Application of the Soil and Water Assessment Tool (SWAT)to predict the impact of alternative management practices on water quality and quantity
Antje Ullrich *,Martin Volk
Helmholtz Centre for Environmental Research –UFZ,Department of Computational Landscape Ecology,Permoserstr.15,D-04318Leipzig,Germany
1.Introduction
Due to insufficient water quality of European streams environmental programs such as the European Water Framework Directive (WFD)were implemented to achieve good ecological and chemical conditions of water quality of groundwater and surface water bodies (EC,2000;Rekolainen et al.,2003).Main nutrient input comes from nonpoint source pollution,mainly forced by intensive agricultural activities (Behrendt et al.,1999).Therefore,alternative land management practices are increasingly used to reduce nonpoint source pollution.Reduction of soil tillage intensity positively affects numerous soil properties,such as aggregate stability,macroporosity,and saturated hydraulic con-ductivity and consequently increases infiltration rates and reduces surface runoff,nutrient loss,and soil erosion (Jones et al.,1969;
Pitka
¨nen and Nuutinen,1998;Schmidt et al.,2001;Kirsch et al.,2002;Pandey et al.,2005;Tripathi et al.,2005).Alternative land management practices may include reduced tillage such as
conservation tillage (e.g.without deep ploughing,field preparation just before planting)or no-tillage (direct drilling)(Sullivan,2003;LfULG,2006).In Germany the implementation of alternative tillage syste
ms is increasingly supported by agro-environmental pro-grams.In the German State of Saxony,for instance,conservation tillage and mulch seeding on arable land has increased from <1%to about 27%during 1994–2004with support from the Saxonian Program for Environmental Agriculture (LfL,2006).
A number of field studies have illustrated the positive effects of conservation tillage and no-tillage practices on water and material fluxes at the field local level (e.g.Sloot et al.,1994;King et al.,1996;Schmidt et al.,2001),but this effect needs to be assessed on the watershed level to guide river basin management programs as WFD claimed (Kirsch et al.,2002;Chaplot et al.,2004;Pandey et al.,2005;Behera and Panda,2006;Bracmort et al.,2006).Therefore,watershed models are useful tools and have been used for decades to evaluate nonpoint source pollution and the short-and long-term impacts of alternative management practices.
In order to fulfill the objectives of the WFD,we have chosen the semi-distributed river basin model,Soil and Water Assessment Tool (SWAT)2005(Neitsch et al.,2002;Arnold and Fohrer,2005)to examine the impact of alternative management practices on water
Agricultural Water Management 96(2009)1207–1217
A R T I C L E I N F O Article history:
Received 3September 2008Accepted 6March 2009Keywords:SWAT
Tillage management practice Conservation tillage Water balance Nutrient Modelling
A B S T R A C T
Alternative land management practices such as conservation or no-tillage,contour farming,terraces,and buffer strips are increasingly used to reduce nonpoint source and water pollution resulting from agricultural activities.Models are useful tools to investigate effects of such management practice alternatives on the watershed level.However,there is a lack of knowledge about the sensitivity of such models to parameters used to represent these conservation practices.Knowledge about the sensitivity to these parameters would help models better simulate the effects of land management.Hence,this paper presents in the first step a sensitivity analysis for conservation management parameters (specifically tillage depth,mechanical soil mixing efficiency,biological soil mixing efficiency,curve number,Manning’s roughness coefficient for overland flow,USLE support practice factor,and filter strip width)in the Soil and Water Assessment Tool (SWAT).With this analysis we aimed to improve model parameterisation and calibration efficiency.In
contrast to less sensitive parameters such as tillage depth and mixing efficiency we parameterised sensitive parameters such as curve number values in detail.In the second step the analysis consisted of varying management practices (conventional tillage,conservation tillage,and no-tillage)for different crops (spring barley,winter barley,and sugar beet)and varying operation dates.Results showed that the model is very sensitive to applied crop rotations and in some cases even to small variations of management practices.But the different settings do not have the same sensitivity.Duration of vegetation period and soil cover over time was most sensitive followed by soil cover characteristics of applied crops.
ß2009Elsevier B.V.All rights reserved.
*Corresponding author.Tel.:+493454722602;fax:+493412351939.E-mail address:antje.ullrich@yahoo (A.Ullrich).Contents lists available at ScienceDirect
Agricultural Water Management
j o u r n a l h o m e p a g e :w w w.e l s e v i e r.c o m /l o c a t e /a g w a t
0378-3774/$–see front matter ß2009Elsevier B.V.All rights reserved.doi:10.1016/j.agwat.2009.03.010
quantity and quality.Gassman et al.(2007)point out that‘‘a key strength of SWAT is aflexible framework that allows the simulation of a wide variety of conservation practices and other BMP,such as fertiliser and manure application rate and timing, cover crops(perennial grasses),filter strips,conservation tillage, irrigation management,flood prevention structures,grassed waterways,and wetlands.The majority of conservation practices can be simulated in SWAT with straightforward parameter changes’’.Many studies have used SWAT(Saleh et al.,2000; Shanti et al.,2001;Vache et al.,2002;Shanti et al.,2003;Chaplot et al.,2004;Pandey et al.,2005;Tripathi et al.,2005;Arabi et al., 2006;Behera and Panda,2006;Rode et al.,2008;Volk et al.,2009) and EPIC(Sloot et al.,1994;King et al.,1996)to evaluate the effects of land use scenarios and management practices.Several studies have analysed the long-term effects of structural Best Management Practices(BMP)on water Kirsch et al.,2002;Chaplot et al.,2004;Tripathi et al.,2005;Pandey et al.,2005or Behera and Panda,2006;Bracmort et al.,2006).Arabi et al.(2007)investigated the impact of modelling uncertainty on evaluation of management practices using a Monte Carlo-based probabilistic approach.
The SWAT model was developed for application to large complex watersheds over long periods of time(Neitsch et al., 2002).Working on the watershed-scale means that required input data are often a
ggregated in terms of temporal daily climate data).In contrast,land management parameters(tillage, fertilisation,crop rotation,etc.)can be included in high resolution and detail,due to its modular structure and its historical development based on the EPIC(Erosion Productivity Impact Calculator)model(Benson et al.,1988;Neitsch et al.,2002;Arnold and Fohrer,2005;Gassman et al.,2007).Furthermore,modelling evaluations of conservation management effects at the watershed-scale are limited by the lack of management operation data.Thus, knowledge is needed of the sensitivity of such models to conservation management parameters and practice to improve the efficiency of model parameterisation and the quality of model calibration.Furthermore,potential simulation uncertainties based on ranges of realistic parameter values and on influences of scale need to be understood because simulated effects often drive financial and political decisions(Onatski and Williams,2003).
water pollutionAs a result,the main objective of this study is to analyse the sensitivity of the SWAT model to selected conservation manage-ment parameters and practices to improve model parameterisa-tion and calibration.To the best of our knowledge,a sensitivity analysis of the model to conservation management parameters and practices has not yet been conducted.But in our opinion this is essential for a more efficient use of models for the implementation of land management practices as tools to improve water quantity and quality.
We used a so-called semi-virtual watershed for which we combined topography and climate information of an existing watershed(Parthe watershed,Saxony,Germany)but with homo-geneous land use and soil information.Homogeneous land use and soil information were utilised because the resulting data sets are concise and manageable and calculation time is reduced. Recommendations are given for the parameterisation of tillage operations under certain conditions.
2.Materials and methods
2.1.Model description
The SWAT model is considered as one of the most suitable models for predicting long-term impacts of land management measures on water,sediment,and agricultural chemical yield (nutrient loss)in large complex watersheds with varying soils,land use,and management conditions(Arnold and Fohrer,2005;Behera and Panda,2006;Gassman et al.,2007).The model has been gained international acceptance as a robust interdisciplinary watershed modelling tool(Gassman et al.,2007).SWAT is a physically based, conceptual,continuous-time river basin model with spatial distributed parameters operating on a daily time step.It is not designed to simulate detailed,single-eventflood routing(Neitsch et al.,2002).The relationship between input and output variables is describ
ed by regression equations.The SWAT model integrates all relevant eco-hydrological processes including waterflow, nutrient transport and turn-over,vegetation growth,and land use and water management at the subbasin scale.Consequently, the watershed is subdivided into subbasins based on the number of tributaries.Size and number of subbasins is variable,depending on stream network and size of the entire watershed.Subbasins are further disaggregated into classes of Hydrological Response Units (HRU),whereby each unique combination of the underlying geographical maps(soils,land use,etc.)forms one class.HRU are the spatial unit where the verticalflows of water and nutrients are calculated,which are then aggregated and summed for each subbasin.Water and material from HRU in sub-watersheds are routed to the sub-watershed outlet.The HRU in SWAT are spatially implicit,their exact position in the landscape is unknown,and it might be that the same HRU covers different locations in a subbasin(Neitsch et al.,2002;Di Luzio et al.,2005).The water balance for each HRU is represented by the four storages snow,soil profile,shallow aquifer and deep aquifer.The soil profile can be subdivided in up to ten soil layers.Soil water processes include evaporation,surface runoff,infiltration,plant uptake,lateralflow and percolation to lower layers(Arnold and Allen,1996;Neitsch et al.,2002).The surface runoff from daily rainfall is estimated with a modification of the SCS curve number method from United States Department of Agriculture-Soil Conservation Service(USDA SCS) (Arnold and Allen,1996;Neitsch et al.,2002).
Nitrogen movement and transformation are simulated as a function of the nitrogen cycle(Neitsch et al.,2002;Jha et al.,2004). The SWAT model monitorsfive different pools of nitrogen in the soils:two inorganic(ammonium(NH4+)and nitrate(NO3À))and three organic(fresh organic nitrogen(associated with crop residue and microbial biomass)and active and stable organic nitrogen (associated with the soil humus)).Nitrogen is added to the soil by fertiliser,manure or residue application,fixation by bacteria,and rain(Neitsch et al.,2002).Nitrogen losses occur by plant uptake, surface runoff in the solution and the eroded sediment(Neitsch et al.,2002;Jha et al.,2004).
Background for the crop growth and the management practices is the EPIC crop growth model,which is a comprehensivefield scale model.EPIC was originally developed to simulate the impact of erosion on crop productivity and has now evolved into a comprehensive agricultural management,field scale,nonpoint source loading model(Benson et al.,1988;King et al.,1996; Neitsch et al.,2002).The management practices are defined by specific management he beginning and end of growing season,timing of tillage operations as well as timing and amount of fertiliser,pesticide,and irrigation application).These management operations take place in every HRU.The operations in turn are defined by specific management tillage depth,biological soil mixing efficiency,etc.)(Neitsch et al., 2002).
2.2.Input data
We used a semi-virtual watershed.Therefore we combined topography and climate information of an existing watershed with virtual land use and soil information.The Parthe watershed was chosen as study area.It is located in the State of Saxony in Central Germany and drains an area of about315km2(Fig.1).It is a
A.Ullrich,M.Volk/Agricultural Water Management96(2009)1207–1217 1208
subbasin of the Weiße Elster catchment in the Elbe River system. The topography of the area isflat with altitudes between106and 230m above sea level.The mean annual precipitation is about 570mm.The model input data are shown in Table1.For the sensitivity analysis,we assumed‘‘arable land’’to be homogeneous land use without any further differentiation.A typical soil profile was used from a soil map(1:25,000)of the Parthe watershed.The use of homogeneous land use and soil(semi-virtual catchment)is advantageous because the resulting data sets are concise and manageable and calculation time is reduced.Daily precipitation data and other climate data are from one weather station in the watershed.This station is part of the environmental monitoring network of the Environmental Operation Agency of the Saxon State Agency for the Environment,Agriculture and Geology.
2.3.Sensitivity analysis
Model sensitivity analysis regarding selected management practices was donefirstly by varying the most important tillage and management parameters:tillage depth,mechanical mixing efficiency,biological mixing efficiency,curve number,and Man-ning’s roughness coefficient for overlandflow,USLE support practice factor,andfilter strip width.Secondly,management practices were parameterised and varied for different crops and dates of operation. Subsequently,the influence of varying these practices on water balance components and nutrients was evaluated.
2.4.Management parameters
The applied tillage operation(plough,stubble cultivation, harrow,etc.)is defined by the parameters tillage depth(DEPTIL) and mechanical soil mixing efficiency(EFFMIX).These parameters also define the fraction of crop residue,nutrients,pesticides,and bacteria for each soil horizon,which are redistributed within the mixed soil depth(Neitsch et al.,2002).The biological soil mixing efficiency(BIOMIX)defines the activity of soil organisms,such as earthworms as representatives of macrofauna,which influence soil porosity and waterfluxes by their grubbing activity(Kladivoka, 2001;Neitsch et al.,2002).The biological soil mixing efficiency can be defined for each HRU.The SCS
curve number(CN)defines soil permeability based on soil characteristics and land cover(land use).This parameter routes the process of infiltration and generation of surface runoff.The parameter CN can generally be defined on the HRU level and more detailed based on tillage operations data(Neitsch et al.,2002).The Manning’s roughness coefficient for overlandflow(OV_N)is a parameter to
estimate
Fig.1.Location of the study area in Germany.
Table1
Input data.
Topography Land use Soil Weather
DEM Homogeneous Homogeneous Daily values
-Area:315km2-Arable land-Cambisol-Precipitation[mm]
-Grid cell size:30m-Mean wind speed(recorded in2.5m height)[m/s]
-Max.and min.air temperature[8C]
-total solar radiation(calculated using global radiation)[MJ/m2]
-Mean relative humidity(recorded in0.5and2m height)[%]
A.Ullrich,M.Volk/Agricultural Water Management96(2009)1207–12171209
overlandflow velocity,which depends on characteristics of the land surface.This parameter can be defined for each HRU(Neitsch et al.,2002).The management parameter USLE support practice factor(USLE_P)defines the ratio of soil loss with a specific support practice(such as contour tillage,strip cropping,and terraces)to the corresponding loss with up-and-down the cultivation.The parameter USLE_P can also be defined on the HRU level(Neitsch et al.,2002).The width of edge offieldfilter strip(FILTERW)defines filter strips with a specific width in meters.Thefilter strips are not differentiated any further.Sediment and nutrient loads in surface runoff and subsurfaceflow are reduced as it passes through the strip.Filter strips can be defined for each HRU(Neitsch et al.,2002).
For each simulation only one parameter was varied within a realistic range(see Table2).The value ranges defined for each model parameter are based on literature studies(Neitsch et al., 2002).SWAT also supplies the user with default parameters values. For the basic settings we used these default values,as we are not working with real conditions where these values would have to be adjusted to.The advantage of this method is that the effect on model output is related to the variability of only the selected parameter,but it does not consider the dependency on settings chosen for the other parameters(Arabi et al.,2007).
For sensitivity analysis a basic management scenario was used. This is a generalised Agricultural Land Close Grown(AGRC) scenario,which uses values for winter wheat.One fertiliser application(70kg N/ha)just after seeding,and one tillage operation(deep ploughing)just after harvesting,was implemen-ted.This basic scenario was not changed during sensitivity analysis of management parameters.The model output parameters investigated are surface runoff,baseflow,total water yield,total sediment loading,organic nitrogen,organic phosphorus,nitrate in surface runoff,nitrate,and phosphorus leached.
The results(see Fig.2)indicate SCS curve number as a very sensitive parameter for water balance components and nutrient and sediment load.This observation is confirmed by Neitsch et al. (2002)as well as by other studies,such as Sloot et al.(1994), Heuvelmans et al.(2004),Bracmort et al.(2006),and Arabi et al. (2007).The CN is most sensitive for values above65.For example surface runoff increased from almost zero up to more than100mm between CN65to95.At the same time,corresponding surface runoff organic nitrogen,organic phosphate,nitrate in surface runoff and total sediment loading increased.In contrast,based on the decrease of baseflow,nitrate and phosphate leached decreased.
Biological soil mixing efficiency and Manning’s roughness coefficient for overlandflow do only modera
tely affect water balance components.The values vary less than5mm for applied parameter ranges.For the value range of biological soil mixing efficiency(0–1)the results for organic nitrogen(simulated range between0.225kg/ha and0.55kg/ha),organic phosphate(0.034–0.069kg/ha),and total sediment loading(0.057–0.318t/ha)increased while phosphate leached(0.318–0.145kg/ha)decreased. For the watershed area(315km2)this represents load values for organic nitrogen between7.1t and17.3t,organic phosphate between1.1t and2.2t,phosphate leached between10t and4.6t, and total sediment loading between1795t and10,017t.Based on these results we assume biological mixing efficiency to be a sensitive parameter regarding the above described nutrient components.
Manning’s roughness coefficient for overlandflow affected organic nitrogen(simulated range between0.248kg/ha and 0.135kg/ha;7.8–4.3t for watershed area)followed by sediment loading(0.088–0.048t/ha; 2.8–1.5t for watershed area)and organic phosphorus(0.038–0.02kg/ha; 1.2–0.6t for watershed area).Therefore,we assume this parameter to be sensitive with respect to the described nutrients and sediment load.
Both the USLE support practice factor and the width of edge of fieldfilter strip do not influence water balance components.But USLE support practice factor is very sensitive to total sediment loading(sim
ulated range between0.007t/ha and0.067t/ha;220–2110t for watershed area),organic nitrogen(0.019–0.187kg/ha, 0.6–5.9t for watershed area),and organic phosphorus(0.003–0.027kg/ha;0.09–0.9t for watershed area)while the width of edge offieldfilter strip affected organic nitrogen and decreased from 0.187kg/ha to0.076kg/ha(see Fig.3).This means that with an extension of thefilter strip width from0.5m to5m the organic nitrogen loss related to the watershed area decreased about50% (from5.9t to2.4t).Furthermore,with increasing width of edge of fieldfilter strip organic phosphate(simulated values:0.027kg/ha and0.011kg/ha;0.85t and0.35t),nitrate load in surface runoff (0.047–0.019kg/ha;1.5–0.6t)and total sediment loading0.067–0.027t/ha;2110–850t)decreased.In this study the variation of tillage depth and mechanical soil mixing efficiency did not influence neither water cycle components nor nutrient and sediment cycle components.
2.5.Management practices
With respect to the management parameters’sensitivity,the tillage operations subject to management practices(conventional (CVT),conservation(CST)and no-tillage(NOT))were parameterised exemplary(see Table3).Here,conventional tillage primarily is distinguished dependent on tillage practices applied after harvest-ing including deep ploughing,previous stubble cultivation and following harrow operation before seeding/planting.For conserva-tion management a multiplicity of m
easures can be taken.For tillage practice we chose altogether three variations:(a)deep ploughing operation is replaced by a less intensive operation(CST_A),(b)deep ploughing operation is left out and not replaced(CST_B)and(c) harrow operation is applied only(CST_C).
Parameters were set as follows.Differentiated by applied tillage operation,we parameterised curve number(CN)values in detail. The curve number adjustment is strongly linked to the soil dependent basic curve number identified within calibration process,planted crop(grain and root crop),applied tillage operation,and residue coverage(defined by applied management practice).The allocation of the SCS curve number is based on the parameterisations suggested by Neitsch et al.(2002)and continuative on the comments of Rawls and Richardson(1983). Rawls and Richardson(1983)recommend lowering the SCS curve number by2%for soils with poor hydrological conditions when applying conservation tillage(compared to conventional tillage). Forfields with good hydrological conditions,the SCS curve number should be lowered by4%compared to conventional tillage.King et al.(1996)applied EPIC using curve number values of87and82 for conventional tillage and for no-tillage practices respectively representing the soil hydrological group D(clay soil).Sloot et al. (1994)used the initial curve numbers:A value of84for
Table2
Basic parameter settings and variation ranges.
Parameter Basic setting Parameter range
SCS curve number(CN)7535–95 Biological soil mixing
efficiency(BIOMIX)
0.20–1.0
Manning’s roughness
coefficient for overland
flow(OV_N)
0.140.01–0.5
Tillage depth(DEPTIL)[cm]300–95 Mechanical soil mixing
efficiency(EFFMIX)
0.50–1.0
USLE support practice
factor(USLE_P)
1.00.1–1.0
Width of edge offield
filter strip(FILTERW)[m]0.00–5.0
A.Ullrich,M.Volk/Agricultural Water Management96(2009)1207–1217
1210
conventional tillage,83for minimum tillage(conservation tillage) and82for no-tillage.Here,the initial CN representing the soil hydrological group A(for soils with good hydrological conditions) was used.
Biological soil mixing efficiency is a sensitive parameter and was parameterised in detail depending on the intensity of the applied management practice.As a result,the biological mixing activity decreas
es with increasing tillage intensity.The
parameter Fig.2.Sensitivity of SWAT model to tillage parameters:CN,BIOMIX and
OV_N.
Fig.3.Sensitivity of SWAT model to management practice parameters USLE_P and FILTERW.
A.Ullrich,M.Volk/Agricultural Water Management96(2009)1207–12171211
发布评论