Embed Size (px)
Citation preview
In
Ba
b
2
a
ARRAA
KWNTA
1
nty1ofi1aftwMdtp
0d
Ecological Modelling 220 (2009) 2325–2334
Contents lists available at ScienceDirect
Ecological Modelling
journa l homepage: www.e lsev ier .com/ locate /eco lmodel
ntegrated biogeochemical modelling of nitrogen load from anthropogenic andatural sources in Japan
in Hea,∗, Taikan Okia, Shinjiro Kanaeb, Goro Mouria, Ken Kodamaa, Daisuke Komoria, Shinta Setoa
Institute of Industrial Science, The University of Tokyo, Be605, 4-6-1 Komaba, Meguro-ku, Tokyo 153-8505, JapanDepartment of Mechanical and Environmental Informatics, Tokyo Institute of Technology,-12-1 O-Okayama, Meguro-ku, Tokyo 152-8552, Japan
r t i c l e i n f o
rticle history:eceived 27 January 2009eceived in revised form 13 May 2009ccepted 14 May 2009vailable online 25 June 2009
eywords:ater pollutionitrogen loaderrestrial ecosystem
a b s t r a c t
This study proposed an integrated biogeochemical modelling of nitrogen loads from anthropogenic andnatural sources in Japan. Firstly, the nitrogen load (NL) from different sources such as crop, livestock,industrial plant, urban and rural resident was calculated by using datasets of fertilizer utilization, popu-lation distribution, land use map, and social census. Then, the nitrate leaching from soil layers in farmland,grassland and natural conditions was calculated by using a terrestrial nitrogen cycle model (TNCM). TheTotal Runoff Integrating Pathways (TRIP) was used to transport nitrogen from natural and anthropogenicsources through river channels, as well as collect and route nitrogen to the river mouths. The forcingmeteorological and hydrological data is a 30-year (1976–2005) dataset for Japan which were obtained bythe land surface model, Minimal Advanced Treatments of Surface Interaction and Runoff (MATSIRO). For
nthropogenic and natural sources the model validation, we collected total nitrogen (TN) concentration data from 59 rivers in Japan. As acomparison result, calculated TN concentration values were in good agreement with the observed ones,which shows the reliability of the proposed model. Finally, the TN loads from point and nonpoint sourceswere summarized and evaluated for 59 river basins in Japan. The proposed modelling framework can beused as a tool for quantitative evaluation of nitrogen load in terrestrial ecosystems at a national scale. Theconnection to land use and climate data provides a possibility to use this model for analysis of climate
ge im
change and land use chan. Introduction
As an integral component of many essential plant nutrients,itrogen (N) is both an essential nutrient and a major pollutant inerrestrial ecosystems and plays important roles in increasing cropields and crop quality (Brady, 1998; Baker, 2003; Oenema et al.,998; Schepers et al., 1995). During the last century, the productionf food and energy has markedly increased the amount of newlyxed N entering terrestrial and aquatic ecosystems. Compared with890, the amount of newly fixed N entering terrestrial systemsnnually had about doubled due to the production of syntheticertilizers, increased biological N fixation associated with agricul-ural crops, and increased atmospheric N deposition associatedith fossil fuel combustion (Galloway et al., 1995; Galloway, 2000).
oreover, excess nitrogen used in fertilization has undoubtedlyisturbed the biogeochemical nitrogen cycle of natural ecosys-ems, resulting in various global, regional, and local environmentalroblems such as stratospheric ozone depletion, soil acidification,
∗ Corresponding author. Tel.: +81 03 5452 6382/6381; fax: +81 03 5452 6383.E-mail address: [email protected] (B. He).
304-3800/$ – see front matter © 2009 Elsevier B.V. All rights reserved.oi:10.1016/j.ecolmodel.2009.05.018
pacts on hydrology and water quality.© 2009 Elsevier B.V. All rights reserved.
eutrophication, and NO3− pollution of ground and surface waters
(Davis and Koop, 2006; Ding et al., 2006; Hantschel and Beese,1997; Rijtema and Kroes, 1991). Especially, water quality associatedwith nitrate (NO3
−) leaching from agricultural soils is an importantenvironmental issue in the globe (Galloway, 1998, 2000; Gallowayand Cowling, 2002; Galloway et al., 1995). The effect of agriculturalnonpoint source (NPS) N pollution on water quality and aquaticecosystems has been the subject of considerable research in recentyears (Howarth et al., 2002; Hudson et al., 2005).
In Japan, the water quality has been improved remarkably dur-ing the past decades but Japanese rivers are still heavily impactedby canalization, loss of most dynamic flood plains, flow regulation,invasion by exotic species, and intensive urbanization (Yoshimuraet al., 2005; Kimura and Hatano, 2007). Japanese agriculture hascreated high N surpluses in agricultural lands due to the increas-ing rate of chemical fertilizer application and disposal of livestockwastes per farmland area (Mishima, 2001; Kimura, 2005). As a
consequence, agricultural activities, including intensive livestockproduction, have been widely criticized for producing environmen-tal pollution (Kimura et al., 2004). It is necessary to take measuresfor sustainable agricultural production in better harmony withthe environment in preserving and improving the natural cyclical
2 dellin
fdicfafcntro(inbtscf
amsciicbdtwa
2
2
tDtaaoASchcassct(swtewfaM(
326 B. He et al. / Ecological Mo
unctions of agriculture. It is considered that excess nitrogen fromomestic wastewater, livestock wastewater, and nonpoint resource
s the major reason that made some rivers nutrient polluted andause some lakes and inner bay eutrophication in Japan. It is knownrom field tests that if water contains high nitrogen concentrationnd flows into paddy fields, high nitrogen removal would be per-ormed (Nakasone et al., 2003). Furthermore, underground waterontamination caused by nitrate nitrogen has begun to make itselfoticed. The causes for this contamination are household wastewa-er discharge, agricultural waste resulting from cattle farming, andunoff from areas that are targeted by intensive farming meth-ds reliant on the consumption of large quantities of fertilizersMinistry of Environment of Japan). Therefore, the water qualitys a big concern among water related issues in Japan. Furthermore,itrogen pollution is one of the major pollutants but difficult toe estimated because its sources are widely spread and relatedo different complex sources including natural and anthropogenicources. So far, the researches on nitrogen cycle study are mainlyonducted at basin scale and the national or global scale study isew.
This study’s objective is to estimate the nitrogen loads in Japant a national scale and identify areas where sever water pollutionight be taking place. The nitrogen loads from point and nonpoint
ources are calculated separately and an integrated biogeochemi-al modelling of nitrogen export to Japanese streams was proposedn this study. Subsequently, we discuss methodology in Section 2ncluding land surface model (Section 2.1), the terrestrial nitrogenycle model (Section 2.2), the river routing model (Section 2.3),iological N fixation and atmospheric N deposition (Section 2.4),enitrification in root zone (Section 2.5), and nitrate leaching (Sec-ion 2.6). The database for the point sources and nonpoint sourcesill be discussed in Section 3. The integrated simulation with the
bove model and database will be discussed in Section 4.
. Methodology
.1. Land surface model
A number of land surface models (LSMs) have been developedo be used in global or regional climate models (Sellers et al., 1996;ickinson et al., 1998). These models incorporate the radiation
ransfer, the evaporation, the transpiration, the snow, the runoff,nd so on considering the effects of vegetation, and solve the energynd water exchange between land and atmosphere as the verticalne-dimensional processes. In this study, we employ the Minimaldvanced Treatment of Surface Interaction and Runoff Model (MAT-IRO) which is projected to be used for long-term simulations oflimate studies (Takata, 2000, 2001; Takata et al., 2003). MATSIROas a single-layer canopy and albedo. The bulk exchange coeffi-ients are evaluated based on a multilayer canopy model. The fluxesre calculated from the energy balance at the ground and canopyurfaces in both snow-free and snow-covered portions that con-ider the subgrid snow distribution. Evaporation of water on theanopy and transpiration parameterized on the basis of photosyn-hesis (Sellers et al., 1996) are included. A simplified TOPMODELBeven and Kirkby, 1979) calculates baseflow runoff, in addition tourface flows. Snow has up to three layers depending on the snowater equivalent, and snow layer temperatures are calculated with
hermal conduction equations. Snowmelt and refreeze are consid-red in this model. There are five soil layers in which energy and
ater movements are treated with physical equations that considerreezing and condensation. The mathematical formulas describingll these processes in detail can be found in Takata (2000, 2001).odel application results are also described in Hirabayashi et al.
2005). It has been validated both at the global scale (Takata, 2000)
g 220 (2009) 2325–2334
and at a local scale (Takata, 2001). A simulation coupled with anatmospheric general circulation model (AGCM) was described inSakamoto et al. (2004). It reproduces well the observed seasonalcycles of the energy and water balance.
2.2. Terrestrial nitrogen cycle model
The TNCM (Fig. 1) is developed to consider the mass balanceof nitrogen in the natural ecosystem integrated with the carboncycle in vegetation and organic soil. It is based on the originalmodel by Lin et al. (2000, 2001). The ecosystem was divided intoan atmospheric and a terrestrial reservoir. The terrestrial nitrogencycle consists in biological processes which depend on the varietyof the environmental factors. The model contains eight variables:nitrogen in vegetation (Nveg, unit: tonne N km−2), carbon in vege-tation (Cveg, unit: tonne N km−2), organic N in detritus (Ndet, unit:tonne N km−2), organic carbon in detritus (Cdet, unit: tonne N km−2),organic nitrogen in humus (Nhum, unit: tonne N km−2), organic car-bon in humus (Chum, unit: tonne N km−2), ammonium (Namm, unit:tonne N km−2), and nitrate (Nnit, unit: tonne N km−2) as below.
∂Cveg
∂t= gpp − ctrr − cf (1 − hvst) (1)
∂Cdet
∂t= cf − cdr − cdh (2)
∂Chum
∂t= cdh − chr − chcar (3)
∂Nveg
∂t= nuptake − nf (1 − hvst) + nfix (4)
∂Ndet
∂t= nf − ndm − ndh (5)
∂Nhum
∂t= ndh − nhm + fert hum + lst (6)
∂Namm
∂t= ndm + nhm + nammd − nuptake × Namm
Namm + Nnit− nnitrif
− nvola + fert amm (7)
∂Nnit
∂t= nnitrif − nnitrgas + nnitd − nuptake × Nnit
Namm + Nnit− ndenitr
− nleach + fert nit (8)
where, gpp is flux of photosynthesis as in gross primary produc-tion (tonne N km−2 day−1), ctrr is flux of respiration of trunk androot (tonne C km−2 day−1), cf is flux of litter-fall from leaf, trunk,and root as in carbon (tonne C km−2 day−1), cdr is flux of detri-tus decomposition as in carbon (tonne C km−2 day−1), cdh is flux ofdetritus huminification as in carbon (tonne C km−2 day−1), chr is fluxof humus decomposition as in carbon (tonne C km−2 day−1), chcaris flux of humus carbonization as in carbon (tonne C km−2 day−1),nuptake is flux of nitrogen uptake by plant (tonne N km−2 day−1),nf is flux of litter-fall from leaf, trunk, and root as in nitrogen(tonne N km−2 day−1), nfix is flux of nitrogen fixation as in nitro-gen (tonne N km−2 day−1), ndm is flux of detritus mineralization asin nitrogen (tonne N km−2 day−1), ndh is flux of detritus humini-fication as in nitrogen (tonne N km−2 day−1), nhm is flux of humusmineralization as in nitrogen (tonne N km−2 day−1), nammd is flux ofnitrogen deposition as in ammonium (tonne N km−2 day−1), Namm
is potential nitrogen storage as in ammonium (tonne N km−2), Nnit
is potential nitrogen storage as in nitrate (tonne N km−2), nnitrif isflux of nitrification (tonne N km−2 day−1), nvola is flux of ammo-nia volatilization (tonne N km−2 day−1), nnitrgas is flux of gaseousemissions during nitrification process (tonne N km−2 day−1), nnitdis flux of nitrogen deposition as in nitrate (tonne N km−2 day−1),
B. He et al. / Ecological Modelling 220 (2009) 2325–2334 2327
errest
nnfoohfcc
2
lnP2fwttaat
2
ebfwfoett
2
tifl
Fig. 1. Flow chart for the t
denitr is flux of denitrification (tonne N km−2 day−1), nleach is flux ofitrate leaching (tonne N km−2 day−1), fert hum is the amount of
ertilizer in humus (tonne N km−2 day−1), fert amm is the amountf fertilizer in ammonium (tonne N km−2 day−1), lst is the amountf fertilizer from livestock (tonne N km−2 day−1), hvst is the ratio ofarvested crops. For natural ecosystem, all of fertilizer amount, i.e.,
ert hum, fert amm, and lst equal to zero. The detailed mathemati-al formulas describing all these processes and parameters in detailan be found in Lin et al. (2000, 2001).
.3. River routing model
The aim of river routing model (RRM) is to give directions forateral water and pollutant movement by creating an idealizedetwork of river channels. In this study, Total Runoff Integratingathways (TRIP) (Oki et al., 1999; Oki and Sud, 1998; Ngo-Duc et al.,007), one of RRMs, was employed. It was used to transport nitrogenrom natural and anthropogenic sources through river channels, asell as collect and route nitrogen to the river mouths. Amounts of
otal nitrogen in direct runoff, lateral subsurface flow and percola-ion are estimated as the products of the volume of water and theverage concentration. Transport or retention factors are taken intoccount through routing of water and nitrogen in the river flow viaransmission losses.
.4. Biological N fixation and atmospheric N deposition
Biological N fixation of atmospheric N in natural ecosystems wasstimated by using TNCM. It was assumed to be the sum of sym-iotic and nonsymbiotic fixations, which can be modelled by theunction in Lin et al. (2000). Nitrogen deposition includes dry andet deposition of ammonia gas, nitrate, and nitrogen compounds
rom the atmosphere to soil by rain, snow, and dust. The depositionf ammonium and nitrate was modelled by using the method in Lint al. (2000), where wet deposition was modelled as a linear func-ion of precipitation, and dry deposition of these components, dueo the lack of data, was modelled as average (Hudson et al., 1994).
.5. Denitrification in root zone
Nitrogen discharging from land surface to rivers was assumedo infiltrate through soil where some fraction was removed by den-trification and organic matter accumulation. The surplus nitrogenows to the river and then to the sea, together with precipitation
rial nitrogen cycle model.
surplus. In this study, nitrogen was consumed to denitrified andaccumulated in the soil by a first-order reaction expressed by thefollowing equation (Shindo et al., 2003).
C = C0exp(−kT · tR) (9)
kT = 2(T−20)/10 · k20 (10)
where, C0 indicates the original nitrogen concentration (mg/L), kT
and k20 are coefficients of denitrification and accumulation at T◦ and20◦ (k20 = 3.0), respectively, and tR is residence time in soil (day).
2.6. Nitrate leaching
Many different models are used to the detailed simulation ofaverage nitrate leaching and denitrification process (Brisson et al.,2003; Johnsson et al., 1987; Shaffer et al., 1991). However, suchmodels are too detailed for the 0.1◦ by 0.1◦ resolution and thesemodels require data on environmental conditions (i.e., daily con-dition of root growth, phenology stage, crop yield, leaf area index,etc.) and agricultural management (i.e., irrigation option, drainageoption, precise planting and cultivation date, fertilizer application),which are not available on the spatial scale of our model. In thepresent stage of this study, the TNCM was applied to estimatenitrate leaching from natural ecosystems such as grassland and for-est with fertilizer application rate as zero. Furthermore, it was usedto estimate nitrate leaching from paddy land and farmland withapplication of fertilizer amount. The nitrate leaching is stronglyrelated to soil water content, soil texture, and NO3
− concentration.For modelling the nitrate leaching flux, the below equation wasemployed:
Nleach = Nnit · Rt
�s· 103 (11)
where, Nleach is the flux of nitrate leaching (tonne N km−2 day−1),Nnit is potential nitrogen storage as in nitrate (tonne N km−2) whichwas calculated by TNCM model, Rt is runoff (tonne km−2 day−1)which was calculated by MATSIRO model, and �s is soil water stor-age (mm) which was calculated by MATSIRO model.
3. Database description
3.1. Hydrometeorological database
Air temperature, precipitation, short wave downward radia-tion are from Sakimura (2007) based on AMeDAs database. The
2328 B. He et al. / Ecological Modellin
Ft
sv23tmawoiw
3
MlsTllealw
3
uiticsdidM1t
ig. 2. Land use map of paddy land in Japan (Value of areal ratio ranges from 0.0%o 100%).
oil temperature, soil water content, and runoff are calculated andalidated by MATSIRO model (Sakimura, 2007; Yoshimura et al.,008). A surface meteorological forcing dataset was prepared using0 years (1976–2005) of observed data that include precipitation,emperature, and wind direction and speed from over 1300 auto-
ated meteorological data-acquisition system (AMeDAS) sites andir pressure, vapor pressure, and cloud cover from 155 in situeather stations. These observation sites cover Japan at intervals
f an average of 17–21 km. The meteorological data were basicallynterpolated to 0.1 degree (about 10 km) using the inverse distance
eighting method.
.2. Land use map
The land use map was generated from the database of Japaneseinistry of Land, Infrastructure, Transport and Tourism (MLIT
anduse data). Its original spatial resolution is 100 m grid. In thistudy, the spatial resolution of 0.1 degree has been applied forNCM, LSM and RRM. In each 0.1 degree grid, the area ratio of eachand use was calculated and accumulated from the 100 m grid. Theand use data were reclassified among paddy field, farmland, for-st, wild land, building, road and rail, river and lake, other land. Therea of each land use type in each 0.1 × 0.1 degree cell was calcu-ated. Fig. 2 shows an example of paddy land distribution for the
hole Japan.
.3. Nitrogen load from fertilizer use
The spatial and temporal distributions of on-ground N fertilizerse from various crops and agricultural practices were quantified
n this study. The assessment of on-ground N fertilizer due to mul-iple land use activities can be complex and the traditional methods to only use the national census data without considering therops’ spatial distribution, fertilization and harvest patterns. In thistudy, the bi-weekly fertilizer application rates and harvest calen-ar associated to 63 crop types selected from 97 total crop types
n 47 prefectures in Japan were calculated by using national censusatabase and each crop’s agricultural manuals (Fertilizer and Farmachinery Division, 1986, 1987, 1992, 2000; MAFF, 1984, 1986,
991, 1998). Then, the land cover map showing the spatial distribu-ion of areal ratio of each crop in each grid with a spatial resolution
g 220 (2009) 2325–2334
of 0.1 × 0.1 degree was generated for the whole Japan. Finally, thespatial distribution map of fertilizer utilization with a temporal res-olution of two weeks in Japan was generated. The planted area ofeach crop in each village or city was collected for 97 crops from thedatabase of Japan Ministry of Agriculture, Forestry and Fisheries(MAFF, 1984, 1986, 1991, 1994, 1998). The crop distribution mapwas then generated from the above collected database of crop area.Then, the fertilizer amount for different land cover was calculatedin each grid.
3.4. Nitrogen load from livestock
The nitrogen load from livestock has two types. One is the live-stock’s excreta in the pasture land. Another is the drainage fromlivestock’s cattlesheds. The number of grazing and non-grazinglivestock was calculated from national livestock census databaseand grazing rate in each prefecture. The nitrate leaching from thepastureland was calculated by the TNCM. The nitrogen load fromlivestock’s cattlesheds was calculated by using the animal numbersand pollutant emission load per animal.
3.5. Nitrogen load from domestic and industrial water use
Point sources of N are primarily associated with human exc-reta and industrial water use (wastewater drainage). As for thecalculation of nitrogen load from domestic water use, the popu-lation provided with sewage plants was calculated by using thedatabase of sewage plants distribution, diffusion rate of publicsewerage, and population distribution. The distribution of pop-ulation provided by wastewater service was calculated by usingthe database of population without sewage plants and populationprovided with wastewater service. The pollutant emission basicunits of wastewater service, small-scale onsite treatment, and otherdomestic sewage were applied to calculate the N load from domes-tic water use.
As for the calculation of nitrogen load from industrial wateruse, nitrogen load in each prefecture was calculated by using thedatabase of the production of each industrial classification, pol-lutant emission basic unit of each industrial classification. Thedistribution of nitrogen load from industry can then be calculatedby land use data and nitrogen load in each prefecture. Furthermore,the distribution of sewage diffusion rate was calculated from totalpopulation distribution and population without sewage plants. Thenitrogen load from industrial water use was calculated by distribu-tion of nitrogen load from industry and the distribution of sewagediffusion rate.
4. Integrated simulation with MATSIRO, TNCM and TRIP
4.1. Design of the integrated simulation
The model operates on a daily time step and at a spatial resolu-tion of 0.1 × 0.1 degree over Japan. After the input parameters areread from files, the three-step modelling procedure is applied. First,water discharge, nitrogen balance, and nitrate leaching are calcu-lated for each grid (0.1 × 0.1 degree) by the MATSIRO and TNCM.Then the outputs from each grid (e.g., lateral water flows, nitrateflow) are summed with point pollution load (e.g., from industrialsource, sewage plant). Finally, the routing procedure TIRP is appliedto transport point and nonpoint pollution along rivers, taking trans-
mission losses into account. Among these, the hydrological moduleis fundamental for all the modelling systems in this study. It wastested and validated in Sakamoto et al. (2004) and Sakimura (2007).It reproduces well the observed seasonal cycles of the energy andwater balance.
B. He et al. / Ecological Modellin
fab
Fm
Fig. 3. All basins in Japan and selected 59 basins for model validation.
Nitrogen modelling is also a very complicated task due to theact that nitrogen has many chemical forms and compounds, whichre very mobile and dynamic both in space and time. In addition,iogeochemical modelling at the global or national scale with large
ig. 4. Variation in the national nitrogen storage in soil and vegetation during theodel spinup calculations.
g 220 (2009) 2325–2334 2329
grid cells, where usually only the vertical flows are considered andpractically no real validation is possible, is difficult to be validated.In this study, the lateral flows are included by using the routing pro-cedure of TRIP in the modelling system so that the chemical fluxes atthe national scale can be validated using the data of measurementsat the river outlet. For nitrogen cycle and routing model’s validation,we collected the available observed total nitrogen (TN) concen-tration in Japanese streams from MLIT (MLIT water informationsystem). Fig. 3 shows all basins in Japan and the selected 59 basinsfor model validation. All the observation dataset of these selected59 basins have long-term records (however, record period are dif-ferent) which are available for our model validation.
4.2. Calculation of the steady state model
Before commencing a long-term nitrogen cycle simulation, it isusually necessary to allow the land surface to adjust to a mutualequilibrium state. Its objective is to bring the model close to a sta-ble equilibrium so that negligible climate drift is experienced in thecontrol run which follows. Some groups have nevertheless appar-ently used spinup techniques successfully to initialize the oceanstate for long climate studies (Manabe et al., 1991, 1992; Stoufferet al., 1994; Thornton and Rosenbloom, 2005; Lin et al., 2000). Forthis study, spinup procedure with a time step of one day was used.In this procedure, the national nitrogen cycle model restarts every1 year with its output as its new initial conditions. The experimentbegins with a 100-year spinup of the model, forced by the repeated1976 annual cycle and climatological data. Variations in national
nitrogen storage in soil and vegetation derived from the round cal-culations of nitrogen cycle model are shown in Fig. 4. It can be seenfrom this figure that the state variable for nitrogen storage in veg-etation and soil reached equilibrium after 40-year calculation. Itsuggests that the steady state of the nitrogen cycle model requireFig. 5. Seasonal comparison of observed and modelled TN concentration (TNcon) inselected 59 sites (year 1995).
2 dellin
aw
4
cmwtrtmAbfm5ucaooitaicI
330 B. He et al. / Ecological Mo
successive calculation of over 40 years. After the spinup time thehole nitrogen cycle model was run for at least 3 years.
.3. TN concentration comparison in 59 river basins
We further tested the model by comparing simulated TN con-entration with observed ones. Because TN concentrations wereeasured monthly, monthly average simulated TN concentrationsere first calculated from daily predictions and then compared
o TN concentrations measured by MLIT. Based on the monthlyesults, the seasonally comparison can be conducted by averaginghe monthly value. Fig. 5 shows the comparison of observed and
odelled seasonal (Spring: March, April, May; Summer: June, July,ugust; Autumn: September, October, November; Winter: Decem-er, January, February) average value of TN concentration in 1995or selected river’s outlets. From the figure, we can find that the
odel captured most of the nitrogen concentration variations for9 rivers even though in some rivers the discrepancy between sim-lated TN concentration and observations exits. It also shows thelear variation of spatial distribution of TN concentration. Over-ll, the simulated TN concentrations were reasonably close to theirbserved ones. Fig. 6 shows the scatter plot of the comparison ofbserved and modelled seasonal average value of TN concentrationn 1995 for selected 59 sites. The squared correlation coefficient of
he observed and predicted TN concentrations were 0.61, 0.71, 0.30,nd 0.77 for spring, summer, autumn, and winter, respectively. Fromt, we can find that the model simulation underpredicted the total Noncentrations in Spring, Autumn and some river basins in Summer.n winter, the model overpredicted the total N concentrations.Fig. 6. Scatter plots of the comparison of observed and
g 220 (2009) 2325–2334
In addition to seasonal comparison, we also compared theobserved and model simulated time series data in some riverswhich have long-term record data. Fig. 7 shows the temporal varia-tion of nitrogen concentration in selected rivers from 1993 to 1996.From this figure, the seasonal variation of nitrogen concentrationcan be found from both modelled and observed TN concentrationdata. In most rivers, the model captured the variability of nitro-gen concentration. Fig. 8 shows the scatter plot of the comparisonof observed and modelled TN concentration in selected sites ofJapan from 1993 to 1996. The squared correlation coefficient of theobserved and predicted TN concentrations was 0.24 for all selectedrivers.
4.4. Nitrogen load evaluation
Based on the database collected, a detailed modelling study wasconducted to estimate both the point and nonpoint source nitro-gen pollution in Japan. The model gave the spatial distributionsof total nitrogen loadings (TNL) from point and nonpoint sources.Fig. 9 demonstrates the spatial distribution of estimated annualamount of TNL for each river basin in 1995 as an example. Theanalysis of the TNL in Japanese river basins shows that the TNLin urban river basins, i.e., Tsurumi River Basin, Sho River Basin,Yodo River basin, and Kumode River Basin, is large. In west part
of Japan, e.g., Kyushu, Shikoku, and Chukoku, the TNL is relativelysmaller than other areas. In Kurobe River basin, the nitrogen load isrelatively small since its high area ratio of forest area and low pop-ulation. However, in Tsurumi River basin, the nitrogen load is quitelarge because of its high population. Moreover, its nitrogen loadmodelled TN concentration in selected 59 sites.
B. He et al. / Ecological Modelling 220 (2009) 2325–2334 2331
Fo
fwpagfsIlbraspTeb
sTapbt
four river basins, the NLs from nonpoint sources are all decreasing.Especially during 1980s, the speed of its decrease is very rapid. Formost of the river basins in Japan, the ratio of NL from point sourceto total NL is less than 0.4, except for Kiso River basin.
ig. 7. Temporal distribution of monthly mean TN concentration in selected 5 sitesf Japan (year 1993–1996).
rom agriculture is almost zero and from 1994 the nitrogen loadas almost transferred from domestic direct drainage to sewagelant drainage because of its urban development. We also find that,s for the Ishikari River basin, the basin area is larger but nitro-en load is smaller than that of Tone River basin. It is because thearmland and grassland area of animal grazing and number of live-tock in the Tone River basin is much greater than those in theshikari River basin. Therefore, the nitrogen load generated fromivestock in Tone River basin is the largest in Japan. In Kiso Riverasin, the nitrogen load from sewage is much smaller than otheriver basins since its low population. In Yodo River basin, its basinrea is smallest but its nitrogen load is just next to Tone River basinince its high population. Fig. 9 shows the capability of the pro-osed model system for assessing the potential risk or impacts ofNL to the sea through rivers from each basin. It can be useful asffective tool for environmental evaluation and planning of riverasins.
For further evaluation of TNL in each river basin of Japan, Fig. 10hows the ratios of modelled nonpoint source TNL and point sourceNL to total TNL in 59 river basins in Japan for 1976, 1985, 1995,
nd 2005. We can find from it that in 1970s the TNL from non-oint sources became the dominating sources in most of riverasins in Japan due to the excessive fertilizer use. It is also clearhat in all basins the nonpoint source nitrogen load is decreas-Fig. 8. Scatter plot of the comparison of observed and modelled TN concentrationin selected 5 sites of Japan (year 1993–1996).
ing. In addition, from 1976 to 1985, most of the nitrogen load fromnonpoint sources decreased a lot since Japanese government triedmany ways to control the nonpoint sources pollution. From 1995to 2005, the nitrogen load from nonpoint sources still decreasedbut the decreasing speed became less than that from 1985 to1995.
Fig. 11 shows the ratios of modelled annual total nitrogen loadin four large river basins in Japan. As for the Ishikari River basin,the ratio of NL from point source is the largest from 1976 to 2005.As for the Yodo River basin, the ratio of NL from nonpoint source isthe largest from 1976 to 2005 since its high population. For all the
Fig. 9. Spatial distribution of annual mean TN load in Japanese river basins (year1995).
2332 B. He et al. / Ecological Modellin
Fi
4
stsooFwf
Fb
ig. 10. Ratios of point source (PS) and nonpoint source (NPS) to total nitrogen loadn 59 river basins.
.5. Uncertainty analysis
From the above model validation results, we may see that theretill remains residual discrepancy between simulated TN concen-ration and the observations over some places. Some uncertaintieshould be one of the reasons. As for nitrogen in agricultural land,
nly rice was considered to be planted in the paddy field. More-ver, the rotational planting and lying fallow was not considered.urthermore, fruit tree was considered as mature or full-grown treehich can produce the most N amount. In the harvest period, Nrom the fruit and other vegetation were considered to be removed
ig. 11. Ratios of nitrogen load (NL) from point source (ps) and nonpoint source (nps) inasin, Kiso R. basin, and Yodo R. basin are 14,367 km2, 16,508 km2, 9232 km2, 8295 km2, r
g 220 (2009) 2325–2334
from their land. The calculated N fertilizer was finally comparedwith national census data which has some uncertainties in them-selves. Nitrogen in livestock was calculated from the number ofanimals and excretion rate per head (Bouwman et al., 1997; Shindoet al., 2003). All the animals were considered as full-grown ani-mals.
As for N deposition data, the simple empirical relationshipbetween N deposition and precipitation was employed. However,long-term range transport of N especially NOx is important and itmust be necessary to adopt a transport model for more precise esti-mation (Shindo et al., 2003). In respect to the denitrification andorganic matter accumulation in the soil, a simple reaction modelconsidering temperature and resident time was applied. Nitrogenremoval due to the instream N retention is affected by complexconditions so that this method also leads to a source of uncer-tainty. The observation data of total N concentration was collectedin the outlet of each river. However, estimates of the absolute val-ues of N concentration in river water had uncertainties becausesufficient data on the ratio of total N export to nitrate export wasnot available and this ratio value would vary spatially according totemperature and other factors (Shindo et al., 2003). The transporta-tion of N was only considered from land surface to rivers and thento sea. The transportation of N in groundwater and the interactionbetween river water and groundwater was not considered yet. Thefurther considering them would improved the better estimationresults.
For conducting the model uncertainty analysis, four scenarioshave been considered in which the nitrogen loads from agriculture,livestock, industrial plant, and domestic sewage were supposed tobe increased by 5%, respectively. The average measured river TNconcentration was shown in the bottom part of Fig. 12. The ratios of
the simulated river TN concentration in four scenarios comparingwith the original TN concentration simulated by models withoutscenarios were demonstrated in the upper part of Fig. 12. We findthat the contribution of NL from livestock to river TN concentrationis the largest for most of river basins in Japan, the agriculture is thefour large river basins of Japan (Year 1976–2005). Areas of Ishikari R. basin, Tone R.espectively.
B. He et al. / Ecological Modellin
Flf
sitr
gTiatcaimeiId
5
eNutcautrrooatticmo
ig. 12. Annual measured total nitrogen concentration (TNcon) and ratios of simu-ated TNcon under four scenarios in which the nitrogen load was increased by 5%or agriculture, livestock, industrial plant and sewage, respectively.
mallest. It means that the river TN concentration will be sensitivelympacted by the NL from livestock source. Accordingly, the uncer-ainty from livestock will have biggest impact on the estimation ofiver TN concentration.
Despite the uncertainties described above, we can still get aood insight in the nitrogen load simulation at a national scale.he constructed nitrogen loads from agricultural land, livestock,ndustrial plant, and domestic water use are good databases for
national scale nitrogen load evaluation. It will be useful to usehe proposed model framework to test future scenarios of climatehanging, social and economic changes in food supply in order tossess the prospects for long-term changes at a national scale. Its reasonable to conclude that the proposed nitrogen cycle model
ay be very helpful in the management of aquatic ecosystems withutrophication problems, especially by estimating nutrient load-ngs from nonpoint sources, which is always difficult to determine.t is possible to estimate TNL using this model system for futureevelopment in the drainage area.
. Conclusion
This paper demonstrated an integrated biogeochemical mod-lling of nitrogen load and its export to Japanese streams. TheL from different sources such as crop, livestock, industrial plant,rban and rural resident were calculated by using datasets of fer-ilizer utilization, population distribution, land use map and socialensus. The nitrate leaching from soil layers in farmland, grasslandnd natural conditions was calculated by using TNCM. The TRIP wassed to transport nitrogen from natural and anthropogenic sourceshrough river channels, as well as collect and route nitrogen to theiver mouths. As the model validation results, there still remainsesidual discrepancy between simulated TN concentration and thebservations over some places. Some uncertainties should be onef the reasons. However, this study of understanding the spatialnd temporal distribution of nitrogen load and its export may con-ribute to the process of developing the appropriate policy, research,
echnology, and education needed to reverse the trend of increas-ng loads of nitrogen to the sea. Due to the direct incorporation oflimate and land use data, there is a possibility to use the proposedodel system for analysis of climate and land use change impactsn hydrology and water quality.
g 220 (2009) 2325–2334 2333
Acknowledgements
This study was supported by the Japan Society for the Promotionof Science (JSPS) Postdoctoral Fellowship (P) 07117, the JSPS Grant-in-Aid for Scientific Research (S) 19106008, and the Data Integrationand Analysis System (DIAS) of Earth Observation Data Integration &Fusion Research Initiative (EDITORIA), Japan. The authors are grate-ful for their supports.
References
Baker, A., 2003. Land use and water quality. Hydrol. Proc. 17, 2499–2501.Beven, K.J., Kirkby, M.J., 1979. A physically based, variable contributing area model
of basin hydrology. Hydrol. Sci. Bull. 24, 43–69.Brady, N.C., 1998. Nitrogen and sulfur economy of soils. In: Brady, N.C., Weil, R.R.
(Eds.), The Nature and Properties of Soils, 12th ed. Elsevier, New York, pp.492–522.
Brisson, N., Gary, C., Justes, E., Mary, B., Roche, R., Ripoche, D., Zimmer, D., Sierra,J., Bertuzzi, P., Burger, P., Bussière, F., Cabidoche, Y.M., Cellier, P., Debaeke, P.,Gaudillère, J.P., Maraux, F., Seguin, F.B., Sinoquet, H., 2003. An overview of thecrop model STICS. Eur. J. Agron. 18, 309–332.
Bouwman, A.F., Lee, D.S., Asman, W.A.H., Dentener, F.J., Van Der Hoek, K.W., Oliver,J.G.J., 1997. A global high-resolution emission inventory for ammonia. Glob. Bio-geochem. Cycl. 11, 561–587.
Davis, J.R., Koop, K., 2006. Eutrophication in Australian rivers, reservoirs andestuaries—a southern hemisphere perspective on the science and its implica-tions. Hydrobiologia 559, 23–76.
Dickinson, R.E., Shaikh, M., Bryant, R., Graumlich, L., 1998. Interactive canopies for aclimate model. J. Clim. 11, 2823–2836.
Ding, C., Ohmori, H., Takamura, H., 2006. Relationship between water quality andland use in the Mamagawa river, Tokyo Metropolitan Area. J. Jpn. Assoc. Hydrol.Sci. 36 (4), 219–233.
Fertilizer and Farm Machinery Division, 1986. Pocket Fertilizer Handbook. NorinToukei Kyoukai, Tokyo, Japan, p. 47 (in Japanese).
Fertilizer and Farm Machinery Division, 1987. Pocket Fertilizer Handbook. NorinToukei Kyoukai, Tokyo, Japan, p. 47 (in Japanese).
Fertilizer and Farm Machinery Division, 1992. Pocket Fertilizer Handbook. NorinToukei Kyoukai, Tokyo, Japan, p. 47 (in Japanese).
Fertilizer and Farm Machinery Division, 2000. Pocket Fertilizer Handbook. NorinToukei Kyoukai, Tokyo, Japan, p. 48 (in Japanese).
Galloway, J.N., Cowling, E.B., 2002. Nitrogen and the World. Ambio 31, 64–71.Galloway, J.N., Schlesinger, W.H., LEVYL, L., Micka, II.A., INooi, J.L., 1995. Nitrogen fixa-
tion: anthropogenic enhancement—environmental response. Glob. Biogeochem.Cycl. 9, 235–252.
Galloway, J.N., 1998. The global nitrogen cycle: changes and consequences. Environ.Pollut. 102, 15–26.
Galloway, J.N., 2000. Nitrogen mobilization in Asia. Nutr. Cycl. Agroecosyst. 57,1–12.
Hantschel, R.E., Beese, F., 1997. Site-oriented ecosystem management: preconditionto reducing the contamination of waters and the atmosphere. In: Rosen, D., Tel-Or, E., Hadar, Y., Chen, Y. (Eds.), Modern Agriculture and the Environment. KluwerAcademic Publishers, Dordrecht, pp. 135–145.
Hirabayashi, Y., Kanae, S., Oki, T., 2005. A 100-year (1901–2000) global retrospectiveestimation of terrestrial water cycle. J. Geophys. Res. 110, D19101.
Howarth, R.W., Sharpley, A., Walker, D., 2002. Sources of nutrient pollution to coastalwaters in the United States: implications for achieving coastal water qualitygoals. Estuaries 25 (4B), 656–676.
Hudson, D., Hite, D., Haab, T., 2005. Public perception of agricultural pollution andGulf of Mexico hypoxia. Coast. Manage. 33 (1), 25–36.
Hudson, R.J.M., Gherini, S.A., Goldstein, R.A., 1994. Modelling the global carbon cycle:nitrogen fertilization of the terrestrial biosphere and the ‘missing’ CO2 sink. Glob.Biogeochem. Cycl. 8 (3), 307–333.
Johnsson, H., Bergström, L., Jansson, P.E., Paustian, K., 1987. Simulated nitrogendynamics and losses in a layered agricultural soil. Agric. Ecosyst. Environ. 18(4), 333–356.
Kimura, S.D., Hatano, R., 2007. An eco-balance approach to the evaluation of historicalchanges in nitrogen loads at a regional scale. Agric. Syst. 94, 165–176.
Kimura SD, 2005. Creation of an eco-balance model to assess environmental riskscaused by nitrogen load in a basin-agroecosystem. PhD thesis. Hokkaido Univer-sity Graduate School of Agriculture, Japan.
Kimura, S.D., Liang, L., Hatano, R., 2004. Influence of N flow change on environmentbetween 1912 and 2002: a case study of one city in Hokkaido. Jpn. Nutr. Cycl.Agroecosyst. 70, 271–282.
Lin, B.L., Sakoda, A., Shibasaki, R., Goto, N., Suzuki, M., 2000. Modelling a globalbiogeochemical nitrogen cycle model in terrestrial ecosystems. Ecol. Model. 135-1, 89–110.
Lin, B.L., Sakoda, A., Shibasaki, R., Suzuki, M., 2001. A modelling approach to
global nitrate leaching caused by anthropogenic fertilisation. Water Res. 35 (8),1961–1968.MAFF Crop Production Division, 1984. Material for Rice Production. MAFF, Tokyo,Japan, pp. 246–248 (in Japanese).
MAFF Crop Production Division, 1986. Material for Rice Production. MAFF, Tokyo,Japan, pp. 236–238 (in Japanese).
2 dellin
M
M
M
M
M
M
MMMN
N
O
O
O
R
S
Model. 189, 25–48.Yoshimura, K., Sakimura, T., Oki, T., Kanae, S., Seto, S., 2008. Toward flood risk pre-
334 B. He et al. / Ecological Mo
AFF Crop Production Division, 1986. Material for Rice Production. MAFF, Tokyo,Japan, pp. 341–345 (in Japanese).
AFF Research Council, 1994. Model Examples of Agricultural Technique Systemsfor Environmental Conservation. MAFF, Tokyo, Japan, pp. 236–238 (in Japanese),http://www.tdb.maff.go.jp/toukei/toukei.
AFF Crop Production Division, 1998. Material for Rice Production. MAFF, Tokyo,Japan, pp. 205–206 (in Japanese).
anabe, S., Stouffer, R.J., Spelman, M.J., Bryan, K., 1991. Transient responses of acoupled ocean-atmosphere model to gradual changes of atmospheric CO2 PartI: annual mean response. J. Clim. 4, 785–818.
anabe, S., Spelman, M.J., Stouffer, R.J., 1992. Transient responses of a coupled ocean-atmosphere model to gradual changes of atmospheric CO2 Part II: seasonalresponse. J. Clim. 5, 105–126.
ishima, S., 2001. Recent trend of nitrogen flow associated with agricultural pro-duction in Japan. Soil Sci. Plant. Nutr. 47 (1), 157–166.
inistry of Environment of Japan, http://www.env.go.jp/en/.LIT landuse data, http://nlftp.mlit.go.jp/ksj2/index.html.LIT water information system, http://www1.river.go.jp/.akasone, H., Kuroda, H., Kato, T., Tabuchi, T., 2003. Nitrogen removal from water
containing high nitrate nitrogen in a paddy field. Water Sci. Technol. 48 (10),209–216.
go-Duc, T., Oki, T., Kanae, S., 2007. A variable stream flow velocity method for globalriver routing model: model description and preliminary results. Hydrol. EarthSys. Sci. Disc. 4, 4389–4414.
enema, O., Boers, P.C.M., Willems, W.J., 1998. Leaching of nitrate from agricultureto groundwater: the effect of policies and measures in the Netherlands. Environ.Pollut. 102, 471–478.
ki, T., Nishimura, T., Dirmeyer, P., 1999. Assessment of annual runoff from landsurface models using Total Runoff Integrating Pathways (TRIP). J. Meteorol. Soc.Jpn. 77 (1B), 235–255.
ki, T., Sud, Y.C., 1998. Design of total runoff integrating pathways (TRIP)—a global
river channel network. Earth Interact., 2.ijtema, P.E., Kroes, J.G., 1991. Some results of nitrogen simulations with the modelANIMO. Fert. Res. 27, 189–198.
akamoto, T.T., Emori, S., Nishimura, T., Hasumi, H., Suzuki, T., Komoto, M., 2004. Far-reaching effects of the Hawaiian Islands in the CCSR/NIES/FRCGC high-resolutionclimate model. Geophys. Res. Lett. 31, L17212.
g 220 (2009) 2325–2334
Sakimura T., 2007. Development of an approach to predict flood risks based onwater and energy balance simulation over Japan. Master thesis. The University ofTokyo.
Schepers, J.S., Varvel, G.E., Watts, D.G., 1995. Nitrogen and water management strate-gies to reduce nitrate leaching under irrigated maize. J. Contam. Hydrol. 20,227–239.
Sellers, P.J., Randall, D.A., Collatz, G.J., Berry, J.A., Field, C.B., Dazlich, D.A., Zhang, C.,Collelo, G.D., Bounoua, L., 1996. A revised land surface parameterization (SiB2)for atmospheric GCMs Part I: model formulation. J. Clim. 9, 676–705.
Shaffer, M.J., Halvorson, A.D., Pierce, F.J., 1991. Nitrate leaching and economic analy-sis package (NLEAP): model description and application. In: Follet, R.F., Keeney,D., Cruse, R.M. (Eds.), Managing Nitrogen for Groundwater Quality and FarmProfitability, pp. 285–322 (Soil Sci. Soc. of Am., Madison, Wis.).
Shindo, J., Okamoto, K., Kawashima, H., 2003. A model-based estimation of nitrogenflow in the food production—supply system and its environmental effects in EastAsia. Ecol. Model. 169, 197–212.
Stouffer, R.J., Manabe, S., Vinnikov, K.Y., 1994. Model assessment of the role of naturalvariability in recent global warming. Nature 367, 634–636.
Takata, K., 2000. Sensitivity study on snowmelt discharge of Lena using a land sur-face model, “MATSIRO”. In: Activity Report of GAME-Siberia, 1999. Nagoya Univ.,Nagoya, Japan, pp. 73–76 (GAME Publication, No. 21).
Takata, K., 2001. Estimation of summer heat and water fluxes at Tiksi using a one-dimensional land surface model. In: Activity Report of GAME-Siberia, 2000.Nagoya Univ., Nagoya, Japan, pp. 207–209 (GAME Publication, No. 26).
Takata, K., Emori, S., Watanabe, T., 2003. Development of the minimal advancedtreatments of surface interaction and runoff. Glob. Planet. Change 38, 209–222.
Thornton, P.E., Rosenbloom, N.A., 2005. Ecosystem model spin-up: estimating steadystate conditions in a coupled terrestrial carbon and nitrogen cycle model. Ecol.
diction: a statistical approach using a 29-year river discharge simulation overJapan. Hydrol. Res. Lett. 2, 22–26.
Yoshimura, C., Omura, T., Furumai, H., Tockner, K., 2005. Present status of rivers andstreams in Japan. River Res. Appl. 21, 93–112.
