THE JOURNAL OF RAPTOR RESEARCH A QUARTERLY PUBLICATION OF THE RAPTOR RESEARCH FOUNDATION, INC.

VoI. 29 JuNw 1995 No. 2 j RaptorRes. 29(2):85-92 ¸ 1995 The Raptor ResearchFoundation, Inc.

FLIGHT STRATEGIES OF MIGRATING : FASTING VS. FORAGING

GRADY L. CANDLER • AND PATRICIA L. KENNEDY 2 Departmentof Fisheryand WildlifeBiology, Colorado State University, Fort Collins, CO 80523 U.S.A.

ABSTRACT.--Wedeveloped energetics models to predictmigration times and fat consumptionrates of osprey (Pandionhaliaetus) migrating south from their breedinggrounds in the IntermountainWest of North America. In thesemodels we simulatedthree migration strategies: fasting, foraging at severalmid-migration stopovers (jump strategy)and frequentforaging at stopovers(hop strategy).Because these piscivores appear to migrate predominantlyover land and are rarely seen along ridges used by othermigrant raptors that exploitdeflection updrafts,we assumedthey primarily used thermal soaring during migration. Our modelpredicts a 1.68-kg ospreywould take 12 d and 0.25 kg of fat (a fat densityof 15% of lean bodymass), to completea fasting migrationof 3780 km (mean of migrationdistances estimated from 21 band recoveriesof ospreynesting in northernIdaho and easternWashington) when wind speedis 0 m s-•. A sensitivityanalysis of this model showedthat changesin wind speed(+5 m s-l) had the greatestinfluence on migrationtime (8-21 d) and fat consumption(0.16-0.45 kg). In the foragingmodel, maximum fat depositionrate was predicted to be 0.04 kg d-•. Giventhis predictionand assumingosprey settle for 1 d at eachstopover, migrations with one,three (jumpstrategies), five or 11 (hopstrategies) stopovers were predicted to take17, 21, 25, or 34 d, respectively. With no settlingtime at stopoversthe predictedforaging migration times only range from 15-17 d. The model predictionsfor boththe foraging(without settling costs) and fasting migrations are consistent with thelimited data availableon fall migrationstrategies of ospreybreeding in the IntermountainWest. Our resultsalso suggestthat, under certain assumptions, nonstop migration may be energeticallypossible for westerninterior osprey. KEYWORDS: aerodynamicmodel;fat deposition; migration strategies; osprey; haliaetus; sensitivity analysis.

Estrategiasde vuelode Pandionhaliaetus migrantes; rapidez vs. forrajeo Rv_,SUMV,N.--Desarrollamos un modeloenerg&ico para predecirlos tiempos de migraci6ny tasasde consumo degrasas de Pandion haliaetus migrando hacia el surdesde sus fireas de reproducci6n en el oestede Norteamgrica. En estosmodelos simulamos tres estrategias de migraci6n:rapidez, forrajeo en variossitios de descansoen el transcursode la migraci6n("jump strategy") y forrajeofrecuente en sitiosde descanso ("hop strategy"). Debido a que estasaves piscivoras parecen migrar predominantemente sobre el campoy raramenteson observadas a lo largode cordones montafiosos, usados por otrasrapaces migratorias que explotan la deflexi6nde corrientes deaire, presumimos que ellas se remontan usando corrientes de aire ascendentes.Nuestro modelo predice que un individuode estaespecie de 1.68 kg podrlatomar 12 diasy 0.25 kg de grasaen recorrerrapidamente 3780 km (distanciamigratoria media, estimada de 21 individuosmarcados), cuando la velocidaddel vientoes de 0 m/s. Un anfilisisde sensibilidaddel modelomostr6 que cambiosen la veloddaddel viento(+5 m/s) tiene graninfluencia sobre el tiempode migraci6n(ocho a 21 dias)y sobreel consumode grasa(0.16-0.45 kg). En el modelode forrajeo, la tasamfixima de depositaci6n degrasa, se estim6 en 0.04 kg/dla. Dada esta predicci6n y presumiendoque P. haliaetusse detiene un dla en cadaparada, se presume que migracionescon una, tres ("jumpstrategies"), dnco o onceparadas ("hop strategies") toman 17, 21, 25 o 34 dias,respectivamente. Sin establecerseen cadaparada, el modelode forrajeopredice tiempos de migrad6ndel rangode 15 a 17 dias.

Current address:New York BotanicalGardens, Bronx, NY 10458 U.S.A. PLK is the correspondingauthor.

85 86 GRADYL. CANDLERAND PATRICIAL. KENNEDY VOL. 29, NO. 2

Las prediccionesdel modelotanto para la migraci6ncon forrajeo (sin costosde establecimiento)como para migracionesrfipidas son consistentes con los limitados datos disponibles sobre las estrategias migracionales de otofiode P. haliaetusnidificantes en el oesteintermontafioso. Nuestros resultados sugieren que, bajo ciertas presunciones,la migraci6n sin detenciones puede ser energ•ticamente posible para esta rapaz proveniente del oeste interior. [Traducci6nde Ivan Lazo]

The migrationstrategies of long-distancemigrants soaringover land but probablyresort to flappingflight have been the focusof extensiveresearch (see Baker when crossingthese water barriers(Kerlinger 1989, 1982, Kerlinger 1989, Alerstam 1990 and Berthold Alerstam1990). Osprey breeding in the Intermountain 1993 for reviews),but it is still unclearwhether many West seemto migrateinland over the semi-aridsouth- speciesof long-distancemigrants make stopoversto westernUnited Statesand CentralAmerica (Henny feed during migration,rely totally on pre-migratory and Van Velzer 1972,Melquist et al. 1978, Pooleand tissuedeposition to fuel their migration,or usea com- Agler 1987) and probablyuse thermal soaring for the binationof both strategies.Even the type of tissue majorityof their migration(Kerlinger 1989). We de- accumulatedfor energyreserves during migration has cidedto modelthe migrationstrategies of ospreyfrom come into questionin recent years (fat vs. protein the Intermountain West because the model would re- [Piersma1990, Lindstr6m and Piersma1993]). quire fewer variablesthan a migrationmodel of Eu- Using a bioenergeticsmodel, Smith et al. (1986) ropean ospreywhich would have to incorporatepa- predictedthat the broad-wingedhawk (Buteoplatyp- rmeters (and assumptions)describing both thermal terus)and the Swainson'shawk (Buteoswainsoni) could soaringand flappingflight. migrate from southernNorth America to northern In this paper we describemodels that predict fall SouthAmerica (4300 km) withouteating if they de- migrationtimes and fat consumptionrates for osprey positedfat amountingto 20-25% of lean bodymass breedingin the IntermountainWest usingthree mi- prior to migration.However, Smith et al. (1986) only gration strategies:(1) fasting;(2) foragingat several evaluatedone migrationstrategy (fasting). Similar to mid-migrationstopovers (,jump strategy [Piersma migrationmodels developed for passerinesand waders 1987]);and (3) frequentforaging at stopovers(hop (Castroand Myers 1989, Gudmundssonet al. 1991, strategy[Piersma 1987]). We presentresults of a sen- LindstriSmand Alerstam1992), they did not conduct sitivityanalysis of the fastingmodel in which the pa- a sensitivityanalysis of their model to evaluatethe rameter estimates were varied in a manner that reflects extentto whichtheir modelpredictions vary with un- their empiricalvariation. This analysisprovides in- certaintiesin parametervalues (see Kirkley 1991 and sightsinto the key factors influencing model predictions Goldstein and Smith 1991 for additional evaluations andindicates the extentto whichthe modelpredictions of the Smith et al. model). varywith the uncertaintiesin the parameterestimates. Furtherinsight could be providedon possibleraptor A preliminaryevaluation of the modelis alsopresented migrationstrategies by creatingmodels that compare in which we comparemodel predictions to empirical fastingand foragingmigration. However, the species estimatesof (1) ospreymigration times from lookout modeledshould be well-studiedsince empirical data observationsand bandingdata and (2) raptorfat den- are necessaryfor parameterestimation and for a pre- sities. liminaryevaluation of the realismof the model'spre- dictions.In the United States,the osprey(Pandion METHODS haliaetus)is oneof the few speciesof largeraptors for Fasting Migration Model. To determineif thesebirds which there are empiricaldata on its flight behavior coulddeposit sufficient fat to fuel a fastingmigration, an (Kerlinger 1989), and migrationdistance and time energeticsmodel was constructedbased on morphometrics, flight characteristics,diurnal activity patterns, energy cost of (Henny and Van Velzer 1972, Melquist et al. 1978, soaringflight, energycost of roosting,and wind speed.This Melquist and Johnson1984, Pooleand Agler 1987). modelassumes that fat depositsare the main sourceof energy Althoughthere is an abundanceof bandingdata on duringmigration since there are no data availableto estimate Europeanosprey migrations (•)ster16ff 1977, Alerstam the role of other fuel resourcesused by ospreyduring hU- gration.Values used as input parametersare shownas nom- 1990), thesemigrations occur across large water bar- inal valuesin Table 1. We assumeda migrationdistance of rierssuch as the Mediterraneanand Baltic seas 3780 km. This distanceis the mean of migrationdistances terl/Sff1977, Alerstam 1990). These osprey use thermal estimatedfrom 21 bandrecoveries of ospreynesting in north- JUNE 1995 OSPREYMIGRATION STRATEGIES 87 ern Idaho and easternWashington (Melquist et al. 1978, still allowingthem to glidefarther and take advantageof Melquist and Johnson1984, Johnsonand Melquist 1991). strongerthermals than if they choseVopt. Therefore, cross- Morphometricvalues were basedon averagemeasure- countryair speedof theosprey [Vxc, (m s-l)] wascalculated mentsof body mass,wing span and wing area for osprey usingan interthermalvelocity that was 2 m s-l lessthan Vopt breedingin New York (Kerlinger 1989) becauseno com- estimatedby Pennycuick'smodel. parabledata setswere ava:.lablefor westernosprey. Average Daily distancetraveled (DDT [km d-l]) was calculated mass(1.67 kg) of 33 adultosprey breeding in centralIdaho as[ (M. Bechardunpubl. data) is similarto the New York av- DDT = Vxc x Tr x (.001 km erage(1.68 kg, Table 1), indicatingthe morphometricdata on easternosprey are a goodapproximation of westernosprey where Tr is daily flight time in sec.Daily fat consumedin morphometrics. flight (FFC [kg d-l]) was calculatedas: Usingradar, Kerlinger (1989) measuredthe flight altitude FFC = (2RMRa x Tr)/e, of ospreyusing thermal soaringto migrate acrosscentral New York (• = 880 m). In our modelthe nominal value for where RMR• is in Watts and e is the energycontent of fat flight altitudewas 2000 m to compensatefor the increased (3.96 x 107j kg-l [Robbins1993]). Although energetic costs elevationof the westernmigration route. Sinceit is difficult are usuallycalculated from lean body massbecause fat is to determinewind velocityand directionat migrationalti- relativelyinert metabolically (Luke and Schoeller 1992), stored tudes,a nominal wind speedof 0 m s-1 was used. fat was included in these calculations of RMRa to account The energycost of soaringflight andcross-country velocity for the costof increasedwing loadingdue to pre-migratory were estimatedwith a modificationof Pennycuick'smodel fat deposits.The amountof fat depositedfor a 3780 km (Program2; 1989) which is derivedfrom theoreticalaero- migrationis unknown,so the modelwas initiated with a body dynamics.We chosePennycuick's model as the templatefor mass 50% >1.68 kg. We used 2 x RMRa as a nominal our model becauseit providesthe mostrealistic estimates of estimateof the costof glidingflight (Baudinetteand Schmidt- migrationflight costsin comparisonwith otheraerodynamic Nielsen 1974). models(Welham 1994). Output from Pennycuick'smodel As in Smith et al.'s (1986) model,we assumedthat fasting predictsthe amount of dailyfat utilized,average cross-country ospreyengage in onlytwo activitiesduring migration, soaring velocity,and daily distanceflown at the predictedvelocity. flight(8 hr d-l) androosting (16 hr d-l). A dailyflight time This model usesclimb rate as an input variable for deter- of 8 hr was used as the nominal value in this model because mining cross-countryvelocity. A climb rate of 3 m s-1 was this approximatesthe lengthof time convectivefields used chosenfor thesecalculations, based on Kerlinger's(1989) by migrating,soaring hawks are availablein temperatelat- radar-trackingresults of ospreymigrating in New York. This itudes(Kerlinger 1989). From this,daily fat consumedduring may be a conservativeestimate of climb rate sincethermal roosting(RFC [kg d-l]) can be calculatedas: updraftsin the IntermountainWest are exceptionallystrong RFC = (RMRr x T•)/e, (Hoffman 1985). Pennycuick'smodel also calculates basal metabolicrate (BMR) usingLasiewski and Dawson's (1967) whereTr is the daily roostingtime in sec.In this equation, allometricequation for nonpasserinesand then usesthis es- RMRr (Watts) is calculatedfrom the unadjustedmass (1.68 timate in the flight-costcalculations. _The accuracy of allo- kg) which is assumedto be a goodapproximation of lean metricequations for predictingBMR of falconiformeshas bodymass (M). The sum of FFC and RFC is the daily fat b•enquestioned byse;eeral investigators {•Vasser • 9•56, l•n- consumed(DFC [kg hedy•-nct t2•essaman 1991). Therefore, we replacedthe pre- The modelwas run iteratively,with DFC subtractedfrom dictedBMR in Pennycuick'smodel with an activephase initial bodymass after eachiteration (one day of migration). restingmetabolic rate (RMRa) basedon energeticsmea- The programterminated when the new masswas < 1.68 kg. surementsof osprey.We estimatedRMRa to be 1.24 x Using the daily output, DDT, DFC and the number of RMRr whereRMRr is the meanresting metabolic rate (3.69 iterationswere individually summedfrom the termination W kg-l) of threecaptive osprey from Florida(Wasser 1986). point until the sumof DDT = 3780 km. The sumsof DFC We increasedRMRr by 24% becauseactive-phase measure- andthe numberof iterationswere usedas the predictionsfor ments (RMRa) average24% higher than resting-phase total fat consumedand total migrationtime, respectively. (RMRr) measurementsin otherraptors (Kennedy and Ges- SensitivityAnalysis of the FastingModel. We compared saman 1991). changesin predictedmigration times and fat consumedunder Pennycuick'smodel uses these inputs of morphometrics fastingconditions to uncertaintiesin the followingmodel and flight characteristicsto calculateglide superpolar (GSP parameters:lean body mass, wingspan, wing area, wind speed, [dimensionless]),which describesthe relationshipbetween climb, altitude, daily flight time, total migration distance, glidingsink speedand air speed,and interthermalvelocity RMRr, and the energeticcost of glidingflight (Table 1). We (Vit [m s-l]). Vit is the velocitythe ospreytravels gliding were interestedin discoveringhow robustthe modelpredic- betweenthermals. Pennycuick's model provides two estimates tionswere to uncertaintiesinherent in theseparameter es- of Vit;Vopt (m s-l) andVt• (m s-l). Vop•is optimalinterther- timatesand thus, identify thoseparameters that must be mal speedwhich maximizes distance traveled per unit time, carefullyestimated empirically in future research. andVt• isa slowerinterthermal speed which maximizes glide Parameteruncertainties were estimated by assigningupper distance.Interthermal velocities of ospreymigrating in central and lower parameterbounds and running the model sepa- New York (Kerlinger1989) suggest they use an interthermal ratelyvarying one parameterat a time to its upper or lower speedapproximately 1-2 m s-l lessthan Vopt.This would bound (Table 1). The boundsfor morphometricparameter allow them to migratefaster than if they choseV•g, while valueswere basedon rangesof measurementsof western 88 GRADY L. CANDI•ER AND PATRICIA L. KENNEDY VOL. 29, NO. 2

Table 1. Resultsof the sensitivityanalyses of all the parametersincluded in the fastingmodel for migratingosprey. Nominal and boundaryvalues are presented.

M•GR•T•ON FAT CONSUM•.D(g) PA•M•.T•.•. (UmTS) NOMn•AL RANGE T•MV. (d) (% FAT Dv.NS•TY) Lean mass(kg) a 1.68 1.35-2.03 12-13 196(14)-293(14) Wing area (m2)a 0.30 0.24-0.33 12 239(14)-248(15) Wing span(m) a 1.49 1.34-1.58 12 243(14)-247(15) Altitude (m)b 2000 1000-3000 12 240(14)-249(15) Climb (m s-•) c 3 2.5-3.5 11-13 225(13)-267(16) Wind speed(m s-•) d 0 -5.0-+5.0 8-21 164(10)-447(27) Flighttime (hr d-•) b 8 6-10 10-16 205(12)-284(17) Migration distance(km) e 3780 2940-4620 9-15 185(11 )-304(18) RMR (W) t 6.2 4.0-6.2 12 163(10)-247(15) Cost of flight (x RMRa)g 2 2-4.5 12 247(15)-416(25) a Nominal value from Kerlinger (1989) and boundsare from M. Bechard(unpubl. data). b Estimatedvalue (see text). c Nominal value from Kerlinger et al. (1985 in Kerlinger 1989). Boundswere basedon climb ratesobserved in other soaringhawks by Kerlingeret al. (1985 in Kerlinger1989). d Estimatedvalue (see text). Positivevalues indicate a headwind and negativevalues indicate a tail wind. e Nominal valueis the meanof 21 bandrecovery distances from ospreybreeding in the IntermountainWest (Melquist et al. 1978, Melquist and Johnson1984, Johnsonand Melquist 1991). Upper and lower boundsare _+1SD. f Nominalvalue is the averageresting metabolic rate (RMR) measuredfor threecaptive osprey during their resting phase (Wasser 1986) The lowerbound is basedon Wasser'sallometric equation for estimatingRMR for Falconiformesduring their activephase. The nominal value was alsoused as the upper boundbecause these measurements are higherthan predictedby any publishedallometric equation for estimatingRMR for nonpasserinesduring their activephase. g Valuesare basedon Baudinetteand Schmidt-Nielsen's(1974) measurementsof energeticcosts of glidingflight in herringgulls (Larus argentatus).The lower boundis basedon the measureddifference between resting metabolic rate and glidingmetabolic rate in the wind tunnel. The upper boundis the differencebetween the measuredgliding metabolic rate and the restingmetablic rate calculatedusing Lasiewskiand Dawson's(1967) equationfor nonpasserines.The lowerbound was usedfor the nominalvalue since metabolic rate during flight was calculatedusing active phase RMR. osprey(M. Bechardunpubl. data). Boundsfor the flight To estimatefat-deposition rates at stopovers,maximum characteristicsand wind speedwere basedon publishedvari- dailyfat depositionrate (FDRm•) wascalculated using Lind- ationsof theseparameters for migratingosprey or otherlarge str6m's(1991) Eq. 1: migratingraptors if dataon ospreywere not available(Ker- (DMEmax- DEEmin) (100 n) linger1989, Kerlinger and Moore 1989).The nominalvalue x-- for RMRr was usedas its upper boundbecause Wasser's FDRmax= M E ' (1986) metabolicmeasurements on ospreywere higherthan where FDRm• is a percentageof lean bodymass, DMEm• the predictionsfrom allometricequations. The lowerbound ismaximum daily metabolizable energy intake in kj, DEE,an for RMRr was estimatedwith Wasser's(1986) equationfor is minimumdaily energy expenditure in kJ, M is leanbody restingphase RMRr in .This equationpredicts massin kilograms,n is the conversionefficiency of metabo- lower valuesfor RMRr than do the equivalentallometric lizedenergy into fat (0.88[Kersten and Piersma 1987]), and equationsfor nonpasserincs(Aschoff and Pohl 1970, Las- E is the energycontent of storedfat (3.96 x 104kJ kg-•). lewski and Dawson 1967). Like Lindstr6m,DMEm• was calculatedusing Kirkwood's ForagingMigration Models.If theseosprey chose not to (1983) allometricequation; however we calculatedDEEr•n fast duringthe entiremigration, they would haveto spend as: timeforaging daily or breakup theirtrip intoseveral segments DEE•. = DEE + (DMEm• x 1%), (Eq. 1) that are separatedby stopoversto replenishtheir fat reserves. To comparefasting migration to foragingmigration, the fast- whereDEE is theminimum daily energy expenditure of 1.5 lng modelwas modified to estimatetotal migration time and BMR predictedby Lindstr/3m(1991). The additionalterm stopoverfat depositionfor jump or hop strategies(Piersma in Eq. 1 (DMEm• x 1%) estimatesthe additionalenergy 1987). Migration modelsusing the jump strategyincluded expendituresincurred by huntingosprey (kJ d-•) (Machmer migrationswith onestopover (after 1890km) andthree stop- and Ydenberg1990). overs(one every 945 km), while the hop strategyincluded Lindstr6m and Alerstam (1992) observedthat at fivestopovers (one every 630 km) and11 stopovers(one every stopoversites lose body mass during the first day at the site 315 km--the daily flight distancepredicted by the foraging and/or there is a time lag beforefat depositionstarts, but model). Moore and Kerlinger (1987) found that weight gain can JUNE 1995 OSPREYMIGRATION STRATEGIES 89

0,6 35

>, 40 X / 0.5

_• 30 0.4 z 0 •_ 7 •o f"h • 15

_• :• lO -.J lO 0 _j 0.1 o 0

o -- 0 0 • 2,000 4,000 6,d00 8,000 o 1 3 DISTANCE TRAVEllED (kin) NUMBER OF STOPOVERS

Figure 1. Fastingmodel predictions of totalmigration time Figure 2. Foraging model predictionsof total migration at three differentwind speeds(dashed lines) and required time with zero, one, three, five or 11 stopovers.The cross- initial fat deposits(solid line) for an ospreybreeding in the hatchedportion of eachbar indicatesthe predictedtime in IntermountainWest to completea 3780 km southwardmi- flight,the hatchedportion indicates the predictedtotal time gration.The dottedlines to the X and Y axesindicate the foragingat stopovers,and the unshadedportion indicates the predictedduration and amount of premigratoryfat deposition predictedtotal settling costs (time spentsettling at stopovers for migrationsunder nominal conditions. [1 d stopover-1] plusthe extra time spentforaging to cover the energyexpenditures of the dayssettling at stopovers).If there are no settlingcosts, then the total migration time is occuron their arrivalday at a stopover.Therefore, we eval- representedby the sum of the cross-hatchedand hatched uatedthe foragingmigration using three stopover strategies: portionsof the bar. (1) stopoverswith one day of settlingbefore weight gain begins;(2) stopoverswith weightgain beginningon the first day of stopover;and (3) foragingdaily without stopovers. In tail wind (Fig. 1). The amountof time a flies thefirst strategy we assumed(1) osprey'senergy expenditures duringthe day and total distance migrated also resulted while settlingwere equal to DEE and (2) they arrivedat in largechanges in modelpredictions. Flying 6 hr d-1 stopoverswith enoughfat to spendone day without foraging. wouldincrease migration time 33% and fat consumed RESULTS 13%, while changingflight time to 10 h d-1 reduces Fasting-modelPredictions. Fasting-model predic- the predictedmigration time by 17%and fat consumed tionsare summarizedin Fig. 1. The solidline in Fig. by 17%. An 840-km variationin migrationdistance 1 representsmodel predictions for totalfat consumption changedmigration time by 25% and fat consumedby using nominal valueslisted in Table 1. Dashed lines 23-25%. Uncertainties associated with estimates of are model predictionsfor distancetraveled with the RMR andenergetic cost of flightdid not changethe nominalvalue and the lower and upper boundsfor predictedmigration time but did vary the predicted fat wind speed.As indicatedby the dottedlines in Fig. 1, consumedby asmuch as 34%and 68%, respectively. we predict that a 1.68-kg ospreywould take 12 d, Uncertaintiesassociated with the other parameters consume0.25 kg of fat (a fat densityof 15% of lean causedpredicted migration times to vary by no more bodyweight), to completea fastingmigration of 3780 than 1 d and predictedfat densityto vary by only 1- km if it migratedwith no head or tail wind. 2%. SensitivityAnalysis of FastingModel. The results Foraging-modelPredictions. We estimatemaxi- of the sensitivityanalysis of the fastingmodel are listed mum daily fat depositionrate (FDRm•_,)to be 2.2% in Table 1. Variation in wind speedcaused the greatest (0.04 kg d-1) and that it would take 7 d of maximum changein predictedfat consumptionand migration energyintake to deposita premigratoryfat densityof time, with a 3780 km migrationtaking 21 d and 0.45 15% of lean bodymass (predicted fat densityfor a kg of fat (27% fat density)in a 5 m s-1 headwind, or fastingmigration of 3780 km). In comparison,based 8 d and 0.16 kg of fat (10% fat density)in a 5 m s-1 on LindstriSm'sEq. 1 (1991) and Lasiewskiand Daw- 90 GRADY L. CANDLERAND PATRICIA L. KENNEDY VOL. 29, NO. 2 son's(1967) allometricEq. E for estimatingpassedfie with no head wind is much less than the shortest times BMR, a 0.015-kg passedfnewould requireonly 2 d observedbetween banding and recovery,but may be of maximum energyintake to deposita comparable comparableto actualmigration times if ospreyleaving premigratoryfat density.Using Machmer and Yden- Idahoin the beginningof Septemberarrive in central berg's(1990) estimatesof 218 kJ averagenet energy Mexico by mid-September. gainper prey,an ospreywould require 11 fishper d, The model'sprediction of 21 d for a fastingmigra- or a total of 77 fish,to providethe fat storesrequired tion with a 5 m s-1 head wind is comparableto em- for a 3780 km fastingmigration. The additionof stop- pirical estimatesof migrationtimes, but the predicted oversto the migrationwould increasethe predicted fat densityof 27% is much higher than measuredfat time investmentto 15, 16, 17, and 18 d with one,three, densitiesin raptors.These predictions suggest that os- five,and 11 stopovers,respectively (Fig. 2). However, prey from the IntermountainWest probablydo not if there is settlingtime at eachstopover, the total mi- migrateentirely into head winds of thisintensity. There grationtime is predictedto increaseto 17, 21, 25, and are no observationsof wind conditionsalong osprey 34 d, respectively. migrationroutes in the IntermountainWest to evaluate Model Evaluation.Although premigratory fat den- our conclusion.However, observations by Hall et al. sitiesare not availablefor osprey,the predictedfat (1992) demonstratethat significantlymore coastal mi- densityof 15%is very dose to themaximum fat density grantsin California migrate southwith no wind or measuredin comparablysized Europeangoshawks tail winds than with head winds. (Accipitergentilis) (16.4% in a 1.67-kgfemale, Marc- The additionof fiveor morestopovers with settling strSmand Kenward1980). This fat densityprediction costsincreased the total migration time to 25-34 d •s alsowell below the premigratoryfat densitiesthat (Fig. 2), which is 2-3 timesthe predictedmigration have been observedin warblers and waders (>50%, time while fastingand slightly longer than the empir- Blem1980), and the predicted fat densitiesof migrating ical migrationtime estimatesbased on bandrecoveries. buteos(Smith et al. 1986). American kestrels(Falco With one or three stopovers,the predictedtotal mi- sparverius)were foundto havefall fat densities2-4% grationtime is 17 or 21 d, respectively(including set- higherthan mid-summervalues (5.3-7%, Gessaman tling costs).It seemsunlikely that ospreywould make 1979),but Kirkley andJones (unpubl. data) point out frequentforaging stops that includeda day of settling that these fat levels are not maximal, and should be and a day of foraging.Settling times are mostlyob- consideredwintering fat sincethey are maintained servedin territorial migrantsthat travel in large, in- throughoutthe winter. Obviously,more quantitative traspecificflocks (LindstrSm and Alerstam 1992). Be- dataon premigratoryfat depositionin raptorswill be causeosprey migrate individuallyor in small flocks necessaryto determineif the fat depositionpatterns (Kerlinger1989), it is possiblethey do not experience predictedby this modeloccur in nature. settlingcosts. Without settlingcosts the total migration Basedon the disappearanceof residentbirds from time with multiplestopovers is predictedto be 15-18 nestingterritories, osprey migrations in northernIdaho d (Fig. 2). beginas early as the first week of September(Melquist Without these settling costs, frequent foraging et al. 1978, Melquist and Johnson1984). Peak num- throughoutmigration could occur with little affecton bersof ospreycounted at a migrationstation in north- migrationtime if ospreyforaged and flew in the same ern Utah occurduring the secondand third week of day. Althougha 1.68-kgosprey may only be able to September(Hoffman 1990). Ospreyfrom the Inter- metabolize11 fish d-1, it may not take them a whole mountainWest are widespreadin Mexicoby late Sep- day to catchthose fish. Osprey have one of the highest tember (Melquist and Johnson1984), with banded capturesuccess rates of any raptor (Newton 1979). first-yearmigrants recovered in centralMexico as early Swenson(1978) observedaverage fish catchrates for as17 September(2700 km frombanding site [Melquist ospreyto be 8.8-19.7 min fish-l, and Machmer and and Johnson1984]) and adultsrecovered as early as Ydenberg(1990) observedan averageof 10.3 min 28 Septembereven further south (4200 km fromband- fish-•. Therefore,with goodforage availability,an ing site[Melquist and Johnson 1984]). Osprey banded ospreymay require (2 hr to catchits metabolicmax- as nestlingsfrom 14 July to 2 Augustwere recovered imum.Alternatively, osprey may only needabout five 46-135 d later, after travellinga maximumof 3500 fish d-1, or 45-100 min of daily foraging,if their km fromthe banding site (Melquist and Johnson 1984). strategyis to coverthe costsof a singlemigration day The predictedtravel time of 12 d for a fastingmigration and not depositfat for subsequentdays. If there are JUNE1995 OSPREYMIGRATION STRATEGIES 91 adequatewater sourcesen route, e.g., reservoirsand to determineif this modeladequately presents the ma- lakes,it is possiblethat an ospreycould forage daily jor factorsinfluencing fat depositionstrategies. Total withoreaffecting its daily time in flight and total mi- bodyelectrical conductivity (TOBEC) has beenused grationtime by hunting at the beginningor end of to estimatefat in live .However, a recenteval- eachday beforethe thermalswere strongenough for uation of this methodologyby Skagenet al. (1993) soaring. indicatesTOBEC accuratelymeasures lean bodymass bm its lipid estimateshave numerous potential errors. CONCLUSIONS Othertechnologies are beingevaluated to estimatelip- Our foraging-(without settlingcosts) and fasting- ids of free-ranginganimals in a non-invasivemanner modelpredictions are consistentwith the limited data (J.A. Gessamanpers. comm.). With thesenew tech- availableon fall migrationstrategies of ospreybreeding nologies,osprey migration strategies can be determined in the IntermountainWest. Our resultsalso suggest empiricallyand thesedata canbe usedto evaluatethe that, undercertain assumptions, nonstop migration may validityof our models. be energeticallypossible for westerninterior osprey. Whether or not stopoversare usedby theseosprey is ACKNOWLEDGMENTS probablya functionof the foodavailability en rome. We would like to thank Thomas Alerstam, Donald R. Hop strategiesare generallythought to be the most Johnson,Paul Kerlinger,John S. Kirkley, Colin J. Pen- nycuick,and SusanK. Skagenfor their usefulcomments on favorablefor conservingenergy (Piersma 1987) bm earlier draftsof this manuscript. probablyrequire plentiful food resources en route.In ecologicallyunfavorable situations, e.g., low foodavail- LITERATURE CITED ability, fasting or jump strategiesare more likely ASCHOFF,J. ANDH. POHL. 1970. Der Ruheumsatzvon (Johnsonand Herter 1990, Berthold1993). Anecdotal Vogelnals Funktion der Tageszeitund derKorpergrosse. observationsof ospreycarrying fish during migration J. Ornithol. 111:38-47. (Kerlinger1989) combinedwith few foragingareas in ALERSTaM,T. 1990. Bird migration. CambridgeUniv. the semi-aridSomhwest and Mexico would suggesta Press,Cambridge, U.K. BAUDmETrE, R.V. AND K. SCHMIDT-NIELSEN. 1974. En- jump strategyas the beststrategy for ospreymigrating ergy costof glidingflight in herringgulls. Nature 248: from the Intermountain West. 83-84. Although our predictionsare consistentwith em- B•KEIt, R.R. 1982. Migration paths through time and pirical observations,our modelwas not evaluatedrig- space.Holmes and Meier Publishers,Inc., New York, orouslybecause of the absenceof observationson mi- NY U.S.A. grationstrategies of individualosprey. Programmable BERTHOLD,P. 1993. Bird migration.A general survey. satellite transmitters are now available that are small Oxford Univ. Press, Oxford, U.K. enough(<30 g) for mediumto large raptors,which BLF_J•t,C. 1980. The energeticsof migration, pages175- allow investigatorsto measuremigration and stopover 224 in S.A. Gauthreaux [ED.], migration,ori- times,and identify migration routes of individualbirds entation,and navigation. Academic Press, New York, NY U.S.A. (R.E. Ambrose, M.W. Britten, P. Howie and P.L. CXSTt•O,G. ANDJ.P. MYEr. 1989. Flight rangeestimates Kennedyunpubl. data). This new technologycan also for shorebirds. Auk 106:474-476. be usedto measurethe daily time in flight, a key GESS•d•, J.A. 1979. Pre-migratoryfat in the American parameterin our model.Based on our sensitivityanal- kestrel. Wilson Bull. 91:625-626. ysis,precise estimates of wind speedsat the locationof GOnDSTmN,D.L. AND N.G. SMrrH. 1991. Responseto a bird in flight are alsonecessary for accuratepredic- Kirkley. J. RaptorRes. 25:87-88. tions,but this type of informationwill probablynot GUDIvlUND$SON,G.A., fk. LINDSTR/•M,AND T. ALERSTAM. be available until anemometers can be attached to the 1991. Optimal fat loads and long-distanceflights by satellitetransmitters. Anemometers on ospreywould migratingknots Calidris canutus, sanderlings C. albaand measureairspeed moving past the bird (Va) while the turnstonesArenaria interpres. Ibis 133:140-152. HALL, L.S., A.M. FISH ANDM.L. MORRISON.1992. The satellitetransmitter would indicate changes in location, influence of weather on hawk movements in coastal north- allowingthe calculationof groundspeed (Vg). Vg - ern California. Wilson Bull. 104:447-461. Va could be usedas an estimateof the wind velocity HENNY,C.J. ANDW.Y. VAN VELZER. 1972. Migration betweenmigration locations. patternsand winteringlocalities of Americanosprey. J. Empirical data on premigratoryand stopoverfat Wildl. Manage.36:1133-1138. deposition(if any) by migratingosprey are alsoneeded HOFFMAN,S.W. 1985. Raptormovements in inlandwest- 92 GRADY L. CANDLER AND PATRICIA L. KENNEDY VOL. 29, NO. 2

ern North America:a synthesis,pages 325-337 /n M. commentson the migration of northernIdaho and eastern Harwood [ED.], Proceedingsof hawk migrationconfer- Washingtonospreys. J. FieldOrnithol. 55:483-485. ence IV. Hawk Migration Assoc.of North America, --, AND W.D. CA_•tIEI•. 1978. Migration Medford, MA U.S.A. patternsof northernIdaho and easternWashington os- 1990. Southwestregion. ]. Hawk Migr. 16:87-89. preys.Bird-Banding 49:234-236. JOHNSON,D.R. AND W.E. MELQUIST. 1991. Wintering MOORE,F. AND P. KERLINGER.1987. Stopoverand fat distributionand dispersalof northernIdaho and eastern depositionby North Americanwood-warblers (Parul'.'- Washingtonospreys. ]. Field Ornithol.62:517-520. nae)following spring migration over the Gulf of Mexico. JOHNSON,S.R. ANDD.R. HEI•TEI•. 1990. Bird migration Oecologia74:47-54. in the arctic:a review.pages 22-43 in E. Gwinner [ED.], NEWTON,I. 1979. Populationecology of raptors.Buteo Bird migration:the physiologyand ecophysiology. Spring- Books,Vermillion, SD U.S.A. er-Verlag, Berlin, Germany. •SSTEWL61•l•,S. 1977. Migration,wintering areas, and site ,•KENNEDY,P.L. AND J.A. GESSAM•. 1991. Diurnal resting tenacityof the Europeanosprey Pandion h. haliaetus(L.). metabolicrates of accipiters.Wilson Bull. 103:101-105. Ornis Scand. 8:61-78. KEgLINGER,P. 1989. Flight strategiesof migratinghawks. PENNYCraCK,C.J. 1989. Bird flight performance.A prac- Univ. ChicagoPress, Chicago, IL U.S.A. tical calculation manual. Oxford Univ. Press, Oxford, -- AND F.R. MOORE. 1989. Atmosphericstructure U.K. and avian migration.Curr. Ornithol.6:109-142. Pml•SMA,T. 1987. Hop, skip, or jump? Constraintson KERSTEN,M. ANDT. PIERSMA.1987. High levelsof energy migrationof arcticwaders by feeding,fattening, and flight expenditurein shorebirds,metabolic adaptations to an speed.Limosa 60:185-194. energeticallyexpensive way of life. Ardea75:175-187. 1990. Pre-migratory"fattening" usually involves KIRKLEY,J.S. 1991. Do migrantSwainsoffs hawks fast en more than depositionof fat alone.Ringing & Migr. 11: routeto Argentina?J. RaptorRes. 25:82-86. 113-115. KInKWOOD,J.K. 1983. A limit to metabolizableenergy POOLE,A.F. ANDB. AGLER. 1987. Recoveriesof ospreys intakein mammalsand birds.Comp. Biochem. Physiol. A bandedin the United States,1914-84. ]. Wildl.Manage. 75:1-3. 51:148-155. LASIEWSKI,R.C. AND W.R. DAWSON. 1967. A re-exami- ROBBINS,C.T. 1993. Wildlife feedingand nutrition. Ac- nation of the relation between standard metabolic rate ademicPress Inc., San Diego, CA U.S.A. and bodyweight in birds.Condor 69:13-23. SK•GEN, S.K., F.L. KNO?I• AND B.S. CADE. 1993. Esti- LINDSTR•3M,fk. 1991. Maximumfat depositionin mi- mationof lipidsand lean massof migratingsandpipers. gratingbirds. Ornis Scand. 22:12-19. Condor 95:944-956. -- AND T. ALEl•STAM. 1992. Optimal fat loads in SMITH, N.G., D.L. GOLDSTEINAND G.A. BARTHOLOMEW. migratingbirds: a testof the time-minimizinghypothesis. 1986. Is long-distancemigration possiblefor soaring Am. Nat. 140:477-491. hawksusing only storedfat? Auk 103:607-612. -- .$,NDT. PIERSMA.1993. Masschanges in migrating SWENSON,J.E. 1978. Preyand foraging behavior of ospreys birds:the evidencefor fat and protein storagere-exam- on YellowstoneLake, Wyoming.J. Wildl. Manage.42. ined. Ibis 135:70-78. 87-90. LUKe,A. AND D.A. SCHOELLER. 1992.Basal metabolic•WAssv. R,J.S. 1986. The relationship ofenergetics offai- rate, fat-free mass,and body cell massduring energy coniformbirds to bodymass and dimate. Condor88:57- restriction. Metabolism 41:450-456. 62. MACHMER, M.M. AND R.C. YDENBERG. 1990. Weather WELHAM,C.V.J. 1994. Flight speedsof migratingbirds. and ospreyforaging energetics. Can. ]. Zool.68:40-43. a test of maximum range speedpredictions from three MARCSTR6M,V. AND R. KENWARD. 1980. Sexual and aerodynamicequations. Behar. Ecol. 5:1-8. seasonal variation in condition and survival of Swedish goshawks.Ibis 123:311-327. MELQtaST,W.E. AND D.R. JOHNSON.1984. Additional Received11 July 1994; accepted17 January 1995