monitoring insect pollinators and flower visitation: the … · 2021. 5. 7. · coverage and...
TRANSCRIPT
Methods Ecol Evol. 2019;10:2129–2140. wileyonlinelibrary.com/journal/mee3 | 2129© 2019 The Authors. Methods in Ecology and Evolution © 2019 British Ecological Society
Received:28November2018 | Accepted:16August2019DOI: 10.1111/2041-210X.13292
R E S E A R C H A R T I C L E
Monitoring insect pollinators and flower visitation: The effectiveness and feasibility of different survey methods
Rory S. O'Connor1,2 | William E. Kunin2 | Michael P. D. Garratt1 | Simon G. Potts1 | Helen E. Roy3 | Christopher Andrews4 | Catherine M. Jones2,5 | Jodey M. Peyton3 | Joanna Savage3 | Martin C. Harvey3 | Roger K. A. Morris6 | Stuart P. M. Roberts1 | Ivan Wright7 | Adam J. Vanbergen4,8 | Claire Carvell3
1CentreforAgri‐EnvironmentalResearch,SchoolofAgriculture,PolicyandDevelopment,UniversityofReading,Reading,UK;2TheFacultyofBiologicalSciences,UniversityofLeeds,Leeds,UK;3CentreforEcology&Hydrology,Wallingford,UK;4CentreforEcology&Hydrology,Penicuik,UK;5Buglife–TheInvertebrateConservationTrust,Peterborough,UK;6CommonsideEast,Surrey,UK;7ShotoverWildlife,Oxford,UKand8Agroécologie,AgroSupDijon,INRA,Univ.BourgogneFranche‐Comté,Dijon,France
CorrespondenceRoryS.O'ConnorEmail:[email protected]
Funding informationBiotechnologyandBiologicalSciencesResearchCouncil,Grant/AwardNumber:BB/I000348/1;WellcomeTrust,Grant/AwardNumber:BB/I000348/1;DepartmentforEnvironment,FoodandRuralAffairs,Grant/AwardNumber:BB/I000348/1andWC1101;ScottishGovernment,Grant/AwardNumber:BB/I000348/1andWC1101;NaturalEnvironmentResearchCouncil,Grant/AwardNumber:NE/R016429/1
HandlingEditor:LuisaCarvalheiro
Abstract1. Thestatusofpollinatinginsectsisofinternationalconcern,butknowledgeofthemagnitudeandextentofdeclines is limitedbya lackof systematicmonitoring.Standardizedprotocolsareurgentlyneeded,alongsideabetterunderstandingofhowdifferentmethodsandrecorders(datacollectors)influenceestimatesofpol-linatorabundanceanddiversity.
2. Wecomparedtwocommonmethodsforsamplingwildpollinating insects (soli-tarybees,bumblebeesandhoverflies),pantrapsandtransects,insurveysof1kmcountryside squares (agricultural and semi‐natural habitats) and flowering cropfieldsacrossGreatBritain,includingtheinfluenceoflocalfloralresources(nectarsugaravailabilityorcropflowerdensity)ontheinsectssampled.Further,wecom-pared theperformanceof recorderswithdifferingexpertise (non‐specialist re-searchstaff,taxonomicexpertsandnon‐expertvolunteers)inapplyingmethods.
3. Pantrapsandtransectsproducedcompositionallydistinctsamplesofpollinatorcommunities.Inthewidercountryside,pantrapssampledmorespeciesofsolitarybeeandhoverfly.Infloweringcrops,transectsrecordedagreaternumberofindi-vidualbumblebees,butfewerspecies.
4. Across all taxonomic groups and countryside and crop samples, transects gen-erallyhadlowerratesofspeciesaccumulationperindividualcollectedthanpantraps.Thisdemonstrates thatdifferencesbetweenmethods inestimating rich-nessarenotduetosamplingeffortalone.However,recorderspossessinggreatertaxonomicexpertisecanproducespeciesaccumulationdatafromtransectsthatarealmostcommensuratewithpantrapping.
5. Theabundanceandspeciesrichnessofpollinators(exceptsolitarybees)ontran-sects in thewider countrysidewaspositively related to theavailabilityofesti-matednectarsugar.Incrops,pollinatorabundanceresponsestoflowerdensities
2130 | Methods in Ecology and Evolu on O'CONNOR et al.
1 | INTRODUC TION
There is international concern about declines in the diversity anddistributionofinsectpollinatorsandtheconsequencesforpollina-tion services (Potts et al., 2016). Research is increasingly demon-stratinghow land‐use change, pesticides, climate change, invasivenon‐nativespecies,pestsanddiseasemayact,andinteract,tocausedeclines in pollinating insects (Vanbergen et al., 2013). However,evidenceis incompleteandimportantgapsremainwithrespecttothemagnitude,geographicandtaxonomicextentofthesedeclines(Pottsetal.,2016).Forexample,ourunderstandingofthepopulationstatusandtrendsinabundanceanddiversityofpollinatinginsectsisseverelylimitedbyaworldwidelackofstandardized,long‐termandlarge‐scaledata(Lebuhnetal.,2013).Thiscreatesanurgentneedformonitoring,andprotocolsthataccommodatebroadtaxonomicandgeographic coverage, account for potential biases in the data andgenerateadequatesamplesizes;allwhileremainingcosteffective.
The most important providers of pollination services globallyare insects, particularly bees and some flies (e.g. hoverflies) (Pottsetal.,2016).Currentbestevidenceforthestatusofwildbeesandhoverflies comes from records of species occurrence collected innationalandglobalbiodiversitydatabases.InGreatBritain(GB),re-cordscollatedbytheBees,WaspsandAntsRecordingSocietyandtheHoverflyRecordingSchemehaveallowedunparalleled insightsintothestatusanddistributionalchangesofbeesandhoverflies inGB(Carvalheiroetal.,2013;Powneyetal.,2019).Toourknowledge,suchverifiedlong‐termoccurrencedataforwildbeesandhoverfliesexistonlyforGB,theNetherlands,Belgium(Carvalheiroetal.,2013)andbumblebees intheUSA(Cameronetal.,2011).Thesedataarecollectedusingunstandardizedorsemi‐standardizedprotocols(Isaac&Pocock,2015)andchangesinrecordingintensity,taxonomicabilityandsamplingstrategiesmeansourcesofbiashavenotbeenconsis-tentovertime.Critically,occurrencerecordsprovidenostandardizedestimates of abundance, which are fundamental to understandingchangesinpopulationsizeandthelinksbetweenpollinatorsandpol-linationservices(Pottsetal.,2016).Identifyingthebestapproachesforpollinatormonitoringiscrucialtoreducetheselimitations.
Different methods for sampling pollinating insects are associ-atedwithdifferentoutputsandchallengeswithregardtotaxonomiccoverage and implementation. Direct observations (transects and
observation plots) and pan traps (sampling within painted water‐filledbowls)arethemostcommonlyusedmethods(Westphaletal.,2008).Transectsandtimedfocalfloralobservationsarestraightfor-wardtoconductandcangeneratedataoninsect–plantinteractionsbut depend on the expertise of the observer (Sutherland, Roy, &Amano,2015)andmaybebiasedtowardsmoreconspicuousspecies(Dennisetal.,2006).Pantrapstendtosamplemorespeciesofbeethanotherstandardizedmethods(Westphaletal.,2008),areinde-pendentofobserverexpertiseandarerecommendedbytheFoodandAgricultureOrganisation (FAO)formonitoringbees inagricul-tural habitats (LeBuhn,Droege, Connor, Gemmill‐Herren, &Azzu,2016).However, pan trap efficacymaybebiasedbecause certaintaxa (e.g. social bees)may be less likely to be caught and effectsoflocalfloralresourcedensityoncatchesarenotwellunderstood(Cane,Minckley,&Kervin,2000;butseeWood,Holland,&Goulson,2015).Similarly,usingnon‐expertvolunteers,or ‘citizenscientists’,presentsanopportunitytocollectlargeamountsofdataandengageawiderangeofindividualsinwildliferecording.However,theseben-efitspotentiallytrade‐offagainstthereducedtaxonomicresolutionthat thesevolunteerscantypicallygatheranddataaccuracy (Roy,Baxter,Saunders,&Pocock,2016),whichisrequiredtoaddresseco-logicalquestionsconcerningthediversityofwildpollinators.
Wecomparedthepotentialofpantrapsandtransectsforsur-veyingpollinatinginsectsin(a)thewidercountrysideand(b)flow-ering crop fields in 38 sites acrossGB. Furthermore, in thewidercountryside, we explored the effect of recorder expertise on thenatureandaccuracyofdatacollectedusingtransectsandfloralob-servationplots.Thereafter,weoutlineoptionsforthedevelopmentof protocols formonitoring pollinator abundance and diversity tofacilitatetheproductionoflong‐term,standardizednationalandin-ternationaldatasetsinaccordwithinternationalscienceandpolicyneedsidentifiedbytheIntergovernmentalScience‐PolicyPlatformonBiodiversityandEcosystemServices(Pottsetal.,2016).
2 | MATERIAL S AND METHODS
2.1 | Wider countryside surveys
We tested three commonly usedmethods for sampling bees andhoverflies(O'Connoretal.,2016;Westphaletal.,2008);
wereidiosyncraticaccordingtocroptype,butoveralltheresponsewaspositiveandnegativefortransectsandpantraps,respectively.
6. Giventhesetaxonomicandcontext‐specificdifferencesinmethodperformance,weassesstheirsuitabilityformonitoringpollinatinginsectcommunitiesandpol-linationservices.Wediscusstherelevanceofthesefindingswithinthecontextofachievingstandardized,large‐scalemonitoringofpollinatinginsects.
K E Y W O R D S
abundance,bees,diversity,expertise,hoverflies,pantraps,pollinatormonitoring,transects
| 2131Methods in Ecology and Evolu onO'CONNOR et al.
(i) Pan traps: a triplet of plastic bowls (350 ml capacity; Salbert,Item Number: 92012A500) sprayed with UV fluorescent paint(1 × white, 1 × yellow, 1 × blue; Sparvar “Leuchtfarbe”) witheachbowl containing100mlofwaterplusadropofunscenteddetergenttobreaksurfacetension.Eachtriplet(hereafterstation)was fixed toawoodenstakeusingwiresupportsandsetat theaverage height of flowers or other surrounding vegetation orsecured to the ground in very short vegetation or bare ground.
(ii) Insectvisitationtransects:Fivetransectsections,each200minlengthandfollowingalinearroute,werewalkedataslowpaceforbetween12and15minallowingforvariationintransectter-rain.Allinsectsseenvisitingflowerswererecordedwithina1m3 samplingboxaheadandtothesideoftherecorderandassignedtooneofthefollowingtaxonomicgroups:bumblebees,honey-bees, solitary bees (including primitively eusocial species) andhoverflies. Individual insectswere recordedonlyonce.Wherespecieslevelidentificationswererequired(seebelow),individu-alswere netted, placed in a labelled tube and frozen for lateridentification,unlesstheycouldbereadilyidentifiedinsitu.Timespenthandlinginsectsforidentificationwasnotincludedinthetransecttime.
(iii) Floralobservationplots:adefinedareaobservedforasettimetorecordinsectflowervisitors.Plotsof50×50cm2 were ob-served for 10min for insect flower visitation on a focal plantspecies,insectswereobservedandrecordedonceandclassifiedinto taxonomic groups, as described above (without specimenidentification).Focalplantspeciesonasitewereselectedfromalistof25nationallycommonfloweringplants(TableS1)or,ifnotpresent,thenalocallyabundantplantspecies.Theplantspeciesandnumberoffloralunitswithineachplotwererecorded.
Thewidercountrysidesurveysusedaone‐dayprotocoltosamplewithina1km2,compatiblewithexistingbiodiversitymonitoringschemesinGB
(e.g.Pescottetal.,2015).Fourteen1kmgridsquares(Brtishnationalgrid)were sampled acrossGB (Figure1a; England=6; Scotland=6;Wales = 2) with half the squares dominated (>50%) by semi‐naturallandcoverandhalfdominatedbyagriculturallandcover(arable,horti-cultureorimprovedgrasslandcollectively).Ineachsquare,wesituatedfive200mtransectsandfivepantrapstationsatapproximately200mintervalsonadiagonallinebisectingthesquare(Figure1b),typicallyfol-lowingboundaryfeaturesor,whereaccessible, followingtractor lineswithincroppedfieldsoredgesofgrassfieldswithlivestock.
Pan trap stationswere deployed at the start of each transect(Figure1b) and left exposed for6–7hr (dependingon terrain andtime taken to complete the other methods) between 10:00 and16:00.Afterpantrapdeployment,each200mtransectsectionwaswalkedtorecord insect flowervisitors.Foreachsection,availablefloralresourceswerequantified.Thenumberoffloralunits(flowerheads,umbelsorspikes)of≥5mostcommonfloweringplantspe-cieswasalsorecordedona5‐pointordinalscale:(1)1–2,(2)2–30,(3)31–300,(4)301–3,000, (5)>3,000.Tostandardizenectaravail-abilitypertransect,thetotalamountofavailablenectarsugarwasestimated for each recorded flowering plant species as µg sugarproducedin24hrperfloralunit(followingBaudeetal.,2016);seeSupplementaryMaterial).Wemultiplied this value by themediancoverageofeachspecies forcategories1–4andby3,001forcat-egory5andconverteditintoanestimateofnectaravailabilityperm2foreachtransect(bydividingthisproductby200).Duetosomeextremeestimatesofflowerdensity,weimposedamaximumlimitof20,000µgsugarperm2per24hr.Two10‐minfocalfloralobser-vationspersitewerealsoconductedduringeachsamplingday.Eachsitewassampledonceduringfoursamplingroundsin2015:(a)27April–10May,(b)1–14June,(c)6–19July,(d)17–30August.
Toexploretheeffectofrecorderexpertiseonthedatacollected,we classified recorders according to their degree of expertise infield surveys and recognizing pollinating insects: (a) non‐specialist
F I G U R E 1 (a)Distributionofstudysites,showingtheagriculturalwidercountrysidesites(browncircles).Semi‐naturalwidercountrysidesites(yellowcircles),strawberrysites(redstars),fieldbeansites(redsquares)andapplesites(redtriangles);(b)Thelayoutofpantrapsandtransectsforthewidercountryside‘one‐day’protocolata1kmsamplingsquare;(c)Thelayoutofpantrapsandtransectsinasamplingplotforfloweringcrops
2132 | Methods in Ecology and Evolu on O'CONNOR et al.
researchstaff–employeesofuniversitiesorresearchinstituteswithpriorexperienceofsurveyingandidentifyinginsectsandplantstoatleastbroadgrouplevels;(b)taxonomicexperts–volunteerorpro-fessional entomologists who submit records to existing biologicalrecordingschemespossessingahighlevelofexpertiseincollectingandidentifyingatleastonebroadtaxonomicgrouptospecieslevel;(c)non‐expertvolunteers–membersofthepublicwhopartake incitizenscienceprojectspossessingvaryinglevelsoffamiliaritywithpollinator identification or ecological surveys. All recorders con-ducted transects, volunteers and researchers conducted focalob-servations,butonlyresearchersconductedpantraps.Allrecordersfollowedthesameprotocolforeachmethodandwereprovidedwithidentificationguidesforbroadinsectgroupsandfocalplantspecies.Researchstaffandexpertscollecteddatatospeciesresolutionasfaraspossible,whereasnon‐expertsonlyclassifiedinsectsintobroadgroups.
Allsitesweresurveyedbyresearchstaff;taxonomicexpertsvis-itedonlythesitesinEnglandandWalesandnon‐expertvolunteerswererestrictedtoroundsthreeandfour,surveyingonthesamedaysastheresearchstaff.Researchstaffandvolunteersundertooktran-sectswithin15minofeachotherandfocalobservationsinparallelonthesamepatchesofflowers.Here,55sitevisitswereachievedbyresearchstaff,25bytaxonomicexperts,and17byvolunteernon‐experts(TableS2).
2.2 | Flowering crop surveys
Tocomparepollinator surveymethods in crops,pan trappingandtransectswerecarriedoutsimultaneouslyindessertapples(Malus domestica, variety Cox's Orange Pippin), strawberries (Fragaria X ananassa,mixedvarieties)andfieldbeans(Vicia faba,varietyWizard)inthespringandsummerof2011(Garratt&Potts,2011).WeusedeightappleorchardsinKent,eightstrawberryfieldsinYorkshireandeightfieldbeanfieldsinOxfordshireandBerkshire(Figure1a),withthreesamplingroundscarriedoutduringstrawberryandfieldbeanflowering and two during apple bloom. Sampling plots containedtwo150msamplingtransects,dividedintothree50msectionsandapantrapstationwasplacedattheendofeachsection,givingsixpseudo‐replicates of eachmethod per field (Figure 1c). Transectswereatleast25mapartandfromthefieldedge(Figure1c)andeach50msectionwaswalkedfor10minatasteadypace.Pantrapswereasspecifiedaboveforwidercountryside,butused460mlbowls,leftoutfor24hrinapplesandstrawberries,and7–10hrinfieldbeans.Appleflowerdensitieswerecountedwithin1×1m2quadratsheldagainst treesatheadheight,whereas forstrawberriesa1×2m2 areawasassessed.Fieldbeanfloweringstemswerecountedwithina1×2m2area,andmultipliedbythemeanflowercountsonfiverandomlychosenstems.
2.3 | Survey conditions and identification
All surveys were carried out between 10:00 and 16:00 in dryweather, with light winds (<29 km/hr, Beaufort 5), and where
minimumtemperaturesexceeded13°Cif<50%cloudcover,or15°Cif>50%cloudcover(although11°Cor13°CwasallowedforsomeuplandlocationsorvisitsinApril).Collectedbeeandhoverflyspeci-menswerestoredin70%ethanolforidentificationtospecieslevelbyexperttaxonomistsandarchivedin99%ethanol.
2.4 | Analysis
All analyseswere performed usingR version 3.3.2 (RCore Team,2016).
2.5 | Similarity of pan trap and transect samples of pollinator communities
Dataweresummarizedatthesite (1kmsquareorcropfield) leveltodemonstratethetypicalsamplesizesachievedbythetwometh-odsandbythedifferentrecordergroupsacrossthefourfocalinsectgroups(Tables1and2;TablesS3andS4).
We assessed the degree of dissimilarity (Morisita–Horn abun-dance‐based dissimilarity index) between the pollinator (bees andhoverflies identified to species) communities sampledby researchstaffusingpantrapsandtransectsinthewidercountrysidedatasetandeach floweringcropdataset (apple, strawberryand fieldbeanseparately).Todetermineifthepantrapandtransectmethodspro-duced significantly dissimilar assemblages,we used permutationalANOVAs (r: vegan: adonis) against random permutations of theoriginal data (countryside=999; FC=255 for each cropdataset)(Oksanenet al., 2015).Data for thewider countryside semi‐natu-raldominatedsiteinWaleswereexcludedduetotoofewrecords.Non‐metricmultidimensionalscaling(NMDS)wasusedtovisualizedissimilarity between sampling methods based onMortista–Horndissimilarity(r:vegan:MetaNMDS;Oksanenetal.,2015).
2.6 | The effects of sampling effort and recorder expertise on estimates of species richness
We used species accumulation curves to understand the influ-enceof samplingefforton theefficacyofmethodsand recorderstoproducespecies richnessestimatesgiven theirdifferentmodesofactionandinherentbiases.Thenumberofindividualssampledisthebasiccurrencywithwhichspeciesrichnessestimatesbetweensamples or datasets can be compared.Using the inext package inr (Hsieh,Ma, & Chao, 2019), we plotted individual‐based speciesaccumulation curves that show interpolated species richness (percumulativeindividualsampled)uptothetotalsamplesizeandthere-after extrapolated species richness. Curves were plotted for pantrapsandtransects,usingsamplesamalgamatedacrossthedatasetforeachbroadtaxonomicgroup in thewidercountrysidedataset,forsolitarybeesinapples,bumblebeesinstrawberriesandbumble-bees and solitary bees in field beans. Further, for a subset of thewidercountrysidedatacoveringsevensites(fourwithsamplesforallfoursamplingrounds,oneforthe2nd,3rdand4thsamplingroundsandtwoforthefirsttwosamplingrounds,totally23samplingvisits)
| 2133Methods in Ecology and Evolu onO'CONNOR et al.
individual‐basedspeciesaccumulationcurveswereplottedforbum-blebees,solitarybeesandhoverfliestocomparepantrapswithtran-sectsconductedbyeitherresearchersortaxonomicexperts.
Correlationanalyses(Spearman'sorKendall'srank)wereusedtocompareestimatesofbumblebee,solitarybee,hoverflyandhoney-bee abundance from transectswalkedby research staff andnon‐expertvolunteers(17sitevisitswithcorrespondingdata)andfromparallelfloralobservationplots.
2.7 | Per sampling unit differences between pan traps and transects
Generalizedlinearmixedmodels(GLMMs)wereusedtotestfordif-ferencesbetweenpantrapsandtransectsatthesamplingunitlevel(individualpantrapstationorcorrespondingtransectsection),alongwiththeeffectsoflocalfloralresourcesandothercovariates,usingthedatasetsforbumblebees,solitarybeesandhoverfliesgeneratedbyresearchstaff(honeybeenumberswereinsufficient).ModelswerefittedandselectedusingtheglmmADMBpackage(Skaug,Fournier,Bolker, Magnusson, & Nielsen, 2015), which allows zero‐inflatedmodels,althoughpoissonornegativebinomialerrorswereappropri-ateforallmodels.Finalmodelswereselectedbystepwiseeliminationof non‐significant variables using log‐likelihood tests (Zuur, Hilbe,& Ieno, 2013). Final models were also run with the lme4 package(Pinheiro,Bates,DebRoy,&Sarkar,2015)tochecktheagreementofmodelfitsbetweenpackages. Inevery instance,theywerecompa-rable,givingthesamequalitativeresultswithonlyslightdifferencesinparameterestimates.Thelsmeanspackage(Lenth,2016)wasusedtocalculateleastsquaremeansandmarginaleffectsplotsfromlme4outputwereproducedusingtheSJPlotpackage(Lüdecke,2017).
Fortheabundanceandspeciesrichnessofbumblebees,solitarybeesandhoverfliessampledonthewidercountrysidesurveys,ini-tial model predictors included sampling method, sampling round,
country (EnglandandWaleswereamalgamated intoone levelduetolowreplicationforWales),logestimatednectarsugaravailabilitypertransect(µgper24hr),maximumdaytimetemperature(°C)fromthenearestUKMETofficerecordingstationanddominantland‐useofthesiteasfixedeffects.Two‐wayinteractionswereincludedbe-tweenmethodandlognectar,methodandsamplinground,lognec-tarandsamplinground,andcountryandsamplinground.Allmodelsincludedan intercept level randomeffectof sample location (1–5)nestedwithinsite(1–14).
ForeachFCdataset,estimatesofabundanceforthedominantinsectpollinatorvisitorgroupweremodelled;solitarybeesforap-ples, bumblebees for strawberries and fieldbeans.Datawerenotsufficient to model the abundance of all groups individually, butmodelsofthetotalabundanceofallbeesandhoverflieswererunforcomparison.Speciesrichnessofallbeesandhoverflieswasalsomodelled.Initialmodelsincludedsamplingmethod,thenaturallogofflowerdensityandtheirinteractionasfixedeffectsandaninterceptlevelrandomeffectofthesamplingsection(1–6)nestedwithinthesite.
3 | RESULTS
Pan traps and transects implemented by research staff on thewider countryside surveys across 14 1 km2 sampled a total of110species (16bumblebee,38solitarybee,55hoverflyspeciesand thehoneybeeApis mellifera)with variations in species rich-nessandabundance foreachmethod (Table1,TableS3). In thewider countryside, 65% of solitary bees, 19% of hoverflies and14%ofbumblebeesrecordedbyresearchstaffwereidentifiedtothegrouplevelonly,becausespecimenswerenotnettedforiden-tification.Taxonomicexpertsrecorded10speciesofbumblebee,21speciesofsolitarybeeand34speciesofhoverflyontransects,
TA B L E 1 Mean±SEabundanceandspeciesrichnesspersamplingsite(n=14)sampledbyresearchstaffacrossthewidercountryside
Method
Abundance Species richness
Bumblebee Solitary bee Honeybee Hoverfly Bumblebee Solitary bee Hoverfly
PanTrap 12.14±3.17 18.36±5.77 3.00±1.03 32.07±70.53 2.36±0.59 2.43±0.74 9.43±1.28
Transect 17.86±3.18 5.86±2.35 4.36±1.39 39.79±16.93 2.64±0.42 0.5±0.24 3.64±0.75
TA B L E 2 Meanabundance±SEandspeciespersamplingsiteforapples,strawberryandfieldbeansites
Crop Method
Abundance Species
Bumblebee Solitary bee Honeybee Hoverfly Bumblebee Solitary bee Hoverfly
Apple Pantrap 2.63±0.46 148.88±53.82 0.88±0.35 0.13±0.13 2.25±0.53 16.88±2.22 0.13±0.13
Transect 4.38±0.98 14.00±3.49 5.88±1.64 1.38±1.10 2.13±0.40 2.00±0.38 0.00±0.00
Strawb Pantrap 15.75±6.01 11.13±2.75 5.25±2.02 3.75±1.29 3.75±0.53 4.13±0.81 0.88±0.23
Transect 147.25±32.28 1.75±0.65 121.00±34.55 40.00±12.30 3.88±0.35 0.38±0.26 0.25±0.16
FieldB Pantrap 16.50±6.35 33.75±4.55 3.50±1.58 2.38±0.46 4.63±0.84 12.25±0.88 1.63±0.26
Transect 65.38±9.43 1.88±0.58 8.75±1.96 1.25±0.45 5.63±0.38 0.88±0.30 0.13±0.13
2134 | Methods in Ecology and Evolu on O'CONNOR et al.
whilstforthesamenumberofsamplingvisitstothesametransectlocations (25, thoughondifferentdays) research staff recorded11,9and18speciesofeach,respectively.Forcrops,werecordeda total of54 species in apples (8bumblebee,44 solitarybee,1hoverflyandthehoneybee),32speciesinstrawberries(12bum-blebee,14solitarybee,5hoverflyandthehoneybee)and55 infield beans (14 bumblebee, 31 solitary bee, 9 hoverfly and thehoneybee)(Table2,TableS4fortotalspeciesrichnessandabun-dancepercrop).
3.1 | Community dissimilarity
Overall,therewasasignificantdissimilaritybetweenthepollinatorcommunities sampled using pan traps and transects in the wider
countryside(R2=0.121,F1,24=3.312,p<.001)drivenbymoresoli-tarybeeandhoverflyspeciesdetectedbypantrapsthantransects,butmore individuals of common bumblebee species on transects(Figure2,TableS3,FigureS1a).Therewasa significantdissimilar-itybetween thepollinatorcommunities sampledbypan trapsandtransectsinallcroptypes;apples(R2=0.51,F1,14=14.309,p=.008);strawberries(R2=0.29,F1,14=5.744,p=.008);fieldbeans(R
2=0.41,F1,14 = 9.58,p = .008). (Figure 3). Transects sampledmuch highernumbersofbumblebee individuals in strawberriesand fieldbeansthandidpan traps (around10and5 times, respectively,TableS4)with samplesmoredominatedby commonspecies thanpan traps(FigureS1c,d).Inappleswerepantrapssamplednearly10timesthenumberofsolitarybees(TableS4).
3.2 | Species accumulation and recorder effects
For bumblebees in the wider countryside, there was a close cor-respondence between the species accumulation rates for eachmethod;although theoverallpan trapsaccumulatedmorespeciesand transects sampledmore individuals (Figure 4a). In crops, thispatternwasaccentuated,withthetransectmethodshowinglowerratesofbumblebeespeciesaccumulationperindividualsampledandreaching an asymptote,whereas the steeper accumulation curvesforpantrapsarepredictedtocontinue(Figure4b). Ingeneral, thespecies accumulation curves for bumblebeeswere broadly similarbetweenpantraps,transectsbyresearchersandtransectsbytaxo-nomicexperts(Figure5a).
Forsolitarybees,thesamegeneralpatternofspeciesaccumu-lationbetweenpantrapsandtransectswasobservedinthewidercountrysideandinapplesandfieldbeans.Itwasdifficulttoconstructmeaningfulspeciesaccumulationcurvesfortransects(Figure4candFigureS2)becausealargeproportionofindividualswasnotidenti-fiedtospeciesresolution(TableS4).However,whilethenumberofindividualsrecordedbytaxonomicexpertsontransectswaslowerthan those sampled in pan traps, species accumulation curves fortransects completed by experts suggest that, per individual, this
F I G U R E 2 Non‐metricmulti‐dimensionalscaling(NMDS)plotofpantraps(largerdarkgreycircles)andtransects(largerlightgreycircles)forallspeciesofbeeandhoverflydetectedinthewidercountrysidebynon‐expertresearchers.Bumblebeesareshownbystars,Apis melliferaasquare,solitarybeesbytrianglesandhoverfliesbycircles.Circleswiththesamenumberareforthesamesiteandthepolygonsconnectingsitesindicatetheoverlapbetweensamples
F I G U R E 3 Non‐metricmulti‐dimensionalscaling(NMDS)plotsofpantraps(largerdarkgreycircles)andtransects(largerlightgreycircles)forallspeciesofbeeandhoverflydetectedin(a)apples,(b)strawberriesand(c)fieldbeans.Bumblebeesareshownbystars,Apis melliferaasquare,solitarybeesbytrianglesandhoverfliesbycircles.Circleswiththesamenumberareforthesamesiteandthepolygonsconnectingsitesindicatetheoverlapbetweensamples
| 2135Methods in Ecology and Evolu onO'CONNOR et al.
wouldachievecomparableorbetterspeciescoveragewithgreatersamplingofindividuals(Figure5b).
Hoverflieswerenotsampledincropsinhighenoughnumbers,butforthewidercountryside,therateofspeciesaccumulationper
individualforpantrapswasarounddoublefortransects(Figure4d).However, it is notable that two species (E. balteatus and S. ribesi)comprised 84% of individual hoverflies sampled on transects andidentifiabletospeciesresolution.Removingthesetwospeciesleadstogreatercorrespondencebetweenpantapsandtransectsinspe-ciesaccumulation (FigureS3a).Correspondencebetweenhoverflyspecies accumulation curves for pan traps and taxonomic expertssuggestthattheyperformcomparablyintermsofsamplingspecies(Figure5c).Removing thehighly abundantE. balteatus and S. ribe-sisimprovedthecorrespondenceofresearchertransectstoexperttransectsandpantraps(FigureS3b).
Estimates of abundance for all taxonomic groupswere signifi-cantly,positivelycorrelatedbetweenresearchstaffandvolunteers,usingtransectandfocalobservations(seeSupplementaryMaterialandFiguresS4andS5forfullresults).
3.3 | Sampling unit level analyses
There were significant differences between sampling methods inboththeabundanceandspeciesrichnessofsolitarybeespersam-plingunit(pantrapstationor200mtransectsection).Pantrapssam-pledgreaternumbersofsolitarybeeindividuals(β=−1.27±0.22,z=−5.77,p<.001;Figure6b)andspecies(β=−2.38±0.27,z=−8.87,p<.001;FigureS7b)thantransects.However,forbumblebeesandhoverflies, significant interactions suggest that the effects of thesamplingmethodonabundanceandspeciesrichnessweredepend-entonboththeestimatednectarsugaravailabilityalongthe200mtransectand,forhoverflies,thetimingofthesamplinground(TablesS5andS6).Ontransects,theincreasingnectaravailabilityhadasig-nificant,positiveeffectcomparedtopantrapsforbumblebeeabun-dance (β =0.28±0.07, z =4.12,p < .001;Figure6a) and speciesrichness(β=2.09±0.34,z=6.09,p<.001;FigureS7a),andhoverflyabundance(β=0.16±0.06,z=2.59,p=.010;Figure6c)andspeciesrichness(β=0.16±0.06,z=2.74,p=.006;FigureS7c).Theeffectsofcountry,samplingroundandmaxtemperature inthemodelsofabundanceandrichnessarereportedintheSupplementaryMaterial(TablesS5andS6).
Inapplesa significant interactionbetweenmethodand flowerdensityshowedanegativeeffectofincreasedflowerdensityonsol-itarybeeabundanceinpantrapsbutapositiveeffectontransects(β=0.87±0.18,z=4.99,p<.001;Figure7a).Themodelforabun-danceofallpollinatinginsectswasqualitativelythesame(TableS7),aswasforspeciesrichness(β=0.51±0.13,z=3.92,p<.001;FigureS7a,TableS8).
In strawberries, bumblebee abundance on transects was sig-nificantly higher than in pan traps regardless of flower density(β=2.27±0.13,z=17.00,p<.001;Figure7b).However,fortheabun-dance of all pollinating insects, estimates from transects increasedsignificantly with flower density compared to those of pan traps(β=0.52±0.13,z=4.10,p<.001;TableS7),asdidthenumberofspe-ciessampled(β=0.38±0.12,z=3.32,p=.001;FigureS7b,TableS8).
In field beans, a significant interaction between method andflowerdensityshowedbumblebeeabundanceincreasedwithflower
F I G U R E 4 Individual‐basedspeciesaccumulationcurvesacrossthewholedatasetspooledfor(a)bumblebeesinthewidercountryside(b)bumblebeesinfieldbeansandstrawberries(c)solitarybeesinthewidercountrysideand(d)hoverfliesinthewidercountryside.Curveswereplottedbasedondatagroupedacrossallsites,usingtheinextpackageinr.Thesolidlineshowspredictionsbasedoninterpolationandthedashedpartshowspredictionsbasedonextrapolation.95%confidenceintervalsareshownasshadedareas
2136 | Methods in Ecology and Evolu on O'CONNOR et al.
F I G U R E 5 Individual‐basedspeciesaccumulationcurvesfromasubsetofdatafromacrosssevenofthewidercountrysitesprovidingcorrespondingdatafrompantraps,transectsconductedbyresearcherandtransectsconductedbyprofessionalexpertsfor(a)bumblebees,(b)solitarybeesand(c)hoverflies.Thesolidlineshowspredictionsbasedoninterpolationdashedlinethepredictionsbasedonextrapolation,95%confidenceintervalsareshownasshadedareas
F I G U R E 6 Plotsforthewidercountrysideof(a)predictionsofthemarginaleffectsofsamplingmethodandnectarsugaravailabilityonbumblebeeabundance(b)theleastsquaremeanpermethodforsolitarybeeabundanceand(c)predictionsofthemarginaleffectsofsamplingmethodandnectarsugaravailabilityonhoverflyabundance.Unbrokenlinesshowpredictedvaluesforpantrapsandbrokenfortransects.95%confidenceintervalsareshowningrey.Errorbarsonpointsshow±SE.Thesamplingunitforpantrapsisatrappingstation(tripletofbowls)andfortransectsisa200msection(Figure1b).ModelresultsarepresentedinTableS4.ModelsforspeciesrichnessarepresentedinFigure4sandTableS5
F I G U R E 7 Plotsshowing(a)predictionsformarginaleffectsofsamplingmethodandflowerdensityonsolitarybeeabundanceinapplecrops(b)meanabundancebumblebeespersamplingmethodinstrawberrycropsand(c)predictionsformarginaleffectsofsamplingmethodandflowerdensityonbumblebeeabundanceinfieldbeancrops.Unbrokenlinesshowpredictedvaluesforpantrapsandbrokenfortransects.95%confidenceintervalsareshowningrey.Errorbarsonpointsshow±SE.Samplingunitforpantrapsisatrappingstation(tripletofbowls)andfortransectsisa50msection(Figure1c).ModelresultsarepresentedinTablesS7.ModelsforthespeciesrichnessofallbeesandhoverfliesareshowninFigureS5andTableS8
| 2137Methods in Ecology and Evolu onO'CONNOR et al.
densityontransects,butdeclinedwithflowerdensityinpantraps(β=0.38±0.12,z=3.32,p=.001;Figure7c).Resultsfortotalpol-linator abundance were qualitatively the same (β = 0.35 ± 0.16,z=2.15,p=.032;TableS7),aswerethoseforthenumberofspeciessampled(β=0.42±0.15,z=2.88,p=.004;FigureS7c,TableS8).
4 | DISCUSSION
Understanding the status and trends of pollinators is an urgentglobal priority requiringdevelopmentof national scalemonitoringusing repeatable and standardized survey methods (Dicks et al.,2016).Ourstudycomparedtheperformanceofdifferentpollinatorsurveymethods insamplingdifferenttaxonomicgroupsandwhenimplementedbydifferentrecordersvaryinginexperience.Wedis-cussour findingswithin the contextof the logistical and financialconstraintspresentedbylarge‐scalebiologicalmonitoring.
Pantrapsandtransectsprovidedadifferentpictureofthepol-linatinginsectcommunity.Overall,theassemblagessampledbythetwomethodswere significantly dissimilar compositionally in boththewider countryside and crop fields. This differencewas drivenbytransectssamplingfewerspecies,particularlyofsolitarybeeandhoverfly,butmorebumblebeeindividuals,particularlyincrops.
Sampling effort dictates the relative performance of methods(Rhoadesetal.,2017),forexample,increasingthedurationofexperttransectsmayresultindatathatconvergesontherichnessestimatesproduced by pan‐traps. Fundamentally different modes of actionmakeitimpossibletoproperlystandardizethesamplingeffort(e.g.samplingduration)betweenpantrapsandtransects.However,usingspecies accumulation curves, we were able to compare estimatesof species richnessproducedby thedifferentmethods andactorstounderstandtheextentthatsamplingeffort(i.e.numbersofindi-vidualscollected)contributestotheobserveddifferentialpatterns.Accumulationofspeciesoccurringatasimilarrateindicatesthatdif-ferencesinrelativesamplingeffortaredrivingdifferencesinspeciesrichness.Wefoundhigherspeciesaccumulationratesforpantraps,exceptforbumblebeesinthewidercountryside,suggestingfactorsotherthansamplesizearedrivingdifferencesbetweenmethods.
In all datasets, transects sampledmore individual bumblebeesthanpantraps,probablydueinparttothestrongpositiveassocia-tionbetweenfloral resourcesandbumblebeecountson transectsandtothebias inpantrapsagainstsampling largerbodiedinsects(Caneetal.,2000).Thatthisdifferencewasofagreatermagnitudeinstrawberryandfieldbeanfieldscomparedtothewidercountry-sidemaybebecausethesecropsarepredominantlybumblebeepol-linated(Kleijnetal.,2015)andduetothecompetitionforbumblebeevisitsfromtheabundantfloraldisplaysofthesecropmonocultures,loweringpantrapcatches.However,pantrapsshowedhigherratesof species accumulation and generally sampled more species ofbumblebee.Oneexplanationisthatthetransectprotocolwascon-strainedtorecordflowervisitorsonly,sospeciesforagingspecialismwillreducethepoolofspeciesbeingsampled,particularlyincrops(whereonlyoneflowertypewassurveyed).
Forsolitarybees,pantrapscollectedmorespeciesandindividu-alsthantransects,andinapplesthelargermagnitudeofdifferenceinnumberscollectedmayrelatetothe24‐hrpantrappingused(asopposedto6–7hr).Projectingspeciesaccumulationwasdifficultfortransectsduetolowratesofspecieslevelidentification.However,whenexpertsundertooktransectsinthewidercountryside,thoughthenumberofsolitarybeesrecordedwasstilllowerthanpantraps,species accumulation rate per individual became higher for tran-sects. These findings highlight a limitationwhen using such ‘real‐time’methods tocollectdataonsolitarybees thataredifficult todetect,identifyorcapture,particularlyforlessexperiencedrecord-ers.Forhoverflies,pantrapsshowedsimilarlyhigherratesofspe-ciesaccumulationperindividualsampledthantransects,butagain,expert recordersmitigated this by providing a convergent rate ofspeciesaccumulationbetweenmethods.
While expertise seems necessary to collect species resolutiondatafromtransects,ourresultssuggesttransectscouldbesuitablefor novices to collect group level abundance data of bumblebeesandpossiblyhoverflies,withbasicinstructions.However,wefoundthe potential for miscounts or misclassifications, particularly forhoverflies.Kremen,Ullman,andThorp(2011),similarlyfoundesti-matesofbeeabundancewerecorrelatedbetweenvolunteerswithfivehourstrainingandexperts.Atransect‐based(1–2km)approachin 373 sites, ‘BeeWalks’, has been developed by the BumblebeeConservationTrust in theUK and is generating data on trends inabundance for bumblebee species (Comont & Dickinson, 2017).However, training, assessment and data validation processes areneededbeforemassparticipationobservationalmethodsarewidelyadoptedformonitoring.
Across all surveys, per sampling unit, estimates of abundanceandspeciesrichnessontransectsincreasedwithestimatednectaravailability or floral density. This effect is intrinsic to themethod(transectsrecordedflowervisitors),butthestrengthofresponsefordifferenttaxonomicgroupstofloralresourcesmayreflecttheirdif-ferentecologies.Socialbumblebeesincreasecolonyforagingactiv-ityinresponsetonectaravailability(Dornhaus&Chittka,2001)andover larger ranges thansmaller, solitarybeespecies (Gathmann&Tscharntke,2012;Osborneetal.,1999).Thismayexplainthestrongresponse of bumblebees to transect floral resources in thewidercountrysidecomparedwithsolitarybeesthatpossesssmaller for-agingrangesandalackofsocialrecruitmentbehaviour.Hoverfliesalsodonot recruit,butarenot restrictedto foragingaroundnestsites,andsoindividualsmayfreelyaggregatearoundhighfloralre-sources.Thisisconsistentwithourresultsshowingapositiverela-tionbetweenhoverflyabundanceandnectaravailability.
Fortransects,abundancerecordsmayreflectpopulationdensitiesinalocationbutalsotheredistributionofindividualsacrossaland-scapeinresponsetotemporaryincreasesinfloralresources(Carvell,Bourke,Osborne,&Heard,2015);however,methodsarenowavail-abletoaddressthis(Kleijnetal.,2018).Thenegativerelationshipbe-tweenlocalfloraldensityandthenumberofindividuals(andspecies)caughtinpantrapsinfloweringcropfieldssuggestthatcropflowerswere‘competing’withpantrapsbydrawingawayinsects(e.g.Caneet
2138 | Methods in Ecology and Evolu on O'CONNOR et al.
al.,2000).Ifpantrappingisconfoundedbyfloraldensities,thiscouldaffect theiruse inmonitoringschemesas itmay leadtoerroneousdetectionofdeclinesifanarea’sfloralresourcesincreaseovertime.However, this inverserelationshipbetweenpantrapcatchandflo-raldensitywasparticulartocrops,likelyduetotheveryhighflowerdensitiesinthesecropmonocultures.Themagnitudeoffloral‘com-petition’withpantrapswillbelowerinflorallyheterogeneouswidercountryside environments. Moreover, our results reflect a seriesofsnapshotsamplesofthedifferentmethods inspace.Structured,longitudinalmonitoringorexperimentsmanipulatingfloraldensitiesareneededtodemonstratehowpantrapcatchesmightrespondtoannualandmultiannualchangesinfloralresourcesatagivensite.Itmustbenotedthatournectarestimatesandpantrapstationswerenotpreciselyspatialcoincidentandquantifyingfloralresourcesinafixedareasurroundingthepantraps(inthewidercountrysidesetting)mayhavegivendifferentresults(Carvelletal.,2016).Previousfind-ingsontheimpactsoffloralresourcesonpantrapcatcheshavealsobeenmixed;withnegativeeffectsonabundance(Roulston,Smith,&Brewster,2007)andspeciesrichness (Baum&Wallen,2011),posi-tiveeffectsonabundance(e.g.Woodetal.,2015),andnoeffect(e.g.Rhoadeset al., 2017).Overall,measuresaccounting for local floralresourceswillbeavitalcovariateforcollectionwithanymethodusedinpollinatorsurveyprotocolsformonitoring.
Pan traps and transects have different utility and efficacy formonitoringdifferentaspectsofpollinatorbiodiversity.Identifyingtheobjectiveofthemonitoringandwhatmetricsofthepollinatorcom-munityare required isessential todeterminingwhichmethodsareemployed.Characterizingplant–pollinatorinteractionsoridentifyingwhichspeciesofinsectaredeliveringpollinationservicetocropsandwildflowersrequiretransects(orotherobservationalmethods)aspantrapsdonotreflectthis(Gibbsetal.,2017;Kleijnetal.,2015).Whilepan trapshave limitations andbiases, theyprovide species resolu-tiondataindependentofexpertiseandrequirelesspersonefforttoachieveequivalentsamplesizeswhencomparedtotransects.Theycouldalsominimizenoiseinthedatafromdifferentlevelsofrecorderknowledgeorchangesinrecordersovertime.Ourresultsshowthat,independentofdifferencesinsamplingeffort,transectsconductedbypeoplewithoutalargedegreeoftaxonomicexpertisewillnotsam-plethesamenumberofspeciesaspantraps,andforsolitarybeesthey require considerablymore sampling effort to detect asmanyindividuals. This could be particularly important when recorderswithappropriateexpertisearealimitingfactor,alongwithlogisticalandresourcingimplications.Forexample,ifspecies‐levelabundanceanddiversityofsolitarybeesweretargeted,ourresultssuggestfivetransectswould require sampling for36–45minby someonewithextensiveexperienceandtaxonomicexpertisetoachieveequivalentsamplesizesandspeciescoverageasfive6–7hrofpantraps.Ifstaffavailabilityorresourcesarelimiting,pantrapsusingnon‐expertre-corderscoupledwithspecies identificationbyexpertscanbeused(LeFéonetal.,2016)andmolecularmethodsmaysoonbeanoption(Creedyetal.,2019).Thoughlethal,pantrapsareunlikelytoreducepollinatinginsectpopulationsatthesamplingintensitiestestedhere(Gezon,Wyman,Ascher,Inouye,&Irwin,2015).
Noonesamplingmethodcanfullycharacterizethepollinatingin-sectcommunityatagivenlocation,butsamplingshouldaimtoprovidenecessarytaxonomiccoverageandkeepbiasasconsistentaspossi-bleover time.Furthermore,combiningdata fromdifferent locationsrequires methods that ensure datasets are at least comparable attheirmostbasic resolution.Anationalpollinatormonitoringschemecouldemploypantrapsandobservationalmethodstoallowthecom-plimentaryrecordingofdifferentfacetsofthepollinatorcommunityincludingabundance,speciesrichness,functionalrolesandpollinationservicepotential.Acrucialcaveat,however, isthedifferentialeffectoflocalfloralresourceavailabilityontheefficacyofthepantrapsandobservationalmethodsandhowthismayinfluencethedataobtainedand the conclusions drawn. This potential complementarity and ca-veatshouldbothbeconsideredcarefullyduringmethod(s) selectionalongsidemonitoringobjectives,desiredmetricsandtheavailabilityoffinancialorhuman resources.Only throughsuchstandardardizationcan monitoring efforts become internationally cohesive. The valueof obtaining standardized datasets on pollinating insects cannot beoverstatedinprovidingrobustevidenceonlong‐termandlarge‐scalepatternsandtrendstoinformnationalandinternationalpolicyneeds.
ACKNOWLEDG EMENTS
TheUKDepartment for theEnvironment,FoodandRuralAffairs,the ScottishGovernment and theWelshGovernment funded thewidercountrysidesurveyunderprojectWC1101.Thecropssurveyswere funded jointly by grant BB/I000348/1 from BBSRC, Defra,NERC,theScottishGovernmentandtheWellcomeTrust,undertheInsectPollinatorsInitiative.ThisworkwassupportedbytheNaturalEnvironmentResearchCouncilawardnumberNE/R016429/1,partoftheUK‐SCAPEprogrammedeliveringNationalCapability.Thankstothefarmers,landownersandlandmanagerswhoallowedusac-cesstotheir land.ThankstoA.Perry,D.Chapman,N.Majlessi,A.Turner,D.Coston,C.Dodson,R.Evans,L.TrusloveandM.Lappagefor undertaking fieldwork and to all the non‐expert volunteers.ThankstoS.Freemanforstatisticaladvice.Thankstothreereview-erswhosinsightsandsuggestionsimprovedthemanuscript.
AUTHORS' CONTRIBUTIONS
R.S.O.–H.E.R,A.J.V.andC.C.concievedanddesignedtheproject.R.S.O., C.A.–M.H. and S.P.M.R.–C.C. collected and collated thewidercountrysidedata,andM.H.–I.W.providedspecimenidenti-fications.M.P.D.Gcoordinatedthecollectionofandprovidedthefloweringcropdata.R.S.O.analyzedthedata.R.S.O.–H.E.R,A.J.V.andC.C.ledthewritingofthemanuscript.Allauthorscontributedcriticallytodraftsandgavefinalapprovalforpublication.
DATA AVAIL ABILIT Y S TATEMENT
Data for the wider countryside surveys are available from theNERC Environmental Information Data Centre: https://doi.org/10.5285/69a0d888‐9f6b‐4e67‐8d29‐402af1412d8e. Data
| 2139Methods in Ecology and Evolu onO'CONNOR et al.
for the flowering crops surveys are available from Data DryadRepository; https://datadryad.org/stash/dataset/doi:10.5061/dryad.31f7ph7ht tps://datadr yad.org/resource/10.5061/dryad.31f7ph7(Garratt&Potts,2011).
ORCID
Rory S. O'Connor https://orcid.org/0000‐0001‐7633‐4304
William E. Kunin https://orcid.org/0000‐0002‐9812‐2326
Michael P. D. Garratt https://orcid.org/0000‐0002‐0196‐6013
Simon G. Potts https://orcid.org/0000‐0002‐2045‐980X
Helen E. Roy https://orcid.org/0000‐0001‐6050‐679X
Christopher Andrews https://orcid.org/0000‐0003‐2428‐272X
Jodey M. Peyton https://orcid.org/0000‐0002‐8313‐6194
Martin C. Harvey https://orcid.org/0000‐0001‐7512‐2449
Adam J. Vanbergen https://orcid.org/0000‐0001‐8320‐5535
Claire Carvell https://orcid.org/0000‐0002‐6784‐3593
R E FE R E N C E S
Baude,M.,Kunin,W.E.,Boatman,N.D.,Conyers,S.,Davies,N.,Gillespie,M.A.K.,…Memmott,J.(2016).Historicalnectarassessmentrevealsthe fall and rise of floral resources inBritain.Nature,530, 80–85.https://doi.org/10.1038/nature16532
Baum,K.A.,&Wallen,K.E. (2011).Potentialbias inpantrappingasafunction of floral abundance. Journal of the Kansas Entomological Society,84(2),155–159.https://doi.org/10.2317/jkes100629.1
Cameron,S.A.,Lozier,J.D.,Strange,J.P.,Koch,J.B.,Cordes,N.,Solter,L. F., … Robinson, G. E. (2011). Patterns of widespread decline inNorthAmericanbumblebees.Proceedings of the National Academy of Sciences of the United States of America,108(2),662–667.https://doi.org/10.1073/pnas.1014743108
Cane,J.H.,Minckley,R.L.,&Kervin,L.J.(2000).Samplingbee(Hymenoptera:Apiformes) for pollinator community studies: Pitfall of Pan‐Trapping.Journal of the Kansas Entomological Society,73(4),225–231.
Carvalheiro,L.G.,Kunin,W.E.,Keil,P.,Aguirre‐Gutiérrez,J.,Ellis,W.N.,Fox,R.,…Biesmeijer,J.C.(2013).Speciesrichnessdeclinesandbiotichomogenisation have slowed down for NW‐European pollinatorsandplants.Ecology Letters,16(7),870–878.https://doi.org/10.1111/ele.12121
Carvell,C.,Bourke,A.F.G.,Osborne,J.L.,&Heard,M.S.(2015).Effectsofanagri‐environmentschemeonbumblebeereproductionatlocalandlandscapescales.Basic and Applied Ecology,16(6),519–530.https://doi.org/10.1016/j.baae.2015.05.006
Carvell,C., Isaac,N. J.B., Jitlal,M.,Peyton, J.,Powney,G.D.,Roy,D.B., …Roy, H. E. (2016). Design and testing of a national pollinatorand pollination monitoring framework. Final summary report tothe Department for Environment, Food and Rural Affairs (Defra),ScottishGovernmentandWelshGovernment:ProjectWC1101.
Comont, R. F., &Dickinson, H. (2017). BeeWalk Annual Report 2017.Retrieved from https://bumblebeeconservation.org/images/uploads/Beewalk/BBCT074_‐_BeeWalk_Annual_Report_2017_03.17_(1).pdf.
Creedy,T.J.,Norman,H.,Tang,C.Q.,Chin,K.Q.,Andujar,C.,Arribas,P.,…Vogler,A.P.(2019).Avalidatedworkflowforrapidtaxonomicassignmentandmonitoringofanationalfaunaofbees(Apiformes)usinghighthroughputDNAbarcoding.Molecular Ecology Resources. https://doi.org/10.1111/1755‐0998.13056
Dennis,R.L.H.,Shreeve,T.G., Isaac,N.J.B.,Roy,D.B.,Hardy,P.B.,Fox,R.,&Asher,J.(2006).Theeffectsofvisualapparencyonbiasinbutterfly recording andmonitoring.Biological Conservation,128(4),486–492.https://doi.org/10.1016/j.biocon.2005.10.015
Dicks, L.V., Viana, B., Bommarco, R., Brosi, B., Arizmendi,M.D.C.,Cunningham, S. A., … Potts, S. G. (2016). Ten policies for polli-nators: What governments can do to safeguard pollination ser-vices. Science, 354(6315), 14–15. https://doi.org/10.1126/science.aai9226
Dornhaus,A.,&Chittka, L. (2001). Foodalert inbumblebees (Bombus terrestris): Possible mechanisms and evolutionary implications.Behavioral Ecology and Sociobiology, 50(6), 570–576. https://doi.org/10.1007/s002650100395
Garratt,M. P. D., & Potts, S. G. (2011). Data from:Monitoring insectpollinators and flower visitation: The effectiveness and feasibilityof different survey methods. Dryad Digital Repository, https://doi.org/10.5061/dryad.31f7ph7
Gathmann, A., & Tscharntke, T. (2012). Foraging ranges of soli-tary bees. Journal of Animal Ecology, 71(5), 757–764. https://doi.org/10.1046/j.1365‐2656.2002.00641.x
Gezon, Z. J.,Wyman, E. S., Ascher, J. S., Inouye,D.W.,& Irwin, R. E.(2015). The effect of repeated, lethal sampling onwild bee abun-dance and diversity.Methods in Ecology and Evolution, 6(9), 1044–1054.https://doi.org/10.1111/2041‐210X.12375
Gibbs,J.,Joshi,N.K.,Wilson,J.K.,Rothwell,N.L.,Powers,K.,Haas,M.,…Isaacs,R.(2017).Doespassivesamplingaccuratelyreflectthebee(apoidea:Anthophila)communitiespollinatingappleandsourcherryorchards? Environmental Entomology, 46(3), 579–588. https://doi.org/10.1093/ee/nvx069
Hsieh,T.C.,Ma,K.H.,&Chao,A.(2019).inext:Interpolationandextrap-olationforspeciesdiversity.
Isaac,N.J.B.,&Pocock,M.J.O.(2015).Biasandinformationinbiologi-calrecords.Biological Journal of the Linnean Society,115(3),522–531.https://doi.org/10.1111/bij.12532
Kleijn,D.,Linders,T.E.W.,Stip,A.,Biesmeijer, J.C.,Wäckers,F.L.,&Bukovinszky, T. (2018). Scaling up effects of measures mitigatingpollinator lossfromlocal‐to landscape‐levelpopulationresponses.Methods in Ecology and Evolution, 9(7), 1727–1738. https://doi.org/10.1111/2041‐210X.13017
Kleijn, D., Winfree, R., Bartomeus, I., Carvalheiro, L. G., Henry, M.,Isaacs,R.,…Potts,S.G.(2015).Deliveryofcroppollinationservicesisan insufficientargument forwildpollinatorconservation.Nature Communications, 6(May), 7414. https://doi.org/10.1038/ncomms8414
Kremen,C.,Ullman,K.S.,&Thorp,R.W.(2011).Evaluatingthequalityofcitizen‐scientistdataonpollinatorcommunities.Conservation Biology,25(3),607–617.https://doi.org/10.1111/j.1523‐1739.2011.01657.x
LeFéon,V.,Henry,M.,Guilbaud,L.,Coiffait‐Gombault,C.,Dufrêne,E.,Kolodziejczyk,E.,…Vaissière,B.E. (2016).Anexpert‐assistedcitizen science program involving agricultural high schools pro-vides national patterns on bee species assemblages. Journal of Insect Conservation, 20(5), 905–918. https://doi.org/10.1007/s10841‐016‐9927‐1
LeBuhn,G.,Droege,S.,Connor,E.,Gemmill‐Herren,B.,&Azzu,N.(2016).Protocoltodetectandmonitorpollinatorcommunities.Guidanceforpractitioners.
Lebuhn,G.,Droege,S.,Connor,E.F.,Gemmill‐Herren,B.,Potts,S.G.,Minckley,R.L.,…Parker,F. (2013).Detecting insectpollinatorde-clinesonregionalandglobalscales.Conservation Biology,27(1),113–120.https://doi.org/10.1111/j.1523‐1739.2012.01962.x
Lenth,R.V.(2016).Least‐squaresmeans:Therpackagelsmeans.Journal of Statistical Software,69(1),1–33.
Lüdecke,D. (2017).Datavisualization for statistics in social science.r packageversion2.3.1.Retrievedfromhttps://CRAN.R‐project.org/package=sjPlot
2140 | Methods in Ecology and Evolu on O'CONNOR et al.
O'Connor,R. S., Jones,C.M.,Carvell,C.,Peyton, J.,Vanbergen,A. J.,Andrews, C., & Kunin,W. E. (2016). Data from:Monitoring insectpollinators and flower visitation: The effectiveness and feasibilityof different survey methods. Dryad Digital Repository, https://doi.org/10.5285/69a0d888‐9f6b‐4e67‐8d29‐402af1412d8e
Oksanen,J.,BlanchetGuillaume,F.,Kindt,R.,Legendre,P.,Minchin,P.R.,O'Hara,R.B.,…Wagner,H. (2015).vegan:Communityecologypackage.rpackageversion2.3‐1.Retrievedfromhttp://cran.r‐project.org/package=vegan
Osborne, A. J. L., Clark, S. J.,Morris, R. J.,Williams, I.H., Riley, J. R.,Smith, A. D., … Edwards, A. S. (1999). A landscape‐scale of bum-ble bee foraging study range and constancy using harmonicradar. Journal of Applied Ecology, 36(4), 519–533. https://doi.org/10.1046/j.1365‐2664.1999.00428.x
Pescott, O. L.,Walker, K. J., Pocock,M. J. O., Jitlal,M., Outhwaite,C.L.,Cheffings,C.M.,…Roy,D.B.(2015).Ecologicalmonitoringwith citizen science: Thedesign and implementationof schemesfor recording plants in Britain and Ireland. Biological Journal of the Linnean Society, 115(3), 505–521. https://doi.org/10.1111/bij.12581
Pinheiro, J., Bates, D., DebRoy, S., Sarkar, D., & R Core DevelopmentTeam.(2015).Linearandnonlinearmixedeffectsmodels.rpackageversion3.1‐122.
Potts,S.G.,Imperatriz‐Fonseca,V.,Ngo,H.T.,Aizen,M.A.,Biesmeijer,J.C.,Breeze,T.D.,…Vanbergen,A.J.(2016).Safeguardingpollinatorsandtheirvaluestohumanwell‐being.Nature,540(7632),220–229.https://doi.org/10.1038/nature20588
Powney, G. D., Carvell, C., Edwards, M., Morris, R. K. A., Roy, H. E.,Woodcock,B.A.,&Isaac,N.J.B.(2019).Widespreadlossesofpolli-natinginsectsinBritain.Nature Communications,10(1),1018.https://doi.org/10.1038/s41467‐019‐08974‐9
RCoreTeam.(2016).R: A language and environment for statistical comput-ing.Vienna,Austria:RFoundationforStatisticalComputing.
Rhoades,P.,Griswold,T.,Waits,L.,Bosque‐Pérez,N.A.,Kennedy,C.M.,&Eigenbrode,S.D.(2017).Samplingtechniqueaffectsdetectionofhabitat factors influencingwild bee communities. Journal of Insect Conservation,https://doi.org/10.1007/s10841‐017‐0013‐0
Roulston, T., Smith, S. A., & Brewster, A. L. (2007). A compari-son of pan trap and intensive net sampling techniques for doc-umentung a bee (Hymenoptera: Apiformes) fauna. Journal of the Kansas Entomological Society, 80(2), 179–181. https://doi.org/10.2317/0022‐8567(2007)80[179:acopta]2.0.co;2
Roy, H. E., Baxter, E., Saunders, A., & Pocock, M. J. O. (2016). Focalplant observations as a standardised method for pollinator
monitoring:Opportunitiesandlimitationsformassparticipationcit-izen science. PLoS ONE, 11(5), 1–14. https://doi.org/10.1371/journal.pone.0155571
Skaug,H.,Fournier,D.,Bolker,B.,Magnusson,A.,&Nielsen,A.(2015).glmmADMB: A generalized linear mixed models using ‘AD ModelBuilder’_.rpackageversion0.8.3.2.
Sutherland,W.J.,Roy,D.B.,&Amano,T.(2015).Anagendaforthefu-tureofbiologicalrecordingforecologicalmonitoringandcitizensci-ence. Biological Journal of the Linnean Society,115(3),779–784.https://doi.org/10.1111/bij.12576
Vanbergen, A. J., Baude, M., Biesmeijer, J. C., Britton, N. F., Brown,M. J. F.,Brown,M.,…Wright,G.A. (2013). Threats to anecosys-tem service: Pressures on pollinators. Frontiers in Ecology and the Environment,11(5),251–259.https://doi.org/10.1890/120126
Westphal, C., Bommarco, R., Carre, G., Lamborn, E., Morison, N.,Petanidou,T.,…Steffan‐Dewenter, I. (2008).Measuringbeediver-sity in different habitats European habitats and BiogeographicalRegions. Ecological Monographs, 78(4), 653–671. https://doi.org/10.1890/07‐1292.1
Wood,T.J.,Holland,J.M.,&Goulson,D.(2015).Acomparisonoftech-niques for assessing farmland bumblebee populations. Oecologia,177(4),1093–1102.https://doi.org/10.1007/s00442‐015‐3255‐0
Zuur,A.F.,Hilbe,J.M.,&Ieno,E.N.(2013).Abeginner'sguidetoGLMandGLMMwithr:AfrequentistandBayesianperspectiveforecolo-gists.HighlandStatistics.
SUPPORTING INFORMATION
Additional supporting information may be found online in theSupportingInformationsectionattheendofthearticle.
How to cite this article:O'ConnorRS,KuninWE,GarrattMPD,etal.Monitoringinsectpollinatorsandflowervisitation:Theeffectivenessandfeasibilityofdifferentsurveymethods.Methods Ecol Evol. 2019;10:2129–2140. https://doi.org/10.1111/2041‐210X.13292