to colorado parks and wildlife commission members, · to colorado parks and wildlife commission...

10
Open letter from scientist and scholars on the proposed CPW mule deer strategy studies – 30 November 2016 1 To Colorado Parks and Wildlife Commission members, The undersigned address the two Colorado Parks and Wildlife (CPW) proposed mule deer strategy studies in the Piceance Basin i and Upper Arkansas River ii . We are concerned about the lack of “gold standard” for scientific inference, which is the “random assignment to control and treatment groups with experimental designs that avoid biases in sampling, treatment, measurement, or reporting,” [1]. Neither of the proposed studies meets the gold standard, because (1) of a lack of proper control (zero cougar killing); (2) there is a risk of selection bias when treatments are not assigned randomly; and (3) the sample size is too small to make robust inferences. These factors preclude scientific conclusions from the studies. Moreover, the designs also raise (4) legal and ethical concerns. While we understand that good experimental design is difficult, we also want to emphasize that poor design invalidates the conclusions and wastes taxpayer resources. The references to the following fundamentals of research design are provided below our summary explanations. 1. Lack of proper control (a) A proper set of controls and treatments would require at least 3 of each to achieve statistical robustness. Furthermore, the control sites must not experience cougar killing (legal and illegal take) and must experience every intrusion except cougars dying, e.g., the same number and intensity of intrusions as in treatment sites but no cougars killed. The current plan to allow ±10% cougar harvest in control areas is indefensible scientifically. Currently the design is flawed, just as if it were a biomedical clinical trial in which the researchers said, “experts don’t know what effect this pill will have, so the control will be a low-dose and in the treatment will be a high dose.” (b) The Piceance Basin study, which involves killing black bears and mountain lions on one parcel of land and then comparing that to an area with no predator control from 2010-2012, is called a pseudo-control or false control. The other area was studied at a different time and place under very different conditions than today. (c) Under the current design, each spatial unit is a single replicate. Events within a unit are not independent of other events within that unit. A more robust design would reverse-treatment within each unit, which receives a treatment by random-assignment, not by researcher selection of sites for treatments. Although the Arkansas River study looks more robust, it remains a sample size of 4 and the lack of a true control will make the results impossible to interpret scientifically. 2. Selection bias The Upper Arkansas River proposal states, “Deer data analysis unit (DAU) D-16 (Figure 1) was identified as an area where cougar suppression could be beneficial to the deer population.” iii This subjective decision will invalidate the scientific value of the proposed study in a single step. When treatments are assigned according to the response variable that one wishes to measure, you have guaranteed a sampling bias that would invalidate the study. Remember, a treatment is a hypothesized solution. If one designs a study with the assumption that the solution will work, one risks intentional bias in measurement and reporting. Random assignment is far easier and more robust to these biases and protects the researcher from claims of intentional bias. 3. Small sample size With fewer than 6 study units (3 control and 3 treatment), there is no statistical test that can reliably confirm or reject the research hypothesis. That requirement for 6 or more arises because each unit is a

Upload: others

Post on 22-May-2020

4 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: To Colorado Parks and Wildlife Commission members, · To Colorado Parks and Wildlife Commission members, The undersigned address the two Colorado Parks and Wildlife (CPW) proposed

OpenletterfromscientistandscholarsontheproposedCPWmuledeerstrategystudies–30November2016

1

ToColoradoParksandWildlifeCommissionmembers,The undersigned address the two Colorado Parks and Wildlife (CPW) proposed mule deer strategystudies in the Piceance Basini and Upper Arkansas Riverii.We are concerned about the lack of “goldstandard” for scientific inference,which is the “random assignment to control and treatment groupswith experimental designs that avoid biases in sampling, treatment,measurement, or reporting,” [1].Neitheroftheproposedstudiesmeetsthegoldstandard,because(1)ofalackofpropercontrol(zerocougarkilling);(2)thereisariskofselectionbiaswhentreatmentsarenotassignedrandomly;and(3)the sample size is too small tomake robust inferences. These factors preclude scientific conclusionsfromthestudies.Moreover,thedesignsalsoraise(4)legalandethicalconcerns.Whileweunderstandthatgoodexperimentaldesign isdifficult,wealsowant toemphasize thatpoordesign invalidates the conclusions and wastes taxpayer resources. The references to the followingfundamentalsofresearchdesignareprovidedbelowoursummaryexplanations.1. Lackofpropercontrol(a) A proper set of controls and treatments would require at least 3 of each to achieve statisticalrobustness.Furthermore,thecontrolsitesmustnotexperiencecougarkilling(legalandillegaltake)andmustexperienceeveryintrusionexceptcougarsdying,e.g.,thesamenumberandintensityofintrusionsas in treatment sites but no cougars killed. The current plan to allow±10% cougar harvest in controlareas is indefensiblescientifically.Currentlythedesignisflawed, justas if itwereabiomedicalclinicaltrialinwhichtheresearcherssaid,“expertsdon’tknowwhateffectthispillwillhave,sothecontrolwillbealow-doseandinthetreatmentwillbeahighdose.”(b)ThePiceanceBasinstudy,whichinvolveskillingblackbearsandmountainlionsononeparceloflandandthencomparingthattoanareawithnopredatorcontrolfrom2010-2012,iscalledapseudo-controlorfalsecontrol.Theotherareawasstudiedatadifferenttimeandplaceunderverydifferentconditionsthantoday.(c) Under the current design, each spatial unit is a single replicate. Events within a unit are notindependent of other events within that unit. Amore robust designwould reverse-treatmentwithineachunit,which receivesa treatmentby random-assignment,notby researcher selectionof sites fortreatments.AlthoughtheArkansasRiverstudylooksmorerobust,itremainsasamplesizeof4andthelackofatruecontrolwillmaketheresultsimpossibletointerpretscientifically.2. SelectionbiasTheUpperArkansasRiverproposalstates,“Deerdataanalysisunit(DAU)D-16(Figure1)wasidentifiedas an area where cougar suppression could be beneficial to the deer population.”iii This subjectivedecisionwillinvalidatethescientificvalueoftheproposedstudyinasinglestep.Whentreatmentsareassigned according to the response variable that one wishes to measure, you have guaranteed asamplingbiasthatwouldinvalidatethestudy.Remember,atreatmentisahypothesizedsolution.Ifonedesigns a study with the assumption that the solution will work, one risks intentional bias inmeasurement and reporting. Random assignment is far easier and more robust to these biases andprotectstheresearcherfromclaimsofintentionalbias.3. SmallsamplesizeWithfewerthan6studyunits (3controland3treatment), there isnostatistical test thatcanreliablyconfirmorrejecttheresearchhypothesis.Thatrequirementfor6ormorearisesbecauseeachunitisa

Page 2: To Colorado Parks and Wildlife Commission members, · To Colorado Parks and Wildlife Commission members, The undersigned address the two Colorado Parks and Wildlife (CPW) proposed

OpenletterfromscientistandscholarsontheproposedCPWmuledeerstrategystudies–30November2016

2

single replicate. Eventswithinaunit, suchas the survivalof amarkedmuledeer, arenot statisticallyindependent of other events (i.e., another mule deer’s survival) in that same unit. They have allexperienced the same treatment and confounding variables associated with that unit.We suggest areverse-treatmentdesigntoincreasethesamplesizebutthatrecommendationMUSTbeaccompaniedbyrandom-assignmentoritcanproduceanotherformofbias(treatmentbias).AlthoughtheArkansasRiver study looksmore robust because of the crossover design (reverse-treatment), it does not haverandomassignmentandthelow-levelofcougarkillingthroughoutbothunitsandthroughoutthestudycreatesapseudo-controlthatinvalidatestheexperiment.Giventhefourunitschosenforthestudies,theCPWcouldachieveasamplesizeof8iftheyarewillingtoassigntreatmentandcontrolrandomlyandthenreversethetreatmentineachunitinthefollowingphaseofthestudy.Referencestoresearchdesignandnarrativeexplainingtheprinciples:In1964,inthejournalScience,Platthypothesizedaboutscientificprogresswiththedeceptivelysimpletitle “Strong Inference” [2]. Platt hypothesized that certain fields advance slowly and others quicklybecausetheirpractitionersvariedintheefficiencywithwhichtheytestedbetweenalternative,opposedhypotheses.Heobservedthattheslowerfieldsofhistimehadbecomeboggeddownbytheperceptionthattheirtopicwastoocomplexforsimpletests.Platt[2]anticipatedtheargumentandcounteredthattheirmodels were becoming too complex to be falsifiable. Falsifiability is a foundational principle ofgoodscience.Plattalsopredictedthatslowerfieldshadbecomeboggeddownbyafocusonmethods,asopposedtorapidlyadvancingfieldsthathadfocusedonincisiveexperimentsthatforcedalternativehypothesesintodivergentpredictions[2].Subsequentwritershaveechoedhisviewsintheirparticularfields(biomedicalresearch,paleo-sciences,andpopulationbiology,amongothers)[3-6].Inecologytoday,weseeexamplesofbothofPlatt’shypothesizedbrakesonprogresswhenonehearsthatecosystemsaretoocomplextomanipulateexperimentally,ratherthancallsforelegantecologicalexperiments as we saw decades ago [7-10]. The field of predator ecology is at that crossroads. Thetraditionalhypothesis is that killingpredatorsequalsmoreprey.That viewhasbeendisputedas longago as Leopold (1949) who proposed the alternative that functional predator populations keepecosystemshealthier.CPWisfacingthisquestiontoday.Howeversalutaryeffortsemergedrecentlybypredator-preyecologistswhohadconductedcarefulexperimentalmanipulationstoexcludeor includepredators from complex ecological systems [11]. We see the salutary effects today in importantarguments over whether wolves – and other large carnivores such as big cats – strongly shapedbiodiversitybyscaringherbivoresandfeedingonherbivores[12-14].Resolvingthatscientificdebatewilldemand strong inference. The strong inference espousedby Platt [2] is best servedby gold standardexperiments using random assignment to control and treatment with sufficient sample sizes toovercomerandomvariationthatmayconfoundaneleganttestofanimportanthypothesis.4. LegalandethicalconsiderationsWildlifeareapublic trustassetand theproposedstudiespreferentially serveanarrowcommunityofmuledeerhuntersandcougarhunters,whileignoringthebroadpublicinterestinhealthyecosystems,unimpaired wildlife populations, and transparent accounting for wildlife assets. If CPW is heldaccountableincourtorbythelegislatureforitsmanagementofcougarsandblackbears,theproposedstudieswillnotsurvivethelegaltestforaprudenttrusteeofthepublicinterestinwildlife.The Colorado Supreme Court characterized the public trust in wildlife, and the privilege of huntingwildlifegrantedbythestate,insimilarlanguage:

Page 3: To Colorado Parks and Wildlife Commission members, · To Colorado Parks and Wildlife Commission members, The undersigned address the two Colorado Parks and Wildlife (CPW) proposed

OpenletterfromscientistandscholarsontheproposedCPWmuledeerstrategystudies–30November2016

3

Theownershipofwildgameis inthestateforthebenefitofall thepeople. Therighttokillgameisaboon or privilege granted, either expressly or impliedly, by the sovereign authority, and is not a rightinhering in any individual. The power of the state tomake regulations tending to conserve the gamewithinitsjurisdictionisbasedlargelyonthecircumstancethatthepropertyrighttothewildgamewithinitsbordersisvestedinthepeopleofthestateintheirsovereigncapacity;and,asanexerciseofitspolicepowersandtoprotectitspropertyforthebenefitofitscitizens,itisnotonlytherightbutitisthedutyofthestatetotakesuchstepsasshallpreservethegamefromthegreedofhunters.ivFor these reasons,weconclude that these twostudiesbedenied in their current stateby theCPWCommissionanddrasticallyreexaminedtoimplementthegoldstandardforscientificinference.Asis,thesestudieswilloffernovalidconclusionsandmisusealreadylimitedfunds.Signatures:AdrianTreves,PhD,UniversityofWisconsin–MadisonBradBergstrom,Ph.D.,ValdostaStateUniversityFranzJ.Camenzind,Ph.D.,Ex.Director,JacksonHoleConservationAlliance,RetiredChrisDarimont,Ph.D.,UniversityofVictoriaJohnC.Emerick,Ph.D.,ColoradoSchoolofMines(Retired)CamillaFox,M.S.,ProjectCoyoteMaureenHackett,M.D.,HowlingforWolvesDr.DavidW.Inouye,Ph.D.,RockyMountainBiologicalLaboratoryandUniversityofMarylandMichelleLute,Ph.D.,WildEarthGuardiansWilliamLynn,Ph.D.,MarshInstitute,ClarkUniversityStephanieMatlock,M.S.,ColoradoMesaUniversityDonaldA.Molde,M.D.,Co-founder,NevadaWildlifeAlliancePaulC.Paquet,AdjunctProfessor,UniversityofVictoriaDavidParsons,M.S.,USFishandWildlifeService,RetiredBradPurcell,Ph.D.,WinstonChurchillMemorialTrustHughRobinson,Ph.D.,LandscapeAnalysisLaboratoryandUniversityofMontanaKirkRobinson,Ph.D.,Ex.Director,WesternWildlifeConservancyFranciscoJ.Santiago-Ávila,Ph.D.Student,UniversityofWisconsin–MadisonMichaelSoule,ProfessorEmeritus,UniversityofCalifornia–SantaCruzJenniferWolch,Ph.D.,UniversityofCalifornia–BerkeleyGeorgeWuerthner,M.S.,PublicLandsMediaihttp://cpw.state.co.us/Documents/Research/Mammals/Piceance-Basin-Predator-Management-Plan-Overview.pdfiihttp://cpw.state.co.us/Documents/Research/Mammals/Upper-Arkansas-River-Predator-Management-Plan-Overview.pdfiiihttp://cpw.state.co.us/Documents/Research/Mammals/Upper-Arkansas-River-Predator-Management-Plan-Overview.pdfivMaitlandv.People,93Colo.59,62,23P.2d116,117(1933).References1. TrevesA,KrofelM,McManusJ.Predatorcontrolshouldnotbeashotinthedark.FrontEcolEnviron.2016;14:380–8.2. PlattJR.Stronginference.Science.1964;146:347–53.3. BiondiF.Paleoecologygrandchallenge.FrontiersinEcologyandEvolution.2014;doi:10.3389/fevo.2014.00050.4. GouldSJ.Thepromiseofpaleobiologyasanomothetic,evolutionarydiscipline.Paleobiology.1980;6:96-118.5. MukherjeeS.TheEmperorofAllMaladies:ABiographyofCancer.MewYork:Scribner;2010.

Page 4: To Colorado Parks and Wildlife Commission members, · To Colorado Parks and Wildlife Commission members, The undersigned address the two Colorado Parks and Wildlife (CPW) proposed

OpenletterfromscientistandscholarsontheproposedCPWmuledeerstrategystudies–30November2016

4

6. Oro D. Grand challenges in population dynamics. Frontiers in Ecology and Evolution. 2013;1 doi:

10.3389/fevo.2013.00002:2.7. HairstonNG.EcologicalExperiments:Purpose,DesignandExecution:CambridgeUniversityPress,Cambridge.;1989.8. BoutinS.Foodsupplementationexperimentswithterrestrialvertebrates:patterns,problemsandthefuture.Canadian

JournalofZoology.1990;68:203-20.9. LendremDW.Sleepingandvigilanceinbirds,IIAnexperimentalstudyoftheBarbarydove(Streptopeliarisoria).Anim

Behav.1984;32:243-8.10. Tapper SC, PottsGR, BrocklessMH. The effect of an experimental reduction in predation pressure on the breeding

successandpopulationdensityofgreypartridges(Perdixperdix).JournalofAppliedEcology.1996;33:965-78.11. KricherJ.TheBalanceofNature:Ecology'sEnduringMyth.Princeton,NJ:PrincetonUniversityPress;2009.12. Ripple WJ, Beschta RL. Wolves and the ecology of fear: Can predation risk structure ecosystems? Bioscience.

2004;54:755-66.13. MiddletonAD,KauffmanMJ,McWhirterDE,JimenezMD,CookJG,CookRC,etal.Linkingantipredatorbehaviortoprey

demographyrevealslimitedriskeffectsofanactivelyhuntinglargecarnivore.EcolLett.2013;16:1023-30.14. RippleWJ, Estes JA, Beschta RL,Wilmers CC, Ritchie EG, HebblewhiteM, et al. Status and ecological effects of the

world’slargestcarnivores.Science.2014;343(6167):1241484.

Page 5: To Colorado Parks and Wildlife Commission members, · To Colorado Parks and Wildlife Commission members, The undersigned address the two Colorado Parks and Wildlife (CPW) proposed

OpenletterfromscientistandscholarsontheproposedCPWmuledeerstrategystudies–30November2016

5

Page 6: To Colorado Parks and Wildlife Commission members, · To Colorado Parks and Wildlife Commission members, The undersigned address the two Colorado Parks and Wildlife (CPW) proposed

OpenletterfromscientistandscholarsontheproposedCPWmuledeerstrategystudies–30November2016

6

Page 7: To Colorado Parks and Wildlife Commission members, · To Colorado Parks and Wildlife Commission members, The undersigned address the two Colorado Parks and Wildlife (CPW) proposed

OpenletterfromscientistandscholarsontheproposedCPWmuledeerstrategystudies–30November2016

7

Page 8: To Colorado Parks and Wildlife Commission members, · To Colorado Parks and Wildlife Commission members, The undersigned address the two Colorado Parks and Wildlife (CPW) proposed

OpenletterfromscientistandscholarsontheproposedCPWmuledeerstrategystudies–30November2016

8

Page 9: To Colorado Parks and Wildlife Commission members, · To Colorado Parks and Wildlife Commission members, The undersigned address the two Colorado Parks and Wildlife (CPW) proposed

OpenletterfromscientistandscholarsontheproposedCPWmuledeerstrategystudies–30November2016

9

Page 10: To Colorado Parks and Wildlife Commission members, · To Colorado Parks and Wildlife Commission members, The undersigned address the two Colorado Parks and Wildlife (CPW) proposed

OpenletterfromscientistandscholarsontheproposedCPWmuledeerstrategystudies–30November2016

10