AnnAGNPS
technical processes. USDA-ARS, National Sedimentation
Laboratory.
Bray, R. H., & Kurtz, L. (1945). Determination of total, organic,
and available forms of phosphorus in soils. Soil Science, 59,
39–46.
Breeuwsma, A., Reijerink, J., & Schoumans, O. (1995). Impact of
manure on accumulation and leaching of phosphate in areas
of intensive livestock farming. Animal Waste and the LandWater Interface, S, 239–249.
Brown, L. R., Renner, M., & Lavin, C. (1997). Vital signs 1997–
1998: the environmental trends that are shaping our future,
Earthscan.
Buczko, U., & Kuchenbuch, R. O. (2007). Phosphorus indices as
risk–assessment tools in the USA and Europe—a review.
Journal of Plant Nutrition and Soil Science, 170, 445–460.
Chai, C., Ding, S., & Shi, Z. (2000). Study of applying USLE and
geographical information system IDRISI to predict soil erosion in small watershed. Journal of Soil and Water
Conservation, 14, 19–24.
Chen, Y., & Zhu, X. (1991). Research on the Yuqiao Reservoir’s
eutrophication and its control measure. Tianjin WaterPower
Press.
Childs, C. (2004) Interpolating surfaces in ArcGIS spatial analyst.
ArcUser, July-September, 32–35.
Cordell, D., Drangert, J.-O., & White, S. (2009). The story of
phosphorus: global food security and food for thought.
Global Environmental Change, 19, 292–305.
Correll, D. L. (1998). The role of phosphorus in the eutrophication
of receiving waters: a review. Journal of Environmental
Quality, 27, 261–266.
CPSC. (2008). The first national census on pollution sources:
coefficient manual of livestock breeding industry pollution
produced and discharged. Beijing: Ministry of
Environmental Protection of China Press.
Drewry, J., Newham, L., & Greene, R. (2011). Index models to
evaluate the risk of phosphorus and nitrogen loss at catchment scales. Journal of Environmental Management, 92,
639–649.
Elliott, H., Brandt, R., Kleinman, P., Sharpley, A., & Beegle, D.
(2006). Estimating source coefficients for phosphorus site
indices. Journal of Environmental Quality, 35, 2195–2201.
Elrashidi, M. (2010). Selection of an appropriate phosphorus test
for soils. Lincoln: USDA NRCS.
Fan, Z., Song, B., Yang, Y., Wang, W., & Chen, Y. (2011). Study
on methods of estimating non-point source pollution load.
Environmental Science Survey, 30, 1–6.
FAO (1950–1995). Fertilizer Yearbook, Rome, Italy.
Hillier, A. (2011). Manual for working with ArcGIS 10.
Huang, J., Hong, H., & Zhang, L. (2004). Study on predicting soil
erosion in Jiulong River watershed based on GIS and USLE.
Journal of Soil and Water Conservation, 18, 75–79.
Imagine, E. (2006). Erdas imagine tour guides. Leica Geosystems,
730.
ISO (2003). Water quality—determination of orthophosphate and
total phosphorus contents by flow analysis (FIA and CFA)—
part 1: method by flow injection analysis (FIA).
Water Air Soil Pollut (2014) 225:2103
JCBS (2011). Ji County Statistical Yearbook 2011. Ji County
Bureau of Statistics.
JCEPB (2012). Ji County Environmental Quality Report (2012). Ji
County Environmental Protection Bureau
Jenson, S., & Domingue, J. (1988). Extracting topographic structure from digital elevation data for geographic information
system analysis. Photogrammetric Engineering and Remote
Sensing, 54, 1593–1600.
Johnes, P., & Heathwaite, A. L. (1997). Modelling the impact of
land use change on water quality in agricultural catchments.
Hydrological Processes, 11, 269–286.
Jokela, B. (1999). Phosphorus index for Vermont: background,
rationale, and questions. Annual meeting of SERA-17
Minimizing P Loss from Agriculture. Quebec City, Canada.
Jokela, W. E., Magdoff, F. R., & Durieux, R. P. (1998). Improved
phosphorus recommendations using modified Morgan phosphorus and aluminum soil tests. Communications in Soil
Science and Plant Analysis, 29, 1739–1749.
Joshi, B. P. (2014). Assessment of phosphorus loss risk from soil—
a case study from Yuqiao reservoir local watershed in north
China. Department of Chemistry, University of Oslo, 95 pp.
Kendy, E., Zhang, Y., Liu, C., Wang, J., & Steenhuis, T. (2004).
Groundwater recharge from irrigated cropland in the North
China Plain: case study of Luancheng County, Hebei
Province, 1949–2000. Hydrological Processes, 18, 2289–
2302.
Lemunyon, J., & Gilbert, R. (1993). The concept and need for a
phosphorus assessment tool. Journal of Production
Agriculture, 6, 483–486.
Li, N., & Guo, H. (2010). Review of risk assessment of phosphorus loss potential from agricultural non-point source: phosphorus index. Progress in Geography, 29, 1360–1360.
Li, Q., Chen, L., & Qi, X. (2007). Catchment scale risk assessment
and critical source area identification of agricultural phosphorus loss. Chinese Journal of Applied Ecology 18.
Ma, C., & Ma, J. (2001). Quantitative assessment of vegetation
coverage factor in USLE model using remote sensing data.
Journal of Soil and Water Conservation, 21, 6–9.
Marjerison, R., Dahlke, H., Easton, Z., Seifert, S., & Walter, M.
(2011). A phosphorus index transport factor based on variable source area hydrology for New York State. Journal of
Soil and Water Conservation, 66, 149–157.
McFarland, A., Hauck, L., White, J., Donham, W., Lemunyon, J., &
Jones, S. (1998). Nutrient management using a phosphorus risk
index for manure application fields. Proceedings of Manure
Management in Harmony with the Environment and Society.
Soil and Water Conservation Society. Ankeny, Iowa, 241–244.
McMaster, R. (1997). In Memoriam: George F. Jenks (1916–
1996). Cartography and Geographic Information Systems,
24, 56–59.
Mehlich, A. (1984). Mehlich 3 soil test extractant: a modification
of Mehlich 2 extractant. Communications in Soil Science and
Plant Analysis, 15, 1409–1416.
MOA (2009). Report of the first national pollution source census
data: emission estimation technique manual for livestock and
poultry breeding source. Ministry of Agriculture of the
People’s Republic of China.
Neitsch, S., Arnold, J., Kiniry, J. & Williams, J. (2011). Soil and
water assessment tool theoretical documentation: version
2009. Texas Water Resources Institute Technical Report
406. Texas A&M Univ. System, College Station.
Page 15 of 16, 2103
Olsen, S. R., Cole, C., Watanabe, F. S., & Dean, L. (1954).
Estimation of available phosphorus in soils by extraction
with sodium bicarbonate. US Department of Agriculture
Washington, DC.
Orderud, G. I.,, & Vogt, R. D. (2013). Trans-disciplinarity required in understanding, predicting and dealing with water
eutrophication. International Journal of Sustainable
Development & World Ecology, 1–12.
Pautler, M. C., & Sims, J. T. (2000). Relationships between soil
test phosphorus, soluble phosphorus, and phosphorus saturation in Delaware soils. Soil Science Society of America
Journal, 64, 765–773.
Pierzynski, G. M. (2000). Methods of phosphorus analysis for
soils, sediments, residuals, and waters. North Carolina State
University Raleigh.
Renard, K. G., Foster, G. R., Weesies, G. A., McCool, D., &
Yoder, D. (1997). Predicting soil erosion by water: a guide
to conservation planning with the revised universal soil loss
equation (RUSLE). Agriculture Handbook (Washington).
Sharpley, A. (1995). Identifying sites vulnerable to phosphorus
loss in agricultural runoff. Journal of Environmental Quality,
24, 947–951.
Sharpley, A. N., Gburek, W. J., Folmar, G., & Pionke, H. (1999).
Sources of phosphorus exported from an agricultural watershed
in Pennsylvania. Agricultural Water Management, 41, 77–89.
Sharpley, A. N., McDowell, R. W., & Kleinman, P. J. (2001a).
Phosphorus loss from land to water: integrating agricultural
and environmental management. Plant and Soil, 237, 287–
307.
Sharpley, A. N., McDowell, R. W., Weld, J. L., & Kleinman, P. J.
(2001b). Assessing site vulnerability to phosphorus loss in an
agricultural watershed. Journal of Environmental Quality,
30, 2026–2036.
Shen, Z., Hong, Q., Chu, Z., & Gong, Y. (2011). A framework for
priority non-point source area identification and load estimation integrated with APPI and PLOAD model in Fujiang
Watershed, China. Agricultural Water Management, 98,
977–989.
Sivertun, Å., & Prange, L. (2003). Non-point source critical area
analysis in the Gisselo watershed using GIS. Environmental
Modelling & Software, 18, 887–898.
Stevens, R., Sobecki, T., & Spofford, T. L. (1993). Using the
phosphorus assessment tool in the field. Journal of
Production Agriculture, 6, 487–492.
Stone, R. P., Ontario. Ministry of Agriculture, F. & Affairs, R.
(2000). Universal soil loss equation, USLE, Ministry of
Agriculture, Food and Rural Affairs.
TMWA (2010). The implementation scheme of pollution source
treatment project around Yuqiao Reservoir. Tianjin municipal water administration.
Tominaga, K. (2013). Lake modelling- an interdisciplinary context. Department of Biosciences, University of Oslo, 162 pp.
Tuo, G., Li, H., Jin, Y., & Li, Y. (2009). Characteristics of phosphorus transfer in farmland under artificial rainfall conditions. Journal of Lake Sciences, 21, 45–52.
Wang, Y. (2010). Indices of phosphorus loss potential from
Ontario agricultural soils to surface waters. The University
of Guelph.
Wang, H., Wang, Q., & Shao, M. (2006). Influence of different
drainage conditions on soil erosion and nutrient loss of slope
land. Science of Soil and Water Conservation, 4, 21–25.
2103, Page 16 of 16
Wang, X., Zheng, L., Zhang, X., & Hao, F. (2010a). Agriculture
non-point source phosphorus loss risk assessment in Yellow
River Basin by modified phosphorus index. Proceedings of
the Annual International Conference on Soils, Sediments,
Water and Energy, p. 29.
Wang, Y., Zhang, T., Hu, Q., Tan, C., Halloran, I., Drury, C., Reid,
D., Ma, B. L., Ball-Coelho, B., & Lauzon, J. (2010b).
Estimating dissolved reactive phosphorus concentration in
surface runoff water from major Ontario soils. Journal of
Environmental Quality, 39, 1771–1781.
Watson, M., & Mullen, R. (2007). Understanding soil tests for
plant-available phosphorus. The Ohio State University
Extension.
Williams, J., Renard, K., & Dyke, P. (1983). EPIC: a new method
for assessing erosion’s effect on soil productivity. Journal of
Soil and Water Conservation, 38, 381–383.
Wischmeier, W. H., & Smith, D. D. (19650). Predicting rainfallerosion losses from cropland east of the Rocky Mountains:
Water Air Soil Pollut (2014) 225:2103
guide for selection of practices for soil and water conservation. Agricultural Research Service, US Department of
Agriculture.
Wischmeier, W. H., & Smith, D. D. (1978) Predicting rainfall
erosion losses—a guide to conservation planning.
Predicting Rainfall Erosion Losses—A Guide to
Conservation Planning, 537–539.
Zhang, S., Chen, L., Fu, B., & Li, J. (2003). The risk assessment of
nonpoint pollution of phosphorus from agricultural lands: a
case study of Yuqiao reservoir watershed. Quaternary
Sciences, 23, 262–269.
Zhou, H., & Gao, C. (2008). Identifying critical source areas for
nonpoint phosphorus loss in Chaohu watershed. Chinese
Journal of Environmental Science, 29, 2696–2702. (in Chinese)
Zhou, H., & Gao, C. (2011). Assessing the risk of phosphorus loss
and identifying critical source areas in the Chaohu Lake
watershed, China. Environmental Management, 48, 1033–
1043.