ulpeak discharge in a small catchment

a c d

ering, Royry Geolog, Norwegimental ReInstitute

n theirn the sictive at

h could be gained by

Science of the Total Environment 466467 (2014) 741754

Contents lists available at ScienceDirect

Science of the Tot

j ourna l homepage: www.e lseThe consequences of heavy rainfall and other extreme weatherevents are greatly inuenced by land use. To avoid damage to road

reducing the size and duration of peak ows. Many factors such asstorm characteristics (size and timing) and catchment characteristics(structure and land use) can inuence the hydrological response to1. Introduction infrastructure and road drainage facilities, muca b s t r a c ta r t i c l e i n f o

Article history:Received 17 April 2013Received in revised form 4 July 2013Accepted 14 July 2013Available online 25 August 2013

Editor: D. Barcelo

Keywords:Extreme rainfallrunoff eventsRoad infrastructureHydrological modelRunoffLand use change

A physically-based, distributed hydrological model (MIKE SHE) was used to quantify overland runoff in re-sponse to four extreme rain events and four types of simulated land use measure in a catchment inNorway. The current land use in the catchment comprises arable lands, forest, urban areas and a streamthat passes under a motorway at the catchment outlet. This model simulation study demonstrates howthe composition and conguration of land use measures affect discharge at the catchment outlet different-ly in response to storms of different sizes. For example, clear-cutting on 30% of the catchment area pro-duced a 60% increase in peak discharge and a 10% increase in total runoff resulting from a 50-year stormevent in summer, but the effects on peak discharge were less pronounced during smaller storms. Refores-tation of 60% of the catchment area was the most effective measure in reducing peak ows for smaller(2-, 5- and 10-year) storms. Introducing grassed waterways reduced water velocity in the stream andresulted in a 28% reduction in peak ow at the catchment outlet for the 50-year storm event. Overall,the results indicate that the specic effect of land use measures on catchment discharge depends ontheir spatial distribution and on the size and timing of storm events.

2013 Elsevier B.V. All rights reserved. Corresponding author. Tel.: +46 8790 7377; fax: +4E-mail addresses: zahrak@kth.se (Z. Kalantari), steve.l

lennart.folkeson@vti.se (L. Folkeson), helen.french@umb.njannes.stolte@bioforsk.no (J. Stolte), pej@kth.se (P.-E. Jans(M. Sassner).

0048-9697/$ see front matter 2013 Elsevier B.V. All rihttp://dx.doi.org/10.1016/j.scitotenv.2013.07.047reducing peak ow and total runoff.ocation within the catchment.st be considered before being implemented. The effect of reforestation was correlated to its l The spatial distribution of land use measures mua Department of Land and Water Resources Engineb Department of Physical Geography and Quaternac Department of Plant and Environmental Sciencesd Norwegian Institute for Agricultural and Environe Department of Land and Water Resources, Royalf DHI Sverige AB, SE-111 29 Stockholm, Sweden

H I G H L I G H T S

The effect of land use changes depends o The effect of land use changes depends o Reforestation proved to be the most effeal Institute of Technology, SE-100 44 Stockholm, Swedeny, Stockholm University, SE-106 91 Stockholm, Swedenan University of Life Sciences, NO-1432 s, Norwaysearch, Bioforsk, Soil and Environment Division, NO-1432 s, Norwayof Technology/KTH, SE-100 44 Stockholm, Sweden

spatial distribution.ze and timing of storm events.Zahra Kalantari , Steve W. Lyon ,Per-Erik Jansson e, Mona Sassner fLennart Folkeson , Helen K. French , Jannes Stolte ,a, bQuantifying the hydrological impact of sim6 8790 6857.yon@natgeo.su.se (S.W. Lyon),o (H.K. French),son), mona.sassner@dhi.se

ghts reserved.ated changes in land use on

al Environment

v ie r .com/ locate /sc i totenvvarious management measures aimed at reducing peak ows (Yeoet al., 2004). Therefore, a better understanding of how different landuses affect the water balance, peak ow and total runoff affecting roadstructures is vital.

Land use inuences the local hydrology near roads, e.g., the amount,intensity and duration of overland runoff approaching a low-lying road.

742 Z. Kalantari et al. / Science of the Total Environment 466467 (2014) 741754The sensitivity of peak discharges as ameasure of hydrological responseto vegetation and forest cover changes has been actively debated andclaried by researchers for many decades now (e.g., Hoyt and Troxell,1934; Mathur et al., 1976). For example, a number of studies haveshown that forest clear-cutting affects hydrological functioning andcatchment response to storms, with impacts on peak ow conditions(e.g., Jones and Grant, 1996; Wilk et al., 2001; Wissmar et al., 2004;Choi and Deal, 2008). Several decades of forest hydrology researchshow that harvesting activities can usually increase runoff by decreasingthe canopy interception and evapotranspiration losses (e.g., Campbelland Doeg, 1989; Stednick, 1996; Robinson et al., 2003; Moore andWondzell, 2005). Reforestation can thus be an efcient way to inuencecatchment hydrological responses (Linde et al., 2010) because it in-creases interception, evaporation and inltration rate and reduces runoffvolumes (Hundecha and Bardossy, 2004). Increasing the hydraulic fric-tion of a catchment by reforestation can also reduce the ow velocityin upstream parts and effectively minimise the peak runoff ow down-stream. However, the effectiveness of reforestation depends on factorssuch as catchment size and topography (FAO, 2005; Bronstert et al.,2008).

In addition to land use measures, on-site and prescribed manage-ment measures such as vegetation buffers and grassed waterways canchange the hydrological response of catchments. Vegetation buffersconsist of narrow strips of grass, trees, shrubs or combinations ofthese, usually located near the ow paths (DNR, 2011). They have thepotential to slow stormwater ow and trap soil particles and chemicalsubstances. Vegetation buffer zones are mandatory on agriculturalland in many countries (e.g., Dworak et al., 2009). The effectiveness ofvegetation buffer measures with regard to controlling high owsdepends on storm size and vegetation buffer properties (Yeo et al.,2004). Grassed waterways are also a commonmeasure to drain surfacerunoff from elds and are used primarily to prevent gully formation.They can be efcient in reducing erosion risk, runoff volume andpeak ow, mainly in relatively small basins in fragmented land-scapes (Fiener and Auerswald, 2005) such that their constructioncost is often offset through subsidy programmes (e.g., ygardenet al., 2006). According to Fiener and Auerswald (2005), the magni-tude of the reduction in ow discharge brought about by grassedwaterways depends on: (i) the size of the catchment overowingto the grassed waterways; (ii) the difference between soil inltra-tion volume plus storage capacity in the grassed waterways andthe runoff volume generated; and (iii) the geometry of waterwaycross-section in the area of concentrated ow.

The myriad aforementioned measures can all potentially affect alandscape's hydrological response to rainfall of different intensities.There remains, however, the classic problem of how to estimate theimpact of such measures and land use changes before they areimplemented on the ground. Hydrological models have widely beenused as tools to explore the potential impact of land use changes onstreamows in unmonitored watersheds (e.g., Isik et al. (2013)). Yanet al. (2013) evaluated the impacts of land use changes on streamowusing both hydrological and statistical modelling. Moreover, with re-spect to hydrological modelling as a tool for exploration, catchment-level changes in hydrology have been shown to be associated with for-est harvesting (e.g., Mathur et al., 1976; Seibert and McDonnell, 2010),forest re (e.g., Hoyt and Troxell, 1934; Hundecha and Bardossy,2004), agriculture (e.g., Schreider et al., 2002; Jaramillo et al., 2013),and a mosaic of land use changes (e.g., Ott and Uhlenbrook, 2004).More recently, Zgre et al. (2013) demonstrated the utility of a simplerainfallrunoff model to conduct non-parametric trend analyses on an-nual hydrological metrics for estimating hydrological response time in acatchment.

Although the potential impacts of changing land use in watershedshave been investigated for several purposes over the past few decades,assessing the potential often coupled inuences of various land uses

and/or management strategies on ooding and other extreme weatherevents near and on road constructions still remains among themore chal-lenging problems for land-use planners and road managers (Kalantariand Folkeson, 2013). This paper, thus, seeks to quantify the impact of dif-ferentmanagementmeasures on streamowsduringprecipitation eventsof different sizes for a catchment in Norway. Such knowledge will help inmanaging actors involved to identify themeasures necessary for reducingweather-related hazard risks in order to potentially prevent/mitigatedamages to roads. Hydrological models can be seen as useful tools inthis regard as they allow quantication of catchment responses tosimulated land use changes or various measures. Models havebeen demonstrated to be useful in assessing the relative effective-ness of adaptation measures to reduce peak ows in response toextreme weather events (e.g., Stolte et al., 2005). Distributed hydro-logical models in particular are capable of linking land use and man-agement effects across a catchment to physical processes (Whitakeret al., 2002). Such hydrological models can also be used to assesshow the catchment-scale storm response affects the total runoff andpeak discharge reaching roads at the catchment outlet (Wemple andJones, 2003). This allows the relative importance of different hydrologi-cal processes to be assessed in terms of water released from a catchmentover various timescales.

This study examined changes in peak ow and total runoffresulting from six simulated land use measures in a catchment,combined with long time series of meteorological data and aphysically-based, distributed hydrological model. Specic objectiveswere: 1) to determine the impact of forest clear-cutting on the sim-ulated amount and intensity of runoff; 2) to determine the effective-ness of simulated upstream measures (reforestation, vegetationbuffers and grassed waterways) in reducing the amount and inten-sity of runoff; and 3) to examine the extent to which the hydrolog-ical response to land use measures differs depending on storm size.The overall aim was to identify the best measures for reducing peakows and total runoff that can cause damage to downstream roadsor other infrastructures.

2. Materials and methods

2.1. Experimental site

The study was carried out on a catchment at Skuterud near s, ap-proximately 30 km south-east of Oslo, Norway. The mean annual tem-perature at s is 5.3 C, with a minimum of 4.8 C in January/February and a maximum of 16.1 C in July. Mean annual precipitationis 785 mm, with a minimum monthly amount of 35 mm in Februaryand a maximum of 100 mm in October (Department of AgriculturalEngineering (IMT), Norwegian University of Life Sciences). The averageannual potential evapotranspiration is 535 mm and varies from 463to 691 mm (Thue-Hansen and Grimenes, 2009; Farkas et al., 2013).Potential evapotranspiration of the site is about 1.5 mm day1 cal-culated using climatological data collected at the IMT station and thePenmanMonteith method to represent evaporation from an openwater surface.

The main soil type in the Skuterud catchment is marine siltyclay loam deposits, with some marine sand and moraine deposits(Deelstra et al., 2005). Marine deposits, occasionally rich in graveland stone, cover most of the catchment. Coarser marine shore de-posits dominate the fringes of the agricultural land both near andin the forested area. The catchment is transected by marginal mo-raine ridges. Soil types have only been explicitly mapped on the ar-able land in the catchment (Fig. 1). In the soil map, reference soilgroups are classied according to the World Reference Base forSoil Resources (WRB), and local soil series names are provided(Kvrn and Stolte, 2012). The soil map for Skuterud contains 34local soil series. The majority of soils in the central and level parts aremarine silt loam and silty clay loam soils (Albeluvisols and Stagnosols

according to the WRB). The texture of the shore deposits is mainly sand

743Z. Kalantari et al. / Science of the Total Environment 466467 (2014) 741754and loamy sand (Arenosols, Umbrisols, Podzols, Cambisols, Gleysolsaccording to the WRB). Lighter clay soils (loam, sandy loam) are foundin the transition zones between marine and shore deposits and on themoraine ridges. The WRB classes with the largest areal extent areEndostagnic Albeluvisols (silt loam and silty clay loam), Luvic Stagnosol(Siltic) (silty clay loam) and Endostagnic Cambisol (Dystric) (loamyne sand) (Kvrn and Stolte, 2012). The underlying bedrock is mainlyPre-Cambrian bedrock with predominant gneiss (Dons, 1977).

In view of ooding risks, the catchment is potentially vulnerableto high intensity rain events being transferred quickly through the

Fig. 1. The Skuterud catchmcatchment. A stream (Skuterudbekken) intersects with the mainmotorway (E18) between Oslo and Stockholm at the catchmentoutlet. As part of the Environmental Agricultural Monitoring Pro-gramme in Norway, total discharge and water quality at the catch-ment outlet have been monitored since 1993 (Srbotten, 2011)although not continuously. Discharge is measured at the outletusing a Crump-weir. Water levels are recorded automatically usinga data-logger in combination with a pressure transducer and dis-charge is calculated on the basis of the existing head-discharge rela-tion (Deelstra et al., 2005).

ent near s, Norway.

a) Land use scenarios: forest clear-cutting

b) Land use scenarios: reforestation (60% )

c) Land use scenarios: reforestation (30%)_downstream

Fig. 2. a. Land use scenarios: forest clear-cutting. b. Land use scenarios: reforestation (60%). c. Land use scenarios: reforestation (30%)_downstream. d. Land use scenarios: reforestation(30%)_upstream. e. Land use scenarios: vegetation buffers. f. Land use scenarios: grassed waterways.

744 Z. Kalantari et al. / Science of the Total Environment 466467 (2014) 741754

n (

745Z. Kalantari et al. / Science of the Total Environment 466467 (2014) 741754d) Land use scenarios:reforestatio2.1.1. Current land useThe Skuterud catchment is 4.5 km2 in area and consists of 2.7 km2

(60%) arable land (mostly used for grain, potato and ley crops),

e) Land use scenarios: vegetation bu

f) Land use scenarios: grassed water

Fig. 2 (cont30%)_upstream 1.3 km2 (30%) (mainly coniferous) forest and 0.5 km2 (10%) urbanarea (Fig. 1). The arable soils in this catchment are tile-drained, with adrain depth of 0.8 m (Deelstra et al., 2010).

ffers

ways

inued).

Table 1Parameter values used inMIKE SHE for six land use patterns (current land use, arable land, clear-cutting, reforestation, vegetation buffers and grassedwaterways). All valueswere initially calibrated (step 0) using the general pre-calibration approachwith additional calibration (bold) for (step 1) the drainage time constant (Tc) using the 2-year event in 2008 and (step 2) the reference evapotranspiration (ETref) using the 5-, 10- and 50-year events from 1995 to 2001 (see Fig. 3 for calibrationprocedure schematic).

Model parameter MIKE SHE Current land use Land use scenarios

Method Clear-cutting Reforestation Vegetationbuffers

Grassedwaterways

Overland ow Surface roughness (M) 2D nite difference-diffusive wave Spatial distribution, Manning M = 5and 6a,b

ManningM = 20 and 6

ManningM = 5

Manning M = 5,6 and 30a,b

ManningM = 5 and 6

River ow River bed roughness (M) 1D st-Venant equation M = 45a,b M = 45 M = 45 M = 45 M = 30Unsaturated zone (UZ) Saturated hydraulic conductivity (Ks) Richards equation Ks = 4.2e 005 to 1e 010Saturated zone (SZ) Horizontal cond. (Kh)

Vertical cond. (Kv)Specic yield (Sy)Drain level

Finite difference Kh = 1e 006 to 1e 008 m/sKv = 1e 006 to 1e 008 m/sSy = rst layer: 0.04 to 0.1, second and third layers: 0.0001Drain level = 0.8 m relative to the ground

Actual evapotranspiration Leaf area index (LAI), root depth (RD) andcrop coefcient (Kc)

MIKE SHE uses the Kristensen and Jensen (1975)method and the two-layer UZ/ET module basedon a formulation presented in Yan and Smith(1994)

Coniferous forest:LAI = 7c,d,e

RD = 0.8 mc,d,e

Kc = 1c,d,e

Grass:LAI = 1.53c,d,e

RD =0.20.6 mc,d,e

Kc = 1c,d,e

Coniferousforest:LAI = 7RD = 0.8 mKc = 1

Coniferous forest:LAI = 7RD = 0.8 mKc = 1

Coniferous forest:LAI = 7RD = 0.8 mKc = 1

Arable land:LAI = 16c,d,e

RD = 0.11 mc,d,e

Kc = 11.3c,d,e

Arable land:LAI = 16RD = 0.11 mKc = 11.3

Forest (mix):LAI = 46.5RD = 0.11 mKc = 11.3

Arable land:LAI = 16RD = 0.11 mKc = 11.3

Arable land:LAI = 16RD = 0.11 mKc = 11.3

Grass strips:LAI = 1.53RD = 0.20.6 mKc = 1

Wetland:LAI = 46c,d,e

RD = 1 mc,d,e

Kc = 1c,d,e

Wetland:LAI = 46RD = 1 mKc = 1

Wetland:LAI = 46RD = 1 mKc = 1

Wetland:LAI = 46RD = 1 mKc = 1

Wetland:LAI = 46RD = 1 mKc = 1

Time step control Time steps Initial time step = 1 hDrainage option Drainage time constant (Tc) Calibrated in the rst calibration step Tc = 5.5e 007 s1Climate Reference evapotranspiration (ETref) Calibrated in the second calibration step With an average of 420 mm year1 or 1.15 mm day1

Current land use Clear-cutting Reforestation (60%) Reforestation (30%) Vegetation buffers Grassed waterways

Land use Agricultural land Area (km2) = 2.7 Agricultural land (60%) Forest (mix) (60%) Agricultural land (30%) Short grass (5 m width) on both sides of stream (60%) Agricultural land (60%)Area (%) = 60 Forest (mix) (30%)

Forest Area (km2) = 1.3 Bare soil with tree stumps (30%) Forest (30%) Forest (30%) Forest (30%)Area (%) = 30

Urban area Area (km2) = 0.5 Urban area (10%) Urban area (10%) Urban area (10%) Urban area (10%)Area (%) = 10

a Chow (1959).b Arcement and Schneider (1989).c Bultot et al. (1990).d DHI Software (2008).e Kelliher et al. (1993).

746Z.Kalantarietal./Science

oftheTotalEnvironm

ent466467

(2014)741

754

747Z. Kalantari et al. / Science of the Total Environment 466467 (2014) 7417542.1.2. Simulated land use scenariosFocusing on land use and physical measures aimed at reducing the

generation of runoff andminimising stormwater ows, six land use po-tential scenarios were developed (Fig. 2af) and tested in this study:

Scenario 1) Clear-cutting of all forest in the catchment (Clear-cutting30%). The effect of this common forestry practice was tested by as-suming clear-cutting of all the area currently forested (30% of thewhole catchment).Scenarios 24) Reforestation (Reforestation 60%, Reforestation 30%upstream and Reforestation 30% downstream). In the Reforestation60% scenario, the entire arable area (comprising 60% of the entirecatchment) was assumed to be replaced by mixed forest. In theReforestation 30% upstream scenario, only half the arable area, lo-cated in the upstream area of the catchment and far from the roaddrainage structure, was changed to mixed forest. In the Reforesta-tion 30% downstream scenario, only half the arable area, locatedin downstream positions of the catchment and immediately besidethe road drainage structure, was changed to mixed forest.Scenario 5) Vegetation buffers along the stream (Vegetation buffers).A 5 m wide buffer strip of permanent vegetation was included oneach side of the stream (Syversen and Bechmann, 2003). The forestin this scenario is the same as in the current condition.Scenario 6) Grassed waterways along the stream (Grassed water-ways). The effectiveness of grassed waterways in reducing runoffwas tested by changing the natural bed of the stream in Skuterudto a grass-covered bed. The forest in this scenario is the same as inthe current condition.

2.2. Simulation systems

Anumber ofmodels are available for simulatingdischarge andwaterow processes in a small catchment. MIKE SHE, a deterministic, dynam-ic, physically-based and distributed model (DHI, 1998) is applicable onspatial scales ranging from a single soil prole to large regions includingdifferent landscapes (Graham and Butts, 2005). Therefore this modelwas applied to quantify the effect of land use scenarios on peak dis-charge and total runoff.

2.2.1. MIKE SHEMIKE SHE describes the main processes in the land phase of the

hydrological cycle. Precipitation can either be intercepted by the canopyor fall to the ground. Thewater on the ground surface can inltrate, evap-orate or form overland ow. Once the water has inltrated it enters theunsaturated soil zone, where it can either be extracted by roots andleave the system as transpiration, or percolate down to the saturatedzone. MIKE SHE is fully integrated with a channel-ow code, MIKE 11(DHI Software, 2008). The model components are described in detail ine.g., Vsquez et al. (2002), Graham and Butts (2005) and DHI Software(2008).

2.2.2. Model parameterisationThe horizontal resolution of the calculation grid in MIKE SHE was set

to 10 m by 10 m in the whole model area and applied on all ow com-ponents, i.e. overland ow, the unsaturated zone and the saturatedzone. The boundary of the catchment was dened based on the topo-graphical divide. The top boundary condition is expressed in terms ofprecipitation and reference evapotranspiration. Precipitation is assumedto be uniformly distributed over themodel area and is given as a time se-ries, while actual evapotranspiration is calculated during the simulation.The model was run for current land use and the previously presentedland use scenarios during different simulated storm events. The simulat-ed land use scenarioswere dened by applying changes in roughness pa-

rameters (roughness coefcient, M) and the vegetation parameters leafarea index (LAI), root depth (RD) and crop coefcient (Kc). MIKE SHEparameter values used for current land use and different land use mea-sures (Table 1) were based on literature values and site-specic esti-mates and partly adopted from previous applications of the model indifferent studies (e.g., Kalantari et al., submitted for publication).

The roughness parameter needs to be specied for calculation ofoverland ow. The roughness coefcient (M) typically has values be-tween 10 (thickly vegetated channels) and 100 (smooth channels)and lower values of M are generally used more for overland ow thanchannel ow. In the present study, the M variations were developedusing the dened relationship with land use/land cover from Chow(1959) and Arcement and Schneider (1989).

Because of seasonal changes, the vegetation has different cropstages, for each of which the vegetation parameters LAI, RD and Kchave to be specied. LAI can vary between 0 and 7 depending on thevegetation type and values were taken from local estimates. RD valuesfor the relevant crops and land uses were adopted from literature. Kcis used to adjust the reference evapotranspiration relative to the poten-tial evapotranspiration for a specic crop. Again, the values consideredin this study to represent the various land use scenarios were takenfrom literature values. Interception is dened as the process wherebyprecipitation is retained in canopies (leaves, twigs, branches andstems) and is determined in the model by multiplying the interceptioncapacity, Cint, by the LAI (leaf area per unit ground surface area). Theintercepted water is assumed to evaporate without adding to the mois-ture storage in the soil.

The reference evapotranspiration (ETref) is the rate of evapotranspi-ration from a reference surface (a hypothetical short-clipped lawn)with no limitation on water availability (Allen et al., 1998). ETref is de-pendent on climate and can be calculated from weather data. In thisstudy, the ETref was determined using the terms in the CoupModel ofJansson and Karlberg (2004) (essentially PenmanMonteith method)where ETref is estimated from a canopy resistance estimated from LAIand canopy conductance (Lohammar equation) (Jansson and Karlberg,2004).

Setting up a saturated zone hydraulic model based on the 3D nitedifference method (Cooley, 1971) involves dening the geologicalmodel, the vertical numerical discretisation, the initial conditions, andthe boundary conditions. The ground surface, as given by the topo-graphicmodel, is the uppermodel boundary whereas the lower bound-ary in the base setup of the model is at 100 m below the surface level.The geological layers are the basis for the model parameterisation,whichmeans that the hydrogeological parameter values such as verticaland horizontal hydraulic conductivity, specic yield and specic storageare assigned to the different geological layers based on existing general(i.e., not site-specic) relationships. As the uxes through the soil sur-face are functions of the vertical hydraulic conductivity of the aquiferand inuence ground-water level and peaks of streamow and the hor-izontal saturated conductivity in the saturated zone can affect the baseow as well as the peak ows and the time ow reaching the stream(Graham and Butts, 2005; DHI Software, 2008), a short sensitivity anal-ysis was carried out in this study in connection with model calibration(see the following sections) to help guide efforts with regards to achiev-ing a more efcient calibration procedure to potentially adjust generalparameter values to more site-specic values.

2.3. Historical events

The intensitydurationfrequency (IDF) curve and recorded maxi-mum short-term precipitation in the Norwegian Meteorological Insti-tute database between 1974 and 2008 were used to dene the valuesof storm size and rainfall intensity for specic return period events.In this study, the return period storm sizes considered were 2-years,5-years, 10-years, and 50-years. The number of events observed acrossthe record equivalent to each specic return period considered varies

(Table 2). One event was chosen for each return period storm size as

numerical stability (DHI Software, 2007). The modelling time stepused in each subsequent calibration step was thusly determined to be

e pefallng tsent

748 Z. Kalantari et al. / Science of the Total Environment 466467 (2014) 741754an hourly time step. In a pre-calibration step (step 0, Fig. 3), the numericalinstabilities were checked as was the validity of the physical meaning oftherstmodel results. In this initial pre-calibration step, parameter valueswere dened as previously outlined and model streamows were com-pared to observations. Once the model was established, additional cali-bration was carried out in two steps in a simple intuitive way withfocus on the most sensitive parameters in the model through a manualcalibration performed rst (step 1) on an event-specic basis followeda representative example formodelling. Themodel was set up to performcontinuous simulations of the hydrology and several events wereextracted for comparison in this study. These were spread across the var-ious seasons in the region to further explore the potential inuence oftime of year on hydrological responses. A storm from the winter seasonof 2008 was chosen to represent the 2-year storm and a 5-year storm in2001 to represent an autumn event. The year 1995 was special, as botha 10-year and a 50-year event were recorded during the summer period.

2.4. Calibration and sensitivity analyses

MIKE SHEwas calibratedwith the aim of dening optimal values forphysically realistic parameters in order to simulate the discharge fromthe Skuterud catchment (Table 1). Fig. 3 schematically summarisesthe steps taken in model calibration, including the main sub-versionsof the model, the main actions taken in each step and the target ofeach calibration step. The model was run for a 14-year simulation peri-od, 1 January 1995 to 30 April 2008. To acquire better numerical accura-cy, the time step and numerical interaction criteria were controlled toobtain a reasonable compromise between actual simulation times and

Table 2Characteristics of the historical events considered including the number of storm events, thdischarge divided by drainage area), fraction of total amount (corresponding to Q-peak rainresponding to runoff volume divided by drainage area) and fraction of rainfall (correspondiMeteorological Institute database from 1974 to 2008 for Skuterud catchment for the repre

2-year storm

Number of events during 19742008 39Event used in simulation 11 Jan 2008Peak rainfall intensity (mm h1) 14.4Q-peak intensity (mm h1) 1.5Fraction of rainfall intensity () 0.10Total rainfall (mm) 10.0Q-total amount (mm) 4.5Fraction of total rainfall () 0.45by (step 2) an annual basis (Table 1).A preliminary sensitivity analysis was, thus, made to determine

which parameters to consider for further calibration. As the focus inthis study's simulations was on describing the dynamics of surfacewater and near-surface groundwater and their interactions in generat-ing discharge and surface runoff, the sensitivity analysis was focusedon the parameters inuencing the hydraulic properties of the soil, theunsaturated zone parameters, vegetation parameters, drainage, andhydroclimatic setting. The sensitivity analysis consisted of adjustingparameter value sets by various multipliers (ranging from 2 to 10depending on the magnitude of the parameter values) and assessingthe relative change these adjustments caused in the streamow simula-tion. Generally, the pre-calibrated model was not sensitive to saturatedhydraulic conductivity (Ks) in the unsaturated zone, the horizontal andvertical hydraulic conductivity (Kh and Kv) and specic yield (Sy) in thesaturated zone, vegetation parameters such as the leaf area index(LAI), the root depth (RD) and the crop coefcient (Kc), the roughnesscoefcient (M) and the drainage level. As such, the parameter valuesassigned in pre-calibration were kept throughout the remaining model-ling calibrations and experiments.

The model setup was, however, rather sensitive to the drainagetime constant (Tc). For example, changing Tc from its base value(1e 007 s1) by a factor of 5 (in either direction) resulted in large(signicant) changes in the maximum peak ows but not the surfacewater discharge volume. Furthermore, the model was sensitive to thereference evapotranspiration (ETref) values assigned with a reductionin ETref creating signicant differences in total annual runoff (but not inpeak discharge). From this simple sensitivity analysis, Tc and ETref wereassumed to be the most sensitive parameters inuencing the model'sability to simulate peak discharge and the total annual runoff,respectively.

Once themain sensitive parameterswere identied, therst calibra-tion step was to calibrate the model to the representative 2-year stormevent during 2008 (Table 2). A simplied tuning of the drainage timeconstant, Tc, was applied in order to obtain a better calibrated modelwith emphasis on getting the peak discharge correct for the 2-yearevent (Kalantari et al., 2011, 2013). The Tc is equivalent to a leakage co-efcient which implies a factor that is used to regulate how quickly thewater can drain from the catchment. In the present case, the subsurfacedrainage may thus be interpreted as additional runoff capacity in thetop soil layer related to the drainage time constant.

In the second calibration step, the model run individually for one-year simulation periods containing the other selected return periodstorm sizes (i.e., January 1995December 1995 for the 10-year and50-year events and January 2001December 2001 for the 5-yearevent). In this second calibration step, due to lack of continuously mea-sured discharge data, the ow simulated byMIKE SHE for the years con-taining these larger storm events (5-, 10- and 50-year) was calibratedby adjusting the ETref such that the total annual runoff simulated t the

ak rainfall intensity (mm h1), Q-peak rainfall intensity (mm h1; corresponding to peakintensity divided by peak rainfall intensity), total rainfall (mm), Q-Total amount (mm; cor-o Q-Total amount divided by total rainfall for different storm sizes) recorded in Norwegianative events used in our modelling experiments.

5-year storm 10-year storm 50-year storm

10 4 116 Sept 2001 15 Jul 1995 31 Jul 199521.8 25.2 34.37 5 80.32 0.20 0.2331.0 34.8 71.815 9 260.48 0.26 0.36total annual runoff calculated by Deelstra and Iital (2008) (606 mm and501 mm for 2001 and 1995, respectively).

2.5. Validation and model evaluation

The nal calibrated model was validated using six rainfallrunoffevents occurring between 4 November and 12 December 2007(Fig. 4). In general, there was agreement between simulated and ob-served discharge over both the calibration considered and the validationperiod. In the present case, the peak error (dened as the difference be-tween simulated and observed peak streamows), NashSutcliffe ef-ciency (Nash and Sutcliffe, 1970) and coefcient of determination(Table 3) were assumed to provide a good combination of likelihoodmeasures to evaluate the accuracy of both the magnitude and timingof predicted discharge (e.g., Andersen et al., 2001; Beven, 2001;Vsquez et al., 2002; Tague et al., 2004). According to Moriasi et al.(2007), the calibration performance for hydrologicalmodels can be con-sidered satisfactorywhenNSE N0.5 and R2 N0.5. The peak error value for

the validation period is the average of the peak errors from all 6 events.Further, the low NSE during the validation (Table 3) may be in part due

with calibration, it is by no means perfect (which no model can everbe). These initial validation results, however, lend condence to our

Fig. 3. Schematic summary of the calibration steps performed in the development of the MIKE SHE model.

749Z. Kalantari et al. / Science of the Total Environment 466467 (2014) 741754to compounded simulation error impacts of which NSE is fairly suscep-tible (Beven, 2001). As such, while the model performance improvedValidation period

Fig. 4. Simulated and observed discharges duringability to use the model as a tool for exploring (to at least a rst-order) impacts associated with land use and management changes.Calibration period

a calibration period and a validation period.

3. Results and discussion

3.1. Simulated annual discharge for different land use measures

Water balance output from MIKE SHE simulations include area-normalised ows (storage depths) and storage changes for individualmodel components (e.g., unsaturated zone, evapotranspiration). Inthis study, the components of the annual average water balance (totalevapotranspiration and total runoff) were dened as follows based onthe output of MIKE SHE:

Total ET E E E T ET 1

Table 4, thus, outlines the components of the annual average waterbalance and the MIKE SHE output from the current catchment landuse and the six simulated scenarios for the period 1 January 1995 to30 April 2008.

Comparison of average total annual runoff values across the entire14 years showed that clear-cutting as a forest management practiceled to a minor increase in the average annual runoff volume by only6% over the amount estimated under present land use conditions,while the scenarios Reforestation 60%, Reforestation 30% down-stream and Reforestation 30% upstream reduced the average annualrunoff volume by 14%, 12% and 8%, respectively. The other landusemea-sures considered, i.e., vegetation buffers and grassed waterways, hadminor decreasing effects on average annual runoff volume (2% and 3%,respectively). These differencesweremainly due to differences in actualevapotranspiration and available water exchange between OL, UZ andSZ. The water balance analysis from the forest clear-cutting scenarioshowed increased inltration of overland ow simulated in the modeland an increase in groundwater levels across the catchment estimatedthrough simulations. Hence, the simulated groundwater contributionto the river increased, thereby increasing the simulated total runofffrom the whole catchment.

The results from the reforestation scenarios showed that on an an-nual basis, a large proportion of precipitation was intercepted by trees,leading to higher evaporation from the canopy than that simulatedthrough the implementation of vegetation buffers and grassed water-ways. The results also indicated that doubling the forest area did nothalve the runoff volume and that the location of the reforestationwithin

Table 3Evaluation statistics (peak error (m3/s) = Q Peak (sim) Q Peak (obs), R2 = coefcientof determination, NSE = NashSutcliffe simulation efciency before and after calibrationof discharge during two periods). Note that the peak error in the validation period representsthe average value from the 6 events considered in Fig. 4.

Periods Statistics Before calibration After calibration

Calibration period11 Jan 2008 Peak error 0.4 0.03

R2 0.4 0.65NSE 0.2 0.5

Validation period04 Nov12 Dec 2007 Peak error 0.45 0.25

R2 0.15 0.7NSE 1.55 0.45

ages, d

Ref(60

8

1

7

750 Z. Kalantari et al. / Science of the Total Environment 466467 (2014) 741754canopy OL soil RZ SZ

where Total ET is total evapotranspiration; Ecanopy is evaporation fromcanopy; EOL is evaporation from overland ow (OL); Esoil is direct evap-oration from soil; TRZ is transpiration from root zone; and ETSZ is evapo-transpiration directly from saturated zone (SZ).

Total Runoff OLriver SZdrainage SZriver 2

where OLriver is overland outow to river; SZdrainage is saturated zonedrainage to river; and SZriver is saturated zone base ow to river.

Table 4Impact of different land use scenarios on average total annual evapotranspiration and averusing MIKE SHE for the Skuterud catchment, SE Norway. Upward ows have positive valuerated zone. UZ: unsaturated zone.

Current Forest clear-cutting(30%)

Precipitation 837 837Canopy interception itemsEvaporation from canopy 89 71

Overland ow itemsCanopy through fall to OL 748 792

Evaporation from OL 5 6Upward ow from SZ to OL 470 476 3Flow from OL to SZ 80 78Outow from OL to boundary of sub-catchment 38 41Overland outow to river 396 413 3

Unsaturated zone itemsInltration from overland to unsaturated zone 699 730 5Direct evaporation from soil 70 80Transpiration from the root zone 173 170 2

Saturated zone itemsRecharge from UZ to SZ 456 480 3Upward ow from SZ to overland 470 476 3Flow from OL to SZ 80 78 Evapotranspiration directly from SZ 0 1SZ drainage to river 50 60SZ base ow to river 16 21Total ET (mm year1) 337 328 4Total runoff (mm year1) 466 494 4the catchment had an effect on simulated runoff volume. The location ofany land use change is highly important, especially if the water avail-ability is controlled by the topography, which appears to be the casein the Skuterud catchment, also supported by Deelstra et al. (2007). Inthe present study, the reduction in total runoff per unit area of refores-tation was most efcient when the area immediately adjacent to theroad was reforested (see Fig. 2c).

The water balance analysis indicated that surface ow and owthrough tile drains contributed substantially more than base ow tototal runoff. Deelstra et al. (2007, 2010) found a very high ashinessindex (FI) and low base ow index (BFI) for the Skuterud catchment.

total annual runoff for the period 1 January 199530 April 2008. Results from simulationsownward ows have negative values. ET: evapotranspiration. OL: overland ow. SZ: satu-

orestation%)

Reforestation (30%)downstream

Reforestation (30%)upstream

Vegetationbuffers

Grassedwaterways

37 837 837 837 837

36 117 110 96 93

01 725 727 741 7443 3 4 3 554 355 377 422 47082 80 81 78 8223 28 30 38 4360 365 387 393 390

87 604 602 651 69454 69 67 70 7018 216 196 174 174

15 319 339 407 45054 355 377 422 47082 80 81 78 820 1 1 1 030 29 26 43 4313 14 16 19 1911 406 378 344 34203 408 429 455 452

This indicates that total runoff in Skuterud is likely dominated by rapidresponse surface ow and subsurface ow rather than base ow consis-tent with the interpretation and modelling results in this current study.

3.2. Simulated peak discharge during four historical events

Considering the simulated discharge hydrographs at the catchmentoutlet during four historical storms with the current land use condi-tions, the simulated peak discharge for the 10-year storm was lowerthan that for the 5-year storm (Fig. 5).

Moreover, the peak discharge intensity and total amount of dis-charge for the 10-year storm were lower than for the 5-year storm(Table 2). This illustrates that it is not only the boundary conditions,i.e., the precipitation event, that explain the discharge, but that thephysical conditions in the catchment during the event are also highlyimportant. The 10-year storm occurred in summer, after initial drynessof the catchment, and the 5-year storm in autumn,with both inltrationexcess and saturation excess overland ow processes likely occurringdue to a higher general wetness of the catchment. The 2-year eventwas generated from rain and snowmelt on frozen soil (Kalantari et al.,submitted for publication). The biggest storm (50-year) occurred twoweeks after the 10-year storm with a peak discharge almost doublethe size of the previous event. Clearly, seasonality and catchment condi-tions affect hydrological response as is typical in Scandinavian systems(e.g., Lyon et al., 2010, 2012).

3.3. Effect of simulated land use measures on the four historical stormevents

Simulation of complete clear-cutting of the forested area produced a60% increase in peak discharge and only a 10% increase in total runoffresulting from the 50-year summer storm event, which followed a peri-od of high evapotranspiration (Figs. 6 and 7). The lower intensity(shorter return period) storms resulted in limited increases in peakow intensity. The magnitude of peak ow and runoff response there-fore appeared to vary seasonally, e.g., due to uctuations in inltrationcapacity as a function of soil moisture.

The reduction in peak ow (Fig. 6) and total runoff (Fig. 7) for all veremediation measures considered indicates that they all have potentialfor reducing peak discharge from a small catchment following storms ofdifferent intensities. Using themodelling setup as a basis, it is possible topostulate on the mechanisms responsible for each measure's ability toinuence peak ow and total runoff. For example, reforestation led toreduced total runoff and peak ow through increased evapotranspira-tion and increased watershed area roughness, respectively. Vegetationbuffers, their design and implementation, increased oodplain arearoughness and, thus, inuenced peak ow and (to a lesser extent)total runoff. Implementation of grassed waterways increased streamchannel area roughness allowing for greater proportional reduction inpeak ow relative to total runoff across all the storms considered.As such, different measures will inuence the hydrology of a catchment(e.g., peakowor total runoff) in differentways implying that, depending

ent

751Z. Kalantari et al. / Science of the Total Environment 466467 (2014) 741754Fig. 5. Simulated runoff under current land use conditions at the outlet of the Skuterud catchm

P = precipitation.)for four historical storm events with return periods of 2, 5, 10 and 50 years. (Q = discharge,

Fig. 6. Percentage of change in simulated peak ow during different historical storm events due to the impact of different land use scenarios (forest clear-cutting, reforestation of 60% and30% of the whole catchment area, vegetation buffers and grassed waterways). Results from simulation using MIKE SHE for the Skuterud catchment, SE Norway.

752 Z. Kalantari et al. / Science of the Total Environment 466467 (2014) 741754on the desired result, care needs to be taken when identifying the mea-sures necessary for reducing weather-related hazard risks in order to po-tentially prevent/mitigate damages to roads.

Increasing the forest area, for example by reforestation of 60% of thecatchment area, was the most effective measure in reducing peak owfor 2-, 5- and 10-year storms (Fig. 6). However, 30% reforestation ofthe catchment area was the most effective measure in reducing thetotal runoff for all storm sizes (Fig. 7). This, of course, is dependent tosome extent on the size of the catchment considered in this study. Theeffects of forest shown in this studymay be relatively smallerwhen con-sidering larger systems and/or longer return period ooding events. Theinuence of reforestation location (spatial distribution of reforestation)was more obvious for the smaller storms. For the largest storm size(50-year), grassed waterways had the highest potential for reducingthe peak ow at the catchment outlet (Fig. 6), while 30% reforestationwas more effective in decreasing total runoff (Fig. 7). The vegetationbuffers had an impact on discharge mainly when the peak dischargewas low, indicating that this type of measure could be more effectiveduring small events generally occurring in summer time. In general,grassed waterways had a better effect than vegetation buffers in reduc-ing peak ow in all storm events.Fig. 7. Percentage of change in total runoff during different historical storm events due to the imwhole catchment area, vegetation buffers and grassed waterways). Results from simulation usAccording to Fiener and Auerswald (2005), grassed waterways ef-ciently reduce runoff volume and peak discharge rates in relativelysmall catchments with a small-patterned landscape. This is consistentwith our nding that grassed waterways were the most efcient mea-sure to reduce the peak ow for the largest storm size (50-year). How-ever, it should be noticed that grassedwaterways are not recommendedfor channelswith sustained baseowasmost vegetation cannot surviveextended saturated root zone. In such a situation, a compound channelhaving small, lined channel in the centre to carry base ows and a veg-etated zone to carry stormowmay be developed (Haan et al., 1994). Inaddition, Markart et al. (2006) suggest placing runoff-preventing bufferstrips along the receiving water course to avoid concentration of inten-sive cultivation on sites.We conrmed that this measure had an impacton catchment runoff characteristics, but found that its effectivenessdepended specically on storm size, as noted previously by Yeo et al.(2004).

3.4. Implications for road design

The hydrological response of a catchment to precipitation events ofvarious sizes is an important factor in the design of hydrotechnicalpact of different landuse scenarios (forest clear-cutting, reforestation of 60% and 30% of theing MIKE SHE for the Skuterud catchment, SE Norway.

753Z. Kalantari et al. / Science of the Total Environment 466467 (2014) 741754installations, such as surface and lateral subsurface drainage systems,culverts and provisions for the intake of surface runoff.

Modelling the hydrological response is useful for analysing the im-pacts of possible measures to reduce peak ow and total runoff inorder to avoid damage to roads downstream. To characterise the hy-drology of the Skuterud catchment, water balance indicators such asevapotranspiration and the available water exchange between overlandow zone, unsaturated and saturated zone were analysed. Comparisonof six simulated land use measures showed that clear-cutting had astrong impact on thewater balance because of changes in evapotranspi-ration, which is a major determinant of runoff in forested catchmentsparticularly during summer periods (e.g., Lyon et al., 2012; van derVelde et al., 2013). Therefore suppression of evapotranspiration bytree harvesting can be expected to increase water yield and peak dis-charge by making more water available for inltration, surface runoffand groundwater recharge. If groundwater storage capacity is limited,the amount of contributing ow from the saturated zone to the over-land ow zone and the stream will increase generating higher surfacerunoff and posing a threat to roads and other infrastructures down-stream (Table 4). This impact due to changes in land use should, thus,be taken into consideration when designing new and maintainingexisting road infrastructure systems.

The specic effect of land usemeasures on catchment discharge alsodepends on the spatial distribution of the measures. Here, for example,the location of reforestation patches in terms of distance from the roaddrainage structure had an effect on peak ow and runoff volume reduc-tion, especially for 5-year and 10-year events (Figs. 6 and 7). While allthe parameters used in this study were chosen based on literature orearlier modelling studies, the results are still associated with uncer-tainties. However, uncertainty analysis in terms of parameters ormodel structure was beyond the scope of this study. Possible sourcesof uncertainty specic to themodelling include e.g., the quality of phys-ical data (i.e., physical characteristics of the catchment), climate dataand ow data used in the analysis. It is difcult to obtain spatially dis-tributed values of all input parameters for a model such as MIKE SHEwhen it is being applied at the scale of a catchment. As a consequence,the input parameters need to be simplied and spatially averaged forthe selected locations. Due to the discrepancy in scales, measurementerrors and lack of spatially distributed data, there are clearly uncer-tainties that would need to be taken into consideration before onecould apply the results of this study to real-world environmental condi-tions relevant for road planning and management (Sahoo et al., 2006;Kvrn and Stolte, 2012). For example, maintenance of roadriverintersections (which is not explicitly considered here) is potentially oflarge importance since it is often the case that entrance points forwater at roadriver intersections are blocked by branches, trunks, andsediment accumulation (Kalantari and Folkeson, 2013).

In spite of these limitations, the study demonstrates the utility ofusing hydrological models to examine the mechanisms regulatingcatchment responses to different land use measures in relation to thespatial distributions of the measures and selected storm characteristics.The traditional (and long-standing) approach in road construction hasbeen to make sure culverts and bridges are large enough for specicstorm events where discharge is calculated from the rational equation(Benzvi, 1989; Maidment, 1993). Further, consider that the 50-yearow is the current design ow for dimensioning of culverts and bridgesin Norway and Sweden. The question for the road authorities in these(and other) countries is whether or not the current 50-year ow esti-mate will still be a correct design ow under different future manage-ment systems. The method and approach outlined in this study canhelp in answering this question based on the current state-of-the-science with regard to landwater interactions (which is simply notpossible under the rational method). The results for the Skuterud catch-ment considered here indicate, for example, that complete clear-cuttingof the forested area produces a 60% increase in peak discharge and a 10%

increase in total runoff resulting from the 50-year summer storm event(Figs. 6 and 7). This implies that changes in land use could lead toooding and damages to road infrastructures as they have directimpacts on streamows at the catchment-scale.

4. Conclusions

Using a physically-based hydrological model, the implementation ofvarious land use measures was shown to have an impact on the hydro-logical response and amount of overland runoff approaching a low-lyingroad in a mixed land use catchment. More importantly, the specic ef-fect of land use changes and implementation of other managementmeasures on catchment discharge was found to depend on their spatialdistribution and on the size and timing of storm events. For example,clear-cutting of the catchment area caused a strong increase in peak dis-charge during a 50-year storm event. Of the possible remedialmeasurestested, reforestation proved to be the most effective at reducing peakow and total runoff, particularly in smaller storms, but the magnitudeof the effectwas connected to its locationwithin the catchment. As such,reforestation near the outlet was more efcient than higher upstream.Therefore, the spatial distribution of planned land use measures mustbe considered before they can be properly implemented.

Conict of interest

There is no potential conict of interest in this paper.

Acknowledgments

This study arose from the collaboration between the Exood projectcoordinated by Bioforsk in Norway and the Adaptation of road drainagestructures to climate change project funded by the Swedish TransportAdministration through the Centre for Research and Education inOperation and Maintenance of Infrastructure (CDU). We thank SigrunKvrn for assistance with nding historical rain intensity data and forpreparing generic soil prole data.

References

Allen RG, Pereira LS, Raes D, Smith M. Crop evapotranspiration. Guidelines for computingcrop water requirements, 56. FAO irrigation and drainage paper; 1998.

Andersen J, Refsgaard JC, Jensen KH. Distributed hydrological modelling of the SenegalRiver Basin model construction and validation. J Hydrol 2001;247:20014.

Arcement Jr GJ, Schneider VR. Guide for selecting Manning's roughness coefcients fornatural channels and ood plains. U S Geol Surv Water Supply Pap 1989:2339.

Benzvi A. Toward a new rational method. J Hydraul Eng ASCE 1989;115:124155.Beven KJ. Rainfallrunoff modeling: the primer. West Sussex, England: John Wiley &

Sons; 2001 [360 pp.].Bronstert A, Bardossy A, Bismuth C, Buiteveld H, Disse M, Engel H, et al. Multi-scale

modelling of land-use change and river training effects on oods in the Rhine basin(vol 23, pg 1102, 2007). River Res Appl 2008;24:353-353.

Bultot F, Dupriez GL, Gellens D. Simulation of land-use changes and impacts on thewater-balance a case-study for Belgium. J Hydrol 1990;114:32748.

Campbell IC, Doeg TJ. Impact of timber harvesting and production on streams a review.Aust J Mar Freshwat Res 1989;40:51939.

Choi W, Deal BM. Assessing hydrological impact of potential land use change throughhydrological and land use change modeling for the Kishwaukee River basin (USA).J Environ Manage 2008;88:111930.

Chow VY. Open channel hydraulics. New York: McGraw-Hill; 1959.Cooley RL. A nite difference method for unsteady ow in variably saturated porous

media: application to a single pumping well. Water Resour Res 1971;7:160725.Deelstra J, Iital A. The use of the ashiness index as a possible indicator for nutrient loss

prediction in agricultural catchments. Boreal Environ Res 2008;13:20921.Deelstra J, Kvrn SH, Skjevdal R, Vandsemb S, Eggestad HO, Ludvigsen GH. A general

description of the Skuterud catchment. Jordforsk Report Norway: Bioforsk2005;6(05).

Deelstra J, Eggestad HO, Iital A, Jansons V. A hydrological characterisation of catchments.Bioforsk Report Norway: Bioforsk 2007;2(2).

Deelstra J, Eggestad HO, Iital A, Jansons V, Barkved LJ. Time resolution and hydrologicalcharacteristics in agricultural catchments, vol. 336. Wallingford: ROYAUME-UNI:International Association of Hydrological Sciences; 2010. p. 313.

DHI. MIKE-SHE v.5.3 user guide and technical reference manual. Denmark: Danish Hy-draulic Institute; 199850.

DHI Software. MIKE-SHE user manual volume 2: Reference guide. Horsholm, Denmark:

DHI Water & Environment; 2007.

DHI Software. MIKE SHE usermanual. Hrsholm, Denmark: DHIWater & Environment;2008.

DNR. Waters website lists area hydrologists. http://mndnr.gov/waters&http://les.dnr.state.mn.us/publications/waters/buffer_strips.pdf2011.

Dons JA. Geologisk frer for Oslo-trakten. Geology of the Oslo area; in Norwegian.Universitetsforlaget; 1977 [173 pp.].

Dworak T, Berglund M, Grandmougin B, Mattheiss V, Holen SN. International review onpayment schemes for wet buffer strips and other type of wet zones along privatelyowned land. Berlin: Ecologic Institute; 2009.

FAO. Forest and oods: drowning in ction or thriving on facts? Indonesia and Thailand:Forest Perspectives, Center for International Forestry Research; Food and AgricultureOrganization of the United Nations; 2005.

Farkas C, Beldring S, Bechmann M, Deelstra J. Soil erosion and phosphorus losses undervariable land use as simulated by the INCA-P model. Soil Use and Management2013;29:12437.

Fiener P, Auerswald K. Measurement andmodeling of concentrated runoff in grassed wa-terways. J Hydrol 2005;301:198215.

Graham DN, Butts MB. Flexible integrated watershed modeling with MIKE SHE. In: Singh

Moore RD, Wondzell SM. Physical hydrology and the effects of forest harvesting in thePacic Northwest: a review. J Am Water Resour Assoc 2005;41:76384.

Moriasi DN, Arnold JG, Van LiewMW, Bingner RL, Harmel RD, Veith TL. Model evaluationguidelines for systematic quantication of accuracy in watershed simulations. TransASABE 2007;50:885900.

Nash JE, Sutcliffe JV. River ow forecasting through conceptual models part I a discus-sion of principles. Journal of Hydrology 1970;103:28290.

Ott B, Uhlenbrook S. Quantifying the impact of land-use changes at the event andseasonal time scale using a process-oriented catchment model. Hydrol Earth SystSci 2004;8:6278.

ygarden L, Lundekvam H, Arnoldussen AH, Norway Brresen T. Soil erosion in Europe.John Wiley & Sons, Ltd.; 2006115.

Robinson M, Cognard-Plancq AL, Cosandey C, David J, Durand P, Fuhrer HW, et al. Studiesof the impact of forests on peak ows and baseows: a European perspective. ForEcol Manage 2003;186:8597.

Sahoo GB, Ray C, De Carlo EH. Calibration and validation of a physically distributed hydro-logical model, MIKE SHE, to predict streamow at high frequency in a ashy moun-tainous Hawaii stream. J Hydrol 2006;327:94109.

Schreider SY, Jakeman AJ, Letcher RA, Nathan RJ, Neal BP, Beavis SG. Detecting changes instreamow response to changes in nonclimatic catchment conditions: farm dam

754 Z. Kalantari et al. / Science of the Total Environment 466467 (2014) 741754Hoyt WG, Troxell HC. Forest and stream ow, 1858. American Society of Civil Engineers;1934. p. 1111.

Hundecha Y, Bardossy A. Modeling of the effect of land use changes on the runoff gener-ation of a river basin through parameter regionalization of a watershed model.J Hydrol 2004;292:28195.

Isik S, Kalin L, Schoonover JE, Srivastava P, Lockaby BG. Modeling effects of changing landuse/cover on daily streamow: an articial neural network and curve number basedhybrid approach. J Hydrol 2013;485:10312.

Jansson P-E, Karlberg L. Coupled heat and mass transfer model for soilplantatmospheresystems. CoupModel manual, land and water resources engineering. Stockholm:Royal Institute of Technology; 2004.

Jaramillo F, Prieto C, Lyon SW, Destouni G.Multimethod assessment of evapotranspirationshifts due to non-irrigated agricultural development in Sweden. J Hydrol 2013;484:5562.

Jones JA, Grant GE. Peak ow responses to clear-cutting and roads in small and large ba-sins, western Cascades, Oregon. Water Resour Res 1996;32:95974.

Kalantari Z, Folkeson L. Road drainage in Sweden: current practice and suggestions for ad-aptation to climate change. J Infrastruct Syst 2013;19:14756.

Kalantari Z, Jansson P-E, Stolte J, SassnerM.Modelling high resolution discharge dynamicsnearby road structure, using data from small catchment and 3 different models. Pro-ceedings, IAHR Conference, 34th Congress, Brisbane; 2011. p. 22632.

Kalantari Z, Lyon WL, Jansson P-E, Stolte J, Folkeson L, French HK, et al. Event-based cali-bration impacts on different hydrologic models for a catchment adjacent to a road inSouth-East Norway; 2013n [submitted for publication].

Kelliher FM, Leuning R, Schulze ED. Evaporation and canopy characteristics of coniferousforests and grasslands. Oecologia 1993;95:15363.

Kristensen KJ, Jensen SE. A model of estimating actual evapotranspiration from potentialevapotranspiration. Nordic Hydrol 1975;6:17088.

Kvrn SH, Stolte J. Effects of soil physical data sources on discharge and soil loss simu-lated by the LISEM model. Catena 2012;97:13749.

LindeAHT, Aerts JCJH, Kwadijk JCJ. Effectiveness ofoodmanagementmeasures onpeakdis-charges in the Rhine basin under climate change. J Flood Risk Manage 2010;3:24869.

Lyon SW, Laudon H, Seibert J, Mrth M, Tetzlaff D, Bishop K. Controls on snowmelt watermean transit times in northern boreal catchments. Hydrol Process 2010;24(12):167284. http://dx.doi.org/10.1002/hyp.7577.

Lyon SW, Nathanson M, Spans A, Grabs T, Laudon H, Temnerud J, et al. Specic dischargevariability in a boreal landscape. Water Resour Res 2012;48.

Maidment DR. Handbook of hydrology. New York: McGraw-Hill; 1993.Markart G, Kohl B, Kirnbauer R, Pirkl H, Bertle H, Stern R, et al. Surface runoff in a

torrent catchment in Middle Europe and its prevention. Geol Geotech Eng2006;24:140324.

Mathur HN, Babu R, Joshik P, Singh R. Effect of clear felling and reforestation on runoff andpeak rates in small watersheds. The Indian Forester 1976;102:21926.development in the MurrayDarling basin, Australia. J Hydrol 2002;262:8498.Seibert J, McDonnell JJ. Land-cover impacts on streamow: a change-detection modelling

approach that incorporates parameter uncertainty. Hydrol Sci J -J Des Sci Hydrol2010;55:31632.

Srbotten L-E, editor. Jord- og vannovervking i landbruket (JOVA), 6(38). Feltrapporterfra programmet i 2009. Bioforsk rapport978-82-17-00768-5; 2011. [54 pp. (InNorwegian)].

Stednick JD. Monitoring the effects of timber harvest on annual water yield. J Hydrol1996;176:7995.

Stolte J, Ritsema C, Bouma J. Developing interactive land use scenarios on the Loess Pla-teau in China, presenting risk analyses and economic impacts. Agric Ecosyst Environ2005;105:38799.

Syversen N, Bechmann M. Vegetative buffer zones as pesticide-lters for simulated sur-face runoff. Pesticide in Air, Plant, Soil & Water System 2003:58797.

Tague C, McMichael C, Hope A, Choate J, Clark R. Application of the RHESSys model toa California semiarid shrubland watershed. J Am Water Resour Assoc 2004;40:57589.

Thue-Hansen V, Grimenes AA. Meteorologiske data for s 20072008. s, Norway:Norwegian University Of Life Sciences; 2009.

van der Velde Y, Lyon SW, Destouni G. Data-driven regionalization of river discharges andemergent land coverevapotranspiration relationships across Sweden. J GeophysRes Atmos 2013;118:257687.

Vsquez RF, Feyen L, Feyen J, Refsgaard JC. Effect of grid size on effective parameters andmodel performance of the MIKE-SHE code. Hydrol Process 2002;16:35572.

Wemple BC, Jones JA. Runoff production on forest roads in a steep, mountain catchment.Water Resour Res 2003:39.

Whitaker A, Alila Y, Beckers J, Toews D. Evaluating peak ow sensitivity to clear-cutting indifferent elevation bands of a snowmelt-dominated mountainous catchment. WaterResour Res 2002:38.

Wilk J, Andersson L, Plermkamon V. Hydrological impacts of forest conversion to agricul-ture in a large river basin in northeast Thailand. Hydrol Process 2001;15:272948.

Wissmar RC, Timm RK, Logsdon MG. Effects of changing forest and impervious landcovers on discharge characteristics of watersheds. Environ Manage 2004;34:918.

Yan B, Fang NF, Zhang PC, Shi ZH. Impacts of land use change on watershed streamowand sediment yield: an assessment using hydrologic modelling and partial leastsquares regression. J Hydrol 2013;484:2637.

Yan J, Smith KR. Simulation of integrated surface water and ground water systems -Model formulation1. J Am Water Res Assoc 1994;30:87990.

Yeo IY, Gordon SI, Guldmann JM. Optimizing patterns of land use to reduce peak runoffow and nonpoint source pollution with an integrated hydrological and land-usemodel. Earth Interact 2004:8.

Zgre NP, Maxwell A, Lamont S. Characterizing streamow response of a mountaintop-mined watershed to changing land use. Appl Geogr 2013;39:515.VP, Frevert DK, editors. Watershed models. CRC Press; 2005. p. 24571.Haan CT, Bareld BJ, Hayes JC. Design hydrology and sedimentology for small catchments.

San Diego: Academic Press; 1994115.

Quantifying the hydrological impact of simulated changes in land use on peak discharge in a small catchment1. Introduction2. Materials and methods2.1. Experimental site2.1.1. Current land use2.1.2. Simulated land use scenarios

2.2. Simulation systems2.2.1. MIKE SHE2.2.2. Model parameterisation

2.3. Historical events2.4. Calibration and sensitivity analyses2.5. Validation and model evaluation

3. Results and discussion3.1. Simulated annual discharge for different land use measures3.2. Simulated peak discharge during four historical events3.3. Effect of simulated land use measures on the four historical storm events3.4. Implications for road design

4. ConclusionsConflict of interestAcknowledgmentsReferences