You are on page 1of 17

This is the html version of the file

http://infolib.hua.edu.vn/Fulltext/ChuyenDe2009/CD239/68.pdf.

Agriculture, Ecosystems and Environment xxx (2009) xxx–xxx

A R T I C L E I N F OArticle history:Received 12 November 2008Received in revised
form 14 June 2009Accepted 15 June 2009Available online xxx

Keywords:Critical nutrient amount, Diffuse (non-point) source pollution, Export
coefficient model, Input–output mass balance, Nitrogen, Phosphorus;

A B S T R A C T: The concept of critical nutrient amounts (CNA) for a watershed was
developed to address eutrophication in surface waters from diffuse (non-point)
source pollution. CNA is defined as the maximum allowable applied or generated
amount (AGA) of a nutrient from natural and human sources that can be emitted
and still allow compliance with water quality standards. The CNA calculation method
is based on properties of diffuse source pollution, including (i) estimation and
analysis of nutrient input–output balances in terrestrial and riverine systems; (ii)
prediction of terrestrial nutrient export loads and AGA using riverine loads; and (iii)
calculation of critical AGA to meet different regulatory compliance locations (e.g.,
end of a reach or for the whole reach). The CNA concept was developed, validated
and applied for total nitrogen (TN) and phosphorus (TP) in the ChangLe agriculture-
dominated watershed (864 km2/27.8 km reach) of eastern China. Results indicated
that CNA was 7174 t N yÀ1 and 5004 t P yÀ1for the reach-end control method and
8290 t N yÀ1 and 4425 t P yÀ1 for the whole-reach control method. Annual TN AGA
exceeded CNA by 53.2–61.3% and 46.0–55.2% for reach-end and whole-reach
control methods in 2004–06, respectively. In contrast, TP AGA values were 90.3–
95.9% and 68.3–73.2% below CNA values for reach-end and whole-reach control
methods, respectively. These calculations provide a target or permissible nutrient
amount that can be used to develop management practices that allow attainment
of water quality objectives at the watershed scale.© 2009 Elsevier B.V. All rights
reserved.* Corresponding author. Tel.: +86 0571 86091535; fax: +86 0571
86091535.E-mail address: jlu@zju.edu.cn (J. Lu).G ModelAGEE-3459; No of Pages 9.
Please cite this article in press as: Chen, D.J., et al., Estimation of critical nutrient
amounts based on input–output analysis in an agriculture watershed of eastern
China. - Agric. Ecosyst. Environ. (2009), doi:10.1016/j.agee.2009.06.011;

Contents lists available at ScienceDirectAgriculture, Ecosystems and
Environmentjournal homepage: www.elsevier.com/locate/agee0167-8809/$ – see
front matter © 2009 Elsevier B.V. All rights
reserved.doi:10.1016/j.agee.2009.06.011
--------------------------------------------------------------------------------

Page 1

Estimation of critical nutrient amounts based on input–output analysis in
anagriculture watershed of eastern China

DingJiang Chen a, Jun Lu b,*, YeNa Shen c, Randy A. Dahlgren d, ShuQuan Jin a

a Department of Natural Resources, College of Environmental Science and Natural
Resources, Zhejiang University, Hangzhou 310029, Chinab China Ministry of
Education Key Lab of Environment Remediation and Ecological Health, Zhejiang
University, Hangzhou 310029, Chinac Zhejiang Provincial Key Laboratory of
Subtropic Soil and Plant Nutrition, Zhejiang University, Hangzhou 310029, Chinad
Department of Land, Air, and Water Resources, University of California, Davis, CA
95616 USA1.

Text:

IntroductionNon-point source (NPS) pollution is a leading cause of waterpollution
throughout the world, with polluted runoff fromagriculture being a primary cause.
Recently, the loss of nitrogen(N) and phosphorus (P) from agricultural lands to
surface watershas increased rapidly in comparison to losses from industrial
andresidential lands (De Wit et al., 2003; Reungsang et al., 2007).Excessive N and P
in surface waters induce eutrophication anddramatic shifts in trophic relationships
in many fresh watersystems (De Wit and Giuseppe, 2001; Markku et al.,
1995),including hypoxia/anoxia induced ‘‘dead zones’’ in more than 400locations
worldwide (Diaz and Rosenberg, 2008). To reduce thenegative impacts of
eutrophication, policymakers and scientists inseveral countries, including China,
have agreed on regulatorystandards for pollutant concentrations in water. In
Europe, the aimis to reduce pollutant loads to nearly zero for human-
derivedsubstances and to near background values for naturally occurringsubstances
(Laane, 2005). In the USA, the goal is to reducepollutants to levels that allow
attainment of the water qualitystandard for the designated use of the water body
(NationalResearch Council, 2001).An analogous problem to NPS water pollution is
atmosphericdeposition of sulfur and nitrogen substances to terrestrial andaquatic
ecosystems. An international policy addressing atmo-spheric deposition is the
critical loads (CL) concept, which isexpressed as: ‘‘A quantitative estimate of an
exposure to one ormore pollutants below which significant harmful effects
onspecified sensitive elements of the environment do not occuraccording to present
knowledge’’ (Malcolm, 2000). Critical loadswere widely used as a strategy for
reducing emissions of acidifyingcompounds from atmospheric deposition using a
mass balanceaccounting method (Barkman and Alveteg, 2001; Duan et
al.,2000).Following the CL concept for atmospheric deposition, Laane(2005) applied
CL to address eutrophication in the Rhine River
--------------------------------------------------------------------------------

Page 2

watershed and its coastal zone. This work examined the feasibilityof CL to deal with
eutrophication and provided an approach forremediation of excessive N and P in
freshwaters. However, theresults of CL could not be directly applied to practical
controls of Nand P since it is difficult to effectively and directly monitor NPSpollution
loads from land to river. Consequently, it was moreefficient and cost effective to
focus on diffuse source restrictions,such as fertilizer (Gassman et al., 2006; Meals et
al., 2008) andlivestock sources, rather than restrictions on the resulting N and
Pexport loads (Elofsson, 2003; Ribaudo et al., 2001), since it isdifficult to track
discharge of NPS pollutants to a specific parcel ofland or land management activity
(Brian et al., 2008). For the totalmaximum daily loads (TMDL) approach to NPS
pollution in theUSA, the focus is primarily on setting a load value for NPSpollutants
from all sources that will allow attainment of a givenwater quality standard
(National Research Council, 2001). How-ever, it does not provide a specific
approach for setting pollutantreduction targets for various pollution sources or land
uses withina watershed. In contrast, the critical nutrient amount proposed inthis
paper establishes a generated or applied quantity from variousnatural and human-
derived sources (i.e., source amounts), which isdistinguished from the concept of
measured river loads orcatchment export loads.The aim of this paper is to apply
and extend the concept of CL toquantitatively address river eutrophication from
human-derivednutrient sources without changing land-use types at thewatershed
scale. The concept of critical nutrient amounts (CNA)for a watershed, referring to
the critical nutrient level in a waterbody, is proposed. The method for CNA
estimation, based on theproperties of NPS pollution, was established, validated and
appliedfor a typical agriculture-dominated watershed in eastern China.This
approach provides a basis for quantitatively guidingwatershed management
activities to reduce NPS pollutantamounts and can be easily developed and applied.
The CNAapproach has the potential for universal application and comple-ments
existing tools, such as the TMDL approach, for controllingNPS pollution at the
watershed scale.2. The study areaThe ChangLe River is located in southwest
Shaoxing of ZhejiangProvince, southeast China (Fig. 1). The river drains a total area
of864 km2 and flows about 70.5 km with 0.36% gradients; it has asandy-gravel
riverbed and is 40–70 m wide. Long-term averageannual discharge is 18.4 m3 sÀ1
with monthly variations rangingfrom 2.4 to 36.0 m3 sÀ1; annual flow is 571 Â 106
m3 yÀ1.The 27.8km reach of the ChangLe River (from CL2 to CL4)examined in this
study has a watershed that includes seven townswith a total population of about
330,000. The primary land-use inthe drainage area is agriculture (including paddy,
dry land andnursery) and rural habitation with a population density of about380
persons kmÀ2. The area represents a typical agriculturalwatershed in eastern China
and is characterized by a subtropicalmonsoon climate. Long-term average annual
rainfall is 1541 mmwith the highest rainfall usually occurring from May to
September(Fig. 2). The watershed has a developed economy and serious
waterpollution from NPS agricultural sources. Water inputs fromheadwater streams
(NanShan Reservoir, Fig. 1) are limited dueto export for drinking water.3. Materials
and methods3.1. River sampling and chemical analysisWater quality at three
sampling stations was monitoredmonthly along the ChangLe River from July 2003 to
June 2006(Fig. 1). Water samples were collected from the upper 30 cm of thewater
column in 2.5 L polyethylene bottles between 9:00 am andFig. 1. Map shows the
sampling sites along the ChangLe River and its catchment.D.J. Chen et al. /
Agriculture, Ecosystems and Environment xxx (2009) xxx–xxx2G ModelAGEE-3459;
No of Pages 9Please cite this article in press as: Chen, D.J., et al., Estimation of
critical nutrient amounts based on input–output analysis in anagriculture watershed
of eastern China. Agric. Ecosyst. Environ. (2009), doi:10.1016/j.agee.2009.06.011

--------------------------------------------------------------------------------

Page 3

18:00 pm local time. Dissolved oxygen (DO), temperature and pHwere measured at
the time of collection using a Multi-parameter350i SETs (The Merck Co Ltd.,
Germany). Instantaneous river flowwas measured in situ using a propeller type
flowmeter (ChongQingHydrology Instrument Co Ltd., China). Total nitrogen
(TN)concentration was measured using the persulfate digestion-
UVspectrophotometric method and total phosphorus (TP) wasmeasured following
persulfate digestion using the ammoniummolybdate spectrophotometric method.
Following filtrationthrough a 0.45 mm filter, ammonium (NH4-N) was measured
bythe Astoria Analyzer System (Brown Rupee Co Ltd., Germany),nitrate (NO3-N)
using the UV spectrophotometric method, andsoluble reactive phosphorus (SRP)
using the standard spectro-photometric method (National Standard Method of China
GB11893-89) (Wei et al., 2002). All laboratory analyses werecompleted within 8 h of
sampling.There were abundant hydrophytes in the ChangLe River,especially during
the period from May to September. However,the majority of hydrophytes were
removed by local people in lateOctober and used as a green-manure for winter
crops. In order todescribe the abundance of hydrophytes in the ChangLe River,
thehydrophyte coverage percentage was measured (Irena, 2002). Thehydrophyte
coverage of the river bed (the dimensionless portion ofareal coverage related to
unity, %) was measured in two 20-mreaches (located after CL1 and CL2,
respectively) in April, June andAugust 2004, and April, May, August and October in
2005 and 2006.These 20-m reaches were selected because they contained
differentregions with high, average and low hydrophyte coverage percen-tages in
different locations during the measurement periods.3.2. Source data collectionData
on TN and TP sources (human population, quantity oflivestock–poultry and chemical
fertilizer use) in the ChangLewatershed were collected for each village in 2004–06.
The TN andTP quantity derived from atmosphere deposition (wet deposition)in
2004–06 was provided by Shaoxing City Environment ProtectionBureau of Zhejiang
Province. The TN and TP generation coefficientsfor human and livestock–poultry,
and the TN and TP exportcoefficients for domestic waste, livestock–poultry waste
anddifferent fertilizer types from different land types to the river,were firstly
adopted from literatures (Johnes, 1996; Alexanderet al., 2000; Lester et al., 2000;
May et al., 2001), and then allcoefficients were determined after local validation
using monthlyflow and water quality data in the ChangLe River in 2004–05 (Chenet
al., 2007) (Tables 1 and 2). Monthly river flows and rainfall in2003–06 were supplied
by the Zhejiang Province Hydrology Office.3.3. Critical nutrient amount estimation
methodBased on the definition of critical loads for atmosphericdeposition, the
critical nutrient amount (CNA) for river systemswas defined as: ‘The maximum
amount of a nutrient that can beapplied or generated from natural and human
sources in awatershed without exceeding the designated critical level for
thatnutrient in a water body’. In order to estimate CNA, it is necessaryto determine
the quantitative relationship between applied andgenerated nutrient amounts (AGA,
tyÀ1) and riverine nutrientconcentrations or loads. The method for determining CNA
wasassembled as several linked components (Fig. 3), which includedwatershed
terrestrial and riverine nutrient input–output massbalance estimation (i.e., input–
output model), prediction ofterrestrial nutrient export loads (EL, tyÀ1) from land to
river,and AGA using riverine loads (i.e., output–input feedback model).3.4. Input–
output mass balance modelThe input–output mass balance model examined two
differentsource/sink processes for the generation and fate of TN and TP inand
between terrestrial and riverine systems in the watershed. Thefirst is the nutrients
that are applied or generated from pollutionsources in the watershed and are then
exported from land to river.The second is nutrients that are transported from
upstream and arethen discharged into the river reach of interest. This
approachrequires quantitative estimation of the nutrient AGA and EL andthe
retention of nutrients (assimilative capacity) within the riverduring downstream
transport.For nutrient fate and transport studies at the watershed scale,process-
level models have been developed, such as SWAT (Arnoldet al., 1998) and AGNPS
(Young et al., 1987). They model theprocesses associated with water and sediment
movement, cropgrowth, and nutrient cycling in terrestrial and riverine
systemswithin a watershed. Because these models are complex andFig. 2. Monthly
average rainfall and river flows of ChangLe watershed for the period2003–06.Table
1TN and TP sources in ChangLe watershed.Nutrient
sourcesSpecies200420052006Paddy fieldDry land andother farm landPaddy fieldDry
land andother farm landPaddy fieldDry land andother farm landFertilizer (t
yÀ1)Multiple
fertilizer82111,390112612,37081111,860Urea397110,342473012,038375611,221S
alvolatile637229597077346065343254Superphosphate4857152156429095043112
1Ca–Mg–P mixture557153161424546102159Domestic wastesHuman
population329,895333,265339,820Livestock–poultry
(head)Cattle432035913557Pig141,370188,142180,858Sheep199,730232,79320,24
6Poultry655,537803,4061,156,726D.J. Chen et al. / Agriculture, Ecosystems and
Environment xxx (2009) xxx–xxx3G ModelAGEE-3459; No of Pages 9Please cite this
article in press as: Chen, D.J., et al., Estimation of critical nutrient amounts based on
input–output analysis in anagriculture watershed of eastern China. Agric. Ecosyst.
Environ. (2009), doi:10.1016/j.agee.2009.06.011

--------------------------------------------------------------------------------

Page 4

contain so many parameters for calibration, they require largeamounts of site-
specific data making their application to othersites and times difficult (Borah and
Bera, 2004). Though thesemechanism-based models can estimate EL and riverine
nutrientconcentrations or loads in a given watershed, they are notapplicable in
reverse to predict the AGA for target levels ofriverine nutrient concentrations.
Therefore, watershed AGA and ELwere separately estimated using an export
coefficient model(Johnes, 1996; Lester et al., 2000; May et al., 2001) in this
study.Export coefficients were derived from literature sources andresults of local
field experiments to determine the AGA of nutrientsfrom each identifiable land-use
source and the rate of loss to thesurface drainage network. In ChangLe watershed,
point sourcepollution (including waste-water treatment plants and industrialsewage)
was negligible compared to NPS; annual discharge was0.19 t TN and <0.01 t TP
from point sources in 2004–06. The AGATable 2TN/TP generation coefficients (kg
headÀ1 yÀ1), and export coefficients (%) are defined as the ratio, annual export
load from land to river: annual applied or generated amountsfrom pollution
sources.Nutrient sourcesTNTPGenerate coefficients(kg headÀ1 yÀ1)Export
coefficients (%)Generate coefficients(kg headÀ1 yÀ1)Export coefficients
(%)Livestock–
poultryCattle45.516.29.202.85Pigs5.1614.51.782.55Sheep0.6917.00.173.00Poultry
0.4115.30.232.70Domestic wastes4.0015.81.002.78Arable land typesChemical
fertilizerNitrogen fertilizerPhosphorus fertilizerFertilization quantity(kg haÀ1
yÀ1)Urea (%)aSalvolatile (%)aCompoundfertilizer (%)aSuperphosphate/Ca–Mg–
Pmixture (%)aPaddy field<30018.024.022.04.0300–
40020.026.524.5>40022.029.027.0Dry land/Garden plot<30010.05.0300–
40011.0>40012.0a Fertilizer types.Fig. 3. Outline for critical nutrient amount
estimation method.D.J. Chen et al. / Agriculture, Ecosystems and Environment xxx
(2009) xxx–xxx4G ModelAGEE-3459; No of Pages 9Please cite this article in press
as: Chen, D.J., et al., Estimation of critical nutrient amounts based on input–output
analysis in anagriculture watershed of eastern China. Agric. Ecosyst. Environ.
(2009), doi:10.1016/j.agee.2009.06.011

--------------------------------------------------------------------------------

Page 5

of TN and TP for NPS nutrients in the watershed included fieldapplication of
fertilizer, animal wastes, domestic wastes andatmosphere deposition. AGA of
nutrients from fertilizer, domesticwastes and livestock–poultry were estimated using
fertilizerquantity (fij, tyÀ1), population (Pd, person) and quantity oflivestock–poultry
(Qk, capita) (Table 1). These values were usedwith corresponding generation
coefficients (aj, ad, ak; Table 2) asfollows:AGA ¼X4n¼1ðG f þ Gd þ Gk þ GaÞ(1)G f
¼Xx;yi¼1; j¼1f i j  aj;Gd ¼ Pd  ad  10À3;Gk ¼Xzk¼1Qk  ak  10À3where Gf is
quantity generated from fertilizer (t yÀ1); Gd is quantitygenerated from livestock–
poultry (t yÀ1); Gk is quantity derivedfrom domestic waste (tyÀ1); and Ga is
quantity derived fromatmosphere deposition (t yÀ1). Export loads from land to river
(i.e.,EL) for different land-use types and pollution sources wereestimated as the
product of AGA and export coefficients (Ef, Ed,Ek; Table 2):EL ¼X4n¼1Xx;yi¼1;
j¼1G f  Ei j þ Gd  Ed þ Gk  Ek þ Ga  ¯Ea0@1A(2)where ¯Ea is the average
export coefficient for the watershed.Rainfall enhances N and P exports by surface
runoff andsubsurface leaching, especially during storm events (Grayson et
al.,1996). If other factors (including natural basin characteristics,hydrological and
meteorological conditions, and land-use andmanagement practices) can be
considered uniform, rainfall is themain driver of seasonal distribution of NPS
pollution loads(Grizzetti et al., 2005; Udawatta et al., 2006). For example,monthly
rainfall amounts strongly influence TN and TP concen-trations in the ChangLe
watershed as demonstrated by thesignificant correlations of rainfall with TN (r =
0.71**, 0.78** and0.72** at sites CL1, CL2 and CL3, respectively) and TP (r =
0.63**,0.70** and 0.69** at CL1, CL2 and CL3, respectively) concentrationsin 2003–
06. Therefore, rainfall was a good predictor for themonthly distribution of TN and TP
export loads from land to riverin the ChangLe watershed. Based on this relationship,
the annual ELwas partitioned to monthly EL by simply allocating the annualloads
according to monthly rainfall.Riverine nutrient retention represents the net effect of
variousbiogeochemical processes, such as organic matter deposition/resuspension,
sorption/desorption by sediments, plant and micro-bial uptake/release, and
denitrification. A mass balance approachwas adopted to estimate riverine TN and TP
retention loads (RRLNand RRLP, t moÀ1 or t yÀ1) (Behrendt and Opitz, 2000;
Lepistö et al.,2006):RRLN ¼ Nbeginning þ NEL À Nend(3)RRLP ¼ Pbeginning þ PEL À
Pend(4)where subscript ‘beginning’ denotes inflow loads at the beginningof the
reach, subscript ‘EL’ denotes export load added along thereach, and subscript ‘end’
denotes outflow loads at the end of thereach. Inflow (i.e., Nbeginning and
Pbeginning) and outflow (i.e., Nendand Pend) TN/TP loads were calculated by
multiplying TN/TPconcentrations with river flow.3.5. Output–input feedback
modelsBased on riverine retention analysis, a riverine TN and TP loadsprediction
model was developed according to regression analysisand was verified using data
from a year not used in modeldevelopment. Two output–input feedback models
(OIFM) werefound to effectively predict nutrient EL and AGA of the watershedusing
riverine nutrients loads (i.e., upstream inflow load anddownstream discharge load),
which determined the directrelationship between terrestrial EL/AGA and riverine
loads. TheOIFM for EL and AGA prediction were both validated using data notused in
model development.3.6. Critical nutrient amount estimationAccording to the OIFMs
for EL and AGA predictions, CNA for TN(CNAN) and TP (CNAP) were estimated and
compared to theestablished critical nutrient levels:CNAN ¼P12m¼1ðRRLmN þ
TDLmN À ILmNÞ¯EN(5)CNAP ¼P12m¼1ðRRLmP þ TDLmP À ILmPÞ¯EP(6)where
TDLmN/TDLmP denotes critical output TN/TP loads in thereach (t moÀ1);
RRLmN/RRLmP is the riverine TN and TP retentionload when achieving critical levels
in the reach (t moÀ1); and ¯ENand ¯EP (%) denote the average annual export
percentage of TN andTP AGA for the watershed, respectively. Monthly upstream
anddownstream river flows for riverine nutrient loads estimationswere adopted as
monthly averages for 2003–06.In China, watershed-based water quantity and
quality manage-ment has been conducted only for selected river systems, such
asthe Yangtse River. A river is usually divided into several reaches foreach district
and they are regulated at a designated water qualitycritical level (i.e., water quality
standard) to satisfy different water-use functions for the local area. The critical
levels for TN and TP inthe ChangLe River were set at 1.0mgNLÀ1 and 0.2 mg P
LÀ1according to the National Surface Water Quality Standard of China(GB2002).
CNA for a given river reach is not only influenced byupstream inflow loads (or
concentrations) and critical concentra-tions, but also by the water pollution
compliance method appliedto the analysis (reach-end versus whole-reach
compliance).Therefore, critical TN and TP amounts were estimated to meetcritical
levels for both the end of a specified reach (reach-end) andthe entire length of the
studied reach (whole-reach). The reach-end(typically corresponding to the
boundaries between districts) iscommonly used as the compliance location for river
water qualitymanagement in China. Based on this method, pollutant concen-trations
must be less than or equal to the critical level only at theend of the designated
reach. However, water quality at severalupstream locations may exceed critical
nutrient levels when usingthe reach-end compliance method. Therefore, the whole-
reachcompliance method was evaluated to determine the CNA thatwould be
necessary for nutrients to be less than or equal to criticalnutrient levels for all points
along the river reach, as is the goal formeeting a water quality standard for the
designated use of a waterbody in the USA (National Research Council, 2001).4.
Results and discussion4.1. Applied and generated amounts and resulting export
loadsTotal applied and generated amounts (AGA) for the ChangLewatershed in
2004–06 varied from 15,339 to 18,523 t N yÀ1 for TND.J. Chen et al. / Agriculture,
Ecosystems and Environment xxx (2009) xxx–xxx5G ModelAGEE-3459; No of Pages
9Please cite this article in press as: Chen, D.J., et al., Estimation of critical nutrient
amounts based on input–output analysis in anagriculture watershed of eastern
China. Agric. Ecosyst. Environ. (2009), doi:10.1016/j.agee.2009.06.011

--------------------------------------------------------------------------------

Page 6

and 2555 to 2627tPyÀ1for TP (Table 3). Fertilizer inputsaccounted for the majority
of AGA N ($70.7%) and P ($64.1%),with livestock–poultry waste (N $ 9.1%; P $
22.5%), domesticwaste(N $ 7.2%; P $ 11.4%) andatmospheredeposition(N $ 13.0%;
P $ 2.0%) contributing lesser amounts. Atmosphericdeposition of N ranged from
21.4 to 31.8 kg haÀ1 yÀ1 and P rangedfrom 0.5 to 0.8kg haÀ1 yÀ1. Land to river
export loads (EL)estimated from export coefficients in 2004–06 varied from 2485
to3174 t N yÀ1 for TN, and 110 to 119 t P yÀ1 for TP. EL represents16.2–17.2% of
TN AGA and 4.3–4.5% of TP AGA. Fertilizer was thelargest contributor to N ($70.8%)
and P ($68.7%) export loads.Thus, a nutrient reduction strategy that focuses on
fertilizermanagement will have the most impact on water quality. AnnualEL was
allocated proportional to monthly rainfall resulting inmonthly EL of 7.8–499 t N
moÀ1 for TN and 0.3–19.7 t P moÀ1 forTP. The maximum monthly EL was in
September during thetyphoon season when the highest river flows occurred. There
weresignificant correlations of monthly TN EL with riverine TN loads(r =0.95** and
0.89** at sites CL2 and CL3, respectively) andmonthly TP EL with riverine TP loads (r
= 0.88** and 0.85** at sitesCL2 and CL3, respectively), which demonstrated the
reliability ofthe adopted export model. Average annual river concentrations ofTN
(NO3-N represented 38.0–57.5% and NH4-N 2.0–8.2% of TN overthe study period)
for the three sampling sites were 6.3, 2.7, and4.2 mg N LÀ1in 2004, 2005, and
2006, respectively; averageconcentrations of TP (SRP represented 28.3–38.2% of TP
over thestudy period) were 0.18, 0.10, and 0.14 mg P LÀ1 in 2004, 2005, and2006,
respectively. The large difference in concentrations betweenyears was due to
dilution resulting from higher river flows in 2005;average annual flow was 8.5, 18.0,
and 12.1 m3 sÀ1 in 2004, 2005and 2006, respectively.4.2. Riverine TN and TP
retentionAnnual riverine retention loads (RRL) ranged from 1349 to1964 t N yÀ1 for
TN and 72.4 to 84.6 t P yÀ1 for TP during 2004–06.These values suggest that a
large portion of the TN (44.6–61.3%)and TP (56.2–71.9%) export load plus
upstream-input load (i.e.,total input load, 2645.4–3236.8 t N yÀ1for TN and 112.9–
128.8 t P yÀ1 for TP) was retained by riverine retention mechan-isms. TN and TP
retention percentages in the ChangLe River were atthe high end of the range
reported in other studies, i.e., 1–80% forTN (Behrend, 1996; Dierk and Michael,
2008; Grizzetti et al., 2005;Seitzinger et al., 2002) and 20–70% for TP (Garnier et
al., 2002;Grizzetti et al., 2005; Svendsen and Kronvang, 1993). This impliesthat the
ChangLe River was a substantial sink for N and Pthroughout the year. However,
many riverine retention processesare labile and may readily change with changes in
environmentalconditions (Ostroumov, 2004). Monthly RRL varied from 22.2 to322.2
t N moÀ1 for TN and 1.3 to 16.8 t P moÀ1 for TP. Averagenutrient retention loads
were highest during summer (May toSeptember) ranging from 164 to 222 t N moÀ1
for TN and 8.6 to9.9 t P moÀ1 for TP. This approach for calculating RRL would be
auseful approach for determining the assimilative capacity termused in the TMDL
approach for NPS pollution (National ResearchCouncil, 2001).Relationships between
monthly RRL and several hydrologicaland biogeochemical variables were examined
to evaluate possiblecontrol mechanisms for riverine nutrient retention: river dis-
charge, flow velocity, TN/TP input loads (sum of EL and upstreaminflow load),
dissolved oxygen, temperature and pH (Table 4).Monthly sunlight duration,
hydrophyte coverage and watertemperatures were positively correlated with
monthly RRL forTN and TP, which is consistent with the role of aquatic
organisms(especially hydrophytes) for nutrient retention. There wereabundant
hydrophytes in the ChangLe River, especially duringthe period from May to
September (hydrophyte coverage was51.2–62.6%). Abundant hydrophytes can alter
hydrological andecological conditions and provide favorable conditions for
riverineTN and TP retention. Though the retention of N and P byhydrophytes is
temporary until they die or are transported bystorm flows, removal of hydrophytes
from the river by local peoplein October for use as a green-manure on winter
agricultural cropsis common and acts as the permanent removal process for N and
Pin the ChangLe River. Denitrification is often considered as theprincipal mechanism
for riverine N retention (Dierk and Michael,2008; Saunders and Kalff, 2001;
Seitzinger et al., 2002). Hydro-phytes stimulate denitrification since they provide
substrate forepiphytic biofilms of denitrifying bacteria (Eriksson and Weisner,1997;
Schaller et al., 2004), and supply degradable organic matterto serve as an electron
donor for denitrification (Bastviken et al.,2007; Weisner et al., 1994). Meanwhile,
microgradients in redoxpotential found in plant beds may stimulate the
couplednitrification–denitrificationprocess(NH4 ! NO3 ! N2/N2O)(Eriksson and
Weisner, 1997). Hydrophytes also increase TPretention by vegetative uptake and
provide favorable conditionsfor sedimentation and adsorption. In general,
sedimentation andbiomass uptake are considered the main processes for riverine
TPretention (Marcus and Kohler, 2006). Monthly variation of RRL forTP was
negatively correlated with monthly river flow velocity. Thistrend indicates high
retention of P during low-flow periods anddecreasing P retention at higher river
flows (House and Denison,1997). Both monthly DO concentrations and pH values
decreasedwith increasing RRL for TN and TP. The release of H+andconsumption of
DO synchronously occur during degradation oforganic N into inorganic forms (Wei et
al., 2002), with subsequentuptake of these soluble forms by hydrophytes and algae
in theriver. Organic N is mineralized to NH4-N, which is usually nitrifiedTable
3Annual TN/TP applied and generated amounts (AGA) for diffuse sources and
exportloads (EL) from land to river in the ChangLe watershed in 2004–
06.YearPollution sourcesTNTPAGA(t yÀ1)EL(t yÀ1)AGA(t yÀ1)EL (t
yÀ1)2004Fertilizers10,775.31,707.11,643.976.6Poultry
waste1,549.3257.0574.816.8Domestic waste1,169.6207.8292.49.2Atmosphere
deposition1,845.1313.543.97.5Total15,339.32,485.42,555.0110.12005Fertilizers13,
356.72,281.61,68880.9Poultry waste1241.9214.9580.717.8Domestic
waste1,176.3210.0295.49.3Atmosphere
deposition2,747.9467.065.411.1Total18,522.83,173.52,629.5119.12006Fertilizers11
,769.02,078.11,652.878.3Poultry waste1,745.8263.5593.017.9Domestic
waste1,317.4207.5301.29.8Atmosphere
deposition2,031.6345.248.48.2Total16,863.82,894.32,595.4114.2Table 4Correlation
of monthly riverine TN/TP retention loads with environmentalconditions in ChangLe
River in 2004–06.TN retention load (t)TP retention load (t)Sunlight time
(h)0.65**0.70**Hydrophyte coverage (%)0.68*0.71*Water temperature
(8C)0.60**0.63**Flow velocity (m sÀ1)À0.30À0.65**pHÀ0.67**À0.60**DO (mg
LÀ1)À0.57**À0.55**R values are significant at *0.01 < P < 0.05 and **P < 0.01.D.J.
Chen et al. / Agriculture, Ecosystems and Environment xxx (2009) xxx–xxx6G
ModelAGEE-3459; No of Pages 9Please cite this article in press as: Chen, D.J., et al.,
Estimation of critical nutrient amounts based on input–output analysis in
anagriculture watershed of eastern China. Agric. Ecosyst. Environ. (2009),
doi:10.1016/j.agee.2009.06.011

--------------------------------------------------------------------------------

Page 7

to nitrate consuming DO and releasing H+ (Dai et al., 2006). It hasbeen found that
DO >1.0 mg LÀ1 was sufficient to keep nitrificationat high levels (Hao et al., 1997),
and that nitrification was a majorprocess controlling river oxygen depletion (Dai et
al., 2006).Abundant TN and TP facilitated the growth of hydrophytes sinceN and P
were considered the limiting nutrients for their growth.Therefore, monthly RRLmN (t
moÀ1) and RRLmP (t moÀ1) were bestdescribed (i.e., based on r2) as a linear
function of TN/TP totalmonthly input loads (i.e., TLmN and TLmP, t moÀ1) using data
from2004 to 2005 (Fig. 4). This relationship is similar to previousstudies in lakes,
rivers and wetlands that showed that N retentionincreases with increasing N loading
in aquatic systems anddemonstrated that N loading was an excellent predictor of
Nretention (Saunders and Kalff, 2001; Windolf et al., 1996).4.3. Output–input
feedback models establishment and validation4.3.1. Output–input feedback model
for predicting export loadsAn output–input feedback model (OIFM) for predicting
exportloads for the river reach was developed using measured input andoutput
loads and riverine retention loads derived from regressionequations in Fig. 4:ELmN
¼ 0:54 Â ðILmN þ ELmNÞ þ 5:88 þ DLmN À ILmN(7)ELmP ¼ 0:79 Â ðILmP þ ELmPÞ
À 0:88 þ DLmP À ILmP(8)where ELmN/ELmP denotes TN/TP export loads along the
reach(t moÀ1); DLmN/DLmP denotes TN/TP output loads at the end of thereach (t
moÀ1); and ILmN/ILmP denotes TN/TP input loads at thebeginning of the reach (t
moÀ1).Validation of this OIFM for EL was necessary before furtherapplication.
According to Eqs. (7) and (8), ELmN/ELmP could bepredicted using riverine input
and output loads. Thus, the 2006data were used to validate the input–output model
(developedusing 2004–05 data for RRL equations) for ELmNand ELmPprediction.
The correlation between export loads predicted bythe OIFM versus those estimated
using the export coefficientmodel for ELmNand ELmPwere r2 = 0.94** and r2 =
0.95**,respectively(Fig.5).AveragepredictionerrorwasÆ37.2 t N moÀ1 (or Æ19.5%)
for ELmN and Æ1.2 t P moÀ1 (Æ19.2%)for ELmP. The results demonstrate the
reliability of the statisticalmodels for RRLN and RRLP prediction and validate the
OIFM for ELprediction.4.3.2. Output–input feedback model for predicting annual
applied orgenerated amountsBased on validated Eqs. (7) and (8), output–input
feedbackmodels (OIFM) for predicting annual applied or generated TN(AGAN,tyÀ1)
and TP (AGAP,tyÀ1) amounts in the watershed usingpredicted EL were further
established:AGAN ¼P12m¼1 ELmN¯EN(9)AGAP ¼P12m¼1 ELmP¯EP(10)where
ELmN/ELmP is TN and TP export loads from land to the river(t moÀ1), which were
estimated by Eqs. (7) and (8) using DLm andILm.According to Eqs. (9) and (10), the
average export percentages(i.e., ¯EN and ¯EP, %) were essential for application of
the OIFM. ¯EN and¯EP are defined as the percentage calculated from the ratio,
annualEL:annual AGA, for the whole ChangLe River watershed. Threefactors were
considered to determine the average exportpercentages for each month during
2004–06: (i) the pattern ofland-use, including farmland, inhabited land, etc., was
similaramong years, (ii) similar fertilizer and animal waste quantitieswere applied to
farmland each year, and (iii) there was littledifference in monthly rainfall
distribution among years. Based onthese assumptions, export rates for each month
during 2004–06were calculated and the average values (¯EN ¼ 16:8% and¯EP ¼
4:4%) were used. In succession, riverine loads for each monthin 2006, including
downstream-output and upstream-input loads,were combined with ¯EN and ¯EP to
validate the OIFM. There wasreasonable agreement for predicted annual AGA
between the OIFMand export coefficient model for the watershed. Prediction
errorwas À1510.7 t (or À9.8%) for TN and 136.0t (or 5.0%) for TPFig. 4.
Relationships of monthly riverine TN/TP retention loads with total TN/TP input loads
(i.e., sum of export load and upstream inflow load) in ChangLe River in 2004–05.Fig.
5. Validated results of output–input feedback model for monthly TN and TPexport
loads (EL) from land to river in ChangLe watershed in 2006. Abscissaindicates the
EL estimated by export coefficient model; ordinate indicates the ELestimated by
output–input feedback model (OIFM) that was developed usingmeasured riverine
downstream-output and upstream-input loads and retentionloads derived from
regression equations.D.J. Chen et al. / Agriculture, Ecosystems and Environment xxx
(2009) xxx–xxx7G ModelAGEE-3459; No of Pages 9Please cite this article in press
as: Chen, D.J., et al., Estimation of critical nutrient amounts based on input–output
analysis in anagriculture watershed of eastern China. Agric. Ecosyst. Environ.
(2009), doi:10.1016/j.agee.2009.06.011

--------------------------------------------------------------------------------

Page 8

(Table 5), which suggested that the founded OIFM was a goodpredictor of AGA.4.4.
Critical TN and TP amountsBased on the OIFM for AGA prediction, the CNA can
beestimated by Eqs. (5) and (6) to achieve the established criticalnutrient levels. In
general, however, there is a delay between thetime that a nutrient is applied or
generated and the time that thenutrient appears in rivers (Sta˚lnacke et al., 2003).
Other studiesreported that this time delay was mainly dependent on ground-water
residence times (De Wit and Giuseppe, 2001), since theresidence times for direct
surface runoff varied from several hoursto several days. In contrast, the
groundwater residence timesvaried from several months to two hundred years as
watershedareas varied from 162 to 224 Â 103 km2 (Hu et al., 2007; De Witet al.,
2003; Reungsang et al., 2007; Wendland et al., 2002).However, more than 80% of
total runoff was delivered to the riverthrough direct surface runoff and shallow
groundwater movementin several hours to about one year (De Wit et al., 2003).
Especiallyimportant with respect to our approach, the export coefficientsgenerally
described export percentages of AGA for differentsources and land-use at the
annual scale. Therefore, it is morereasonable to estimate CNA at the annual rather
than monthly timescale.4.4.1. Scenario I: reach-end compliance methodThe annual
CNAN for the reach-end compliance method was7174 t N yÀ1and CNAPwas 5004 t P
yÀ1for the watershed.Comparing the mean concentrations at reach-end (CL3)
withcritical levels for TN and TP in 2004–06, the AGA for TN would needto be
reduced 8165 (53.2%), 11,349 (61.3%), and 9690tNyÀ1(57.5%) in 2004, 2005, and
2006, respectively, to meet the criticalTN level (1.0 mg LÀ1) at the reach end. When
TN AGA was reducedby 53.2–61.3% from the existing values, the method predicted
anearly proportional decrease (50.1–60.1%) in riverine TN load atthe river outlet
(CL3). Those predictions are supported by resultsfrom Hu et al. (2007) that found
there was a 10–41% decrease inriverine NO3-N load when N fertilizer rates were
hypotheticallyreduced by 10–50% from the existing values. In contrast, AGA for
TPwas 2449, 2375, and 2409 t P yÀ1 below the critical TP amount in2004, 2005,
and 2006, respectively, indicating that current TP AGAcould be increased by about
90.3–95.9% for the watershed withoutexceeding the critical TP level (0.2 mg LÀ1) at
the reach end.4.4.2. Scenario II: whole-reach compliance methodIn the reach of the
ChangLe River examined in this study most ofthe monthly TN concentrations for
upstream inflow (CL1)exceeded the critical TN level (1.0 mg LÀ1) ranging from 1.1
to7.5 mg N LÀ1 during 2003–06. On the other hand, monthly TPconcentrations were
always less than the critical TP level(0.2 mg LÀ1) ranging from 0.04 to 0.19 mg LÀ1
during 2003–06.Obviously, in order to meet critical levels along the entire reach,the
TN and TP concentrations of upstream-input waters must beless than or equal to the
critical nutrient level. To calculate CNA forthe whole-reach compliance method,
monthly TN and TPconcentrations of upstream inflows (CL1) were set at their
criticallevels for calculation purposes.For the whole-reach compliance method, the
annual CNAN forthe watershed area associated with the studied reach was8290 t N
yÀ1 and the annual CNAP was 4425 t P yÀ1. ComparingCNAN with AGA indicated a
required reduction in AGA of 7049(46.0% reduction), 10,233 (55.2% reduction), and
8574tNyÀ1(50.8% reduction) in 2004, 2005, and 2006, respectively, to meetcritical
nutrient levels for TN along the entire river reach. Incontrast, AGA for TP was 1870,
1796, and 1830 t P yÀ1 below theCNAP, in 2004, 2005, and 2006, respectively. This
means thatcurrent TP AGA could be increased by 68.3–73.2% for thewatershed area
without exceeding the critical TP level (0.2 mg LÀ1)1) along the entire reach.CNAN
for the reach-end method was less than for the whole-reach method, while the
opposite was true for CNAP. Thisdifference primarily resulted from differences in
upstream inflowTN and TP concentrations used in the CNA estimations.
CNANincreased about 1116tNyÀ1(a 15.6% increase of reach-endcompliance value
7174 t N yÀ1) when upstream inflow TN con-centration was decreased 1.5 mg LÀ1
to the 1.0 mg LÀ1 standard (a53.0% decrease of the current TN concentration 2.5
mg LÀ1); CNAPdecreased about 579tPyÀ1(a 11.6% increase of reach-
endcompliance value 5004 t P yÀ1) when upstream inflow TP con-centration was
increased 0.11 mg LÀ1 to the 0.2 mg LÀ1 standard (a53.9% increase of the current
TP concentration 0.09 mg LÀ1). Inother words, CNA decreased when upstream
inflow nutrientconcentrations increased. In practice, total AGA from pollutionsources
in the watershed must be controlled to achieve CNA, whichmeans that agricultural
nutrient applications to farmlands shouldbe reduced appropriately since the sources
of atmospheredeposition have no borders and fertilizer N- and P-use efficiencyis
generally very low (Zhang et al., 2007; Wang et al., 2007).5. ConclusionsThe CNA
approach offers a quantitative method to control waterpollution with a focus on
pollution sources, assuming no change incurrent land-use types at the watershed
scale. In the ChangLeagriculture-dominated watershed of eastern China, CNA
was7174 t N yÀ1 and 5004 t P yÀ1 for the reach-end control method,and 8290 t N
yÀ1 and 4425 t P yÀ1 for the whole-reach controlmethod. Annual TN AGA exceeded
CNA by 53.2–61.3% and 46.0–55.2% for reach-end and whole-reach control
methods, respec-tively. In contrast, TP AGA values were 90.3–95.9% and 68.3–
73.2%below CNA values for reach-end and whole-reach control
methods,respectively.The method developed for CNA in this study can be extended
toany hazardous substance originating from NPS pollution. Itrequires information
about rainfall and pollution sources in thebasin along with monthly chemical and
hydrological data.Reliability of CNA is related to the data collection
frequency,watershed land-use characteristics, fate and transport character-istics of
individual pollutants, water flow, and the uncertainty inthe pollutant export and
riverine pollutant retention due to a non-homogeneous distribution of pollution
sources, land-use types,landscapes, etc. In future studies, seasonal CNA and its
estimateapproach require further testing as more is learned about timedelay
(transport dynamics) between changes in applied andgenerated nutrient amounts
and nutrient export to the riversystem, since NPS pollution usually demonstrates
remarkabletemporal variations. It is feasible to further reduce the criticalamounts of
chemical fertilizer, population, livestock–poultry andothers using corresponding
allocation methods based on CNA.Table 5Output–input feedback model validation
results for annual TN/TP applied andgenerated amounts (AGA) and export loads (EL)
prediction using riverine loads inChangLe watershed in 2006.TNTPEL(t yÀ1)AGA(t
yÀ1)EL(t yÀ1)AGA(t yÀ1)Export coefficient
model2894.316,863.8114.22595.4Output–input feedback
model2584.415,353.1120.52731.3Relative error (%)À10.7À9.85.35.0D.J. Chen et
al. / Agriculture, Ecosystems and Environment xxx (2009) xxx–xxx8G ModelAGEE-
3459; No of Pages 9Please cite this article in press as: Chen, D.J., et al., Estimation
of critical nutrient amounts based on input–output analysis in anagriculture
watershed of eastern China. Agric. Ecosyst. Environ. (2009),
doi:10.1016/j.agee.2009.06.011

--------------------------------------------------------------------------------

Page 9

These calculations provide a target or permissible nutrient amountthat can be used
to develop management practices that allowattainment of water quality objectives
at the watershed scale.Thus, the CNA approach provides a quantitative basis for
theregulatory community to harmonize water quality regulation andagricultural
ecosystem protection at the watershed scale.AcknowledgementsThis work was
supported by National High TechnologyResearch and Development Program of
China (2007AA10Z218)and National Natural Science Foundation of China
(40871104).

References

Alexander, R.B., Smith, R.A., Schwarz, G.E., 2000. Effect of stream channel size on
thedelivery of nitrogen to the Gulf of Mexico. Nature 403, 758–761.Arnold, J.G.,
Srinivasan, R., Muttiah, R.S., Williams, J.R., 1998. Large area hydrologicmodeling
and assessment part I: model development. J. Am. Water Resour.Assoc. 34, 73–
89.Barkman, A., Alveteg, M., 2001. Identification potentials for reducing
uncertaintyin critical load calculations using the profile model. Water Air Soil Pollut.
125,33–54.Bastviken, S.K., Eriksson, P.G., Ekstrom, A., Tonderski, K., 2007. Seasonal
denitrifica-tion potential in wetland sediments with organic matter from different
plantspecies. Water Air Soil Pollut. 183, 25–35.Behrend, H., 1996. Inventories of
point and diffuse sources and estimated nutrientloads: a comparison for different
river basin in central Europe. Water Sci.Technol. 33, 99–107.Behrendt, H., Opitz, D.,
2000. Retention of nutrients in river systems: dependence onspecific runoff and
hydraulic load. Hydrobiologia 410, 111–122.Borah, D.K., Bera, M., 2004. Watershed-
scale hydrologic and nonpoint-sourcepollution models: reviews of application. Trans.
Am. Soc. Agric. Eng. 47,789–803.Brian, M.D., Daniel, P., Marc, L.H., 2008.
Agricultural nonpoint source water pollu-tion policy: the case of California’s Central
Coast. Agric. Ecosys. Environ. 128,151–161.Chen, D.J., Lu, J., Jin, S.Q., Shen, Y.N.,
2007. Estimation and allocation of waterenvironmental capacity in nonpoint source
polluted river (in Chinese, withEnglish abstract). Environ. Sci. 28, 1409–1415.Dai,
M.H., Guo, X.H., Zhai, W.D., Yuan, L.Y., Wang, B.W., Wang, L.F., Cai, P.H., Tang,T.T.,
Cai, W.J., 2006. Oxygen depletion in the upper reach of the Pearl Riverestuary
during a winter drought. Mar. Chem. 102, 159–169.De Wit, M., Giuseppe, B., 2001.
Nutrient fluxes in the Po basin. Sci. Total Environ.273, 147–161.De Wit, M., Meinardi,
C., Wendland, F., Kunkel, R., 2003. Modelling water fluxes forthe analysis of diffuse
pollution at the river basin scale. Hydrolog. Process. 14,1707–1723.Diaz, R.J.,
Rosenberg, R., 2008. Spreading dead zones and consequences for
marineecosystems. Science 321, 926–929.Dierk, W., Michael, R., 2008. Modelling
the impact of river morphology on nitrogenretention—a case study of the Weisse
Elster River (Germany). Ecol. Model. 21,224–232.Duan, L., Hao, J.M., Xie, S.D., Du,
K., 2000. Critical loads of acidity for surface waters inChina. Sci. Total Environ. 246,
1–10.Elofsson, K., 2003. Cost-effective reductions of stochastic agricultural loads to
theBaltic Sea. Ecol. Econ. 47, 13–31.Eriksson, P.G., Weisner, S.E.B., 1997. Nitrogen
removal in a wastewater reservoir:the importance of denitrification by epiphytic
biofilms on submersed vegeta-tion. J. Environ. Qual. 26, 905–910.Garnier, J., Billen,
G., Hannon, E., Fonbonne, S., Videnina, Y., Soulie, M., 2002.Modelling the transfer
and retention of nutrients in the drainage network ofthe Danube River. Estuar.
Coast. Shelf Sci. 54, 285–308.Gassman, P.W., Edward, O., Saleh, A., Rodecap, J.,
Norvell, S., Williams, J., 2006.Alternative practices for sediment and nutrient loss
control on livestock farmsin northeast Iowa. Agric. Ecosys. Environ. 117 (2, 3), 135–
144.Grayson, R.B., Finlayson, B.L., Gippel, C.J., Hart, B.T., 1996. The potential of
fieldturbidity measurements for the computation of total phosphorus and sus-
pended solids loads. J. Environ. Manage. 47, 257–267.Grizzetti, B., Bouraoui, F.,
Marsily, G.D., Bidoglio, G.A., 2005. Statistical method forsource apportionment of
riverine nitrogen loads. J. Hydrol. 304, 302–315.Hao, X.D., Doddema, H.J., Van
Groenestijn, J.W., 1997. Conditions and mechanismsaffecting simultaneous
nitrification and denitrification in a Pasveer oxidationditch. Bioresour. Technol. 59,
207–215.House, W.A., Denison, F.H., 1997. Nutrient dynamics in a lowland stream
impactedby sewage effluent: Great Ouse, England. Sci. Total Environ. 205 (1), 25–
49.Hu, X., McIsaac, G.F., David, M.B., Louwers, C.A.L., 2007. Modeling riverine
nitrateexport from an east-central Illinois watershed using SWAT. J. Environ. Qual.
36,996–1005.Irena, P., 2002. Initial impact of low stocking density of grass carp on
aquaticmacrophytes. Aquat. Bot. 73, 8–19.Johnes, P.J., 1996. Evaluation and
management of the impact of land use change onthe nitrogen and phosphorus load
delivered to surface waters: the exportcoefficient modeling approach. J. Hydrol. 183,
323–349.Laane, R.W.P.M., 2005. Applying the critical load concept to the nitrogen
load of theriver Rhine to the Dutch coastal zone. Estuar. Coast. Shelf Sci. 62, 487–
493.Lepistö, A., Granlund, K., Kortelainen, P., Räike, A., 2006. Nitrogen in river
basins:sources, retention in the surface waters and peatlands, and fluxes to
estuaries inFinland. Sci. Total Environ. 365, 238–259.Lester, J.M., Bradley, D.E.,
Shahanat, H.H., 2000. Transport and retention of nitrogenand phosphorus in the
sub-tropical Richmond River estuary, Australia–a budgetapproach. Biogeochemistry
50, 241–278.Malcolm, S.C., 2000. The critical loads concept: milestone or millstone
for the newmillennium? Sci. Total Environ. 249, 51–62.Marcus, S., Kohler, J., 2006. A
simple model of phosphorus retention evoked bysubmerged macrophytes in
lowland rivers. Hydrobiologia 563, 521–525.Markku, Y.-H., Helinä, H., Petri, E., Eila,
T., Markku, P., Kari, K., 1995. Assessment ofsoluble phosphorus load in surface
runoff by soil analyses. Agric. Ecosys.Environ. 56 (1), 53–62.May, L., House, W.A.,
Bowes, M., McEvoy, J., 2001. Seasonal export of phosphorusfrom a lowland
catchment: upper River Cherwell in Oxfordshire, England. Sci.Total Environ. 269,
117–130.Meals, D.W., Cassell, E.A., Hughell, D., Wood, L., Jokela, W.E., Parsons, R.,
2008.Dynamic spatially explicit mass-balance modeling for targeted
watershedphosphorus management. II. Model application. Agric. Ecosys. Environ.
127(3/4), 223–233.National Research Council (NRC), 2001. Assessing the TMDL
Approach to WaterQuality Management. U.S. National Academy Press, Washington,
DC.Ostroumov, S.A., 2004. On the biotic self-purification of aquatic ecosystems: ele-
ments of the theory. Dokl. Biol. Sci. 396, 206–211.Reungsang, P., Kanwar, R.S., Jha,
M., Gassman, P.W., Ahmad, K., Saleh, A., 2007.Calibration and validation of SWAT
for the Upper Maquoketa River watershed.Int. Agric. Eng. J. 16, 35–48.Ribaudo,
M.O., Heimlich, R.E., Claassen, R., Peters, M., 2001. Least-cost managementof
nonpoint source pollution: source reduction versus interception strategiesfor
controlling nitrogen loss in the Mississippi Basin. Ecol. Econ. 37, 183–197.Saunders,
D.L., Kalff, J., 2001. Nitrogen retention in wetlands, lakes and rivers.Hydrobiologia
443, 205–212.Schaller, J.L., Royer, T.V., David, M.B., Tank, J.L., 2004. Denitrification
associatedwith plants and sediments in an agricultural stream. J. N Am. Benthol.
Soc. 23,667–676.Seitzinger, S.P., Styles, R.V., Boyer, E.W., Alexander, R.B., Billen,
G., Howarth, R.W.,Mayer, B., Van Breemen, N., 2002. Nitrogen retention in rivers:
model devel-opment and application to watersheds in the northeastern U.S.A.
Biogeochem-istry 57/58, 199–237.Stålnacke, P., Grimvall, A., Libiseller, C., Laznik,
M., Kokorite, I., 2003. Trends innutrient concentrations in Latvian rivers and the
response to the dramaticchange in agriculture. J. Hydrol. 283, 184–205.Svendsen,
L.M., Kronvang, B., 1993. Retention of nitrogen and phosphorus in aDanish lowland
river system: implications for the export from the watershed.Hydrobiologia 51, 123–
135.Udawatta, R.P., Motavalli, P.P., Garrett, H.E., Krstansky, J.J., 2006. Nitrogen
losses inrunoff from three adjacent agricultural watersheds with claypan soils.
Agric.Ecosys. Environ. 117 (1), 39–48.Wang, G.H., Zhang, Q.C., Witt, C., Buresh, R.J.,
2007. Opportunities for yield increasesand environmental benefits through site-
specific nutrient management in ricesystems of Zhejiang province, China. Agric.
Syst. 94, 801–806.Wei, F.X., Qi, W.Q., Sun, Z.G., 2002. Water and wastewater
monitoring and analysismethod, 4th ed. China Environmental Science Press, Beijing
(in Chinese) pp,238–240.Weisner, S.E.B., Eriksson, P.G., Graneli, W., Leonardson, L.,
1994. Influence ofmacrophytes on nitrate removal in wetlands. Ambio 23, 363–
366.Wendland, F., Kunkel, R., Grimvall, A., Kronvang, B., Mü ller-Wohlfeil, D.I., 2002.
TheSOIL-N/WEKU model system—a GIS-supported tool for the assessment
andmanagement of diffuse nitrogen leaching at the scale of river basins. Water
Sci.Technol. 45, 285–292.Windolf, J.E., Jeppesen, J.P., Jensen, P.K., 1996. Modelling
of seasonal variation innitrogen retention and in-lake concentration: a four-year
mass balance study in16 shallow Danish lakes. Biogeochemistry 33, 25–44.Young,
R.A., Onstad, C.A., Bosch, D.D., Anderson, W.P., 1987. AGNPS, Agriculturalnonpoint-
source pollution model: a watershed analytical tool. ConservationResearch Report
No. 35. USDA, Washington, DC.Zhang, Q.L., Shi, X.Z., Huang, B., Yu, D.S., O¨born, I.,
Blombäck, K., Wang, H.J.,Pagella, T.F., Sinclair, F.L., 2007. Surface water quality of
factory-based andvegetable-based peri-urban areas in the Yangtze River Delta
region, China.Catena 69, 57–64.

D.J. Chen et al. / Agriculture, Ecosystems and Environment xxx (2009) xxx–xxx9G
ModelAGEE-3459;

No of Pages 9

Please cite this article in press as: Chen, D.J., et al., Estimation of critical nutrient
amounts based on input–output analysis in anagriculture watershed of eastern
China. Agric. Ecosyst. Environ. (2009), doi:10.1016/j.agee.2009.06.011