planning best management practices to reduce sediment

Planning Best Management Practices to Reduce Sediment Delivery from Forest Roads Using WEPP:Road Erosion Modeling and Simulated Annealing Optimization

Efta, J. A., & Chung, W. (2014). Planning Best Management Practices to Reduce Sediment Delivery from Forest Roads Using WEPP: Road Erosion Modeling and Simulated Annealing Optimization. Croatian Journal of Forest Engineering, 35(2), 167-178.

University of Zagreb

Version of Record

http://cdss.library.oregonstate.edu/sa-termsofuse

http://survey.az1.qualtrics.com/SE/?SID=SV_8Io4d9aAYR1VgGx

http://cdss.library.oregonstate.edu/sa-termsofuse

Originalscientificpaper

Croat. j. for. eng. 35(2014)2 167

Planning Best Management Practices

to Reduce Sediment Delivery from Forest Roads Using WEPP:Road Erosion Modeling

and Simulated Annealing Optimization

James A. Efta, Woodam Chung

Abstract

Planning and implementation of road BMPs on a watershed scale can be a difficult task because of the need to prioritize locations while accounting for multiple constraints, such as the avail-able budget, continuous maintenance, and equipment scheduling. Using simulated annealing (SA) as its heuristic optimizer, BMP-SA accounts for sediment being delivered to the stream network through incorporation of modeled road erosion predictions and alternative BMP op-tions and scheduling for problematic road segments. BMP-SA was applied to the Glenbrook Creek watershed in the Lake Tahoe Basin in Nevada, US. WEPP:Road predictions were used to identify road segments posing an erosion risk and appropriate BMPs were identified for problematic segments. Using BMP-SA, modeled road-related sediment leaving the forest buf-fer, thus entering streams, was minimized over the course of the planning horizon while considering budget constraints and equipment scheduling concerns. BMP-SA can be applied to any watershed but relies heavily on the perceived accuracy of road erosion predictions.

Keywords: forest roads, BMPs, simulated annealing, WEPP:Road, erosion modeling, road management, budget planning

sionriskisevaluated,notallpotentialBMPoptionsmaybeexploredatagivensite,forreasonsrangingfrominexperienceoftheprescribertobudgetlimita-tions.WhileaBMPorsetofBMPsmaybeidealforagivensite,selectionofBMPsattheone-tofew-seg-mentscalemaynoteffectivelyminimizeerosionandsedimentationatthewatershedscale.Optimizationstrategieshavebeenemployedsince

the1970stoaddressmultiplemanagementgoalsandenvironmentalconstraints in forestplanning (Rön-nqvist2003,Weintraubetal.1995,Weintraub2006).Applicationofheuristicoptimizationspecificallytoenvironmental concerns, including sedimentation as-sociatedwithroads,hasonlyoccurredmorerecentlyduetothecomplexityofsuchspatially-explicitplan-ning problems (Coulter et al. 2006,Contreras andChung2009,Weintraubetal.2000).Multipleprojectshave incorporatedBMPsand/orassociatederosion

1. IntroductionTominimize sediment-related impactsof forest

roads, BestManagement Practices (BMPs) are fre-quentlyimplementedonforestroadnetworks.WhileBMPsmayconsistofaplanningpracticeormitigationstrategy(e.g.maintainingasetbufferdistancefromastreamchannel),thetermisalsobroadlyusedinrefer-encetospecificstructuresorroadnetworkattributesthataddresssedimentationissues.Examplesinclude,butarenotlimitedto:sedimenttraps;draindips;veg-etatedorrock-linedditches,and/orroadsurfacing.FieldresearchsupportstheeffectivenessofspecificBMPsandthephysicsbehindthem(e.g.,ClintonandVose2003,FoltzandTruebe2003,LuceandBlack2001,MegahanandKetcheson1996).Inpractice,physicalBMPsaregenerallyprescribed

usingexpertjudgmentinthefield,inevitablyunderlimitedbudgetconditions.Regardlessofwhetherero-

J. A. Efta and W. Chung Planning Best Management Practices to Reduce Sediment Delivery from Forest Roads ... (167–178)

168 Croat. j. for. eng. 35(2014)2

potentialintocost-benefitanalysesusingheuristicsforroadmanagementplanning(e.g.,Arugaetal.2005,Madejetal.2006,RackleyandChung2008).Oftheseprojects,however,nonehasdirectlyaddressedBMPimplementationandmaintenanceonanexistingroadnetwork.Theresearchpresentedherefillsthisneedwitha

decisionsupporttooldesignedtoassistmanagersinformulatingwatershed-scaleroadBMPinstallationandmaintenanceplans.Thedecisionsupport tool-calledBMP-SAusessimulatedannealing(SA)asitsheuristicsolvertominimizesedimentcontributiontodownstreamwaterbodiesbyprioritizingroadBMPinstallationswhileaccountingforbudgetconstraints,maintenancerequirements,andequipmentschedul-ing concerns.

2. Study areaThestudyareaforthisprojectislocatedintheLake

TahoeBasin.Losingitsfamedclarityatarateofap-proximatelyone-quartermeterperyearforthepast25years,thelakeiscurrentlydesignatedasanimpairedwaterbodyunderSection303(d)oftheCleanWaterAct(RobertsandReuter2007).Sedimentfromtheba-sinroadnetworkhasbeen identifiedasanegativecontributor to the lakewater clarity (Murphy andKnopp2000).

TheGlenbrookCreekwatershedencompassedthemajorityofthestudyarea(Fig.1).GlenbrookCreekliesapproximately24kmwestofCarsonCity,NVand32kmnorthofSouthLakeTahoe,CAontheeastsideoftheLakeTahoeBasin.ElevationsintheGlenbrookCreekwatershedrangefromapproximately1900mto2,700m.Soilsarevolcanicandgraniticinorigin(Gris-merandHogan2004).AverageannualprecipitationattheMarletteLakeSNOTELweatherstationsite,whichliesat2,400m5.6kmnorthoftheGlenbrookwater-shedboundary,isapproximately84cm.Mostprecip-itation in the Glenbrook watershed falls as snow(Roweetal.2002).Monthlyaveragemaximumtem-peratureattheMarletteLakeSNOTELsitebetween1989and2008was11°Candmonthlyaveragemini-mumtemperaturewas–1°C.

3. Materials and methods

3.1 Problem formulationSimulatedannealing,developedbyKirkpatrickand

others(1983),usesamodifiedMonteCarlosimulationthatisanalogoustoametalcooling,orannealing,afterleavingaforge.Initialtemperatureandcoolingratearevariableswhichcontrolthenumberofiterationsandrangeofacceptablesolutionvalues.Thisoptimizationtechniqueiswellsuitedtothisproblemtypeformul-tiplereasons:1)Itcanreadilybescaledtolargeand

Fig. 1 Map of Glenbrook Creek watershed vicinity, Nevada USA

Planning Best Management Practices to Reduce Sediment Delivery from Forest Roads ... (167–178) J. A. Efta and W. Chung

Croat. j. for. eng. 35(2014)2 169

smalldatasets;and2)itisarelativelysimpleyeteffi-cientheuristic solution technique forcombinatorialoptimizationproblems(Kirkpatricetal.1983,TarpandHelles1997).Duringcomparisonofneighbourhoodsolutions,ifthecurrentsolutionissuperiortothealter-native,anacceptanceprobabilityiscalculatedinordertodeterminewhether thealternative shouldbeac-cepteddespite its inferiority. This uniqueheuristiccomponentislinkedtothetemperaturevariable:

( )current new

tempp new e−

= (1)

Wherep(new)istheprobabilityofacceptingthenewsolution,currentistheobjectivefunctionvalueofthe current solution, new is the objective functionvalueofthenewsolution,andtempisthevalueofthetemperaturevariableatthetimeofcomparison.Whencomparedagainstarandomlygeneratedvalue,thesolutionmaybeacceptedasthe»new«currentsolu-tion,fromwhichasubsequentneighborhoodsolutionwillbeformulated.Indoingso,anear-optimalsolu-tionmaybereachedfasterthanifsolutionswerefor-mulatedrandomly,sincethe»worse«solutionmayprovideabridgetoasuperiorsolutionmorequickly.Astemperaturedecreases(moreiterationsarerun),sotoowilltheprobabilityofacceptanceofaninferiorsolution,therebyreducingsolutionvariability(Fig.2).Toapplythisheuristicframeworktotheissueat

hand,theplanningproblemcanbeformulatedasaminimizationproblemwiththeamountofsedimententeringstreamsastheobjectivefunction(Equations2and3,Fig.3).Theformulationbelowassumesthat

sedimentisdiscountedovertimeinordertopromoteearlyBMPinstallationandsubsequentsedimentre-duction.Adiscountrateoffourpercentwasusedasitisstandardpracticeinnaturalresourceeconomicanal-ysisinvolvingU.S.ForestServiceinvestments(Rowetal.1981).Minimize

Zse ent

i

N

j

H

= −==∑∑

dim i,jj1 04 1

11 . ( ) (2)

Subjectto

cost budgeti

N

i,j j=∑ ≤

1j H∈ (3)

WhereZ Totalsedimentleavingthebufferthrough

thecourseoftheplanninghorizonj PlanningperiodH Totalnumberofplanningperiodsi SegmentnumberN Totalnumberofsegmentsontheroadnet-

worksedimen ti,j Sedimententeringthenearestwater-

wayfromsegmentiduringplanningpe-riod j

costi,j Cost ofBMP treatmentormaintenancescheduledforsegment i inplanningpe-riod j.

budgetj Budgetforplanningperiodj 1.04(j–1) Discountterm

Fig. 2 Changes in objective function value of the current solution during the simulated annealing optimization process


170 Croat. j. for. eng. 35(2014)2

Fig. 3 Flowchart describing adapted simulated annealing optimization


Croat. j. for. eng. 35(2014)2 171

Estimatesofsedimententeringthestreamnetworkfromeachroadsegmentof interestmustbeestab-lishedpriortoexecutionoftheheuristic.Currently,thereareasuiteofempirically-basedandprocess-basederosionmodelsbeingusedforestimatingroaderosion,eachwithvaryingdegreesofuseraccessibil-ity (Fu et al. 2010).Among these,WEPP:RoadhasgainedtractionamongUSlandmanagementagenciesincluding theLakeTahoeBasinManagementUnit(LTBMU) and continues to gain widespread useamongacademic andnonprofitorganizations (e.g.

Briebartetal.2007,Contrerasetal.2008,Inlanderetal.2007).WEPP:Road,auser-friendlyinterfacetotheWaterErosionPredictionProject(WEPP)Model(Fla-naganandNearing1995),providesawebbrowser-basedinterfacethatrequiresfewinputparameters.Inthisstudy,weusedWEPP:Roadtoassesspoten-

tialerosionriskandthreatofsedimentationfromeachroadsegmentinawatershedaswellasappropriateBMPstoaddresstheerosionriskfactor(s)forthatseg-ment.WEPP:Roadestimatesrunoffandsoillossonthreeoverlandflowelements:theroadbeditself,afillslope,andthebuffer(hillslopeareabetweenthebaseofthefillslopeandthenearestwatersource)(Elliotetal. 1999). Four soil types can be modeled byWEPP:Road,alongwithfourroaddesigns,threeroadsurfacetypes,andthreetrafficlevels(Table1).Avari-etyofotherparametersarealsorequiredbythemod-el,someofwhicharebestgatheredinthefieldandsomeofwhicharebestcollectedusingaGISorotherdata sources.

3.2 WEPP:Road data preparationAtotalof173roadsegmentswereidentifiedon

12.5kmofroad.Segmentsweredelineatedbetweentwoexistingdrainagestructures,fromaslopebreakorhighpointtoadrainagestructure,fromahighpointtoalowpoint,orbetweenadrainagestructureandalowpoint.WEPP:Roadinputparametersweredeter-minedormeasuredthroughacombinationoffielddatacollectionandgeoprocessingusingdatasetsac-quiredfromtheLakeTahoeGISDataClearinghouse(http://tahoe.usgs.gov/).»From«nodescomprisedtheentranceorbeginningsegmentlocationsforrunoffandsedimententrainment.»To«nodesweredeliverypoints,ortheperceivedsegmentoutletforrunoffandsediment.Analysisofcoarserockcontentandsoiltex-turingwereperformedonsoiladjacenttotheroadgradeitself.AsWEPP:Roadonlyacceptsoneoffoursoiltextures(Table1),soiltexturesevaluatedinthefieldwerematchedascloselyaspossible tooneofthose four textures available in the standardWEPP:Roadinterface.TheTahoeCASNOTELsite,theclosestavailable

long-termclimatestationwasusedastheclimatein-put forWEPP:Road.While themodel is running,WEPP:RoadusestheCLIGENweathergeneratortostochasticallygeneratedailyclimatedataforthede-siredsimulationtime(Elliotetal.1999).Thirtyyearsofdailyclimatedataweregeneratedforthesesimula-tions.PerWEPP:RoadDocumentation,thirtyyearsofsimulationisgenerallyadequateforobtainingreason-ableerosionestimates(Elliotetal.1999).Roadtrafficlevelwasheldconstantat»low«forallsegments(Brie-bartetal.2007).

Table 1 WEPP:Road input parameters and possible values or pa-rameter ranges

WEPP:Road input parameter

Possible values/ allowable range

Climate N/A

Soil type

Silt loam

Sandy loam

Clay loam

Loam

Road design

Insloped, bare ditch

Insloped, vegetated or rocked ditch

Outsloped, unrutted

Outsloped, rutted

Surface type

Native

Graveled

Paved

Traffic level

None

Low

High

Road width 1 ft – 300 ft

Road length 1 ft – 999 ft

Road gradient 3% – 99%

Fill slope length 1 ft – 999 ft

Fill slope gradient 3% – 99%

Buffer length 1 ft – 999 ft

Buffer gradient 3% – 99%

Coarse rock content 0% – 100%

Years of simulation time 1 yr – 200 yrs


172 Croat. j. for. eng. 35(2014)2

EachroadsegmentwasmodeledmultipletimesusingWEPP:Road,firsttosimulateerosionunderex-isting conditions thenunder the range of possibleBMPsthatwerethenincorporatedintoBMP-SAmod-elinputs.

3.3 BMP-SA model input formulationDuringafieldvisitwithaLakeTahoeBasinroad

engineer,site-specificBMPoptionswereprescribedforthosesegmentspredictedtoyieldthemostsedi-ment.Fromthisvisit,alongwithpersonalcommuni-cationwithotherengineers,guidelineswereestab-lishedforinstallingBMPsonthosesitesnotvisitedinthefield(Table2)(CatherineSchoenandPaulPotts,pers.comm.,USFSLakeTahoeBasinManagementUnit,July2008).BMPinstallationcosts,maintenancecosts,andmaintenancefrequenciesassociatedwithagivenBMPwerealsoobtainedthroughpersonalcom-munication(PaulPotts,pers.comm.,USFSLakeTahoeBasinManagementUnit,July2008)alongwiththeRe-gionFourCostEstimatingGuideforRoadConstruc-tion(Table3;USDAFS2009).ForeachpotentialBMPscenarioonagivensegment,WEPP:RoadwasusedtopredicttheeffectivenessofeachpotentialBMPinstal-lation.UptofourBMPoptionswereassignedtoeachsegment, including no treatment.To account for equipment scheduling costs (in

doingsofavoringsolutionswhereBMPsareinstalledincloseproximityinthesametimeperiod),acluster-ingsubroutinewasdevelopedasapartofthemodel.

Table 2 Priority of BMP assignment for a given road segment

Condition Priority 1 Priority 2 Priority 3

Buffer slope > fill slope Outslope* Drain dips† Pave‡

Road slope > 17% Pave‡ Drain dips† Outslope*

Notes: *Delivery point reassigned to center of road segment. Not applicable on paved or graveled segments. †Drain dips applicable on any segment greater than 46 meters (150 feet) in length. Segment length iteratively divided in half until segment length is less than 46 meters or sediment leaving buffer is zero. Not applicable on outsloped segments. ‡ If paving is already installed on a segment, no further BMPs can be installed

Table 3 Installation costs, maintenance costs, and maintenance frequencies associated with assigned BMPs

BMPInstallation cost

Equipment move-in costs

Maintenance costEquipment move-in

costs for maintenanceMaintenance

frequency

$ $ $ $ yrs

Outsloping 1,865/km 1,000 622/km 500 3

Drain dips 100/each 500 100/each 500 5

Pavement 15,2269/km 1,500 9,323/km 500 7

Fortestingpurposes,clusterdiameterwasfixedat305 meters. If the same type of treatments were scheduledonroadsegmentsthatarelocatedwithinthe cluster diameter, equipmentmove-in costwascountedonlyonceforthosetreatments.Planninghorizonwasassumedtobe20yearswith

aplanningperiodofoneyear.Threebudgetexpendi-turescenariosforBMPimplementationandmainte-nanceweremodeled:agivenannualbudgetisusedfor1)newBMPinstallationonly,2)newBMPinstalla-tionandmaintenance,and3)existingBMPmainte-nancealongwithnewBMPinstallationandmainte-nance.Inallmodelingscenarios,BMPswereassumedtobemaintainedinperpetuityattheirassignedfre-quencies (Table 3).When included, existingBMPswereassumedtostarttheirmaintenancecycleinpe-riodone.Eachscenariowasmodeledatmultipleinitialbudgetstoassessmodelbehaviorunderdifferentbud-getlevels.Ateachlevel,itwasassumedthatthean-nualbudgetwasconstantthroughouttheplanninghorizonandunspentbudgetwasnotcarriedoverintofutureyears.

4. Results and Discussion

4.1 WEPP:Road resultsOfthe173segmentsanalyzedinthestudyarea,74

ofthem(accountingfor6.7km,or53percent,ofroadsinthestudyarea)werepredictedtoproducesedimentleavingthebufferoverthe30-yearmodelingperiod.


Croat. j. for. eng. 35(2014)2 173

Per-segmentsedimentoutputsrangedfrom0Mgyr–1 to0.6Mgyr–1withameanoflessthan0.1Mgyr–1. WEPP:Roadpredictedatotalof49.9Mgyr–1ofsedi-mentleavingtheroadand2.7Mgyr–1 sediment leav-ingthebufferfromthestudyarea(Table4).Thewell-maintainedexistingBMPinfrastructure

onGlenbrookCreekforestroads,aswellasthewater-shed dry climate, partially explains the minimalamountofsedimentpredictedtobeleavingthebuffer.Predictedaverageerosionratefromnativesurface

roadsinsandyloamsoils-thepredominantsoiltex-turefoundwithintheGlenbrookCreekwatershed-was8.1Mgha-1 yr-1.Incomparisontoregionalem-piricalvalues,onthewestslopeoftheSierraNevadarangeCoe(2006)observedanerosionrateof8.1Mgha-1 yr-1onnativesurfaceroadsduringonewetseason(OctoberthroughJune)ofdatacollection.Annualpre-cipitationduringthiswetseasonwasnearthelong-termaverageof1,300mm.Coe’sstudysegmentswereinprimarilyloamsoils.Averageroadgradients,seg-mentlengths,andparentmaterialswerecomparableforbothstudies.

4.2 Alternative BMP assignmentOfthe74roadsegmentsproducinggreaterthan

zerosedimentleavingthebufferperyear,38wereas-signedtreatments.Thirty-sixsegmentscouldnotbeassignedBMPsbecausetheywereeitherpaved(as-sumedtobean»endpoint«BMP)orhadsomecom-binationofconditionswhichpreventedassignmentofBMPtreatments.Forexample,outslopingwasnotcon-sideredanappropriateBMPforgraveledsegmentsanddraindipswerenotappliedtosegmentsalreadyoutsloped.TheeffectivenessofBMPs,definedinthisstudyas

predictedreductioninsedimentdelivery,variedde-pendingonthecharacteristicsoftheroadsegmenttowhichaBMPwasassigned.Draindipsshowedanexponentialincreaseineffectivenessassegmentsweredivided into two(onedraindip), four (threedraindips),andeight(sevendraindips),respectively(Table5).Outslopingwasmodeledasbeingthreetimesmore

effectiveatreducingsedimentthanonedraindip,butanorderofmagnitudelesseffectivethanthreedraindipsandtwoordersofmagnitudelesseffectivethansevendraindips.Therewerenoinstanceswherepav-ingwaschosenasanapplicablenewBMP.IneveryinstancewherepavementwasapotentialBMPoption,WEPP:Roadmodeloutputsindicatedthatpavementincreasedsedimentleavingthebufferaboveexistinglevelspotentiallyduetoincreasedrunoffexacerbatingditchandbuffererosion(Table5).OtherresearchershavehadsimilarresultswhenapplyingWEPP:RoadtopavedroadsegmentsintheLakeTahoeBasin(Brie-bartetal.2007).

4.3 New BMP installation onlyInthismodelingscenario,itwasassumedthatthe

annualbudgetcanbeusedfornewBMPinstallationonly.Theresultsshowthatsedimentleavingthebufferproducedanegativeexponentialtrendwithincreas-ingannualbudget(Fig.4).Theminimumsedimentdeliverysolutionwasachievedwhentheannualbud-getreached$20,000.Inthissolution,all38segmentshadBMPsappliedtotheminperiodone(Fig.5).Dis-countedsedimentwasreducedfrom37.8Mgwhennotreatmentwasappliedoverthe20yearplanninghori-zonto10.4Mg.Increasingbudgetbeyondthislevelyieldednoreductioninsedimentleavingthebuffer.Proportionofsegmentswithoutsloping,chosenas

anappropriateBMP,alsoincreasedwiththebudget.Inseveralinstances,solutionswithtwodifferentbud-getsyieldeddecreasesinsedimentleavingthebufferwhilehaving the samenumberor fewerBMPs in-stalledinperiodone.Inalloftheseinstances,thenum-berofsegments,whereoutslopingwasinstalledasaBMP,wasgreaterinthesolutionproducinglesssedi-mentleavingthebuffer.Outslopingisahighlyeffec-tiveBMPforreducingsedimentleavingtheroadandbufferbutalsotendstobemoreexpensivethansingledraindips (LuceandBlack2001,Elliot et al. 2009,USDAFS2009).Inaddition,therearelimitationsforwhereandwhenoutslopingmaybeanapplicableBMP.Fig. 6 shows thenumberof segmentswhereBMPswereinstalledineachperiodforallthreemod-

Table 4 Predicted sediment leaving road and sediment leaving buffer in Mg yr–1 and Mg ha–1 yr–1 across the study area, Glenbrook Creek, NV. Average road width across each area of interest was used to calculate Mg ha–1 yr–1 values

Study areaSediment leaving road Sediment leaving buffer

Mg yr–1 Mg ha–1 yr–1 Mg yr–1 Mg ha–1 yr–1

Entire study area 49.9 13.6 2.7 0.7

Glenbrook watershed 19.0 7.0 1.4 0.5


174 Croat. j. for. eng. 35(2014)2

elingscenarios.At$3,000,allBMPswereinstalledinthefirstsevenperiodsunderthenewBMPinstallationonly scenario.

4.4 New BMP installation and maintenanceWithmaintenancecostsincorporatedintothemod-

el,agreaterinitialbudgetwasrequiredtoachievethesamereductioninsedimentasthatfoundunderthe

newBMPinstallationonlyscenario.Atabudgetlevelof$6,000,themodelfailedtoproducefeasiblesolu-tionsbecausenopossiblecombinationofBMPsex-istedbelowthisbudgetthreshold.Minimumsedimentleavingthebufferthroughthe

courseoftheplanninghorizonwas10.4Mg,thesamesolutionasthatfoundinthepreviouslymodeledsce-nario.At$6,000annualbudget,twosegmentshadno

Fig. 4 Sediment leaving buffer through the 20 year planning horizon at various budgets under three modeling scenarios

Fig. 5 Number and type of BMP installed in period one at varying initial budget per period resulting from the new BMP installation only sce-nario


Croat. j. for. eng. 35(2014)2 175

Fig. 6 With varying initial budget, number of periods required by BMP-SA to install all new BMPs. Black bars represent the new BMP instal-lation only scenario, white bars represent the new BMP installation and maintenance scenario, and gray bars represent the existing BMP maintenance, new BMP installation, and new BMP maintenance scenario


176 Croat. j. for. eng. 35(2014)2

treatmentchosenasthebestpossibleoption.ThisresultindicatesthatthebudgetwassolimitedthatneitherBMPinstallationnormaintenancewasfeasibleforthesetwosegments.ThenumberandtypesofBMPsinstalledinperiodoneatvaryingbudgetsforthenewBMPin-stallationandmaintenancescenariowasverysimilartothatseenwiththenewBMPinstallationonlyscenario.

4.5 Existing BMP maintenance and new BMP installation and maintenance The lowest sediment delivery solution was

achievedwithahigherannualbudgetthantheprevi-oustwomodeledscenarios(Fig.4).MaintenanceofexistingBMPsrequiredapproximately$35,000mini-mumannualbudget.Period13requiredthegreatestannualbudgetasaresultofnumerouspreexistingBMPshavingmaintenancefrequenciesoftwo,three,orfouryears.Asaresult,anynewBMPswithathree-year maintenance frequency (such as outsloping)couldnotbeinstalledinperiodoneuntilthebudgetwasincreasedbeyondthisminimumlevel.Abudgetof$57,000wasnecessaryforallBMPsto

beinstalledinperiodone.Asaresultofaccountingformaintenance costs, the solution becamemore con-strained,making thisBMP installationscenario lessvariablethanthepreviousscenario.Ingeneral,thenum-berofperiodsrequiredtoinstallBMPsonallsegmentsdecreasedwiththeincreaseinthebudget(Fig.6).

4.6 Discussion of BMP-SA modeling resultsOptimizedsolutionsbyBMP-SAfordifferentbud-

getsshowthatthemodelwasabletoproducecost-efficientBMP locations, types and implementationperiodsinreducingsedimentdeliveryunderlimitedbudgets.HighcostefficiencyofBMPswasrealizedatlowbudgetlevels,butincreasesinthebudgetyieldeddiminishingreturnsinsedimentreduction(Fig.4).Whensolutionsareconstrained(asinthesetypes

ofscenarios),BMP-SAhasthepotentialtoprovidethe

mostbenefit.ABMPmaybechosenthatisnotneces-sarilythebestforagivenlocationduetothebudgetconstraintbutservestomaximizesedimentsavingsacrosstheareaofinterest.Thishighlightstheimpor-tanceoftheclusteringsubroutineinBMP-SA.Whileasensitivityanalysisofclustersizeisbeyondthescopeofthisstudy,itprovidesonefutureavenueofresearchusingthistool.Discountrateplaysaroleindictatingthebudget

atwhichBMPscanbemostcost-effectively imple-mented(indicatedbyaninflectionpointinbudget-sedimentreductionplotsforthethreemodelscenariosinFig.4).Ifdiscountratewereincreased,aninflectionpointwouldalsobereachedmorequicklybutwouldrequireagreaterinitialbudget.Conversely,alowerdiscountratewouldresultinalessrapidreductioninsedimentleavingthebuffer.Themaximumpossiblesedimentreductionwas

achievedwithallthreemodelingscenarios,albeitathigherbudgetlevelswhenmaintenanceofnewandexistingBMPswasaccounted forwithin thegivenbudget.Underthebestpossiblesolution,sedimentleavingthebufferwasreducedby72%ifcomparedtobuffersedimentoutputsshouldnotreatmentsbein-stalledthroughthecourseoftheplanninghorizon.Withrespecttothe38treatedsegments,sedimentwasreducedbynearly93%ifcomparedtobuffersedimentoutputswithnotreatment.Estimatedsedimentsav-ingswithnewBMPinstallationwasconsiderable,inpartbecauselittlesedimentwaspredictedtoleavethebufferfromtheGlenbrookCreekwatershed.Ifpre-dictedsedimentleavingthebufferwasgreater,per-centdecreaseinsedimentleavingthebufferasaresultofBMPinstallationcouldbelower.Inawatershedwithoutawell-developedBMPinfrastructure,how-ever, therewouldbemorepotential fora tool likeBMP-SAthatcanassistfieldmanagersinprescribingeffectiveBMPstominimizesedimentleavingthebuf-fer,especiallywhenthebudgetisconstrained.

Table 5 Predicted effectiveness of BMPs (in terms of sediment savings) assigned to problematic road segments. Average road width across all segments with the same BMP assignment was used to calculate erosion rates

BMP Number assignedMinimum effectiveness Maximum effectiveness Mean effectiveness

Mg ha–1 yr–1 Mg ha–1 yr–1 Mg ha–1 yr–1

One drain dip 33 0.00 0.11 0.02

Three drain dips 7 0.02 1.51 0.53

Seven drain dips 3 0.77 2.56 1.65

Outslope segment 20 0.00 0.59 0.06

Pave segment 5 -1.37 -0.01 -0.52


Croat. j. for. eng. 35(2014)2 177

5. ConclusionsThisresearchdevelopedadecisionsupporttool

designedtoincreasetheefficiencyofBMPplanningonaforestroadnetwork.Modeledroad-relatedsedi-mentleavingtheforestbufferwasminimizedoverthecourseofaplanninghorizonwhileaccounting forbudgetconstraintsaswellasequipmentschedulingconsiderations. The solutions presented here usedmodeledWEPP:Road erosion estimates aswell asguidelines forprioritizing appropriateBMPs for agivenroadsegment.Tominimizesedimentleavingthebuffer,thesedatawereinputintoamodelusinganadaptedsimulatedannealingoptimizationalgo-rithm.Underlimitedbudgets,themodelwasabletoprioritizeBMPplacementsandtypesthroughatrade-offanalysisbetweencostsandeffectivenessofBMPs.WhilethedatausedhereisfromtheLakeTahoe

Basin,BMP-SAcanbeappliedtoanywatershed.Themodelisalsoapplicableatascalegreaterthanasinglewatershedandcanbeeasilymodifiedtoaccommodatenon-linearspatialconstraints,suchasschedulingofequipmentandBMPmaintenance,thoughproblemcomplexitymaysubstantiallyincreaseifmoreBMPlocationsandoptionsexist.TherearetwocriticalassumptionsimplicitinBMP-

SA.OneisthatBMPsmustbemaintainedatappropri-ate intervals inperpetuity, otherwisemoney spentinstallingBMPsisnotworthwhile.Inaddition,thismodelingprocessreliesontheaccuracyofroadsedi-mentation prediction for determining problematicroadsegmentsandtheeffectsofBMPinstallationonsedimentsavings.BMPimplementationisoftensite-specificinnature.Forthatreason,someroadsegmentsmaynotbeabletobetreatedusingoneofonlyahand-fulofgenericBMPs;onlyprofessionaljudgmentinthefieldmayprovidetheidealoptioninsuchsituations.

AcknowledgementsWewouldliketothankBillElliot,RandyFoltz,and

JunRheeattheRockyMountainResearchStationinMoscow,ID,fortheircooperationandinsightthroughthecourseofthiscollaboration.PaulPotts,CatherineSchoen,andJimHarrisoftheUSFSLakeTahoeBasinManagementUnitalsohaveourthanksfortheirtimeonthephone,inmeetings,andinpersonthroughthedurationofthisproject.ThisstudywasfundedbytheTahoeScienceProgramandadministeredbytheUSDAForestServicePacificSouthwestResearchStationincooperationwiththeRockyMountainResearchStation(ProjectNo.P010)alongwithsignificantcostmatchandin-kindcontributionfromTheUniversityofMontana.

6. ReferencesAruga,K.,Sessions,J.,Miyata,E.S.,2005:Forestroaddesignwithsoilsedimentevaluationusingahigh-resolutionDEM.JournalofForestResearch10:471–479.Briebart,A.,Harris,J.,Norman,S.,2007:ForestRoadBMPUpgradeMonitoringReport2003-2005.USDAForestService,Lake Tahoe BasinManagementUnit.Available online athttp://www.fs.usda.gov/Internet/FSE_DOCUMENTS/fsm9_045815.pdf;lastaccessedAug.8,2012.Clinton,B.D.,Vose,J.M.,2004:DifferencesinSurfaceWaterQualityDrainingFourRoadSurfaceTypesintheSouthernAppalachians.SouthernJournalofAppliedForestry27:100–106.Coe,D.,2006:SedimentproductionanddeliveryfromforestroadsintheSierraNevada,California.M.Sc.thesis,ColoradoStateUniversity,FortCollins,CO.,110p.Coulter,E.D.,Sessions,J.,Wing,M.G.,2006:Schedulingfor-estroadmaintenanceusingtheanalytichierarchyprocessandheuristics.SilvaFennica40(1):143–160.Contreras,M.A.,Chung,W.,Jones,G.,2008:Applyingantcolonyoptimizationmetaheuristictosolveforesttransporta-tionplanningproblemswithsideconstraints.CanadianJour-nalofForestResearch38:2896–2910.Contreras,M.A.,Chung,W.,2009:Designingskid-trailnet-workstominimizeskiddingcostandsoildisturbances.InProc.ofthe32ndAnnualMeetingoftheCouncilonForestEngineering,June15–18,2009,KingsBeach,CA.Elliot,W. J.,Hall,D.E.,Scheele,D.L.,1999:WEPP:Road:WEPPinterfaceforpredictingforestroadrunoff,erosionandsedimentdelivery.Availableonlineathttp://forest.moscowfsl.wsu.edu/fswepp/docs/wepproaddoc.html;lastaccessedJun.2,2012.Elliot,W.E.,Foltz,R.B.,Robichaud,P.R.,2009:Recentfind-ingsrelatedtomeasuringandmodelingforestroaderosion.InProc.ofthe18thWorldIMACS/MODSIMCongressonInternationalCongressonModellingandSimulation,Ander-ssen,R.S.,R.D.Braddock,L.T.H.Newham(eds.).Cairns,Aus-tralia,July13–17,2009.Flanagan,D.C.,Nearing,M.A.(eds.),1995:USDA-waterero-sionpredictionproject:hillslopeprofileandwatershedmod-eldocumentation.NSERLReport#10,USDA-ARSNationalSoilErosionResearchLaboratory,WestLafayette,Indiana,298p.Foltz,R.B.,Truebe,M.A.,2003:LocallyAvailableAggregateandSedimentProduction.TransportationResearchRecord1819,PaperNo.LVR8-1050.Fu,B.,Newham,L.T.H.,Ramos-Scharron,C.,2010:Areviewofsurfaceerosionandsedimentdeliverymodelsforunsealedroads.EnvironmentalModellingandSoftware25:1–14.Grismer,M.E.,Hogan,M.P.,2004:SimulatedRainfallEvalu-ationofRevegetation/MulchErosionControlintheLakeTa-hoeBasin–1:MethodAssessment.LandDegradationandDevelopment15:573–588.Inlander,E.,Clingenpeel,A.,Crump,M.A.,VanEpps,M.,Formica,S.,2007:InventoryandSedimentModelingofUn-


178 Croat. j. for. eng. 35(2014)2

pavedRoadsforStreamConservationPlanning.P.156–165inProc.ofthe2007InternationalConferenceonEcologyandTransportation,LeroyIrwin,C.,D.Nelson,andK.P.McDer-mott.(eds.).CenterforTransportationandtheEnvironment,NorthCarolinaStateUniversity,Raleigh,NC.Kirkpatrick,S.,Gelatt,C.D.,Vecchi,M.P.,1983:OptimizationbySimulatedAnnealing.Science220:671–680.Luce,C.H.,Black,T.A.,2001:SpatialandTemporalPatternsinErosionfromForestRoads.P.165–178inInfluenceofUrbanandForestLandUsesontheHydrologic-GeomorphicRe-sponsesofWatersheds,Wigmosta,M.S.andS. J.Burges.(eds.).WaterResourcesMonographs,AmericanGeophysicalUnion,Washington,D.C.Madej,M.,Eschenbach,E.A.,Diaz,C.,Teasley,R.,Baker,K.,2006:Optimizationstrategiesforsedimentreductionprac-ticesonroadsinsteepforestedterrain.EarthSurfacePro-cessesandLandforms31:1643–1656.Megahan,W.F.,Ketcheson,G.L.,1996:PredictingDownslopeTravelofGraniticSedimentsfromForestRoadsinIdaho.Wa-terResourcesBulletin32:371–382.Murphy,D.D.,Knopp,C.M.,2000:LakeTahoeWatershedAssessment:VolumeI.USDAFor.Serv.Gen.Tech.Rep.PSW-GTR-175,753p.Rackley,J.,Chung,W.,2008:Incorporatingforestroaderosionintoforestresourcetransportationplanning:acasestudyintheMicaCreekwatershedinNorthernIdaho.TransactionsoftheASABE51:115–127.Roberts,D.M.,Reuter, J.E.,2007:DraftLakeTahoeTotalMaximumDailyLoadTechnicalReportCaliforniaandNe-vada.Availableonlineathttp://terc.ucdavis.edu/publications/

LakeTahoeTMDLTechnicalReport.pdf;lastaccessedAug.8,2012.Rönnqvist,M.,2003:OptimizationinForestry.MathematicalProgramming97:267–284.Row,C.,Kaiser,H.F.,Sessions,J.,1981:DiscountRateforLong-TermForestServiceInvestments.JournalofForestry79:367–369.Rowe,T.G.,Saleh,D.K.,Watkins,S.A.,Kratzer,C.R.,2002:StreamflowandWater-QualityDataforSelectedWatershedsintheLakeTahoeBasin,CaliforniaandNevada,throughSep-tember1998.U.S.GeologicalSurveyWater-ResourcesInves-tigationsReport02-4030,118p.Tarp,P.,Helles,F.,1997:SpatialOptimizationbySimulatedAnnealingandLinearProgramming.ScandinavianJournalofForestResearch12:390–402.USDAForestService,2009:CostEstimatingGuideforRoadConstruction.Availableonlineathttp://www.fs.usda.gov/Internet/FSE_DOCUMENTS/fsbdev3_015406.pdf; last ac-cessedJuly5,2012.Weintraub,A.,2006:Integerprogramminginforestry.AnnalsofOperationsResearch149:209–216.Weintraub,A.P.,Church,R.L.,Murray,A.T.,Guignard,M.,2000:Forestmanagementmodelsandcombinatorialalgo-rithms: analysis of state of the art. Annals of Operations Re-search96:271–285.Weintraub,A.,Jones,G.,Meacham,M.,Magendzo,A.,Mal-chauk,D.,1995:Heuristicproceduresforsolvingmixed-inte-gerharvestscheduling-transportationplanningmodels.Ca-nadianJournalofForestResearch25:1618–1626.

Received:March4,2014Accepted:July11,2014

Authors’addresses:

JamesA.Efta,ForestHydrologist*e-mail:jefta@fs.fed.usUSDAForestService-CusterNationalForest.1310MainStreetBillingsMT59102USA

*FormerlyGraduateResearchAssistantTheUniversityofMontana32CampusDriveMissoulaMT59812USA

Assoc.Prof.WoodamChung,PhD.e-mail:[email protected] ResourcesandManagementOregonStateUniversity267PeavyHallCorvallis,OR97331-5703USA

*Correspondingauthor

planning best management practices to reduce sediment

Documents