8 Lee Et Al 2023

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

Catena 225 (2023) 107028

Contents lists available at ScienceDirect

Catena
journal homepage: www.elsevier.com/locate/catena

Applicability evaluation of agricultural Best Management Practices to


estimate reduction efficiency of suspended solids
Gwanjae Lee a, Seoro Lee b, Jiyeong Hong c, Yonghun Choi a, Jonggun Kim b, Kyoung Jae Lim b,
Jae E. Yang d, Won Seok Jang e, *
a
ILEM Research Institute, 1 Gangwondaehak-gil, Chuncheon-si, Gangwon-do 24341, Republic of Korea
b
Department of Regional Infrastructure Engineering, Kangwon National University, 1 Gangwondaehak-gil, Chuncheon-si, Gangwon-do 24341, Republic of Korea
c
Department of Earth and Environment, Boston University, Boston, MA 02215, USA
d
Department of Biological Environment, Kangwon National University, 1 Gangwondaehak-gil, Chuncheon-si, Gangwon-do 24341, Republic of Korea
e
Department of Ecological Landscape Architecture Design, Kangwon National University, 1 Gangwondaehak-gil, Chuncheon-si, Gangwon-do 24341, Republic of Korea

A R T I C L E I N F O A B S T R A C T

Keywords: Nonpoint source (NPS) pollution affects water quality, and the path of leakage and discharge of NPS pollutants is
Nonpoint-source pollution not clear. NPS pollution is heavily influenced by weather conditions such as precipitation, making NPS pollution
Best management practices difficult to manage. The Soyang River watershed, in the upper reaches of the Han River, is dominated by
MUSLE
mountain farmland, which causes the water to become turbid due to soil erosion during rainfall. Therefore, South
SWAT
Korea’s Ministry of Environment has designated Mandae District as an NPS pollution management area to install
and manage various NPS pollution reduction facilities (NPRF). However, research on the NPS reduction effi­
ciency is insufficient. Therefore, a Soil and Water Assessment Tool (SWAT) model was used to evaluate soil loss
and suspended solids (SS) reduction efficiency by NPRF installed in the Mandae District. In this study, hydrologic
response units (HRUs) were modified to spatially consider agricultural fields in steep-sloping areas in South
Korea, and soil loss, as well as SS reduction efficiency, were calculated reflecting temporal/spatial characteristics
of best management practices (BMPs). In order to accurately evaluate the NPRF, this study digitized all fields in
the Mandae District and applied them to land use maps where slope and slope length were entered for each
agricultural field. To this end, the digitized boundaries of all agricultural fields have been updated into land use
maps. In addition, the estimated soil loss was corrected by calibrating the exponent and coefficient of the MUSLE
(Modified Universal Soil Loss Equation) formula. Reduction efficiency was assessed by comparing the amount of
soil loss and SS before and after the installation of the NPRF with a scenario-based analysis. As a result, the soil
loss reduction efficiency of the NPRF in Mandae District was 7.8% and the SS reduction efficiency was 5.8%. Both
soil loss and SS were sensitive to slope length in evaluating the effectiveness of reduction. The results showed that
the longer the slope length, the greater the soil loss or SS.

1. Introduction surface runoff, soil erosion, and air pollution) are the main sources of
NPS pollution (Ding and Liu, 2019). NPS pollution has been identified as
Climate change causes various changes in the water ecosystem and a major problem that has a significant impact on water quality world­
has a huge impact on the hydrological cycle, such as runoff. (IPCC, 2014; wide (Chen et al., 2009; Keller et al., 2014; Kourakos et al., 2012; Zhang
Ritter and Shirmohammadi, 2001; Ding and Liu, 2019; U.S.EPA, 2016; et al., 2020). NPS pollution management is inherently uncertain as the
Carpenter et al., 1998; Tao et al., 2003). Changes in the watershed water route of runoff and discharge of pollutants is not clear and depends
cycle have a significant impact on the behavior of non-point source heavily on weather conditions such as precipitation (Carpenter et al.,
(NPS) pollution, as the movement of NPS pollution is closely related to 1998; MOE, 2016). Moreover, the extent of occurrence and difficulties in
rainfall and runoff processes that are affected by climate change (Ritter monitoring NPS pollution make it difficult to accurately calculate NPS
and Shirmohammadi, 2001). Various sources of water pollution (e.g., pollution loads (Ye et al., 2017).

* Corresponding author.
E-mail address: wsjang@kangwon.ac.kr (W.S. Jang).

https://doi.org/10.1016/j.catena.2023.107028
Received 25 August 2021; Received in revised form 15 February 2023; Accepted 16 February 2023
Available online 24 February 2023
0341-8162/© 2023 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
G. Lee et al. Catena 225 (2023) 107028

In particular, soil erosion is a significant environmental problem that combination of all similar land use, soil, and slope within a sub-basin
negatively affects agricultural productivity, water quality, aquatic (Sophocleous and Perkins, 2000; Ouyang et al., 2009; Ouyang et al.,
ecology and river morphology (Lin et al., 2016; Peter Heng et al., 2010). 2012). The current HRU concept in SWAT has limitations because the
Land use and landscape type play an important role in soil erosion (Qi same slopes calculated as average slopes in a sub-basin are applied to all
et al., 2012; Wang et al., 2009), and various researchers have studied the HRUs in the same sub-basin. To overcome this limitation, Jang et al.
impact of land use type on soil erosion and analyzed and reported (Jang et al., 2009) developed Spatially Distributed-HRU (SD-HRU)
landscape patterns (Da Silva et al., 2015; Huang et al., 2013; Xiao and Ji, modules capable of applying actual slope and slope length, and
2007; Zhang et al., 2017). compared flow and soil erosion responses, considering the SD-HRU
In Korea, areas where there are severe water quality problems caused module. The SD-HRU module was developed to apply the actual slope
by NPS pollution, or where NPS pollution needs to be prevented in and slope length by assigning individual IDs to the HRU taking into
advance, are designated as NPS management areas. NPS management account the spatial distribution of soil maps when overlaying land use
areas include areas that require careful water quality management, such and soil data. However, since the SD-HRU was developed based on a
as priority management reservoirs, marine-protected areas, and land use map, it has the disadvantage of not reflecting the exact location
groundwater conservation districts. The Soyang River watershed located of the agricultural field at that point in time. A disadvantage made it
in the upper reaches of the Han River is densely populated with highland challenging to apply the NPRF and estimate the exact mitigation effi­
fields, which causes frequent problems with muddy water due to soil ciency of arable land where multiple NPRFs were applied to each agri­
erosion during extreme rainfall events. In 2006, due to heavy rainfall, a cultural field to the model. A large amount of soil erosion has occurred
large amount of muddy water flowed into the Soyanggang Dam, causing with rainfall in the Mandae District located in the Soyang River water­
environmental, ecological, and economic damage, such as the destruc­ shed, which adversely affects water quality and the ecosystem (MOE,
tion of aquatic habitats, damage to fisheries, death of fish in fish farms, 2017a). South Korea’s Ministry of Environment has designated the
decrease in local revenues due to the declining number of tourists and an Mandae District as an NPS pollution management area to install and
increase in water treatment costs. The aforementioned economic losses manage various NPRFs. However, since measured data has been estab­
were estimated at around KRW 9 billion (USD 6.9 million). Soil erosion lished in 2009 in the Mandae District, the measured data before the
from agricultural fields has generated various harmful substances and installation of the NPRF is insufficient. There are limitations to NPRF
also caused various water pollution problems when it enters water efficiency analysis due to the small amount of flow and suspended solid
bodies (e.g., streams and rivers) (MOE, 2017a). For the sustainable (SS) data measured. In addition, there are rare studies on watersheds
management of the Soyang River watershed, the Ministry of Environ­ where highland fields are dense and many NPRFs have been installed,
ment has made efforts to reduce soil erosion from arable land by such as in Mandae district. In this study, SWAT was used to evaluate the
applying NPS pollution reduction facilities (NPRF), also known as Best soil erosion and SS reduction efficiency of the NPRF installed in the
Management Practices(BMPs)(gabion, filter strips, grassed waterway, Mandae District. For a more accurate estimation, the slope and slope
sediment trap, drainage water management, etc.) to each arable land length of each arable land were calculated using Google Earth and
(NIER, 2009). BMPs prevent soil erosion and reduce water pollution by Geographic Information System (GIS) software (ArcGIS)(ESRI, 2012),
similar and other pollutants in agricultural fields.(Himanshu et al., existing land use was updated by digitizing all the actual arable land
2019; Lam et al., 2011; Mwangi et al., 2015). BMPs are divided into boundaries using satellite images, and NPRF were applied to the actual
structural BMPs and non-structural BMPs. Structural BMPs are natural arable land. In addition, on the basis of literature reviews, for each HRU,
or man-made structures installed in the target area, non-structural BMPs the reduction efficiency of individual NPS pollution reduction facilities
reduce sediment and pollution through agricultural practices such as was investigated.
fertilizer reduction.(Nepal and Parajuli, 2022). In order to evaluate the reduction efficiency of NPRF in the water­
However, using the NPRF, it is difficult to calculate the reduction shed, both the temporal and spatial characteristics of the NPRF were
efficiency because various conditions (i.e., area of arable land, slope and considered. In Mandae District, the number of NPRF changes every year
slope length) have to be considered. In order to effectively manage NPS from 2006 to 2017 due to the installation and loss of NPRFs, temporal
pollution, it is necessary to address both the possible sources of pollution changes should be considered. In addition, it is necessary to consider the
and the watershed management. In particular, watershed management, spatial characteristics of the NPRF because spatial characteristics such as
which is a form of land management, is very important because NPS area, slope, slope length, and altitude are different for each agricultural
pollution is transported to water reservoirs through various channels field where the NPRF is installed.
during rainfall. Therefore, it is important to investigate the effect of However, in most studies that have evaluated the NPS pollution
NPRF installed on each arable land in the watershed (Yu and Zhen, reduction efficiency using a watershed model, the reduction efficiency
2004). A watershed-scale hydrological model should be required, to has been evaluated by applying variations in the amount of pesticides
assess the impact of reduced efficiency of NPRF applied to arable lands used, vegetative filter strips (Zhang and Zhang, 2011), crops, and slopes
in the watershed (Borah and Bera, 2004). Watershed scale hydrological instead of applying various NPRFs (Briak et al., 2019; Merriman et al.,
models such as Soil and Water Assessment Tool (SWAT)(Arnold et al., 2018). Therefore, in this study, NPRFs were applied to changes every
1998), Hydrological Simulation Program – Fortran (HSPF)(Johanson, year in Mandae District from 2006 to 2017. In addition, a modification
1980), and Agricultural Policy/Environmental eXtender Model (APEX) of the HRU distribution method and MUSLE (Modified Universal Soil
(Williams et al., 1998) are widely used for watershed-scale evaluation. Loss Equation) calibration was applied to evaluate soil erosion and the
Lee et al. (Lee et al., 2010) estimated the effect of reducing NPS pollution effectiveness of SS NPRF reduction installed in Mandae District,
by applying the NPRF using the SWAT model in the Gyeongancheon considering the temporal and spatial characteristics of the NPRF.
Watershed. Mishra et al. (2009) used HSPF to evaluate the modeling of The novelties of this study compared to other SWAT applications are;
NPS pollution impact on small-scale watersheds and Tuppad et al. 1) consideration of spatially distributed HRUs based on agricultural field
(Tuppad et al., 2010) used APEX to study the effects of NPRF on long- boundaries, instead of the traditional combination of land use, soil and
term surface runoff, soil erosion and nutrient runoff in fields and small topography in the conventional SWAT HRU delineation method, 2) in­
watersheds.However, the HSPF has limitations in considering storm dividual NPRFs installed each year (2006–2 NPRF), (2007–18 NPRF),
sewer networks or pollutant reduction facilities (Mohamoud et al., (2008–31 NPRF), (2009–29 NPRF), (2010–1 NPRF), (2011–6 NPRF),
2010). SWAT can simulate various NPRF for agricultural areas taking (2012–11 NPRF), (2013–58 NPRF), 2017–41 NPRF) are included in
into account climate, land use and soil, and the effect of NPRF on stream SWAT calibration/validation to reflect the spatial and temporal impact
water quality. SWAT simulates responses of the hydrologic system based of the NPRF on hydrology and sediment yield, 3) instead of the default
on Hydrologic Response Units (HRU) which are the smallest unit and soil erosion model at HRU level, MUSLE, MUSI (MUSLE approach using

2
G. Lee et al. Catena 225 (2023) 107028

input coefficients) was calibrated using local data and then used to 2. Materials and methods
estimation of sediment yield from each spatially distributed HRU.
MUSLE was developed using rainfall-runoff data for 778 single rainfall 2.1. Study area
events observed in 18 watersheds in Nebraska and Texas with a plain
watershed area of 15–1,500 ha, a slope of 0.9–5.9%, and a slope length Mandae District was designated as an NPS pollution management
of 78.6–173.7 m. Therefore, the use of MUSLE is not appropriate for area in 2015 to reduce soil erosion in the Inbukcheon Watershed and
field-level sediment estimation. These three approaches will provide includes Haean-myeon, Yanggu-gun and Seohwa-myeon, Inje-gun,
more accurate sediment behaviors from field level to watershed level Gangwon province in South Korea (Fig. 2). The area of Mandae District
when applied together to simulate what is happening in the watershed is about 64 km2, the maximum / minimum / average altitude is 1,300 /
from calibration/validation to scenario analysis (Fig. 1). 348 / 640 m, respectively. The maximum and average slope in the
In addition, the method of this study proposes a method to improve Mandae District is approximately 113% and 28%, and the average
the accuracy of SWAT modeling in an area with a large number of steep- annual precipitation over the past10 years (2009–2018) is about 1,196
slope farmlands around the world. In addition, a method for estimating mm. Forests and agricultural fields are the main land use in Mandae
soil loss and SS is presented considering the efficiency of BMPs in areas District, with forests and agricultural fields accounting for approxi­
where a large number of NPRFs are applied. These proposals can be used mately 62.4 and 35.8 %, respectively. The main crops are ginseng, po­
to study land use changes (growth and reduction of agricultural land) tatoes, beans, radishes and green onions (MOE, 2017b). There are
and SWAT modeling uncertainty reduction before and after BMP approximately 446 structural/non-structural NPRFs installed in over
application. 230 locations (more than 100 HRUs) in Mandae District (MOE, 2018)
(Table 1).

Fig. 1. The novelty of this study. There are three new approaches such as 1) application of modified HRU distribution method, 2) application of measured BMPs, and
3) calibration of soil erosion using modified MUSLE. These three approaches will provide more accurate sediment behaviors from field to watershed scales.

3
G. Lee et al. Catena 225 (2023) 107028

Fig. 2. Location of the study area. Mandae District was designated as an NPS pollution management area in 2015 for the soil erosion reduction in the Inbukcheon
Watershed and includes Haean-myeon, Yanggu-gun and Seohwa-myeon, Inje-gun, Gangwon province in South Korea.

Table 1 Table 2
List of abbreviations and symbols used in the manuscript. Type and number of NPRF installed in Mandae district.
Abbreviation Definition Type of NPRF Number of Scheduled Management Operations
NPRF (reduction efficiency)
BMPs Best Management Practices
NPS Non-point source Open channel 30 Generic conservation practice(30%)
NPRF NPS pollution reduction facilities Gabion 87 Generic conservation practice(30%)
SWAT The Soil and Water Assessment Tool Ditch restoration 2 Generic conservation practice(30%)
SS Suspended solids Drop fall structure 5 Generic conservation practice(30%)
MUSLE Modified Universal Soill Loss Equation Roads pavement 1 Generic conservation practice(30%)
HRU Hydrology Response Units Spillway 1 Generic conservation practice(30%)
SD-HRU Spatially Distributed-HRU Mattress 1 Generic conservation practice(30%)
GIS Geographic Information System Drainage water 53 Generic conservation practice(30%)
MUSI MUSLE approach that uses input coefficients management
VFS Vegitation Filter Strip Grade stabilization 96 Generic conservation practice(30%)
.OPS file The Scheduled Management Operations structure
MSDH Modified Spatially Distributed-HRU Irrigation reservoir 1 Generic conservation practice(30%)
NSE Nash-Sutcliffe Model Efficiency Waterway 30 Generic conservation practice(26%)
USLE Universal Soil Loss Equation Vegetative mattress 35 Generic conservation practice(30%)
MOE Ministry of Environment (South Korea) Vegetated ridge 3 Generic conservation practice(74.5%)
WAMIS Water Resources Management Information System Vegetated block 2 Generic conservation practice(30%)
ASF Alaska Satellite Facility (US) Grassed waterway 19 Generic conservation practice(41.7%)
RDA Real Development Administration (South Korea) Vegetated gabion 12 Generic conservation practice(30%)
KMA Korea Meteorological Administration (South Korea) wall
R __ Multiply by value Filter strips 4 Filter strip
V__ Replaced by value Vegitation soil back 29 Generic conservation practice(30%)
NPS Unit Load 1. Amount of pollutants generated from unit area per unit time Under drainage 2 Generic conservation practice(30%)
method 2. It was calculated with the event mean concentration Diversion dike 16 Generic conservation practice(26%)
measured during rainfall and the annual pollutant loads for Drainage inlet 8 Generic conservation practice(30%)
each land use Sediment trap 7 Generic conservation practice(54.1%)
Concrete barrier 2 Generic conservation practice(30%)
Summary 446
In this manuscript, abbreviations and symbols used throughout the
manuscript are listed in Table 1.
The Scheduled Management Operations(.ops) file was used to apply 2.2. Evaluation of SS reduction efficiency
NPRF in the SWAT model. Of the 10 options, we used two options in the.
ops file, which are generic conservation practice and the filter strip. In In this study, the soil loss and SS NPRF reduction efficiency applied
the case of reduction facilities without monitored reduction efficiency, to each of the arable lands were calculated. For this purpose, all arable
the reduction efficiency was assumed to be 30%, considering the un­ land in the study area was digitized and the model was calibrated and
certainty (Table 2). validated applying Modified Spatially Distributed-HRU (MSDH)

4
G. Lee et al. Catena 225 (2023) 107028

method, reflecting digitized arable land for land use. In addition, soil even though the location of the HRU in the study area is different
loss was calculated by modifying the MUSLE coefficients and indices (Ramos et al., 2015). So it is challenging to consider accurate topo­
(Williams, 1975) to reflect domestic land use (South-Korean generated graphic information of the actual agricultural field in the current SWAT.
land use data) to calculate the exact soil loss and soil loss reduction SWAT applies the same slope length of 9.1 m for slopes greater than 25
efficiency. Finally, soil loss and SS reduction efficiency at the watershed %. Therefore, it is difficult to accurately calculate the amount of soil loss
outlet were calculated to see the efficiency of the NPS reduction facilities in areas with steep slopes. The land use (30x30m) provided by WAMIS
applied to each of the arable lands in the watershed (Fig. 3). (Water Resources Management Information System) represents one
agricultural area, regardless of how many small pieces of agricultural
fields are in one agricultural area, so it is impossible to confirm the
2.3. SWAT model and input data
boundaries of each agricultural field. As such, there are many cases
where the HRU does not reflect the boundaries of an actual agricultural
The Soil and Water Assessment Tool (SWAT) model is a semi-
field and is created with the size of more than one actual agricultural
distributed model that is widely used for long-term hydrologic pre­
field. This not only makes it difficult to accurately calculate the amount
dictions in the watershed and evaluates the temporal and spatially
of soil loss but also has the disadvantage that it is difficult to apply the
altered characteristics of rainfall and runoff based on various inputs data
NPRF for each cultivated area. For example, when three NPRFs are
related to environmental impacts (e.g., precipitation, soil, topographic
installed on each of the three agricultural fields where HRUs are created
and land use change) in the watershed. SWAT can simulate the function
in an area with the same soil types, the following differences exist. 1) If
of various NPS agricultural facilities and evaluate the runoff character­
the existing HRU method is used, three NPRFs must be applied to one
istics of water, sediment, and chemicals through NPS agricultural fa­
HRU because the three agricultural fields are created as one HRU
cilities. Thus, this study used a SWAT model to evaluate the impact of
because the soil type is the same. 2) If the MSDH method is used, the
NPS agricultural facilities (i.e., vegetation mats, water pipelines and
NPRF can be applied to different HRUs because the three agricultural
gabion retaining walls) on agricultural fields.
fields are created with different HRUs even if the soil type is the same.
Essential SWAT model input data include soil, land use, topographic
Due to the above differences, it is necessary to use the MSDH method to
and weather data. Since topographic data for the study area is not
consider the actual environment when calculating the NPRF efficiency
available, we used a digital elevation model (DEM, 12.5 m resolution)
in agricultural fields.
provided by the Alaska Satellite Facility (ASF), and land use data was
In order to overcome these limitations, as mentioned above, we
acquired by the Water Resources Management Information System
digitized the boundaries of all actual agricultural fields in the study area
(WAMIS, 30 m resolution). Soil data was provided by the Real Devel­
using remotely sensed satellite images (Fig. 5). Using this digitized
opment Administration (RDA, 1:50,000) (Fig. 4). Weather data was
agricultural field boundary information, it was verified whether each
garnered by the Korea Meteorological Administration (KMA)(e.g.,
actual agricultural field was cultivated in the highland field GIS con­
temperature, precipitation, wind speed, relative humidity and daylight
struction project (MOE, 2017b) for the NPS pollution management area
hours) and the weather station in the study area is Inje station with an
selected by the Ministry of Environment of the Republic of Korea. The
average annual rainfall of 1,196 mm over the last decade, daily average
digitized agricultural field boundaries were checked and corrected to be
high temperature of 16.7℃, daily average low temperature of 5.4℃,
the same as the actual agricultural field boundaries, and the corrected
average daylight hours of 5.6 h, average relative humidity of 67 %, and
boundaries were applied in this study. Agricultural fields in the land use
daily average wind speed of 1.8 km/h.
map that were not actually cultivated have been classified as bare lands
in the updated land use map to more accurately estimate soil loss from
2.4. Modified Spatially Distributed-HRU (MSDH) method agricultural fields. The land use map was updated to reflect the modified
agricultural field boundaries of the agricultural fields (Fig. 6(a)). Each
In the SWAT model, the amount of soil loss is calculated for each farm has been assigned unique HRU IDs (e.g., A001, A002, and A003) so
HRU using the MUSLE equation (Behera and Panda, 2006; Jeong et al., that they could be distinguished even when creating an HRU within
2011). The HRU, which is the SWAT smallest simulation unit, is SWAT. The slope and slope length calculated from actual measurements
considered the same HRU if the land use, soil and slope are the same

Fig. 3. Procedure of evaluation of BMPs reduction efficiency. 1) Simulation of SWAT by applying the MSDH method, 2) calibration of soil erosion by applying the
modified MUSLE, and 3) SS calibration and SS reduction efficiency estimation.

5
G. Lee et al. Catena 225 (2023) 107028

Fig. 4. Input data of Soil and Water Assessment Tool (SWAT) for the Mandae watershed ((a) DEM, (b) Land Use Map, (c) Soil Map). *The abbreviations have the
following meaning. Land uses: URMD = Urban, Medium Density; AGRR = Agricultural land, row crops; FRST = Forest, mixed; AGRC = Small Grains; WATR =
Water. Soil types: An = Complex of soils, narrow valleys; Ma = Lithosols, siliceous crystalline; odF = sandy loam soil; Ra = Red-yellow podzolic soils, Re = Lithosols,
severely eroded, siliceous; Rocky = Rocky lands(No soils).

Fig. 5. Digitized agricultural field map. All agricultural fields in Mandae district were digitized using remotely sensed satellite images.

6
G. Lee et al. Catena 225 (2023) 107028

Fig. 6. Updated land use map and HRUs reflecting actual agricultural field. *The abbreviations have the following meaning. Land uses: URMD = Urban, Medium
Density; AGRR = Agricultural land, row crops; FRST = Forest, mixed; AGRC = Small Grains; WATR = Water; UPLAND = agricultural fields reflecting digitized
agricultural field. BMPs = Best Management Practices; HRUs = Hydrologic Response Units.

for each agricultural field and currently installed NPRF were applied to water flow data.
the HRU reflecting the actual agricultural field boundary (Fig. 6(b)). The For the parameters used for calibration and validation of the model,
slope of each agricultural field was provided by the Ministry of Envi­ 9 variables sensitive to water flow simulation were selected taking into
ronment (GIS system of the alpine region) (MOE, 2018), and the slope account the sensitivity analysis from previous studies (Arnold et al.,
length was measured using Google Earth (Google, 2001). The NPRF type 2012; Park et al., 2014; Zhang et al., 2011; Khalid et al., 2016; Me et al.,
and location information used in this study was collected from the 2015). The parameters selected were CN2, SOL_AWC, ESCO and CANMX
Wonju Regional Environmental Office, and the tabular data was trans­ which affects the direct runoff, LAT_TTIME sensitive to the lateral flow,
formed into geospatial data. and GW_DELAY, GW_REVAP, GWQMN, ALPHA_BF, REVAPMN, and
SOL_K which affect the baseflow. There are three methods to adjust
2.5. SWAT calibration and validation parameter values in SWAT-CUP: ‘Add to value’, ‘multiply by value (%)’
and ‘replaced by value’. ‘Add to value’ is used to add or subtract a
Proper model calibration and validation is very important to reduce parameter value to or from the initial value of each parameter. It is
uncertainty and increase the accuracy of hydrologic modeling (Moriasi adjusted by adding or subtracting the initial parameter value. ‘Replaced
et al., 2007; Engel et al., 2007). The SWAT-CUP was developed at the by value’ is used to adjust the parameter by replacing the initial
Eawag Institute in Switzerland to facilitate the calibration and valida­ parameter value. ‘Multiply by value’ is a modification of a parameter by
tion process in hydrologic modeling using SWAT (Abbaspour et al., multiplying the initial value of the parameter by a value within the
2007). SWAT-CUP is a window-based independent program that pro­ allowable range set (Table 3).
vides several model calibration and validation algorithms such as SUFI- The optimization algorithm for calibration and validation is SUFI-2
2, GLUE, PARASOL, MCMC and PSO(Abbaspour et al., 2007). (Sequential Uncertainty Fitting ver.2), which is the easy-to-use among
Observed water flow and SS data obtained from the measured data the five algorithms provided by SWAT-CUP and has been validated for
and evaluation project (2009–2017) for the management of NPS pollu­ its usefulness in various previous studies (Abbaspour et al., 2007;
tion in the Soyang river watershed were used to calibrate and validate Akhavan et al., 2010; Rostamian et al., 2008). The uncertainty in the
the SWAT in this study. Daily runoff was simulated from 2006 to 2017 SWAT calibration and validation process is indicated by various model
using SWAT, and calibration and validation of water flow were per­ evaluation methods (Lee et al., 2014; Moriasi et al., 2007). In this study
formed using SWAT-CUP. The warm-up period of the model was from the coefficient of determination (R2) and Nash-Sutcliffe Model Effi­
2006 to 2009, and the calibration and validation periods were from ciency (Nash and Sutcliffe, 1970) (NSE) were used among the various
2009 to 2014 and from 2015 to 2017, respectively. There are 105 model evaluation methods. R2 ranges from 0 to 1, and is an index that
measured water flow data for a rainfall event from 2009 to 2017. 43 indicates the degree of the linear relationship between the measured
measured water flow data exist from 2009 to 2014 and there are 62 flow data and the simulated results. The equation is as follows (Equation 1).
measurement data from 2015 to 2017. Calibration and validation pe­
riods were determined in consideration of the number of measured

7
G. Lee et al. Catena 225 (2023) 107028

Table 3 ( )0.56
SWAT parameters for streamflow calibration (Streamflow). MUSLE or Sediment = 11.8 × Qsurf × qpeak × areahru × KUSLE
Parameter Description Calibration Range × CUSLE × PUSLE × LSUSLE × CFRG (3)
Method

CN2 NRCS runoff curve number for Multiply by 35 to 98 where MUSLE is the sediment yield on a given day, Qsurf is the surface
moisture Condition II value runoff volume, qpeak is the peak runoff rate, areahru is the area of the
ESCO Soil evaopration compensation factor Replaced by 0 to 1 HRU, KUSLE is the USLE soil erosion factor, CUSLE is the USLE coverage
value
and management factor, PUSLE is the USLE support practice factor,
LAT_TTIME Lateral flow travel time Replaced by 0 to 180
value
LSUSLE is the USLE topographic factor and CFRG is the coarse fragmen­
CANMX Maximum canopy storage Replaced by 0 to 100 tation factor.
value MUSLE uses total runoff (Qsurf) and peak runoff (qpeak) (Qsurf ×
ALPHA_BF Baseflow alpha factor Replaced by 0 to 1 qpeak × areahru ) as runoff factors instead of USLE rainfall energy factor
value
GW_DELAY Groundwater delay time Replaced by 0 to 500
(RUSLE), and MUSLE runoff coefficient and exponent are explained in
value Equation (3), which are 11.8 and 0.56, respectively.
GWQMN Threshold depth of water in the Replaced by 0 to The MUSLE coefficients and exponents suggested by Williams (Wil­
shallow aquifer required for return value 5000 liams, 1975) were developed for the Great Plains of the United States, so
flow to occur
it is difficult for MUSLE to reflect the characteristics of a forest basin
GW_REVAP Groundwater “revap” coefficient Replaced by 0.02 to
value 0.2 (Kim et al., 2015). Therefore, in order to calculate the amount of soil loss
REVAPMN Threshold depth of water in the Replaced by 0 to 500 in a watershed in the Republic of Korea where forest and steep slope
shallow aquifer for “revap” to occur value areas predominate, it is necessary to modify the MUSLE coefficient and
SOL_AWC Available water capacity of the soil Multiply by 0 to 1 exponent in order to accurately estimate the soil loss in the watershed in
layer value
SOL_K Saturated hydraulic conductivity Multiply by 0 to
the republic of Korea. Won and Lee (Won and Lee, 2012) proposed co­
value 2000 efficients and exponents of 5.33 and 0.91, respectively. Kim et al. (Won
and Lee, 2012) estimated coefficients and exponents in a small forest
catchment and they presented them as 0.0002 and 0.81, respectively.
( ∑n ( )( ) )2
yobs,i − Y obs ysim,i − Y sim Lee et al. (Lee et al., 2019) calculated the amount of soil loss by modi­
R2 = ∑n (i=1 (1)
)2 ∑n ( )2 fying the MUSLE coefficient and exponents according to rice paddies,
i=1 yobs,i − Y obs i=1 ysim,i − Y sim
alpine fields and forests in the watershed using SWAT.
where _Yobs, i is the i-th observation, _Ysim, i is the i-th simulation, _Yobs is Based on Lee et al. (Lee et al., 2019)’s study, the amount of soil loss
the mean of the observations, _Ysim is the mean of the simulations and n was calculated by modifying the MUSLE coefficient and exponents for
is the total number of observations. each land use. To this end, the SWAT source code was modified so that
NSE is widely used to compare differences between measured and the MUSLE coefficients and exponents could be applied for any land use.
predicted values in hydrologic modeling. The NSE is a normalized sta­ In addition, the coefficients and exponents of MUSLE were calculated to
tistic that determines relative magnitude by comparing the residual the extent that the difference between the result values of the basic unit
variance and measured data variance (Nash and Sutcliffe, 1970). The data quantified for each land use and the average amount of soil loss per
NSE value ranges from − ∞ to 1, where 1 means that the observed value unit area of the highland fields is within 5%. The average amount of soil
and the predicted value are perfectly matched (Equation 2). loss per unit area for rice paddies and forests was calculated on the basis
( ∑n ( )2 ) of the new base unit of the National Institute of Environmental Research
i=1 yi
obs
− ysim (rice paddy: 0.168 ton/ha/yr, forest: 0.055 ton/ha/yr) (RIG, 2014). The
NSE = 1 − ∑n i
(2)
obs
− ymean )
2 amount of soil loss in alpine fields was calculated based on the amount of
i=1 (yi
soil loss (55 tons/ha/yr) suggested in the report on NPS pollution
where Yobs sim
i is the ith observation for the evaluated constituent, yi is the reduction in highland fields (MOE, 2004).
ith simulated value for the evaluated constituent, ymean is the mean of Two scenarios were simulated and compared to analyze the reduc­
observed data for the evaluated constituent, and n in the total number of tion efficiency of NPRF. The same flow rate correction parameters and
observations. One NSE value indicates perfect agreement between ob­ MUSLE coefficients and exponents were used in all scenarios. Scenarios
servations and simulations. 1 and 2 were simulated without and with applying NPRF, respectively.
The reduction efficiency of the NPRF was analyzed in terms of the
2.6. Analysis of the effectiveness of BMP reduction amount of soil loss and SS. To evaluate the reduction efficiency of the
amount of soil loss, the amount of soil loss generated in all HRUs in
To assess the soil loss reduction efficiency for the NPRF, the total Scenario 1 and Scenario 2 was divided into 9 sub-watersheds (defined by
amount of soil loss in the watershed was estimated when the NPRF was the Ministry of Environment in Korea) (Fig. 7) and compared to 9 sub-
installed, and after installation, the amount of soil loss caused by rainfall watersheds.
and runoff was estimated using MUSLE (Williams, 1975) in SWAT. In order to effectively manage NPS pollution, watershed-level man­
MUSLE is a modified version of USLE (Universal Soil Loss Equation) agement is important, so the impact of NPRF on water quality was
developed by Wischmeier and Smith(Wischmeier and Smith, 1965). evaluated using SWAT. This study is an area where muddy water is a big
USLE uses the function of rainfall energy when simulating average environmental issue. The Ministry of Environment in Korea manages SS
annual erosion but MUSLE uses total runoff to simulate erosion and which causes turbidity through continuous monitoring (MOE, 2017a).
sediment yield (Jang and Kim, 2017). MUSLE was derived from runoff In this study, since the amount of soil loss for each land use was calcu­
data for 778 single rainfall events observed in 18 watersheds in lated using the improved MUSLE, the SS calibration was conducted
Nebraska and Texas with a watershed area of 15-1,500 ha, a slope of using only the PRF (Table 4) that affects routes, except for variables such
0.9–5.9%, and a slope length of 78.6–173.7 m. Equation (3) describes as USLE_K and USLE_P that directly affect MUSLE.
MUSLE, and the calculation method for any factor other than the runoff
factor is the same as for USLE.

8
G. Lee et al. Catena 225 (2023) 107028

reduction efficiency of BMPs was evaluated by calculating the sediment


yield (Merriman et al., 2019). evaluated the efficiency of complex BMPs
by applying various BMPs scenarios, and Strauch et al. (2013) evaluated
the BMPs reduction efficiency of SSs. However, as in this study, there
have been few studies evaluating soil loss and SSs reduction efficiency
after accounting for the spatial location of the various NPRFs actually
applied and soil loss calibration.. The measured flow and simulated flow
by SWAT modeling at watershed outlets were calibrated and validated
using SWAT-CUP, and the accuracy of simulated flow the was evaluated
using R2 and NSE. CN2 was the most sensitive parameter in flow cali­
bration, and GW_DELAY, SOL_K, and GWQMN were found to be sensi­
tive parameters after CN2. CN2 is a function of the SCS runoff curve
number and is considered the most sensitive variable as it is directly
related to the amount of infiltration. Since GW_DELAY, SOL_K, and
GWQMN are variables related to baseflow in SWAT (Abraham et al.,
2007), they are also considered sensitive variables in flow simulation.
The parameters optimized using SWAT-CUP are shown in Table 5. The
higher the value of CN2, the greater the initial runoff occurs and the the
lower amount of infiltration. Since the area of the watershed is not large
and the average slope is about 28%, there can be a large amount of direct
runoff due to rainfall, and the CN is calibrated as high. GW_REVAP can
be described as the closer to 0, the more restricted the movement of
water from the shallow aquifer to the root zone. If ALPHA_BF is between
0.1 and 0.3, it means that the recharge reaction is slow. According to Gee
(1987), the groundwater recharge rates in areas covered with deep-
rooted vegetation such as forests are lower than those with thin-
rooted vegetation such as grass. All forests in this study area are
broad-leaved forests (FRSD) and broad-leaved forests create deciduous
forest layers that prevent soil infiltration (Gerrits et al., 2007). There­
fore, ALPHA_BF was calculated as 0.165 because the groundwater
recharge rate is slow due to infiltration by vegetation and fallen leaves
during rainfall. Fig. 8(a) and Fig. 8(b) show a graph of the correlation
between the measured and simulated flow at the outlet. The R2 and NSE
calibration results analyzed were 0.75 and 0.71, respectively, and the
Fig. 7. Sub-watershed presented by the Ministry of Environment. Study area
was divided into 9 sub-watersheds from No. 101 to No. 109, and the reduction
validation results showed that the R2 and NSE were 0.84 and 0.79,
efficiency was compared with each sub-watersheds. respectively. According to Saleh et al. (2000), NSEs of 0.55 ~ 0.65 was
evaluated as “adequately simulated”, and an NSE greater than 0.65 was
evaluated as “extremely well simulated”. Based on Moriasi et al. (2015),
Table 4 if R2 is 0.75 ~ 0.85, it can be regarded as “Good”, and if NSE is 0.5 ~ 0.7,
SWAT parameter for SS calibration. it can be defined as “Satisfactory”. If an NSE is between 0.7 and 0.8, it
Parameter Description Calibration Range can be evaluated as “Good”. Therefore, the results of calibration and
Method validation of the SWAT simulation were a good representation of natural
PRF Peak rate adjustment factor for Replaced by 0 to 2 phenomena, as both R2 and NSE met the SWAT performance criteria.
sediment routing in the main channel value In order to apply modified MUSLE to SWAT modeling, this study
used a new SWAT engine with modified source codes. At this point, the.
txt file containing each variable information can be loaded so that the
3. Results and discussion modified MUSLE variables can be applied, and the part where the
amount of soil loss is calculated for each HRU consists of two parts, as
3.1. SWAT calibration and validation analysis shown in Fig. 9. The first part (Soil_phys.f) of the pseudocode for the
modified MUSLE below is the process of calculating the amount of soil
The following description outlines the key differences between the
current and previous studies. In Jang et al. (2009) study measured slope
Table 5
and slope length were applied by the SD-HRU method, but the location
Parameters calibrated by SWAT-CUP(Streamflow).
of the actual agricultural fields was not considered and the amount of
Parameter Fitted value
soil loss by each land use was not calibrated. In the study by Lee et al.
(2019), the MUSI method was used to calibrate soil loss from land use R__CN2* 0.18
and evaluate soil loss and SSs, but the NPRF was not included. In this R__ESCO 0.955
V__LAT_TTIME 2
study, the MSDH method was used to consider both actual agricultural V__CANMX 9
fields location and NPRF, and the MUSI method was used to calibrate V__ALPHA_BF 0.165
soil loss for each land use. In addition, the flow rate, the amount of soil V__GW_DELAY 107.5
loss and SSs were further calibrated considering the spatial and temporal V __GWQMN 150
V __GW_REVAP 0.00623
NPRF, and the reduction efficiency of the NPRF was evaluated using the
V __REVAPMN 432.5
calibrated results. R __SOL_AWC − 0.1
There were various studies on the efficiency of BMP applications R __SOL_K 0.12
using the SWAT model. In the studies of Nepal and Parajuli (2022),
*CN2 varies based on each HRU.
Uniyal et al. (2020), Boufala et al. (2022), and Briak et al. (2019), the R __: Multiply by value; V__: Replaced by value.

9
G. Lee et al. Catena 225 (2023) 107028

Fig. 8. Comparison of observed and simulated streamflow. (a) Calibration period is from 2009 to 2014, (b) validation period is from 2015 to 2017.

loss by first multiplying K, C, P, LS and CFRG among the MUSLE factors Fig. 10(a) and Fig. 10(b) show the SS calibration and validation re­
(Equation (3)). The second part (ysed.f) consists in calculating the sults simulated by the modified MUSLE. The R2 and NSE of the cali­
amount of soil loss by changing the coefficients and exponents set in bration results indicated 0.86 and 0.65, respectively and the R2 and NSE
MUSLE at 11.8 and 0.56 to values modified by users and then multi­ of the validation results showed 0.76 and 0.76, respectively. According
plying it by the previously calculated part (Soil_phys.f)(Fig. 9). to Moriasi et al. (2015), if sediment simulation results show that R2
Table 6 shows the coefficients and exponents for each land use ranges from 0.4 to 0.65, the simulation can be regarded as “Satisfac­
calculated using modified MUSLE. The coefficient and exponent of the tory”. If R2 ranges from 0.65 to 0.8, the simulation is classified as “Good”
runoff factors of the current MUSLE were set at 11.8 and 0.56, respec­ and if NSE is between 0.45 and 0.7, the simulation result represents
tively. However, in this study, the coefficient and exponent of runoff “Satisfactory”. If NSE ranges from 0.7 to 0.8, the simulation result is
factors were applied differently for each land use. Therefore, the co­ “Good”. Therefore, both SS calibration and validation replicated natural
efficients and exponents were calibrated to minimize the difference in phenomena and “Satisfactory” or “Good” simulation results.
the amount of soil loss by comparing the average amount of soil loss per
unit area for each land use calculated from the measured value infor­
mation for each land use and MUSLE. Table 7 shows the difference be­ 3.2. Application of modified Spatially Distributed-HRU distribution
tween the results of the amount of soil loss per unit area calculated with method
the modified MUSLE and NPS unit load method (Table 1). The reason
why annual soil loss per unit area in the highland agricultural fields is MSDH has been improved and applied to the SWAT model. enables
greater than in other agricultural fields in average slope areas is that it the application of measured-based calculated slope, slope length, and
reflects the characteristics of highland agricultural fields. Highland NPRF to real agricultural fields. Since slope and slope length affect not
agricultural fields are located in areas of high altitude and steep slopes, only soil erosion but also flow rate, it is necessary to apply a more ac­
so a large amount of soil erosion occurs even with a small amount of curate slope and slope length to accurately estimate flow rate and soil
rainfall. In addition, as a result of frequent topdressing, soil cohesion has erosion. Table 8 shows a comparison of the average slope and slope
been weakened, and the decomposed granite used for topdressing is length for each subwatershed in the HRU calculated with the current
susceptible to soil erosion due to its poor moisture absorption or coag­ HRU method and MSDH method. The average slope lengths calculated
ulation capacity. The difference between the result of the modified with the current HRU method and the modified HRU method were 75.1
MUSLE method and the unit load method was calculated to be 1%, so it m and 40.9 m, respectively. The slope length in the MSDH method was
is judged that land use characteristics of soil erosion in Korea are well reduced by about 34 m, i.e. by 45%. When using slope length and slope
reflected. According to Lee et al. (2019), the difference between the with the current HRU method, soil erosion and suspended solids are
average soil erosion per unit area for each land use calculated with the usually underestimated in areas with steep slopes and long slopes length,
current MUSLE and the unit load method for each land use was analyzed such as Mandae District. Therefore, the use of measured slopes and slope
to be about 46 ~ 5,477%. Therefore, the difference in the results lengths by the MSDH method results in more accurate soil erosion and
calculated within 1% in this study means that the modified MUSLE is a SSs estimation.The average slopes for the watershed were calculated to
good representation of the extent of soil erosion. be 14.9% according to the current HRU method and 12.7% according to
The parameter used to calibrate and validate the SS was PRF and the MSDH, indicating that the average slope in the current HRU was
optimized value was 0.95. The PRF adjusts the peak rate of sediment calculated 2.3% higher than in the modified HRU. The sub-catchments
routing in the main channel and the current MUSLE it is judged to have with the longest average slopes were sub-catchments No. 103 (64.7 m
overestimated the amount of soil erosion. After applying the modified according to the current HRU method) and No. 105 (89.4 m according to
MUSLE, the amount of soil erosion in forests and farmland was signifi­ MSDH) and the sub-catchments with the steepest average slopes were all
cantly reduced. However, the simulated SS is still overestimated because sub-watersheds No. 109, which were calculated at 27% according to the
the amount of soil erosion calculated by the current MUSLE was current HRU method and respectively 18.7% according to MSDH. These
significantly overestimated than the actual soil erosion. results show that the slope length was underestimated and the slope
overestimated in SWAT using the current HRU method. Because slope

10
G. Lee et al. Catena 225 (2023) 107028

Fig. 9. Pseudo code for modified MUSLE. 1) in soil phys.f file, soil loss by land use was calculated 2) in ysed.f file, the modified coefficients and indices of MUSLE
were applied to calculate the amount of soil loss by land use.

and slope length affect runoff and soil erosion, it is important to accu­ National Agencies such as the Korea Environment Corporation, the
rately estimate slope and slope length for each HRU in the watershed. Wonju Regional Environmental Agency, and the Han River Watershed
The reduction efficiency values of the NPRF applied in this study Management Committee. Facilities that do not have a reduction effi­
were taken from the arithmetic average of the measured data by Korea ciency value except for slope protection (78.5%), vegetation ridge

11
G. Lee et al. Catena 225 (2023) 107028

Table 6 NPRFs. Although 37, 22, and 29 NPRFs were installed in sub-watersheds
Result of MUSLE parameter estimation. No. 101, 102, and 106, where the total number of HRUs was high, the
Landuse MUSLE runoff factor reduction efficiency was lower than the average reduction efficiency of
the watershed, with the exception of sub-watershed No. 101. In the case
Current MUSLE Modified MUSLE
of subwatershed No. 106, the average reduction efficiency of NPRF was
Coefficient Exponent Coefficient Exponent

Paddy 11.8 0.56 0.132 0.56


Agricultural field 11.8 0.56 12 0.56 Table 8
Forest 11.8 0.56 0.081 0.56
Change on Slope length and slope after applying MSDH (each sub-watershed).
Sub- w/o MSDH w/ MSDH
watershed
Average slope Average Average slope Average
Table 7 #
length(m) slope(%) length(m) slope(%)
Applicability assessment of MUSLE parameter for landuse.
1 101 48.0 12.0 81.9 11.2
Landuse Area Soil erosion (t h− yr− 1) Percentage Error
102 51.4 11.1 76.2 10.0
(ha) (%)
Unit Modified 103 64.7 8.2 73.4 6.0
load MUSLE 104 41.5 13.2 87.6 12.3
105 37.4 14.8 89.4 14.2
Paddy 614.1 0.168 0.168 0.2
106 47.4 12.6 75.8 10.7
Agricultural 1,542.8 55.00 55.04 − 0.1
107 37.7 15.6 78.1 12.8
field
108 26.8 19.9 70.9 18.0
Forest 3,710.1 0.055 0.055 − 0.6
109 13.3 27.0 42.5 18.7
Total 40.9 14.9 75.1 12.7

(74.5%), diversion ditch (26%), vegetated swale (41.7%) and sedi­ MSDH: Modified Spatially Distributed-HRU distribution method.
mentation basin (54.1%) were referred to the results report of the w/o: without.
Ministry of Environment in 2015 and 30% reduction efficiency was w/: with.
adopted. According to the results report, 62.5% of the NPRF operated by
the Ministry of Environment showed a reduction efficiency below 40%,
Table 9
especially in the case of particulate matter, the reduction efficiency was Agricultural area and average reduction efficiency for NPRF (each sub-
22.2% in 2012, 27.3% in 2013, and 45% in 2014 and the average watershed).
reduction efficiency was 31.5% (National Assembly Secretariat, 2016).
Sub- Agricultural HRU Average Number of Total
Table 9 shows the area of agricultural fields and the average watershed Area(Km2) Area reduction HRU with number
reduction efficiency of NPRF. The average area of agricultural fields by # (Km2) efficiency of NPRF of HRUs
sub-catchment was about 0.39 km2, and the smallest was the agricul­ NPRF (%)
tural area of sub-catchment No. 109, which is the end of the watershed. 101 0.53 15.8 70.0 37 895
The upper reaches of the watershed (sub-catchments No. 101, 102, 104, 102 0.34 10.6 55.2 22 896
105 and 106) were above the average agricultural area. It was found that 103 0.13 1.4 52.0 11 253
the average reduction efficiencies of NPRF in the study area amounted to 104 0.59 10.7 58.5 17 565
105 0.59 4.9 70.9 22 393
approximately 62.5%, and in sub-watersheds No. 101, 106 and 107 106 0.44 6.8 61.1 29 725
respectively 70%, 70.9%, and 65.1%, which was indicated as higher 107 0.55 3.3 68.4 5 276
than average reduction efficiency. On the other hand, sub-watersheds 108 0.35 6.8 65.1 4 240
No. 102, 103, and 104 showed 55.2%, 52%, and 58.5%, respectively, 109 0.03 4.0 – – 22
Total 3.57 64 62.5 147 4,265
which was lower than the average reduction efficiency of all applied

Fig. 10. Comparison of observed and simulated SS. (a) Calibration period is from 2009 to 2014, (b) validation period is from 2015 to 2017.

12
G. Lee et al. Catena 225 (2023) 107028

low despite having about 200% more HRUs and 25 more NPRFs Table 11
installed than in the sub-watershed No. 108, which has a similar Soil erosion and Surface runoff, reduction efficiency for scenarios (each sub-
watershed area. It is estimated that sub-watershed No. 106 is affected by watershed).
the inflow and sediment from sub-watersheds No. 104 and 105. The Sub- Soil erosion (kt/yr) Surface runoff (Mt/yr) Reduction
average reduction efficiency of sub-watersheds No. 101, 105, and 108, watershed
Scenario Scenario Scenario Scenario
efficiency for
which were not affected by the inflow and sediment from other sub- # soil erosion
1 2 1 2
(%)
watersheds, was evaluated to be high. In Table 9, the number of HRUs
in sub-watersheds No. 101, 105 and 107 with high reduction efficiency 101 17.17 15.48 5.9 5.7 9.9
102 20.21 19.16 4.3 4.2 5.2
was 895, 393, and 276, respectively, which indicates a large difference
103 2.39 2.26 0.8 0.7 5.5
in the number of HRUs. The number of installed NPRFs varies; 37 in sub- 104 9.62 9.11 3.0 2.9 5.2
watershed No. 101, 22 in sub-watershed No. 105 and 5 in sub-watershed 105 11.35 9.40 1.7 1.6 17.1
No. 107. It was found that the reduction efficiency of SS was affected by 106 10.83 9.89 3.0 2.9 8.6
slope and slope length, regardless of the number of HRUs or NRPFs. 107 5.52 5.50 1.6 1.6 0.5
108 4.14 4.08 1.8 1.7 1.5
Table 10 shows the statistics of the average slope, slope length and 109 0.35 0.35 1.2 1.2 0.0
NPRF (updated in 2017) by classifying the reduction efficiency at in­ Total 81.58 75.23 5.9 5.7 7.8
tervals of 10%. The efficiency of the NPS pollution reduction facility
kt: Kilo ton.
shown in Table 11 is not the efficiency of a single NPS pollution
reduction facility, but the efficiency applied to a single agricultural field
(HRU). Thus, three or more NPRFs were applied to one HRU, resulting in was found to occur the most, even though the average slope was not
an efficiency in excess of 80%. The average slope length and slope of all steep compared to other sub-watersheds. For the USLE P value in this
HRUs were 75.1 m and 12.7% (Table 8), respectively, but the average study area, all HRUs were equally allocated as 1, and for the USLE C
slope length and slope of the agricultural field to which NPRF was value, all agricultural fields were equally assigned as 0.2. However, in
applied were shown as 122.84 m and 13.5%, individually. The average the case of the USLE K value, the average USLE K value for sub-
slope length and slope of all agricultural fields were 91.1 m and 12%. watershed No.101 was 0.161, and for sub-watershed No.102 was
0.167, which was higher than the overall average of 0.158. For surface
runoff, sub-watershed No.102 was 4.2 Mt (Megaton) which was the
3.3. NPRF performance(reduction efficiency for soil erosion and SS) second largest after sub-watershed No.101 (5.7 Mt) (Table 11). There­
fore runoff factors (e.g., surface runoff, peak runoff and HRU area) and
The efficiency of soil erosion reduction in all HRUs in the sub- USLE K are assessed to have contributed to a large amount of soil erosion
catchments was analyzed and the reduction efficiency of SS at each in sub-watershed No.102.
sub-watershed outlet was calculated. As shown in Table 11, it can be It was found that the reduction efficiency of soil erosion according to
seen that there is a difference in the reduction efficiency (7.8%) that the NPS pollution reduction facility is higher with a higher slope and
affected the amount of soil erosion in the basin and the reduction effi­ longer slope length. In terms of reduction efficiency, sub-watershed No.
ciency (5.8%) that affected the SS at the outlet of the basin by the NPRF 105 showed the greatest reduction efficiency, followed by sub-
installed on every farmland. This is because when the amount of soil watersheds No. 101 and 106. For sub-watershed No.105, the average
erosion is transported in the river, it would be affected by various factors slope was the steepest slope (14.2%) among all sub-watersheds, and the
such as river flow rate, sedimentation and suspension. The reduction average slope length was the longest and amounted to 89.4 m. As shown
efficiencies for each sub-watershed are similar and it can be seen that in Table 10, the reduction efficiency was high because a certain number
managing pollution sources with NPRF installed on cultivated land is of NPRFs were applied in agricultural fields with a long slope length and
very important in managing the NPS pollution at the basin level. While steep slope.
the reduction efficiency applied to each HRU ranges from 20 to 99%, the Fig. 11 shows changes in soil erosion reduction efficiency at various
soil erosion reduction efficiency (7.8%) and SS reduction efficiency slopes (Fig. 11(a)) and slope lengths (Fig. 11(b)) in each sub-watershed.
(5.8%) are 7.8% and 5.8%, respectively. This is because the area where As shown in Fig. 11(a), it is found that the higher the slope, the higher
the NPRF was applied is 1.3 km2 which is only about 2% of the entire the reduction efficiency, but sub-watersheds No.107, 108 and 109
watershed area (64 km2) and about 36.4% of the total agricultural area indicated low reduction efficiency despite the high slope. Since there is
(3.57 km2). almost no cultivated land in the downstream area and no NPRFs are
Table 11 shows the annual soil erosion and reduction efficiency for installed, this result was shown. Out of all 194 NPRFs, 9 were installed in
each sub-watershed in scenarios 1 and 2. As for the annual soil erosion, sub-watershed No.107 and 5 in sub-watershed No.108, and nothing was
sub-watershed No.102 turned out to be the largest, followed by sub- installed in sub-watershed No.109. In the case of sub-watersheds
watersheds No.101 and 105. The average slope of the entire water­ No.107, 108 and 109, low reduction efficiency was found despite the
shed was 12.7%, but the average slope of sub-watershed No.102 was high slope. As shown in Fig. 11(b), the longer the slope, the higher the
10%, which was the second lowest average slope after subwatershed reduction efficiency. Therefore, the reduction efficiency of the amount
No.103. In the case of sub-watershed No.102, the amount of soil erosion of soil erosion can be proportional to the slope length.
The reduction efficiency of soil loss according to the soil physical
Table 10 characteristics of each sub-watershed was examined. Sub-watersheds
Slope length, slope and NPRF for reduction efficiency (divided by 10% range). with a large portion of Ra and OdF soils showed mostly high reduc­
Reduction Average slope Average slope Number of HRU tion efficiency, especially sub-watersheds No. 105, 106, 101, and 102
efficiency(%) length(m) (%) with NRPF with a high portion of Ra soil showed high reduction efficiency of soil
20–29 52.14 8.6% 1
loss (Table 11 and 12). Ra soil in South Korea is normally comprised of
30–39 116.50 14.0% 39 fine loamy with good drainage, which is suitable for farming. Therefore,
40–49 132.87 14.0% 9 it is judged that the regions where there are a lot of agricultural fields
50–59 130.87 12.3% 21 and NPRFs installed and where Ra soil makes up the majority have a
60–69 126.31 13.8% 7
high efficiency in reducing soil loss. The low reduction efficiency of soil
70–79 129.68 14.0% 25
80–89 104.93 12.2% 22 loss in sub-watershed No. 108 with 32% Ra soil occurs due to the small
90–99 134.05 14.6% 23 number of NPRFs installed (only 4 NPRFs). Therefore, the reduction
Total 122.84 13.5% 147 efficiency of soil loss by sub-watershed is judged to be the result of a

13
G. Lee et al. Catena 225 (2023) 107028

Fig. 11. Relationship between reduction efficiency and slope/slope length (Soil erosion). (a) Changes in soil erosion reduction efficiency at various slopes were
presented (b) changes in soil erosion reduction efficiency at various slope lengths were shown.

Table 12
Agricultural area and average reduction efficiency for NPRF (each sub-watershed).
Sub-watershed An Ma OdF Ra Re Rocky Total
#

101 274.97 494.95 0 504.67 34.87 272.57 1,582.02


102 5.93 335.17 4.39 233.81 169.50 312.08 1,060.87
103 62.72 0 0 2.34 71.95 0 137.02
104 55.72 0 925.91 91.13 0 0 1,072.76
105 2.53 29.16 264.66 172.04 1.23 21.44 491.06
106 191.11 0 235.35 239.03 11.37 0 676.85
107 208.50 0 0 63.29 56.92 0 328.71
108 62.69 136.72 262.54 216.13 0 0 678.08
109 168.49 214.62 0 21.63 0 0 404.73
Total 1,032.66 1,210.61 1,692.85 1,544.06 345.83 606.09 6,432.11

of annual SS and the reduction efficiency at outlets for each sub-


Table 13
watershed. The reduction efficiency of SS was similar to the soil
The amount of SS and reduction efficiency for two scenarios (each
erosion reduction efficiency in Table 11, and in the sub-watersheds 101
subwatershed).
and 105, the upstream watershed, the reduction efficiency of SS was
Sub-watershed SS (kt/yr) Reduction efficiency (%) slightly increased or the same compared to the soil loss reduction effi­
#
Scenario 1 Scenario 2 ciency. In sub-watersheds 102, 104 and 108, the SS reduction efficiency
101 4.73 4.23 10.6 was slightly reduced despite the upstream watershed, as the number and
102 4.03 3.82 5.1 area of NPRF were smaller than in sub-watersheds 101 and 105. In the
103 10.00 9.37 6.3 sub-watersheds 107 and 109, the SS reduction efficiency was signifi­
104 4.91 4.69 4.5 cantly increased compared to the soil erosion reduction efficiency shown
105 1.99 1.65 17.1
106 8.59 7.94 7.6
in Table 11 because the sub-watersheds 107 and 109 were located in
107 15.66 14.42 7.9 downstream of the watershed. The SS reduction efficiency is affected by
108 0.68 0.68 1.0 the amount of soil erosion reduced by NPRFs installed in agricultural
109 19.16 18.05 5.8 fields, so it is important to manage pollution sources such as NPRFs in
kt: Kilo ton. agricultural fields.
Fig. 12 shows the relationship between the SS reduction efficiency
complex effect of the soil physical characteristics of the sub-watershed and slope or slope length for each sub-watershed. Similar to the rela­
and the type and number of NPRFs. tionship between the soil erosion reduction efficiency and the slope or
The Sediment Delivery Ratio (SDR) is the ratio of mobile soil parti­ slope length (shown in Fig. 11), Fig. 12 shows sub-watershed No.108
cles eroded to the total amount of soil eroded from areas to a specific indicating low reduction efficiency even though the slope was high. In
point occurring in a watershed, and the SDR is affected by various fac­ the case of sub-watersheds No.107 and 109, the SS reduction efficiency
tors such as the area and slope of the watershed. Since these SDRs vary was higher than the soil erosion reduction efficiency. Sub-watersheds
from each watershed due to different environmental conditions (USDA, No. 107 and No. 109 are located in the downstream area, and SS
1983), this study analyzed the reduction efficiency of the SS considering reduced by the NPRF installed in the upper sub-watershed flowed down
the SDR as well as the soil erosion reduction efficiency of the NPS into sub-watersheds No. 107 and No. 109. Therefore, it is estimated that
pollution reduction facility. the SS reduction efficiency was higher than the soil loss reduction effi­
Table 13 shows the results of scenarios 1 and 2 indicating the amount ciency. Comparing sub-watersheds No. 107 and No. 109, the SS

14
G. Lee et al. Catena 225 (2023) 107028

Fig. 12. Relationship between reduction efficiency and slope/slope length (SS). (a) Changes in SS reduction efficiency at various slopes were presented (b) changes
in SS reduction efficiency at various slope lengths were shown.

reduction efficiency of sub-watershed No. 107 was found to be higher appears that the number and type of NPRF installations have a signifi­
than that of sub-watershed No. 109. The reason is that sub-watershed cant impact on the SS reduction efficiency.
No. 107 has 5 NPRFs installed and there is no NPRF installed in sub-
watershed No. 109. Similar to the soil loss reduction efficiency, the SS 4. Conclusions
reduction efficiency tends to be higher with increasing slope length. As a
result, it was found that the reduction efficiency of the NPS pollution The soil erosion and SS reduction efficiency of NPRF in the water­
reduction facility was higher because the slope length was longer for shed were analyzed using the SWAT watershed model. The average area
both the amount of soil erosion and SS. As mentioned earlier, more or of agricultural fields for each sub-watershed was about 0.39 km2, and
higher efficient NPRFs were installed where the slope length is long. the average efficiency of the NPRF was about 62.5%. It was challenging
As shown in the Fig. 12, the R2 value of the slope and SS reduction to find a strong correlation between the efficiency of NPRF applied to
efficiency is 0.018 and the R2 value of the slope length and SS reduction each sub-watershed and the reduction efficiency of soil erosion and SS
efficiency is 0.1711, respectively. The reason why the R2 values are low because the number and area of HRUs to which the NPRF was applied
is the soil loss reduction efficiency is greatly affected by the NPRF. In were different for each sub-watershed. Considering the relationship
Fig. 13, the R2 value between the slope and the generation of SS is 0.58 between the slope length, the slope and the reduction efficiency, it was
and the R2 between the slope length and the generation of SS is 0.34, found that the greater the slope and the longer the slope length, the
which is statistically significant. The slope and slope length are impor­ greater the soil erosion and SS reduction efficiency. However, some sub-
tant factors that influence the generation of SS, but there was no sig­ watersheds (No. 107, 108, and 109) had high slopes but low soil erosion
nificant correlation with soil loss reduction efficiency. Therefore, it reduction efficiency as agricultural fields and NPRF were rare in the sub-

Fig. 13. Relationship between SS amount and slope/slope length. (a) Changes in SS at various slopes were presented (b) changes in SS at various slope lengths
were shown.

15
G. Lee et al. Catena 225 (2023) 107028

watersheds. However, in the case of the SS, sub-watershed 108 had a Acknowledgements
high slope and low reduction efficiency, while sub-watershed 107 had a
high reduction efficiency due to the impact of the reduced SS from up­ This research was funded by the Ministry of Environment of Korea as
stream. The longer the slope, the greater the soil erosion and SS The SS (Surface Soil conservation and management) projects
reduction efficiency. It was estimated that the longer the slope, the [2019002820003].
greater the reduction efficiency or the greater the number of NPRF in­
stallations. The results section indicated that the SS reduction efficiency References
is strongly affected by the reduction in the amount of soil erosion. So
source management is very important to reduce SS. Accordingly, the Abbaspour, K.C., Yang, J., Maximov, I., Siber, R., Bogner, K., Mieleitner, J., Zobrist, J.,
Srinivasan, R., 2007. Modelling hydrology and water quality in the pre-alpine/alpine
reduction efficiency of NPRF analyzed in this study can be used as Thur watershed using SWAT. J. Hydrol. 333, 413–430.
important information for the installation of reduction facilities when Abraham, L.Z., Roehrig, J., Alamirew, C.D., 2007. Calibration and Validation of SWAT
planning an NPS pollution reduction project in the future. Currently, Hydrologic Model for Meki Watershed, Ethiopia, in: Conference on International
Agricultural Research for Developement. pp. 50–53.
there is no previous research in Korea, on how NPRF performance Akhavan, S., Abedi-Koupai, J., Mousavi, S.F., Afyuni, M., Eslamian, S.S., Abbaspour, K.
changes depending on various conditions such as time and seasons. This C., 2010. Application of SWAT model to investigate nitrate leaching in Hamadan-
study analyzed the efficiency of the NPRF by considering the exact Bahar Watershed. Iran. Agric. Ecosyst. Environ. 139, 675–688. https://doi.org/
10.1016/j.agee.2010.10.015.
location, slope and slope length of the NPRF. It is believed that if more Arnold, J.G., Srinivasan, R., Muttiah, R.S., Williams, J.R., 1998. Large area hydrologic
diverse conditions are considered, it will be possible to accurately assess modeling and assessment part I: Model development. J. Am. Water Resour. Assoc.
soil erosion and the SS reduction efficiency of NPRF in the watershed. https://doi.org/10.1111/j.1752-1688.1998.tb05961.x.
Arnold, J.G., Moriasi, D.N., Gassman, P.W., Abbaspour, K.C., White, M.J., Srinivasan, R.,
Nonlinearly increasing soil erosion and sediment under the influence
Santhi, C., Harmel, R.D., Van Griensven, A., Van Liew, M.W., Kannan, N., Jha, M.K.,
of changes in climate and rainfall regimes are important factors in the 2012. SWAT: Model use, calibration, and validation. Trans. ASABE. 10 (13031/
assessment of water pollution in watersheds. The MSDH and modeling 2013), 42259.
techniques enable the evaluation of reduction efficiency at a watershed Behera, S., Panda, R.K., 2006. Evaluation of management alternatives for an agricultural
watershed in a sub-humid subtropical region using a physical process based model.
scale by considering the temporal and spatial characteristics of the Agric. Ecosyst. Environ. 113, 62–72. https://doi.org/10.1016/j.agee.2005.08.032.
NPRF. In particular, MSDH was used to evaluate the Mandae District, Borah, D.K., Bera, M., 2004. Watershed-scale hydrologic and nonpoint-source pollution
where highland fields are dense and a number of NPRFs have been models: Review of applications. Trans. Am. Soc. Agric. Eng. 10 (13031/2013),
16110.
installed. MSDH is expected to be very useful in evaluating NPS pollu­ Boufala, M.H., El Hmaidi, A., Essahlaoui, A., Chadli, K., El Ouali, A., Lahjouj, A., 2022.
tion in various areas where non-point pollution reduction facilities are Assessment of the best management practices under a semi-arid basin using SWAT
installed. model (case of M’dez watershed, Morocco). Model. Earth Syst. Environ. 8, 713–731.
https://doi.org/10.1007/s40808-021-01123-6.
Turbid water generated in highland agricultural fields located in Briak, H., Mrabet, R., Moussadek, R., Aboumaria, K., 2019. Use of a calibrated SWAT
upstream of the Han River watershed affects water quality and aquatic model to evaluate the effects of agricultural BMPs on sediments of the Kalaya river
ecosystems in downstream. Therefore, the management of water quality basin (North of Morocco). Int. Soil Water Conserv. Res. 7, 176–183. https://doi.org/
10.1016/j.iswcr.2019.02.002.
and aquatic ecosystems in the upstream watershed is very important. Carpenter, S.R., Caraco, N.F., Correll, D.L., Howarth, R.W., Sharpley, A.N., Smith, V.H.,
Since 2007, the Korean government has been installing and operating 1998. Nonpoint pollution of surface waters with phosphorus and nitrogen. Ecol.
structured and unstructured NPRFs to reduce water turbidity. However, Appl. https://doi.org/10.1890/1051-0761(1998)008[0559:NPOSWW]2.0.CO;2.
Chen, D.J., Lu, J., Shen, Y.N., Dahlgren, R.A., Jin, S.Q., 2009. Estimation of critical
it has been difficult to establish an environmental policy for NPS
nutrient amounts based on input-output analysis in an agriculture watershed of
pollution reduction projects because a thorough evaluation of the eastern China. Agric. Ecosyst. Environ. 134, 159–167. https://doi.org/10.1016/j.
reduction efficiency has not been performed. Outside of watersheds that agee.2009.06.011.
include the management areas of NPS pollution reduction, it is difficult Da Silva, A.M., Huang, C.H., Francesconi, W., Saintil, T., Villegas, J., 2015. Using
landscape metrics to analyze micro-scale soil erosion processes. Ecol. Indic. 56,
to understand the effect of turbid water generated in the NPS control 184–193. https://doi.org/10.1016/J.ECOLIND.2015.04.004.
area due to the lack of measurement data such as the flow rate and water Ding, X., Liu, L., 2019. Long-term effects of anthropogenic factors on nonpoint source
quality during rainfall. Therefore, as in this study, if the reduction effi­ pollution in the upper reaches of the Yangtze river. Sustainability. https://doi.org/
10.3390/su11082246.
ciency of NPRF is accurately evaluated using a watershed model, the Engel, B., Storm, D., White, M., Arnold, J., Arabi, M., 2007. A hydrologic/water quality
results of the study can be practically used to establish environmental model application protocol. J. Am. Water Resour. Assoc. https://doi.org/10.1111/
policy for NPS pollution reduction projects. Based on this research j.1752-1688.2007.00105.x.
ESRI, 2012. ArcGIS Desktop 10.1. Redlands CA.
approach, the effect of turbid water on an unmeasured watershed can be Gee, G.W., 1987. Recharge at the Hanford Site: Status report.
evaluated. Gerrits, A.M.J., Savenije, H.H.G., Hoffmann, L., Pflster, L., 2007. New technique to
This study has several shortcomings that should be addressed in measure forest floor interception - An application in a beech forest in Luxembourg.
Hydrol. Earth Syst. Sci. https://doi.org/10.5194/hess-11-695-2007.
further research. There is some doubt as the reduction efficiency of 30% Google, n.d. Google Earth [WWW Document]. U.S.A. URL https://www.google.
has been uniformly applied to NPRFs for which there is no observation com/earth/.
value for the reduction efficiency of various NPRFs. In addition, the Himanshu, S.K., Pandey, A., Yadav, B., Gupta, A., 2019. Evaluation of best management
practices for sediment and nutrient loss control using SWAT model. Soil Tillage Res.
reduction efficiency of NPRF decreases with time after the installation,
192, 42–58. https://doi.org/10.1016/J.STILL.2019.04.016.
but the NPRF degradation efficiency over time after installation was not Huang, J.Y., Liao, I.E., Chung, Y.F., Chen, K.T., 2013. Shielding wireless sensor network
considered in this study. Therefore, in future studies, 1) additional using Markovian intrusion detection system with attack pattern mining. Inf. Sci. (Ny)
measurements of the reduction efficiencies of the various NPRFs are 231, 32–44. https://doi.org/10.1016/J.INS.2011.03.014.
IPCC, 2014. Climate Change 2014, Climate Change 2014: Synthesis Report. 10.1017/
required, and 2) it is necessary to consider the degradation of the CBO9781107415324.
reduction efficiencies over the period that the NPRFs are installed. Jang, S.S., Kim, S.J., 2017. Assessment of Climate Change Impact on Highland
Agricultural Watershed Hydrologic Cycle and Water Quality under RCP Scenarios
using SWAT. J. Korean Soc. Agric. Eng. 59, 41–50. https://doi.org/10.5389/
Declaration of Competing Interest ksae.2017.59.3.041.
Jang, W., Yoo, D.S., Chung, I.M., Kim, N.W., Jun, M.S., Park, Y.S., Kim, J., Lim, K.J.,
The authors declare that they have no known competing financial 2009. Development of SWAT SD-HRU Pre-processor Module for Accurate Estimation
of Slope Length of Each HRU Considering Spatial Topographic Characteristics in
interests or personal relationships that could have appeared to influence SWAT. J. Korean Soc. Water Environ. 25, 351–362.
the work reported in this paper. Jeong, J., Kannan, N., Arnold, J.G., Glick, R., Gosselink, L., Srinivasan, R., Harmel, R.D.,
2011. Development of sub-daily erosion and sediment transport algorithms for
SWAT. Trans. ASABE. 10 (13031/2013), 39841.
Data availability Johanson, R.C., Imhoff, J.C., Davis, H.H., 1980. Users manual for hydrological
simulation program - FORTRAN (HSPF). Final report, Nov 76-Nov 78.
No data was used for the research described in the article.

16
G. Lee et al. Catena 225 (2023) 107028

Keller, A.A., Chen, X., Fox, J., Fulda, M., Dorsey, R., Seapy, B., Glenday, J., Bray, E., Ouyang, W., Huang, H., Hao, F., Shan, Y., Guo, B., 2012. Evaluating spatial interaction of
2014. Attenuation coefficients for water quality trading. Environ. Sci. Technol. 48, soil property with non-point source pollution at watershed scale: The phosphorus
6788–6794. https://doi.org/10.1021/es500202x. indicator in Northeast China. Sci. Total Environ. https://doi.org/10.1016/j.
Khalid, K., Ali, M.F., Rahman, N.F.A., Mispan, M.R., Haron, S.H., Othman, Z., Bachok, M. scitotenv.2012.06.017.
F., 2016. Sensitivity Analysis in Watershed Model Using SUFI-2 Algorithm, in: Park, J.Y., Ahn, S.R., Hwang, S.J., Jang, C.H., Park, G.A., Kim, S.J., 2014. Evaluation of
Procedia Engineering. 10.1016/j.proeng.2016.11.086. MODIS NDVI and LST for indicating soil moisture of forest areas based on SWAT
Kim, J., Choi, H.T., Lim, H., 2015. Evaluation on MUSLE Runoff Energy Coefficient in modeling. Paddy Water Environ. https://doi.org/10.1007/s10333-014-0425-3.
Small Forest Watershed. J. Korea Soc. Environ. Restor. Technol. 18, 191–200. Peter Heng, B.C., Chandler, J.H., Armstrong, A., 2010. Applying close range digital
10.13087/kosert.2015.18.6.191. photogrammetry in soil erosion studies. Photogramm. Rec. 25, 240–265. https://doi.
Kourakos, G., Klein, F., Cortis, A., Harter, T., 2012. A groundwater nonpoint source org/10.1111/j.1477-9730.2010.00584.x.
pollution modeling framework to evaluate long-term dynamics of pollutant Qi, S.S., Hao, F.H., Ouyang, W., Cheng, H.G., 2012. Characterizing landscape and soil
exceedance probabilities in wells and other discharge locations. Water Resour. Res. erosion dynamics under pipeline interventions in Southwest China. Procedia
https://doi.org/10.1029/2011WR010813. Environ. Sci. 13, 1863–1871. https://doi.org/10.1016/j.proenv.2012.01.180.
Lam, Q.D., Schmalz, B., Fohrer, N., 2011. The impact of agricultural Best Management Ramos, M.C., Benito, C., Martínez-Casasnovas, J.A., 2015. Simulating soil conservation
Practices on water quality in a North German lowland catchment. Environ. Monit. measures to control soil and nutrient losses in a small, vineyard dominated, basin.
Assess. 183, 351–379. https://doi.org/10.1007/S10661-011-1926-9. Agric. Ecosyst. Environ. 213, 194–208. https://doi.org/10.1016/j.
Lee, S., Lee, G., Yang, D., Choi, Y., Lim, K.J., Jang, W.S., 2019. Occurrence and Behavior agee.2015.08.004.
Analysis of Soil Erosion by Applying Coefficient and Exponent of MUSLE Runoff RIG, 2014. Policy Changes in Non-point Source Management and Gangwon-do.
Factor Depending on Land Use. J. Wetl. Researh 21, 98–106. Ritter, W., Shirmohammadi, A., 2001. Agricultural nonpoint source pollution: Watershed
Lee, M., Park, G., Park, M., Park, J., Lee, J., Kim, S., 2010. Evaluation of non-point source management and hydrology. Boca Raton, Florida CRC Press.
pollution reduction by applying Best Management Practices using a SWAT model Rostamian, R., Jaleh, A., Afyuni, M., Mousavi, S.F., Heidarpour, M., Jalalian, A.,
and QuickBird high resolution satellite imagery. J. Environ. Sci. https://doi.org/ Abbaspour, K.C., 2008. Application of a SWAT model for estimating runoff and
10.1016/S1001-0742(09)60184-4. sediment in two mountainous basins in central Iran. Hydrol. Sci. J. 53, 977–988.
Lee, G., Shin, Y., Jung, Y., 2014. Development of web-based RECESS model for Saleh, A., Arnold, J.G., Gassman, P.W., Hauck, L.M., Rosenthal, W.D., Williams, J.R.,
estimating baseflow using SWAT. Sustain. 6, 2357–2378. https://doi.org/10.3390/ McFarland, A.M.S., 2000. Application of SWAT for the Upper North Bosque River
su6042357. Watershed. Trans. Am. Soc. Agric. Eng. https://doi.org/10.13031/2013.3000.
Lin, B.S., Thomas, K., Chen, C.K., Ho, H.C., 2016. Evaluation of soil erosion risk for Sophocleous, M., Perkins, S.P., 2000. Methodology and application of combined
watershed management in Shenmu watershed, central Taiwan using USLE model watershed and ground-water models in Kansas. J. Hydrol. https://doi.org/10.1016/
parameters. Paddy Water Environ. 14, 19–43. https://doi.org/10.1007/s10333-014- S0022-1694(00)00293-6.
0476-5. Strauch, M., Lima, J.E.F.W., Volk, M., Lorz, C., Makeschin, F., 2013. The impact of Best
Me, W., Abell, J.M., Hamilton, D.P., 2015. Effects of hydrologic conditions on SWAT Management Practices on simulated streamflow and sediment load in a Central
model performance and parameter sensitivity for a small, mixed land use catchment Brazilian catchment. J. Environ. Manage. 127, S24–S36. https://doi.org/10.1016/j.
in New Zealand. Hydrol. Earth Syst. Sci. https://doi.org/10.5194/hess-19-4127- jenvman.2013.01.014.
2015. Tao, F., Yokozawa, M., Hayashi, Y., Lin, E., 2003. Future climate change, the agricultural
Merriman, K.R., Daggupati, P., Srinivasan, R., Hayhurst, B., 2019. Assessment of site- water cycle, and agricultural production in China. Agric. Ecosyst. Environ. 95,
specific agricultural Best Management Practices in the Upper East River watershed, 203–215. https://doi.org/10.1016/S0167-8809(02)00093-2.
Wisconsin, using a field-scale SWAT model. J. Great Lakes Res. 45, 619–641. https:// Tuppad, P., Santhi, C., Wang, X., Williams, J.R., Srinivasan, R., Gowda, P.H., 2010.
doi.org/10.1016/j.jglr.2019.02.004. Simulation of conservation practices using the apex model. Appl. Eng. Agric.
Merriman, K.R., Daggupati, P., Srinivasan, R., Toussant, C., Russell, A.M., Hayhurst, B., USDA, 1983. National Engineering Handbook: Sedimentation, USDA.
2018. Assessing the impact of site-specific BMPs using a spatially explicit, field-scale U.S.EPA, 2016. National Nonpoint Source Program - A catalyst for water quality
SWAT model with edge-of-field and tile hydrology and water-quality data in the improvements. Environ. Prot. Agency. EPA 841-R-16-009.
Eagle Creek Watershed, Ohio. Water (Switzerland). https://doi.org/10.3390/ Uniyal, B., Jha, M.K., Verma, A.K., Anebagilu, P.K., 2020. Identification of critical areas
w10101299. and evaluation of best management practices using SWAT for sustainable watershed
Mishra, A., Kar, S., Raghuwanshi, N.S., 2009. Modeling nonpoint source pollutant losses management. Sci. Total Environ. 744, 140737. https://doi.org/10.1016/j.
from a small watershed using HSPF model. J. Environ. Eng. https://doi.org/ scitotenv.2020.140737.
10.1061/(ASCE)0733-9372(2009)135:2(92). Wang, K., Wang, H.J., Shi, X.Z., Weindorf, D.C., Yu, D.S., Liang, Y., Shi, D.M., 2009.
MOE, 2004. Comprehensive Measures for Reducing Pollution in Upland Fields. Landscape analysis of dynamic soil erosion in Subtropical China: A case study in
MOE, 2017a. Monitoring and evaluation of Non-point source management areas in Xingguo County, Jiangxi Province. Soil Tillage Res. 105, 313–321. https://doi.org/
Mandae, Gaa and Jaun districts (2nd). 10.1016/J.STILL.2008.08.013.
MOE, 2017b. A Study on GIS System Construction for Highland Agricult ural Fields in Williams, J.R., Arnold, J.G., Srinivasan, R., Ramanarayanan, T.S., 1998. APEX: A New
Mandae and Gaa Non-point Source Pollution Management Areas. Tool for Predicting the Effects of Climate and CO2 Changes on Erosion and Water
MOE, 2018. Monitoring and evaluation of Non-point source management areas in Quality, in: Modelling Soil Erosion by Water. 10.1007/978-3-642-58913-3_33.
Mandae, Gaa and Jaun districts (3rd). Williams, J.R., 1975. Sediment routing for agricultural watersheds. JAWRA J. Am. Water
MOE, n.d. Agricultural Land in Alpine Region Management [WWW Document]. URL Resour. Assoc. 10.1111/j.1752-1688.1975.tb01817.x.
https://www.nonpoint.or.kr/. Wischmeier, W.H., Smith, D.D., 1965. Predicting Rainfall-Erosion Losses From Cropland
MOE, 2016. Installation, Management, and Operation Manual of Non-point source east of the Rocky Mountains: Guide for Selection of Practices for Soil and Water
Pollution Reduction Facilities. MOE. Conservation, Agriculture handbook no. 282.
Mohamoud, Y.M., Parmar, R., Wolfe, K., 2010. Modeling best management practices Won, J.Y., Lee, J.S., 2012. An Improvement on Estimation of Runoff Factor Equation for
(BMPs) with HSPF, in: Watershed Management Conference 2010: Innovations in Mountain Area in MUSLE. J. korean Soc. hazard Mitig. 12, 207–214. https://doi.
Watershed Management under Land Use and Climate Change - Proceedings of the org/10.9798/kosham.2012.12.5.207.
2010 Watershed Management Conference. pp. 892–898. 10.1061/41148(389)81. Xiao, H., Ji, W., 2007. Relating landscape characteristics to non-point source pollution in
Moriasi, D.N., Arnold, J.G., Liew, M.W.V., Bingner, R.L., Harmel, R.D., Veith, T.L., 2007. mine waste-located watersheds using geospatial techniques. J. Environ. Manage. 82,
Model evaluation guidelines for systematic quantification of accuracy in watershed 111–119. https://doi.org/10.1016/J.JENVMAN.2005.12.009.
simulations. Trans. ASABE 50, 885–900. Ye, H., Yuan, X., Han, L., Marip, J.B., Qin, J., 2017. Risk assessment of nitrogen and
Moriasi, D.N., Gitau, M.W., Pai, N., Daggupati, P., 2015. Hydrologic and water quality phosphorus loss in a hilly-plainwatershed based on the different hydrological period:
models: Performance measures and evaluation criteria. Trans. ASABE. https://doi. A case study in tiaoxi watershed. Sustainability. https://doi.org/10.3390/
org/10.13031/trans.58.10715. su9081493.
Mwangi, J.K., Shisanya, C.A., Gathenya, J.M., Namirembe, S., Moriasi, D.N., 2015. Yu, S.L., Zhen, J.X., 2004. Development of a best management practice (BMP) placement
A modeling approach to evaluate the impact of conservation practices on water and strategy at the watershed scale, in: Proceedings of the Watershed Management
sediment yield in Sasumua Watershed, Kenya. J. Soil Water Conserv. 70, 75–90. Symposium. 10.1061/40706(266)16.
https://doi.org/10.2489/JSWC.70.2.75. Zhang, S., Fan, W., Li, Y., Yi, Y., 2017. The influence of changes in land use and
Nash, J.E., Sutcliffe, J.V., 1970. River flow forecasting through conceptual models part I - landscape patterns on soil erosion in a watershed. Sci. Total Environ. 574, 34–45.
A discussion of principles. J. Hydrol. https://doi.org/10.1016/0022-1694(70) https://doi.org/10.1016/j.scitotenv.2016.09.024.
90255-6. Zhang, X., Srinivasan, R., Arnold, J., Izaurralde, R.C., Bosch, D., 2011. Simultaneous
Nepal, D., Parajuli, P.B., 2022. Assessment of Best Management Practices on Hydrology calibration of surface flow and baseflow simulations: A revisit of the SWAT model
and Sediment Yield at Watershed Scale in Mississippi Using SWAT. Agric. 12 https:// calibration framework. Hydrol. Process. https://doi.org/10.1002/hyp.8058.
doi.org/10.3390/agriculture12040518. Zhang, T., Yang, Y., Ni, J., Xie, D., 2020. Construction of an integrated technology system
NIER, 2009. Monitoring and evaluation of NPS pollution priority management region - for control agricultural non-point source pollution in the Three Gorges Reservoir
Soyanggang dam watershed -. Areas. Agric. Ecosyst. Environ. 295, 106919 https://doi.org/10.1016/j.
Ouyang, W., Wang, X., Hao, F., Srinivasan, R., 2009. Temporal-spatial dynamics of agee.2020.106919.
vegetation variation on non-point source nutrient pollution. Ecol. Modell. https:// Zhang, X., Zhang, M., 2011. Modeling effectiveness of agricultural BMPs to reduce
doi.org/10.1016/j.ecolmodel.2009.06.039. sediment load and organophosphate pesticides in surface runoff. Sci. Total Environ.
409, 1949–1958. https://doi.org/10.1016/j.scitotenv.2011.02.012.

17

You might also like