Access to this full-text is provided by Springer Nature.
Content available from Scientific Reports
This content is subject to copyright. Terms and conditions apply.
Scientic Reports | (2021) 11:4635 |
www.nature.com/scientificreports
Coupling ITO3dE model and GIS
for spatiotemporal evolution
analysis of agricultural non‑point
source pollution risks in Chongqing
in China
Kang‑wen Zhu1, Zhi‑min Yang1, Lei Huang1, Yu‑cheng Chen1*, Sheng Zhang2*,
Hai‑ling Xiong3, Sheng Wu3 & Bo Lei2
To determine the risk state distribution, risk level, and risk evolution situation of agricultural non‑
point source pollution (AGNPS), we built an ‘Input‑Translate‑Output’ three‑dimensional evaluation
(ITO3dE) model that involved 12 factors under the support of GIS and analyzed the spatiotemporal
evolution characteristics of AGNPS risks from 2005 to 2015 in Chongqing by using GIS space matrix,
kernel density analysis, and Getis‑Ord Gi* analysis. Land use changes during the 10 years had a
certain inuence on the AGNPS risk. The risk values in 2005, 2010, and 2015 were in the ranges of
0.40–2.28, 0.41–2.57, and 0.41–2.28, respectively, with the main distribution regions being the
western regions of Chongqing (Dazu, Jiangjin, etc.) and other counties such as Dianjiang, Liangping,
Kaizhou, Wanzhou, and Zhongxian. The spatiotemporal transition matrix could well exhibit the
risk transition situation, and the risks generally showed no changes over time. The proportions of
‘no‑risk no‑change’, ‘low‑risk no‑change’, and ‘medium‑risk no‑change’ were 10.86%, 33.42%, and
17.25%, respectively, accounting for 61.53% of the coverage area of Chongqing. The proportions of
risk increase, risk decline, and risk uctuation were 13.45%, 17.66%, and 7.36%, respectively. Kernel
density analysis was suitable to explore high‑risk gathering areas. The peak values of kernel density
in the three periods were around 1110, suggesting that the maximum gathering degree of medium‑
risk pattern spots basically showed no changes, but the spatial positions of high‑risk gathering areas
somehow changed. Getis‑Ord Gi* analysis was suitable to explore the relationships between hot and
cold spots. Counties with high pollution risks were Yongchuan, Shapingba, Dianjiang, Liangping,
northwestern Fengdu, and Zhongxian, while counties with low risks were Chengkou, Wuxi, Wushan,
Pengshui, and Rongchang. High‑value hot spot zones gradually dominated in the northeast of
Chongqing, while low‑value cold spot zones gradually dominated in the Midwest. Our results provide a
scientic base for the development of strategies to prevent and control AGNPS in Chongqing.
Agricultural non-point source pollution (AGNPS) refers to water pollution caused by nitrogen and phosphorus,
pesticides, and other contaminants through farmland water runo and percolation1. In China, in the preven-
tion and control of water pollution, point sources have mainly been taken into consideration, while non-point
sources are largely being ignored2. However, along with the eective control of point source pollution, AGNPS
has gradually attracted more and more attention. Currently, the premise of eectively solving the problem of
AGNPS lies in the accurate evaluation of the risk state distribution, the risk level, and the risk evolution situ-
ation of AGNPS. In particular, the integration of numerous technologies and methods, such as models for the
calculation of AGNPS3, the universal soil loss equation (USLE)4, and GIS technology5, has greatly promoted the
research on AGNPS in China6.
According to the comparison of the results of the two pollution source surveys released by China in 2020,
the total amount of water pollutant emissions decreased signicantly in the past ten years (2007–2017)7. e
OPEN
Chongqing Academy
College of Computer & Information Science,
*
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Vol:.(1234567890)
Scientic Reports | (2021) 11:4635 |
www.nature.com/scientificreports/
proportion of agricultural sources in chemical oxygen demand, nitrogen pollutants and total phosphorus emis-
sions increased from 43.71%, 41.88%, 67.27% to 49.77%, 40.73%, 67.22% respectively. It shows that the contribu-
tion of AGNPS to water pollution is very high8. e city of Chongqing is characterized by a hilly and mountain-
ous landform, a shattered topography, a high proportion of rural areas in the urban and rural dual structure, a
hot and rainy season, and a high and concentrated precipitation, which result in a high potential threat, a wide
coverage area, and a high driving energy of AGNPS in Chongqing9,10. e application amount of chemical
fertilizer, pesticide and agricultural lm in Chongqing is increasing constantly. In 2014, the application level of
chemical fertilizer, pesticide and agricultural lm reached 411kgha, 9.5kgha and 81.9kgha respectively, which
was far higher than the international standard upper limit. In addition, due to the inuence of topography and
climate on soil erosion, the total amount of soil erosion is 146 million tons per year. In addition, there are other
problems such as high multiple cropping index of agricultural land in Chongqing11. Chongqing is located in the
center of the ree Gorges Reservoir area and represents the connection point of the "Belt and Road" and the
Yangtze River economic belt. Hence, Chongqing has an important status in the national regional development
pattern and is an important ecological barrier of the upstream of the Yangtze River; in this sense, strict water
management strategies are indispensable. To ensure the ecological security of the Yangtze River basin, the issue
of AGNPS needs to be resolved. In view of this, it is of great importance to understand the spatiotemporal evo-
lution situation of AGNPS risks in districts and counties of Chongqing. To sum up, there are several problems
to be solved in research elds and risk assessment of AGNPS. (1) Research on the spatiotemporal evolution of
AGNPS in Chongqing or other large regions are scarce. (2) In terms of risk evaluation, on an international level,
the most commonly used technologies and methods include the export coecient approach12, the pollution
index method (phosphorus index method)13, the multi-factor index evaluation method14, and the model evalu-
ation method of non-point source pollution (NPS)15, and there are few studies on the integration of agronomy
and geography. (3) ere are many studies on input and output dimensions in AGNPS risk assessment16, but
few studies consider translate dimension.
is research adopts the method of combining multi factor index and GIS spatial analysis, which was more
advanced in AGNPS risk analysis at present, and could also enrich the research results in this eld. And Wu built
N and P load calculation models in the ree Gorges Reservoir area by combined with RUSLE equation and
GIS technology, and also conrmed the advantages of the combination of the two methods17. erefore, in our
study, based on the comprehensive analysis of previous studies and regional characteristics of Chongqing6,10, and
combining the advantages of GIS technology and the multi factor index comprehensive evaluation method. An
‘Input-Translate-Output’ three-dimensions evaluation (ITO3dE) model was built to evaluate the risks of AGNPS.
ere were 12 factors involved in the ITO3dE model, including input dimension (fertilizer use intensity, pesticide
use intensity, livestock intensity), translate dimension (erosion caused by rainfall, slope length and gradient, soil
erodibility, sloping eld, distance from water area), and output dimension (water quality, water capacity, water
network density, degree of paddy eld retention), selected by expert consultation and system analytic hierarchy
processes. On this basis, the weights of these factors were determined through multiple assignment by combina-
tion with the Delphi method18.
e spatial overlay analysis function of GIS was adopted to conduct a comprehensive evaluation and com-
bined with the results of related research on safe reference values of various factors. e spatiotemporal variation
of the AGNPS risks in Chongqing from 2005 to 2015 was analyzed by using space matrix analysis, kernel density
analysis, and Getis-Ord Gi*. To reect the characteristics of each factor in the large-scale spatialization process,
it was assumed that the factors fertilizer use intensity and pesticide use intensity were evenly dispersed in the
farmland area, that the livestock intensity factor was evenly dispersed in the suitable breeding area, and that the
paddy eld retention factor was evenly dispersed in the paddy eld area.
erefore, the spatiotemporal changes of the AGNPS risk on a large scale are studied in this research. AGNPS
was mainly caused by the application of chemical fertilizer, pesticides and livestock farming, which were prevalent
in the world. e research method takes into account the three dimensions of input, translate, and output. e
research results have the advantages of a high visibility of risk results, identiable risk levels, and analyzable risk
changes. In addition, the study combines regional land use change and links some factors with a certain type of
land use. Compared with other studies, it can better reect the spatial dierences and could therefore solve the
problems of a wide coverage area of AGNPS and of the diculty of accurately identifying high-risk areas19,20.
Research methods
Data sources. Data types were mainly divided into panel data, remote sensing data, and statistical data, at a
resolution of 30m. e remote sensing data included land use data for 2005, 2010 and 2015, soil type data, digital
elevation model (DEM) terrain data, slope data, river data and zones suitable for breeding. e land use data
were derived from China’s Eco-environmental Remote Sensing Assessment Project with a resolution of 30m21.
e DEM data were obtained from the resource and environment data cloud platform with a resolution of 30m
(http://www.resdc .cn/). e slope data were calculated by DEM data in GIS soware, while the river data were
extracted from high-resolution remote sensing images. Data on zones suitable for breeding were obtained from
the delimited projects about the three livestock and poultry breeding zones (i.e., non-breeding zones, breeding-
restricted zones, and zones suitable for breeding), issued by the Chongqing Agriculture and Rural Aairs Com-
mittee.
e statistical data included fertilizer use, pesticide use, crop planting area, livestock and poultry breeding
data, COD, TN, and TP data, and water generation modulus data. Data on fertilizer use, pesticide use, and crop
planting area were derived from the Chongqing data system (http://www.cqdat a.gov.cn/). Livestock and poultry
breeding data were obtained from the Chongqing Agriculture and Rural Aairs Committee (http://nyncw .cq.
gov.cn/), while data on the levels of COD, TN, and TP were provided by the Chongqing Bureau of Ecology and
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Vol.:(0123456789)
Scientic Reports | (2021) 11:4635 |
www.nature.com/scientificreports/
Environment (h ttp://sthjj .cq.gov.cn/). Water generation modulus data were obtained from the Chongqing Water
Resource Environment Bulletin (http://slj.cq.gov.cn/).
Study area. Chongqing is located in southwestern China and is one of China’s four municipalities directly
under the central government. e current fertilizer application level in Chongqing is 411kgha, while the inter-
nationally recognized safe upper limit is 225kgha. More seriously, the fertilizer use rate is only about 35%. Pes-
ticide use intensity is 9.5kgha, with a use rate of only 30%. Livestock and poultry stocks are currently increasing
signicantly. In addition, the hilly and mountainous areas in Chongqing account for 94% of the total area, while
the cultivated land area with a slope greater than 25 degrees accounts for 16% of the total cultivated land area,
which is 11 percentage points higher than the national average level. Rainfall in Chongqing is large and con-
centrated, and the area subjected to soil erosion accounts for 48.6% of the total area11. Based on the schematic
illustration of the land use types in Chongqing from 2005 to 2015, the articial surface area displays signicant
external diusion in the main urban area of Chongqing, the western region of Chongqing, Changshou County,
Kaizhou County, and Wanzhou County. In contrast, the area of paddy elds shows a signicantly decreasing
trend, while the area of dry land is relatively stable; farmland is mainly distributed in the western region as well
as in Wanzhou County, Kaizhou County, Liangping County, and Dianjiang County in the northeastern region.
According to the data analysis results, the water area accounts for 1.59–1.96% of the coverage area of Chongqing.
e area of paddy elds decreased from 14.8% in 2005 and 2010 to 7.64% in 2015, while the area of dry land
is relatively stable at 25%. e articial surface area has increased from 1.42 to 3.18%, while the proportion of
forested land showed a steady increase (Fig.1). Overall, the farmland area of Chongqing showed a signicantly
decreasing trend, with a reduction from 39.35% to 33.03%. Farmland is the main source of AGNPS. To sum up,
the actual regional conditions of land use types, topography, climate, fertilizer and pesticide application lead to
the aggravation of AGNPS in Chongqing.
Figure1. Land use map of Chongqing in 2005, 2010 and 2015.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Vol:.(1234567890)
Scientic Reports | (2021) 11:4635 |
www.nature.com/scientificreports/
Research methods. e risk assessment of AGNPS in Chongqing was carried out by constructing the
ITO3dE model, and the specic content was analyzed by using GIS technology. e overall framework of the
study was as follows (Fig.2).
ITO3dE model building. e ITO3dE model was built from three dimensions of “Input-Translate-Output”
(Table1). Here, the I dimension (A1) includes fertilizer use intensity (I1)22, pesticide use intensity (I2)23, and
livestock intensity (I3)24, the T dimension (A2) includes erosion caused by rainfall (I4), slope length and gradient
Figure2. e research technology frame.
Table 1. Calculation methods and reference values of the indices in the ITO3dE model.
Dimension Index Calculation method Reference value Source of reference value Direction of index
Input dimension (A1)
I1Chemical fertilizer use intensity per
unit of farmland area 250kgha Building indicators of national eco-
logical town/ county/ city/ province Positive
I2Pesticide use intensity per unit of
farmland area 2.5kgha Internationally recognized secure
usage level Positive
I3
Calculating the livestock pollution load ratio, the deciency ratio of environmentally protective facilities for large-scale
breeding, and the ratio of large-scale breeding to small-scale breeding by the Nemerrow index23 environmental quality
comprehensive evaluation method Positive
Translate dimension (A2)
I4Simplied calculation method of R
value, proposed by Zhou etal.58 100Jcmhah Provisional regulations of ecological
function zoning Positive
I5
Calculating the index by a 7 × 7 win-
dow through ArcGIS neighborhood
statistics 50 Provisional regulations of ecological
function zoning Positive
I6
Calculating the K values (soil erod-
ibility) of various land use types based
on previous related research and the
second overall soil examination results
of Chongqing59
Grading standard in provisional regulations of ecological function zoning Positive
I7
Calculating the index according to the
grading standard of soil erodibility
classication by combining with
farmland data and slope data
Field slope of 15° Grading standard of soil erodibility
classication Positive
I8
Calculating the index by the spatial
distance analysis function in GIS,
based on the data of three-level above
rivers and lakes in the study area
1500m Existing literature and document 60 Negative
Output dimension (A3)
I9Calculating the index by computing pollution indices of COD, NH3-N, and TP, using the Nemerow index Positive
I10
e ratio of overall runo generation
modulus to regional runo generation
modulus in the study area 61.95 Long-term average value during
2005–2015 in Chongqing Negative
I11
Calculating the index according to
the lake and reservoir density and the
river density according to the ecologi-
cal environment status evaluation61
0.025km^2km^-2 and
0.404kmkm^-2 are the reference val-
ues of the lake and reservoir density
and the river density respectively
Long-term average value during
2005–2015 in Chongqing Negative
I12 Proportion of paddy eld area in
farmland area 0.3339 Long-term average value during
2005–2015 in Chongqing Negative
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Vol.:(0123456789)
Scientic Reports | (2021) 11:4635 |
www.nature.com/scientificreports/
(I5), soil erodibility (I6), sloping eld (I7), and distance from water area (I8), and the O dimension (A3) includes
water quality (I9)25, water capacity (I10)26, water network density (I11)27, and degree of paddy eld retention (I12).
e computational formula of positive indices is as follows:
e computational formula of negative indices is as follows:
In the above equations, Ii denotes the calculation result of a certain index of i, Ci is the true value of the i index,
and Ei is the reference value. e calculation method, reference value and source of reference value involved in
ITO3dE model calculation were described in Table1. e calculation results for each index can be classied
into ve grades of no risk, low risk, medium risk, high risk, and extremely high risk, corresponding to the val-
ues of ≤ 0.7, 0.7–1.0, 1.0–3.0, 3.0–5.0, and ≥ 5.0. is grading standard refers to the relevant technical planning
of China’s agricultural sector28, and the grading standard of this study was written based on Chongqing local
standard "Specication for evaluation risk of AGNPS in Chongqing".
Delphi method. e weights of the evaluation indices of AGNPS risks were determined by the Delphi method18.
We distributed 120 questionnaires to the leaders of related departments engaged in agricultural and rural work,
environmental work, and water conservancy programs, as well as to experts at colleges, universities, and research
institutes, and to technical personnel at the grass-roots level; in total, we received 115 valid questionnaires. Aer
the rst round of weights assignment, we returned the calculation results to the experts for adjusting the index
weights, representing the second round of weights assignment. According to the evaluation opinions of the
experts, we again revised the evaluation indices and conducted the third round of weights assignment to obtain
the nal weights results. Based on these, we could obtain the multifactor comprehensive evaluation relational
expressions as follows:
In the above equations, the letter A represents the comprehensive risk result of AGNPS, while the other letters
have the same meanings as in Table1, and indicating the calculation results of each index.
Spatiotemporal transition matrix of risk index. Transition matrices are widely used in analyzing the spatiotem-
poral variations and can clearly exhibit the variations of risk indices at dierent periods29. In this study, we
assigned the ve risk grades to the values of 1, 2, 3, 4, and 5, respectively, and subsequently employed the follow-
ing formula to analyze the risk status:
where B2005, B2010, B2015 is the operation layer and BM is the result layer. In the calculation results, "123" represents
the risk transition process in a certain region from no risk in 2005 to low risk in 2010 and to medium risk in
2015; the remaining results can be interpreted in the same manner.
Kernel density analysis of risk index. Kernel density is mainly used to analyze the spatial concentration of an
event and is widely used in the distribution of buildings, schools, and criminal activities30. e kernel density
can reect the concentration degree and agglomeration location of AGNPS above a high risk level. Kernel den-
sity estimation is a spatial analysis method based on nonparametric testing30,31. e basic idea of kernel density
estimation for certain elements is to assume that there always exists an element intensity at an arbitrarily posi-
tion within a particular region. e density intensity of geographical elements in a specic location can then be
estimated by measuring the number of elements per unit area. By determining the element density at dierent
locations and the spatial dierences, the relative concentration degree of the spatial distribution of elements
can be depicted, and the hotspot distribution regions can be identied. Subsequently, using the kernel density
analysis tool in the ArcGIS soware, we can analyze the grids at risk grades of high risk and extremely high risk
and explore the extreme point position of AGNPS in our study area, Chongqing.
Getis‑Ord Gi* analysis. e Getis-Ord Gi* analysis is widely used in crime analysis, epidemiology and eco-
nomic geography, and is used to identify spatial gathering of high values (hot spots) and low values (cold spots)
with statistical signicance32. In Getis-Ord Gi* analysis, the z score, p values, and condence intervals (Gi_Bin)
are employed to create a new output class for each element in the input element class. Here, the z score and p
values can help to judge whether the null hypothesis can be rejected, while the Gi_Bin eld is used to identify
statistically signicant hot and cold spots. e elements in the condence interval of [+ 3, − 3] have a statistical
signicance with a condence level of 99%, while those in the condence interval of [+ 2, − 2] have a statistical
signicance with a condence level of 95%, and those in the condence interval of [+ 1, − 1] have a statistical
(1)
Ii
=
Ci/Ei
(2)
Ii
=
Ei/Ci
(3)
A=0.430A1+0.231A2+0.339A3
(4)
A1
=
0.547I1
+
0.339I2
+
0.114I3
(5)
A
2=
0.290I
4+
0.098I5
+
0.182I
6+
0.267I
7+
0.163I8
(6)
A3=0.347I9+0.293I10 +0.153I11 +0.207I12
(7)
BM
=
B2005
×
100
+
B2010
×
10
+
B2015
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Vol:.(1234567890)
Scientic Reports | (2021) 11:4635 |
www.nature.com/scientificreports/
signicance with a condence level of 90%; when the element gathering of Gi_Bin eld is 0, there is no statistical
signicance.
Statement. We conrm that all experimental protocols were approved by College of Resources and Envi-
ronment of Southwest University, all methods were carried out in accordance with relevant guidelines and regu-
lations of questionnaires, and conrming that informed consent was obtained from all subjects or, if subjects are
under 18, from a parent and/or legal guardian.
Results
Results of risk assessment by the ITO3dE model. e results in the I dimension show that, overall,
the distribution was high in the west and low in the northeast and the southeast in all three periods (Fig.3, I, II,
III; Table2), and this tallies with the topography of Chongqing. e northwestern and central regions of Chong-
qing are mainly hilly and slightly mountainous, while the southeastern and northeastern regions represent the
Dabashan Mountain system and the Daloushan Mountain system, respectively. us, farmland in Chongqing
is mainly distributed in the western regions as well as in regions with extensive at areas, such as Dianjiang and
Liangping. Some regions in Dianjiang, Yongchuan, Dazu, Shapingba, Wansheng, and Jiangbei show relatively
high risks, but the risk level is still medium. Hence, it can be concluded that the risk level in the I dimension
during 2005–2015 is, overall, not high. Considering there are too many single-factor graphs, we omitted these
graphs, but provide the following description: Among the three single factors, I1 has the highest value, and I1 and
I2 both present a rst increasing and then decreasing trend (the maximum values of I1 in 2005, 2010, and 2015
were 3.38, 4.08, and 2.78, respectively, and those of I2 in 2005, 2010, and 2015 were 2.71, 3.37, and 2.48, respec-
tively). For the I1 results, the risk levels of the regions with higher levels in 2005, such as Yongchuan, Fuling, and
Liangping, showed a certain decrease in 2015, but the risk levels of some regions such as Pengshui, Qianjiang,
and Xiushan showed an increasing trend. e risk grade of I2 was relatively lower than that of I1, but overall, the
spatiotemporal variation trend was consistent with that of I1, except for the increasing trend of the risk level of
Figure3. Result distribution map of I, T, and O dimensions of Chongqing in 2005, 2010, 2015.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Vol.:(0123456789)
Scientic Reports | (2021) 11:4635 |
www.nature.com/scientificreports/
Qianjiang. Basically, the risk grade of I3 was zero; only the risk level of Bishan was in the medium risk status,
while those of Hechuan and Fengdu were low.
Spatially, the results in the T dimension presented, overall, an opposite distribution pattern when compared to
the I dimension, that is, with low levels in the western regions and high levels in the northeastern and southeast-
ern regions (Fig.3, IV, V, VI; Table2). e annual dierences in the T dimension data are mainly determined by
the variations in the factors I4 and I7, which showed relatively higher risk levels in all three periods. e values of
I4 in the years 2005, 2010, and 2015 were 1.42–5.78, 0.84–6.12, and 0.14–6.93, respectively, while those of I7 were
0, 0–5.38, and 0–5.06, respectively. Because Chongqing is a typical mountainous city with purple soil33, high-risk
and extremely high-risk regions, I5 and I6, are widely distributed across the city. In addition, due to the introduc-
tion of the factor I8, the water areas had a higher risk level, which is consistent with the actual situation of AGNPS.
e results in the O dimension showed a smaller interannual variation, with a low overall risk level (Fig.3,
VII, VIII, IX; Table2). e O dimension levels were mainly aected by the spatial changes in the paddy eld
area. As mentioned above, during the 10years, the area of paddy elds in Chongqing was nearly reduced
by half, which led to the decrease in the spatial distribution of I12 and an increased risk in counties such as
Kaizhou, Fengjie, Liangping, and Changshou. Spatially, Yongchuan, Shapingba, Bishan, Dianjiang, Changshou,
and Kaizhou showed higher risk levels, and the risk levels of Kaizhou, Fengjie, Wanzhou, Liangping, and Chang-
shou showed a signicantly increasing trend. e high risk values of I9 were mainly distributed in Yongchuan,
Shapingba, Jiangbei, Changshou, Dianjiang, and Liangping, with Shapingba showing the highest value of 3.75,
while Chengkou, Wushan, Fengjie, Shizhu, and Xiushan had lower values. e high risk values of I10 were mainly
distributed in the western regions and were below the medium risk levels. e risk values in 2010 were higher
than those in 2005 or 2015, but did not surpass 3.0, and the high values were mainly distributed in the western
regions as well as in Dianjiang, Wanzhou, and Liangping. e risk values of I11 were all below 3.0, and the highest
value of 2.78 was found for Fengjie; higher values were mainly distributed in the northeastern and southeastern
counties. e high risk values of I12 were mainly distributed in the northeastern and southeastern counties, which
mostly have only small areas of paddy elds.
Figure4 shows the data on AGNPS risks during 2005–2015 in Chongqing. e risk distribution trends in
2005, 2010, and 2015 were basically consistent and in the ranges of 0.40–2.28, 0.41–2.57, and 0.41–2.28, respec-
tively. e maximum risk values were all below 3.0 for the three periods. Regions with medium levels were
mostly distributed in the western regions of Chongqing (Dazu, Jiangjin, etc.) as well as in the counties Dianjiang,
Liangping, Kaizhou, Wanzhou, and Zhongxian. Larger spatial dierences were observed among dierent coun-
ties or dierent parts of a certain county; for example, the middle atland part and the mountain systems at the
two sides in Liangping or the northwestern and southeastern parts in Shizhu.
Table 2. Statistical results of I, T, and O dimensions in 2005–2015.
Dimension Yea r R ange of value Districts and counties with high value
Input
2005 0–2.44 Yongchuan, Shapingba, Jiangbei, Jiulongpo, Dadukou, Dazu, Tongliang, Diangjiang, Nanchuan
2010 0–3.02 Yongchuan, Dadukou, Shapingba,Fuling, Liangping, Dianjiang
2015 0–2.40 Yongchuan, Liangping, Dadukou, Nan’An, Wansheng, Jiangjin, Jiulongpo, Shapingba, Dianjiang
Transl ate
2005 0.79–4.55 Chengkou, Wuxi, Wushan, Fengjie, Yunyang, Kaizhou, Wanzhou, Shizhu, Zhongxian, Fuling,
Fengdou, Wulong, Pengshui, Qianqiang, Youyang, Xiushan
2010 0.71–4.58
2015 0.66–4.57
Output
2005 0.47–2.01 Yongchuan, Shapingba, South of Wulong,
2010 0.50–2.14 Yongchuan, Tongliang, Bishan, Shapingba, Dadukou, Jiangbei
2015 0.38–2.08 Yongchuan, Shapingba, Rongchang, Bishan, Dianjiang, Jiangbei
Figure4. Spatiotemporal distribution graph of the evaluation results of agricultural NPSP risks in Chongqing
during 2005–2015: (a) 2005; (b) 2010; (c) 2015.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Vol:.(1234567890)
Scientic Reports | (2021) 11:4635 |
www.nature.com/scientificreports/
Spatiotemporal change results of risk by transition matrix analysis. By assigning no risk, low
risk, and medium risk levels with 1, 2, and 3, respectively, in GIS, we can obtain the spatiotemporal transition
matrix according to the formula of the transition matrix. Figure5 shows the spatiotemporal transition situation
of the AGNPS risk evaluation in Chongqing. Basically, high levels show no changes, and the proportions of ‘no-
risk no-change’, ‘low-risk no-change’, and ‘medium-risk no-change’ situations were 10.86%, 33.42%, and 17.25%,
respectively, accounting for 61.53% of the total area of Chongqing. Among these, the ‘no-risk no-change’ situa-
tion was mainly distributed in Rongchang, the east of Nanchuan, Shizhu, Pengshui, and Qianjiang; the ‘low-risk
no-change’ situation was widely distributed in Wulong, the southeast of Fengdu, the south of Nanchuan, and
the northeastern counties of Chongqing, while the ‘medium-risk no-change’ situation was mainly distributed in
Shapingba, Yongchuan, Dianjiang, the north of Nanchuan, and Kaizhou.
During 2005–2015, the proportions of risk increase, risk decline, and risk uctuation were 13.45%, 17.66%,
and 7.36%, respectively. Risk increases mainly occurred in central Jiangjin, central Fengdu, Pengshui, Qianjiang,
the midwest of Yunyang, central Liangping, Wuxi, Wushan, and Chengkou, while risk declines were mainly
observed for the main urban area of Chongqing, northern Tongliang, Dazu, Youyang, and Xiushan. Risk uctua-
tion was concentrated in Jiangjin, Bishan, Fuling, and Youyang.
Results of risk concentration degree by Kernel density analysis. Figure6 shows the kernel density
analysis results of the medium-risk regions. As seen in the gures, the peak values of the kernel density at these
three periods were all around 1,110, suggesting that the maximum gathering degree of medium-risk pattern
spots basically showed no changes. e spatial distribution of kernel density at these three periods showed a
consistent trend, but the distribution dierences at dierent periods were signicant. In 2005, medium-risk
regions were mainly concentrated in Shapingba, southern Dazu, central Yongchuan, eastern Beibei, Dianjiang,
central Kaizhou, northwestern Shizhu, northern Nanchuan, central Wanzhou, southwestern Zhongxian, and
southeastern Xiushan, while in 2010, such regions mainly occurred in Shapingba, eastern Jiangjin, southeastern
Beibei, northern Nanchuan, northeastern Changshou, Dianjiang, northern Fuling, northern Fengdu, northeast-
ern Shizhu, northeastern Liangping, central Kaizhou, Wanzhou, northeastern Pengshui, and eastern Xiushan.
In 2015, medium-risk regions were mainly concentrated in Shapingba, Yongchuan, central Jiangjin, northwest-
ern Nanchuan, northeastern Beibei, Dianjiang, Liangping, the junction of Fuling and Fengdu, central Kaizhou,
northern Yunyang, eastern Pengshui, southeastern Qianjiang, and central Xiushan.
To further explore the distribution of regions with the high-risk gathering zones (Table3), we conducted a
separate analysis on the regions with kernel density values higher than 1,000 (the kernel density values of these
regions were divided into 10 grades with equal intervals, and the 10th grade had values from 1,000 to 1,110).
Figure5. Spatiotemporal transition situation of agricultural NPSP risks in Chongqing during 2005–2015.
Figure6. Kernel density graphs of medium-risk areas in Chongqing during 2005–2015: (a) 2005; (b) 2010; (c)
2015.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Vol.:(0123456789)
Scientic Reports | (2021) 11:4635 |
www.nature.com/scientificreports/
Results of hot and cold spots by Getis‑Ord Gi* analysis. Applying Getis-Ord Gi* analysis is help-
ful to clearly identify high-value hot spots (Hot Spot-99% Condence) and low-value cold spots (Cold Spot-
99% Condence). Figure7 shows the Getis-Ord Gi* analysis results; the overall variation trends of high-value
hot spots and low-value cold spots were consistent in all periods, with signicant distribution dierences. e
regions located in the high-value hot spot zones in all three periods were Yongchuan, Shapingba, Dianjiang,
Liangping, northwestern Fengdu, and Zhongxian, while those located in the low-value cold spot zones were
Chengkou, Wuxi, Wushan, Pengshui, and Rongchang. roughout the 10years, the high-value hot spot zones
showed signicant diusion in Fengjie, Yunyang, Kaizhou, central Qianjiang, and northern Nanchuan, while
the low-value cold spot zones showed signicant diusion in some parts of the midwestern counties such as
central Fuling and southern Yubei. ese high-value hot spots or low-value cold spots were mainly distributed
in the above-mentioned regions and their surrounding areas and showed signicant “gathering trends”. e spa-
tiotemporal variation trend of the distribution of these high-value hot spots or low-value cold spots can reect
the variation tendencies of hot spots or cold spots in dierent regions. Over time, the high-value hot spot zones
gradually migrated towards the northeastern counties of Chongqing, while the low-value cold spot zones in the
midwestern counties presented an obvious diusion trend. e low-value cold spot zones in the northeastern
regions gradually decreased, while those in the southeastern regions tended to become more fragmented. ese
results indicate that the high-value hot spot zones gradually dominated the northeastern regions, while the low-
value cold spot zones gradually dominated the midwestern regions.
Discussion
The I, T, and O dimensions could better comprehensively analyze the risk situation of
AGNPS. ere were many achievements in AGNPS risk research using dierent methods. For instance,
Huang has analyzed the risk of NPS in Taihu Lake from multiple perspectives, and the method was only applica-
ble to small-scale studies35. In addition, it focused on the distribution of pollution sources, and on waste disposal,
but did not consider the process from pollution sources to water bodies. Blankenberg has analyzed the relation-
ship between wetland and pesticide concentration in streams from the perspectives of pesticide spraying and
changes in pesticide concentration in water bodies36. e research focused on the relationship between wetland
and pesticide concentration in small watersheds, while the land use types, topographic conditions, rainfall, and
other factors that aect concentration changes were not considered. is research therefore comprehensively
considers input factors, translate factors and output factors.
e chemical fertilizers, pesticides, livestock and poultry factors selected in the I dimension of this study were
recognized as the main sources of AGNPS37–39. In addition, the factors of the T dimension such as rainfall, slope
length and gradient, soil erodibility, sloping eld, and distance from water area have gradually increased in recent
Table 3. Distribution of regions with high-risk gathering zones.
Years e high-risk gathering zones
2005
e junction of Youting town and Longshui town in Dazu, the central part of Longshui town in Dazu, the east of Shapingba, the
junction of Liuyin town and Sansheng town, as well as the junction of Liuyin town and Longshui town in Beibei, the southwest of
Shiyan town in Changshou, the central part of Baijia town in Dianjiang, the junction of Yantai town and Chengxi town in Dianji-
ang, the junction of Gaofeng town and Gaoan town, as well as the junction of Gaofeng town and Gangjia town in Dianjiang, and
the west of Nantuo town in Fuling
2010
e eastern part of Shapingba, the west of Jiangjin, the north of Youting town in Dazu, the south of Liuyin town and the junction
of Jingguan town and Sansheng town in Beibei, the junction of Gelan town and Shiyan town in Changshou, Yihe town and Nantuo
town in Fuling, the central part of Lidu subdistrict in Fuling, the central part of Baijia town in Dianjiang, the southeast of Chengxi
town in Dianjiang, the junction of Gaofeng town and Gangjia town in Dianjiang, and Linjiang town in Kaizhou
2015 e eastern part of Shapingba, the north of Youting town in Dazu, the central part of Baijia town in Dianjiang, the southeast of
Chengxi town in Dianjiang, the junction of Gaofeng town and Gangjia town in Dianjiang, and the western Linjiang town and
central Wenquan town in Kaizhou
Figure7. Getis-Ord Gi analysis results in Chongqing during 2005–2015.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Vol:.(1234567890)
Scientic Reports | (2021) 11:4635 |
www.nature.com/scientificreports/
years. For instance, Zhang has analyzed the eects of rainfall intensity and slope on the loss of suspended solids
and phosphorus in runo40, which showed that the slope dierence of sloping farmland was also an important
factor aecting pollutant migration41, and the distance from the water reected the diculty degree of pollut-
ants entering the water body42. e inuence factors of the O dimension increased the water network density
and the degree of paddy eld retention factor compared with other studies. Most of the studies mainly focused
on risk the analysis of AGNPS from water quality and water capacity43,44. In Wang etal.34 used the minimum
cumulative resistance model to calculate the risk of ANSP of cultivated land in the ree Gorges Reservoir area,
and believed that the topography, hydrology, soil and vegetation had a great impact on the risk of ANSP. In
fact, high water network density and paddy eld ratio could eectively reduce the pollution. In this study, the
I, T, and O dimensions were considered comprehensively, and factors aecting AGNPS risk were added to the
large-scale framework.
Geographical methods were well applied in the eld of agriculture. e use of geographic meth-
ods in this study makes the results of AGNPS risk analysis more intuitive and reects the temporal and spatial
changes of research results from dierent perspectives. e spatiotemporal transition matrix quanties the vari-
ation in the dierent risk levels and can accurately identify the regions with increasing, decreasing, or unchanged
risk levels. e spatiotemporal transition matrix analysis was one of the most widely used methods in geo-
graphic research and oen used to study the change of the same spatial event over time. For instance, Shawul
and Chakma have used this method to analysis the conversion of land use types in dierent periods45. e kernel
density analysis could eectively identify the spatial agglomeration area of an event and was widely used in geo-
graphic, medical event analysis and case analysis46. e advantage of the Getis-Ord Gi* analysis is to judge the
hot spots of high and low values at the same time and to analyze the evolution trend of hot spots of high and low
values. is method was usually applied in studies that require the analysis of high-value and low-value changes
simultaneously47. Zhang etal.48 used the hot spot analysis method in geography to analyze the spatial and tem-
poral pattern of ANSP in Chongqing section of the ree Gorges Reservoir area, and achieved good results. In
our research, we need to consider both high-risk and low-risk changes in AGNPS.
Land variations inuence AGNPS risks. e analysis of land use changes indicates that the land area
in Chongqing has changed greatly during the 10-year-period from 2005 to 2015. In particular, the paddy eld
area was reduced by nearly 50%. Our observations are in agreement with previous ndings49. e regions with
declining farmland areas were mainly located in the urban areas of Chongqing and in the surroundings of the
town, which goes hand in hand with the economic development and the urban expansion of Chongqing in
recent years, reecting the direct inuences of urban development on AGNPS. is tallies with the decrease in
chemical fertilizer use intensity and pesticide use intensity, as farmland reduction inevitably lead to a decrease
in the application of these substances50, complying with the “one regulatory, two reduction, three basic” policy
promoted by the Chinese government. A previous study has analyzed the relationship between the riparian
forest buer zone and pollution mitigation in Chesapeake bay watershed and found that the wide buer zone
could eectively reduce water pollution51. Other authors have analyzed the relationship between climate, land
use change, and water quality in the Pike river watershed and found that land use changes had an impact on
water quality52. e water quality factors in this study also reected that the water quality in the southeast and
northeast areas with high forest coverage was better, indicating that land use change had a great impact on water
quality, especially on the cultivated land, which was closely related to the use of fertilizers and pesticides.
In this study, we analyzed the spatiotemporal variation in AGNPS in three dimensions of "Input-Transition-
Output", with the conclusion that the risk levels in the Input and Output dimensions were lower, while that in
the Translate dimension was higher. Our conclusion accords with the characteristics of Chongqing as a moun-
tainous city; the widely distributed purple soil, the larger dierences in slope gradient and slope length, and
the widely distributed sloping elds are indeed conducive to the generation of AGNPS53. PENG etal.54 analysis
of the driving factors of ANSP in the ree Gorges Reservoir Area (Chongqing Section) showed that land use
type, agricultural production and living intensity have a greater impact on ANSP, which was consistent with our
research conclusion. From 2005 to 2015, the counties with higher risk levels in all three periods were basically
those national basic farmland demonstration counties (Jiangjin, Dazu, Tongnan, Tongliang, Liangping, Dianji-
ang), with higher risk levels in the Input and Output dimensions. ese areas therefore require planting activities
under the conditions of ensuring soil safety, water quality safety, and food security.
Recommendations. Focus on risk increase area in the analysis of spatiotemporal transfer matrix. In 62%
of our study area, the risk levels remained unchanged, indicating that the inuencing factors of AGNPS in most
regions of Chongqing are relatively stable. For about 14% of the study area, the risk levels showed an increasing
trend, and these regions, especially food security zones such as Jiangjin and Liangping as well as important eco-
logical shelter zones of the ree Gorges Reservoir areas such as Fengdu, Yunyang, Wuxi, and Wushan, need to
be considered in management programs. AGNPS in these areas might seriously threaten the ecological security
of the Yangtze River economic belt. Regional managers should control the use of chemical fertilizers and pesti-
cides and keep them within the international limits55. As Kesavan stated in their research, the use of chemical
fertilizers and pesticides results in enhanced productivity over short periods, but leads to the degradation of soil
health, freshwater, and biodiversity in the long term. In addition, the layout of farms in prohibited and restricted
aquaculture areas should be strictly controlled, and strict total quantity control should be implemented in suit-
able areas. e prohibited and restricted livestock breeding areas in all districts and counties have been revised
by Chongqing Environmental Protection Bureau in 2019, aiming to better manage the layout of livestock farms
and reduce the risk of water environment pollution.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Vol.:(0123456789)
Scientic Reports | (2021) 11:4635 |
www.nature.com/scientificreports/
Focus on high‑risk agglomeration area in nuclear density results. e high values of kernel density changed only
slightly over time, indicating that the maximum scope of gathering zones showed no signicant changes. e
spatial position variation of kernel density at dierent periods is closely related to the changes of the factors in
the three dimensions. e main gathering zones of medium risk levels showed some spatiotemporal dierences
in the three periods, but there are high-risk gathering zones in Shapingba, Yongchuan, Nanchuan, Dianjiang,
Liangping and Kaizhou, and Xiushan. ese phenomena are closely related with the agricultural development
degree of these counties. e analysis results of such zones aer further grading more clearly display the distri-
bution of high-risk gathering zones, which, to a certain extent, reects the distribution of medium-risk pattern
spots in the vicinity of the high-risk gathering zones. ese regions, such as Youting town in Dazu, Baijia town
and Chengxi town in Dianjiang, the junction of Gaofeng town and Gangjia town, and Linjiang town in Kaizhou,
require signicant attention from the respective prevention and control departments. e cultivation system
in high-risk agglomeration areas should be rationally adjusted. e crop rotation system could be adopted to
restore soil fertility while controlling the amounts of chemical fertilizers and pesticides. In addition, increasing
the vegetation coverage of ridges could be considered, and previous studies have proved that these measures
could eectively reduce AGNPS56. Chongqing government has constructed river regulation works and ecologi-
cal corridors in several districts and counties of the Yangtze River Basin, with the purpose of reducing soil ero-
sion near water bodies, increasing the interception capacity of vegetation near water bodies to AGNPS, and
improving the prevention and control ability of AGNPS.
Focus on high‑value hot spots area in Getis‑Ord Gi* analysis. e results of the Getis-Ord Gi* analysis clearly
reect the spatiotemporal evolution situation of high-value hot spot zones and low-value cold spot zones in the
three periods in Chongqing. Over time, the dominance of high-value hot spot zones in the Midwest gradu-
ally became lower than that of the low-value cold spot zones, with a concentrated distribution in Dianjiang,
Zhongxian, Fengdu, Wanzhou, Liangping, Kaizhou, Yunyang, and Fengjie. In the Midwest, the coverage areas
of high-value hot spot zones and low-value cold spot zones were generally neck and neck; by contrast, in the
southeast, low-value cold spot zones dominated. is trend indicates that zones with high AGNPS in northeast-
ern Chongqing have the tendency of further gathering and diusing, and the high-risk areas migrate toward
northeastern Chongqing. Against this background, these regions deserve special attention by decision makers.
Due to the considerable amount of sloping farmlands in these areas, the cultivation of such farmland should be
strictly controlled57. In addition, soil and water conservation measures should be strengthened in these areas. In
recent years, Chongqing Planning and Natural Resources Bureau has changed a large number of slope farmland
into non cultivated land in order to reduce the carrying capacity of soil and water loss to pollutants, and reduce
pollution risk from the source and transmission process of AGNPS.
AGNPS risk assessment is dierent from pollution load measurement. e accuracy of the risk results does
not depend on the experimental or measured data, and may not be positively correlated with the measured data.
is is similar to the multi index system assessment of NPS carried out by Huang for Taihu Lake. And according
to the risk values calculated from multi-angle indicators, the risk dierences in various regions are analyzed and
countermeasures are proposed35. For example, if the I dimension value of an area is high, but the T dimension
is low, that is, the measured data value is high, this does not automatically mean a high risk. e results of risk
assessment are based on the reliability of the data source, factor weight and grading reference value of risk fac-
tors. It is based on the historical data to judge the trend of regional pollution risk, so as to prevent and control
the risk, and this is the advantage of risk assessment in pollution prevention. erefore, in the future, we should
increase investment in data monitoring and acquisition of I, T and O dimensions and scientically analyze the
dierences in the reference values of each factor in dierent regions.
Conclusions
We built an ITO3dE model, analyzed the land use change during 2005–2015 in Chongqing by the combination of
various spatial analysis functions including the spatiotemporal transition matrix in GIS, kernel density analysis,
and Getis-Ord Gi* analysis, and claried the inuences of farmland change on AGNPS risks. e spatiotemporal
variation of the pollution risk indicates that the T dimension had the highest risk level in dierent regions of
Chongqing. is phenomenon was closely related with the widely distributed purple soil, the large dierences
in slope gradient and slope length, and the widely distributed sloping elds. e risk evaluation results obtained
via spatiotemporal transition matrix, kernel density analysis, and Getis-Ord Gi* analysis specically quantied
the regions that require close attention in the prevention and control of AGNPS from the perspectives of risk
level variation, the distribution of risk highly-concentrated zones, and the shi of high-value hot spot zones and
low-value cold spot zones; we therefore provide data to support land use planning strategies as well as measures
to prevent and control AGNPS.
Taken together, there would be signicant inuences on the Input and Output dimensions of AGNPS that
promote the “one regulation, two reductions, and three basics” policies promulgated by the Chinese govern-
ment, indicating areas suitable for livestock breeding in Chongqing and implementing regulatory requirements
in 2018. e implementation of related policies would eectively reduce the risk levels of these two dimensions.
In this study, the results suggest that there is a higher risk level in the Translate dimension. Among its inuenc-
ing factors, eld slope is important and can be improved by land use planning or articial handling. Hence,
related departments should focus on the farmland protection in the high-risk gathering regions, avoid farmland
reclamation as far as possible, and reduce the proportion of sloping elds. At the same time, the migration of
pollutants into water bodies is mainly aected by the degree of soil erodibility, so there would a positive impact
that strengthening the control measures of regional soil erosion on reducing AGNPS.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Vol:.(1234567890)
Scientic Reports | (2021) 11:4635 |
www.nature.com/scientificreports/
Received: 25 November 2020; Accepted: 10 February 2021
References
1. Smith, L. E. D. & Siciliano, G. A comprehensive review of constraints to improved management of fertilizers in China and miti-
gation of diuse water pollution from agriculture. Agr. Ecosyst. Environ. 209, 15–25. https ://doi.org/10.1016/j.agee.2015.02.016
(2015).
2. Liu, G. D., Wu, W. L. & Zhang, J. Regional dierentiation of non-point source pollution of agriculture-derived nitrate nitrogen in
groundwater in northern China. Agr. Ecosyst. Environ. 2, 211–220. https ://doi.org/10.1016/j.agee.2004.11.010 (2005).
3. Hou, X. Y., Ying, L. L., Chang, Y. Y. & Qian, S. S. Modeling of non-point source nitrogen pollution from 1979 to 2008 in Jiaodong
Peninsula, China. Hydrol. Process. 8, 3264–3275. https ://doi.org/10.1002/hyp.9886 (2014).
4. Shen, Z. Y. et al. A comparison of WEPP and SWAT for modeling soil erosion of the Zhangjiachong watershed in the three gorges
reservoir area. Agric. Water Manag. 10, 1435–1442. https ://doi.org/10.1016/j.agwat .2009.04.017 (2009).
5. Basnyat, P., Teeter, L. D., Lockaby, B. G. & Flynn, K. M. e use of remote sensing and GIS in watershed level analyses of non-point
source pollution problems. For. Ecol. Manage. 128, 65–73. https ://doi.org/10.1016/S0378 -1127(99)00273 -X (2000).
6. Wang, G., Li, J., Sun, W. & Xue, B. Non-point source pollution risks in a drinking water protection zone based on remote sensing
data embedded within a nutrient budget model. Water Res. 157, 238–246. https ://doi.org/10.1016/j.watre s.2019.03.070 (2019).
7. Xu, C. X. et al. Emission inventory and characteristics of anthropogenic air pollution sources based on second pollution source
census data in Sichuan Province. Environ. Sci. 10, 4482–4494. https ://doi.org/10.13227 /j.hjkx.20200 2162 (2020).
8. Zhu, Z. P., Dong, H. M., Wei, S., Ma, J. Z. & Xue, P. Y. Impact of changes in livestock manure management on greenhouse gas
emissions in China. J. Agro‑Environ. Sci. 4, 743–748. https ://doi.org/10.11654 /jaes.2020-0095 (2020).
9. Li, J. et al. Spatiotemporal variations of ambient volatile organic compounds and their sources in Chongqing, a mountainous
megacity in China. Sci. Total Environ. 627, 1442–1452. https ://doi.org/10.1016/j.scito tenv.2018.02.010 (2018).
10. Zhou, J. et al. Divergent patterns of soil phosphorus discharge from water-level uctuation zone aer full impoundment of ree
Gorges Reservoir, China. Environ. Sci. Pollut. Res. 3, 2559–2568. https ://doi.org/10.1007/s1135 6-018-3805-1 (2019).
11. Hu, J. F. Measurement and spatial feature analysis of agricultural non-point source pollution in Chongqing. Chin. J. Agric. Resour.
Region. Plann. 38, 135–144. https ://doi.org/10.7621/cjarr p.1005-9121.20170 121 (2017).
12. Cai, Y. P., Rong, Q. Q., Yang, Z. F., Yue, W. C. & Tan, Q. An export coecient based inexact fuzzy bi-level multi-objective program-
ming model for the management of agricultural nonpoint source pollution under uncertainty. J. Hydrol. 557, 713–725. https ://doi.
org/10.1016/j.jhydr ol.2017.12.067 (2018).
13. Ouyang, W., Huang, H. B., Hao, F. H., Shan, Y. H. & Guo, B. B. Evaluating spatial interaction of soil property with non-point
source pollution at watershed scale: e phosphorus indicator in Northeast China. Sci. Total Environ. 16, 412–421. https ://doi.
org/10.1016/j.scito tenv.2012.06.017 (2012).
14. Mao, Z. P., Yin, C. Q., Shan, B. Q. & Li, G. B. Multi-factor model for analyzing the regulation eect of wetland area on storm runo.
J. Hydr. Eng. 7, 57–63. https ://doi.org/10.3321/j.issn:0559-9350.2002.07.010 (2002).
15. Strehmel, A., Schmalz, B. & Fohrer, N. Evaluation of land use, land management and soil conservation strategies to reduce non-
point source pollution loads in the three gorges region, China. Environ. Manag. 5, 1–16. https ://doi.org/10.1007/s0026 7-016-0758-3
(2016).
16. Xu, Z. A., Li, T., Bi, J. & Wang, C. Spatiotemporal heterogeneity of antibiotic pollution and ecological risk assessment in Taihu
Lake Basin China. Sci. Total Environ. 643, 12. https ://doi.org/10.1016/j.scito tenv.2018.06.175 (2018).
17. Wu, L., Long, T. Y., Liu, X. & Ma, X. Y. Modeling impacts of sediment delivery ratio and land management on adsorbed non-point
source nitrogen and phosphorus load in a mountainous basin of the ree Gorges reservoir area, China. Environ. Earth Sci. 70,
1405–1422. https ://doi.org/10.1007/s1266 5-013-2227-0 (2013).
18. Strand, J., Carson, R. T., Navrud, S., Ortiz-Bobea, A. & Vincent, J. R. Using the Delphi method to value protection of the Amazon
rainforest. Ecol. Econ. 131, 475–484. https ://doi.org/10.1016/j.ecole con.2016.09.028 (2017).
19. Karandish, F. & Simunek, J. J. Two-dimensional modeling of nitrogen and water dynamics for various N-managed water-saving
irrigation strategies using HYDRUS. Agric. Water Manag. 193, 174–190. https ://doi.org/10.1016/j.agwat .2017.07.023 (2017).
20. Rabotyagov, S. S., Valcu, A. M. & Kling, C. L. Reversing property rights: Practice-based approaches for controlling agricultural
nonpoint-source water pollution when emissions aggregate nonlinearly. Am. J. Agr. Econ. 2, 397–419. https ://doi.org/10.1093/
ajae/aat09 4 (2012).
21. Ouyang, Z. Y., Wang, Q., Zheng, H., Zhang, F. & Hou, P. National ecosystem survey and assessment of China (2000–2010). Bull.
Chin. Acad. Sci. 29, 462–466. https ://doi.org/10.3969/j.issn.1000-3045.2014.04.009 (2014).
22. Liu, Q.P. Distribution of fertilizer application and its environmental risk in dierent provinces of China. Scientia Agricultura Sinica
47, Scientia Agricultura Sinica. https://doi.org/https ://doi.org/10.3864/j.issn.0578-1752.2014.18.008 (2014).
23. Liu, S. C., Wang, L. W., Li, X. X., Chen, Y. C. & Fu, M. M. Potential risk assessment of heavy metal in biogas slurry irrigation. J.
Plant Nutr. Fertil. https ://doi.org/10.11674 /zwyf.2014.0622 (2014).
24. Cai, J. Z. et al. Sources analysis of agricultural non-point source pollution in the three gorges reservoir Area of Hubei Province,
China. J. Agro‑Environ. Sci. 7, 1421–1430 (2012).
25. Lai, Y. C., Yang, C. P., Hsieh, C. Y., Wu, C. Y. & Kao, C. M. Evaluation of non-point source pollution and river water quality using
a multimedia two-model system. J. Hydrol. 3, 583–595. https ://doi.org/10.1016/j.jhydr ol.2011.08.040 (2011).
26. Chen, D. J. et al. Uncertainty analysis of water environmental capacity in the nonpoint source polluted river. Environ. Sci. 31,
1215–1219. https ://doi.org/10.1631/jzus.A1000 244 (2010).
27. Qiu, Z., Geng, Q. W. & Zang, G. M. Statistics and countermeasures of non-point source pollution in plain river network of Taihu
Basin. Eng. Sci. 12, 113–116. https ://doi.org/10.3724/SP.J.1088.2010.00432 (2010).
28. Chen, Y. & Zhang, D. Historical evidence for population dynamics of tibetan migratory locust and the forecast of its outbreak*.
Insect Sci. 2, 135–145. https ://doi.org/10.1111/j.1744-7917.1999.tb001 60.x (2010).
29. Zhu, K. W., Li, Y. C. & Zhou, M. T. Land use scenario simulation of the main city of Chongqing based on the CLUE-S model.
Resour. Environ. Yangtze Basin 5, 789–797. https ://doi.org/10.11870 /cjlyz yyhj2 01505 011 (2015).
30. Okabe, A., Satoh, T. & Sugihara, K. A kernel density estimation method for networks, its computational method and a GIS-based
tool. Int. J. Geogr. Inf. Sci. 193, 7–32. https ://doi.org/10.1080/13658 81080 24754 91 (2009).
31. Kristan, M. & Leonardis, A. Online discriminative kernel density estimator with gaussian kernels. IEEE Trans. Cybern. 3, 355–365.
https ://doi.org/10.1109/TCYB.2013.22559 83 (2017).
32. Gómez-Barroso, D. et al. Spatial clustering and risk factors of malaria infections in Bata district Equatorial Guinea. Malar. J. 1,
146. https ://doi.org/10.1186/s1293 6-017-1794-z (2017).
33. Wang, J., Huang, J., Wu, P. & Zhao, X. N. Application of neural network and grey relational analysis in ranking the factors aecting
runo and sediment yield under simulated rainfall. Soil Res. 3, 291. https ://doi.org/10.1071/sr150 68 (2016).
34. Wang, J. L., Xie, D. T., Shao, J. A., Ni, J. P. & Lei, P. Identication of source-sink risk pattern of agricultural non-point source pollu-
tion in cultivated land in ree Gorge Reservoir Area based on accumulative minimum resistance model. Trans. Chin. Soc. Agric.
Eng. 16, 206–215. https ://doi.org/10.11975 /j.issn.1002-6819.2016.16.029 (2016).
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Vol.:(0123456789)
Scientic Reports | (2021) 11:4635 |
www.nature.com/scientificreports/
35. Huang, L., Ban, J., Han, Y. T., Yang, J. & Bi, J. Multi-angle indicators system of non-point pollution source assessment in rural areas:
A case study near Taihu Lake. Environ. Manag. 4, 939–950. https ://doi.org/10.1007/s0026 7-013-0024-x (2013).
36. Blankenberg, A. G. B., Braskerud, B. & Haarstad, K. Pesticide retention in two small constructed wetlands: Treating non-point
source pollution from agriculture runo. Int. J. Environ. Anal. Chem. 3–4, 225–231. https ://doi.org/10.1080/03067 31050 02474 70
(2006).
37. Nopmongcol, U. et al. Impact of dedicated E85 vehicle use on ozone and particulate matter in the US. Atmos. Environ. 39,
7330–7340. https ://doi.org/10.1016/j.atmos env.2011.07.057 (2011).
38. Partt, R. L., Yeates, G. W. & Ross, D. J. Eect of fertilizer, herbicide and grazing management of pastures on plant and soil com-
munities. Appl. Soil. Ecol. 3, 175–186. https ://doi.org/10.1016/j.apsoi l.2010.03.010 (2010).
39. X ia, L., Lam, S. K., Yan, X. & Chen, D. How does recycling of livestock manure in agroecosystems aect crop productivity, reactive
nitrogen losses and soil carbon balance?. Environ. Sci. Technol. 13, 7450–7457. https ://doi.org/10.1021/acs.est.6b064 70 (2017).
40. Zhang, R. R., Li, M., Yuan, X. & Pan, Z. C. Inuence of rainfall intensity and slope on suspended solids and phosphorus losses in
runo. Environ. Sci. Pollut. Res. 1, 1–13. https ://doi.org/10.1007/s1135 6-018-2999-6 (2018).
41. He, X., Xu, Y. & Zhang, X. Traditional farming system for soil conservation on slope farmland in southwestern China. Soil Tillage
Res. 1, 193–200. https ://doi.org/10.1016/j.still .2006.07.017 (2007).
42. Birge, T., Toivone, M., Kaljonen, M. & Herzona, I. Probing the grounds: Developing a payment-by-results agri-environment scheme
in Finland. Land Use Policy 61, 302–315. https ://doi.org/10.1016/j.landu sepol .2016.11.028 (2017).
43. Ahmed, I. U., Shou, W. & Gang, D. D. Nonpoint source pollution. Water Environ. Res. 82, 1875–1894. https ://doi .o r g/10.2175/10614
3010x 12756 66880 2012 (2014).
44. Heiderscheidt, E., Leiviskä, T. & Kløve, B. Chemical treatment response to variations in non-point pollution water quality: Results
of a factorial design experiment. J. Environ. Manag. 150C, 164–172. https ://doi.org/10.1016/j.jenvm an.2014.10.021 (2015).
45. Shawul, A. A. & Chakma, S. Spatiotemporal detection of land use/land cover change in the large basin using integrated approaches
of remote sensing and GIS in the Upper Awash basin Ethiopia. Environ. Earth Sci. 5, 141. https ://doi.org/10.1007/s1266 5-019-
8154-y (2019).
46. Chrisinger, B. W. et al. Characteristics of automated external debrillator coverage in Philadelphia, PA, based on land use and
estimated risk. Resuscitation 109, 9–15. https ://doi.org/10.1016/j.resus citat ion.2016.09.021 (2016).
47. Stopka, T. J. et al. Identifying and characterizing hepatitis C virus hotspots in Massachusetts: a spatial epidemiological approach.
Bmc Infect. Dis. 1, 294. https ://doi.org/10.1186/s1287 9-017-2400-2 (2017).
48. Zhang, G. N., Shao, J. A., Wang, J. L., Ni, J. P. & Xie, D. T. Spatial and temporal variations of agricultural non-point source pollution
in the three gorges reservoir area of Chongqing. J. Nat. Resour. 7, 1197–1209. https ://doi.org/10.1 1849 /zrzyx b.2015.07.012 (2015).
49. Liu, C. Y., Zhu, K. W. & Liu, J. P. Evolution and prediction of land cover and biodiversity function in Chongqing section of ree
Gorges Reservoir Area. Trans. Chin. Soc. Agric. Eng. 33, 258–267. https ://doi.org/10.11975 /j.issn.1002-6819.2017.19.034 (2017).
50. Zaragozí, B. et al. Modelling farmland abandonment: A study combining GIS and data mining techniques. Agr. Ecosyst. Environ.
155, 124–132. https ://doi.org/10.1016/j.agee.2012.03.019 (2012).
51. Carlington, W., Gregory, M. C., Sangchul, L. & Robert, P. B. Evaluating concentrated owpaths in riparian forest buer contributing
areas using LiDAR imagery and topographic metrics. Remote Sens. Inf. 10, 614. https ://doi.org/10.3390/rs100 40614 (2018).
52. Mehdi, B. et al. Simulated impacts of climate change and agricultural land use change on surface water quality with and without
adaptation management strategies. Agr. Ecosyst. Environ. 213, 47–60. https ://doi.org/10.1016/j.agee.2015.07.019 (2015).
53. Yang, M. C. et al. Non-point pollution of three gorges reservoir region in Chonqing: Assessment and aecting factors. Environ.
Sci. Technol. 35, 179–183. https ://doi.org/10.3969/j.issn.1003-6504.2012.05.039 (2012).
54. Peng, T., Shao, J. A., Wang, J. L., Ni, J. P. & Xie, D. T. Analysis of driving factors aecting agricultural non-point source pollution in
chongqing section of the three gorges reservoir area. J. Southwest Univ. 3, 126–135. h t t ps ://do i .o r g/10.13718 /j .cnki.xdzk.2016.03.021
(2016).
55. Kesavan, P. C. & Malarvannan, S. Green to evergreen revolution; ecological and evolutionary perspectives in pest management.
Curr. Sci. 99, 908–914. https ://doi.org/10.1371/journ al.pone.00132 19 (2010).
56. Braskerud, B. C. Factors aecting phosphorus retention in small constructed wetlands treating agricultural non-point source
pollution. Ecol. Eng. 19, 41–61. https ://doi.org/10.1016/S0925 -8574(02)00014 -9 (2002).
57. Zhao, P., Tang, X. Y., Zhao, P., Wang, C. & Tang, J. L. Tracing water ow from sloping farmland to streams using oxygen-18 isotope
to study a small agricultural catchment in southwest China. Soil Tillage Res. 134, 180–194. https ://doi.org/10.1016/j .still .2013.08.005
(2013).
58. Zhou, F.J., Huang, Y.H. e R value of rainfall erosion force index in Fujian province. Journal of Soil and Water Conservation,
13–18 (1995).
59. Wu, C. G. et al. Spatiotemporal distribution characteristics of rainfall erosivity in ree Gorges Reservoir Area. Chin. J. Appl. Ecol.
22, 151–158 (2011).
60. Lv, C., Zhang, H. Y. & Qi, L. Critical source area identication and risk assessment of agricultural non-point source pollution of
the source areas of liaohe river watershed. Acta Agric. Univ. Jiangxiensis https ://doi.org/10.13836 /j.jjau.20141 10 (2014).
61. Yao, R., Wang, S. X., Zhou, Y., Liu, R. & Han, X. D. e application of ecological environment index model on the national evalu-
ation of ecological environment quality. Remote Sens. Inf. 27, 93–98. https ://doi.org/10.3969/j.issn.1000-3177.2012.03.016 (2012).
Acknowledgements
is work was funded by Chongqing Science and Technology Commission (cstc2018jszx-zdyfxmX0021,
cstc2018jxjl20012 and cstc2019jscx-gksbX0103). We are thankful to all experts involved in the Delphi survey.
anks are expressed to the Chongqing Agricultural and Rural Committee for providing information on ferti-
lizers, pesticides, and livestock and to the Chongqing Eco-environment Bureau for providing information on
water-related issues. We also thank the and Ecological Restoration Engineering Technology Center of the Water
Level Fluctuation Zone in the ree Gorges Reservoir Area in Chongqing for providing respective information.
Author contributions
K.Z. and Y.C. wrote the main manuscript text, Z.Y. and L.H. were responsible for the questionnaire, H.X. and S.W.
were responsible for processing data, and S.Z. and B.L. prepared gures. All authors reviewed the manuscript.
Competing interests
e authors declare no competing interests.
Additional information
Correspondence and requests for materials should be addressed to Y.C.orS.Z.
Reprints and permissions information is available at www.nature.com/reprints.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Vol:.(1234567890)
Scientic Reports | (2021) 11:4635 |
www.nature.com/scientificreports/
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional aliations.
Open Access is article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the
Creative Commons licence, and indicate if changes were made. e images or other third party material in this
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/.
© e Author(s) 2021
Content courtesy of Springer Nature, terms of use apply. Rights reserved
1.
2.
3.
4.
5.
6.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:
use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
use bots or other automated methods to access the content or redirect messages
override any security feature or exclusionary protocol; or
share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at
onlineservice@springernature.com
