J. Field Ornithol., 66(4):590-608

THE ALTITUDE OF MIGRATION IN EAST-CENTRAL ALASKA: A RADAR AND VISUAL STUDY

BRtAN A. COOPER ABR, Inc. P.O. Box 249 ForestGrove, Oregon 97116 USA

ROBERTJ. RITCHIE ABR, Inc. P.O. Box 80410 Fairbanks, Alaska 99708 USA

Abstract.-- altitudesof were monitored in east-centralAlaska during spring and fall migration periodswith visual (1987-1989) and radar (1988-1989) methods.Visual ob- servationsindicated that diurnal flightsoccurred primarily below 300 m aboveground level (agl). Radar observationsindicated that nocturnal generallyoccurred below 500 m agl. Flight altitudeswere significantlyhigher in fall than in spring,and there were interannual differencesin both seasons.There was high night-to-nightvariability in nocturnal flight al- titudesduring both seasons.This night-to-nightvariability probably was not related to daily changesin the magnitude of migration: there was a very low correlation between mean nightly flight altitudesand mean nightly migration rates. Nocturnal flight altitudeswere higher from April to earlyMay than from mid- to late May and higher from late Augustto early Septemberthan from mid-Septemberto mid-October.These seasonaldifferences in flight altitudeslargely were due to changingspecies composition within a season.When springradar data were partitionedamong light conditions,flight altitudeswere found to be similarbetween daylight and crepuscularperiods and betweendaylight and nocturnalperi- ods, but birds flew significantlyhigher in crepuscularthan nocturnal periods.In fall, noc- turnal altitudeswere significantlyhigher than crepuscularand daylightaltitudes, which were similar.Within a night, there was no near-midnightpeak in flight altitudesin either spring or fall.

LA ALTITUD DE LA MIGRACION DE AVES EN LA ALASKA CENTRO-ORIENTAL: UN ESTUDIO VISUAL Y DE RADAR $inopsis.--$e monitorearonlas altitudesde vuelo de avesen la regi6n centro-orientalde Alaskadurante los perVodosde migraci6nde la primaveray el otofio con m•todos visuales (de 1987 a 1989) y de radar (en 1988 y 1989). Observacionesvisuales indicaron que los vuelosdiurnos ocurrieron principalmentebajo los 300 m sobre el nivel del terreno (snt). Observacionesde radar indicaron que los vuelosnocturnos generalmente ocurrieron sobre los 500 m snt. Las alturas de los vuelosfueron significativamentem5_s altas en otofio queen primavera,y existierondiferencias interanuales en ambosperiodos. Se document6una alta variabilidaden las alturasde vuelo de noche a noche en ambasestaciones. La probabilidad de esta variabilidadentre noche y noche no estuvorelacionada a cambiosdiarios en la magnitud de la migracitn: se obtuvouna correlacitn muy pobre entre el promedio de las altitudesde vuelo nocturnosy el promediode las tasasde migracitn nocturna.Las altitudes de vuelonocturno fueron mayoresentre abril y principiosde mayoque entre mitad y finales de mayo y fueron mayoresentre finales de agostoy principios de septiembre que entre mediadosde septiembrey mediadosde octubre. Estasdiferencias estacionales en altitudes de vuelo fueron causadasprincipalmente a los cambiosen la composicitnde especiesdentro de una estacitn. Cuando los datos primaveralesde radar se dividieron entre las condiciones de luz, se hall6 que las alturasde vuelo eran similaresentre luz de dfay los per/odoscre- puscularesy entre luz de dia y per/od0snocturnos, pero pocasaves volaron significativamente m5_salto en periodos crepuscularesqueen periodosnocturnos. En otofio, las altitudesnoc-

590 Vol.66, No. 4 Altitudeof in Alaska [591 turnasfueron significativamentemayores que las altitudescrepusculares y diurnas,las cuales fueron similares.Dentro de una noche no se evidenciaronpicos en altitudesde vuelo cer- canosa la medianocheen primaverao en otofio. Several speciesof birds are prone to collisionswith structures (Avery et al. 1980, Banks1979, Weir 1976). Collectively,these collisions may be a significantmortality factor for some populations:approximately 1.25 million birds are killed each year in collisionswith tall structures(Banks 1979). Declinesof severalspecies of Neotropicalmigrants in recent de- cades (Robbins et al. 1989, Terborgh 1989) suggestthat reducing colli- sion-causedmortality in these speciesmay even be more critical than was previouslythought. Consequently,information on the altitude at which birds fly is important, not only for understandingbird migration but when planning the constructionof tall structuresand utility lines, to minimize bird collisions. Unfortunately, the altitude of migration is extremely variable, depend- ing upon species,location, geographicfeature, season,time of day and weather condition (Alerstam 1992, Kerlinger 1989). Despite numerous studies,our knowledgeof species-specific,diel, seasonaland geographic variation in migration altitude is limited. Further, few data on the migra- tion altitude of birds are available for Alaska. The purpose of this paper is to summarizeseasonal, diel and species- specificpatterns of flight altitude of birds during springand fall migration in east-centralAlaska, determined from visualobservations during 1987 to 1989 and from radar observations in 1988 and 1989. The visual obser- vationsprovided data on flight altitudes for individual species,whereas the radar data providedmore precisealtitudinal information for both day and night (when visual observationswere not possible).These data were collected as part of a study designed to assessthe potential impact of a large antenna array on birds during migration. Detailed analysesof the effectsof weather and other variableson flight altitude will be presented in another paper.

STUDY AREA The studyarea waslocated in the upper Tanaria River Valley,near Tok, Alaska (63ø20'N, 142ø59'W;Fig. 1). The Tanana River Valley,which ex- tends northwestward from the Yukon Territory of Canada into central Alaska,is a major migration corridor for many speciesof birds (e.g., Bell- rose 1976, Gabrielsonand Lincoln 1959, Irving 1961, Kessel1984, Sladen 1973, Spindler and Hall 1991, West et al. 1968). The studyarea is part of the Northway-TanacrossLowland (Wahrhaftig 1965) and is nearly fiat, ranging in elevationfrom 530 to 560 m asl (above sea level). In the studyarea, the valley is approximately15 km wide and is bordered to the north by broad hills (the Yukon-TananaUplands) that crestat approximately750 m asl and to the south by mountainsof the Alaska Range that crest at 2000-3000 m asl. The area is forested with large tractsof black spruce (Piceamariana), white spruce (Piceaglauca), and quaking aspen (Populustremuloides). 592] B.A. Cooperand Pt. J. Ritchie j. FieldOrnithol. Autumn 1995

AL•SVO• ß TOK STUDY AREA

GENERAL FLYWAY •"-- BOUNDARIES • MIGRATION ROUTES

km

g-•tl00 200I'--I 300 400I .500I

Ficum•1. Mapof Alaskashowing location of studyarea and major migration pathways. Locationsof pathwaysdetermined from Bellrose (1976), Gabrielson and Lincoln (1959), Irving (1961), Kessel(1984), Sladen(1973), Spindlerand Hall (1991),West et al. (1968),S. Ambrose (USFWS, pers. comm.),J. King (USFWS, retired, pers. comm.) and R. King (USFWS,pers. comm.).

METHODS VisuaL--Visualobservations were conducteddaily 10 Apr.-24 May 1987,16 Aug.-6 Oct. 1987, 6 Apr.-21May 1988, 16 Aug.-17 Oct. 1988, 5 Apr.-25May 1989, and 16 Aug.-15Oct. 1989.Observers equipped with binocularsand a variable-powerspotting scope made observations from a 6-m-hightower or fromnatural vantage points for the entiredaylight periodduring peak migration (i.e., approximately 20 Aprilto 20 Mayand 15 Augustto 5-15 October)and for approximately6-10 h/d at other times. Systematicscans from zenith to horizon in all directionswere made when searchingfor birds.Data were recordedas each flock of birds (mi- grantspecies only) crossed an imaginaryNorth/South transect extending from the observer (or an East/West line for birds headed to the North or South;note that the predominantdirection of flight is westwardor northwestwardin springand eastwardor southeastwardin fall). For each flock,the followingdata were recorded: time, species, number of birds, flightaltitude category above ground level (agl; 0-15 m, 16-30 m, 31- Vol.66, No. 4 Altitudeof Bird Migration in Alaska [593

150 m, 151-300 m, and •300 m agl), distance to bird, and flight direc- tion. Flight altitude and distance to birds were estimated with the aid of severalreference points of known height and distance:trees marked with colored flagging,human-made structures (e.g., 220-m-tallLoran towers and 15-m-tall telephone poles) and nearby geological features (e.g., mountains and hills). When possible,visual estimatesof altitude were checkedagainst altitudes measured with radar. For all statisticalanalyses, "flocks" and not birds were used as the sam- pling unit, becauseof problemsof independenceamong individuals with- in a flock. A "flock" wasdefined as two or more birds flying in proximity, or any birds flying singly.Thus, individual birds and different-sizedflocks had equal weight for statisticaltests. Further, our data set was restricted to migrant speciesthat flew •500 m horizontal distancefrom the observ- er, becauseit was difficult to detect low-flyingbirds at greater distances. This lack of detection would bias the altitude data and, hence, would inflate our estimatesof mean flight altitudes. Mean flight altitudeswere computedby usingthe midpoint of the flight altitude categoryin which a flock wasobserved. The upper flight altitude category (•300 m) was problematic becauseit did not have an upper boundary.To solvethis problem, we computed a median flight altitude for targetsabove 300 m that were observedduring daylight hours on radar. Half of all flights measuredon radar above300 m during daylight hours occurred below 350 m in spring and below 650 m in fall. Thus, 350 m was used as the midpoint of the •300 m altitude category for springobservations and 650 m wasused as the midpoint for fall. Radar.-•Three models of Furuno X-band marine radars (FR-1900, FR- 8050 and FR-8100) also were used to measureflight altitudes.Different modelswere used in different seasons,depending upon their availability. Eachwas modified with a fixed, vertically-alignedparabolic antenna. Peak power output was3 kW, 5 kW and 10 kW for thesethree models,respec- tively.The maximal range of detection for birds on these radars was un- known;however, small individual could be detectedto approx- imately 1 km, and large speciesand flocksof passerinescould be detected at higher altitudes.For a descriptionof the radar systems,see Cooper et al. (1991). We operated the radar 2-15 h/d during 6 Apr.-21 May 1988 (FR-1900), 17 Aug.-17 Oct. 1988 (FR-1900),12 Apr.-23 May 1989 (FR-8050)and 16 Aug.-17 Oct. 1989 (FR-8100).During periodsof peakmigration, we op- erated the radar from dusk to dawn and for at least 2 h during daylight, to provide information on diel patternsof flight altitude. Data could not be collectedduring periodsof rain or wet snow.We recordedthe follow- ing information for each radar target: time, altitude and light condition. Information on light condition (daylight, crepuscular[the dawn/dusk period between sunrise/sunsetand nocturnal periods], and nocturnal [the periodswhen it was dark enough that one could not read a typed 594] B. A. Cooperand tL J. Ritchie j. FieldOmithol. Autumn 1995 page at arms' length]) was collected at the beginning of each hour of sampling. We used the 1.4-km range setting (actual range = 1900 m) on the FR- 8100 and FR-8050 radars and the 0.9-km range setting (actual range -- 1000 m) on the FR-1900 radar, which did not have a 1.4-km setting.As 0.2% (51) of the targetsobserved in spring 1989 and 0.1% (7) of the targets observedin fall 1989 (when we used the FR-8050 and FR-8100 models, respectively)flew above 1000 m (the vertical range of the FR- 1900 that we used in 1988 was 1000 m), all analyseswere restricted to include only thoseobservations below 1000 m agl, so that data from both 1988 and 1989 could be combined into a single data set. The speciescomposition and size of a flock of birds observedon the radar usually was unknown. Therefore, the term "target," rather than "flock" or "individual" is used to describebirds detected by the radar. For data summariesand analyses,flight altitude data were correctedbe- causethe vertical radar had a beam that was approximatelycone-shaped at the altitudeswe sampled,which meant that the area sampledincreased with increasingaltitude. Refer to Blokpoel(1971) for a detailedexpla- nation of correction factorsfor a vertically-alignedradar. As the lower limit of detectionfor the verticalradar wasapproximately 30 m agl, we sampledthe 0-29-m agl zone at night with a 5 X Noctron V night-visionscope from a 6-m-high tower. Observationswere conducted for 1-2 h/night within the seasonalpeak of migration for 30 nights in 1988 (16 nights in spring and 14 in fall) and 32 nights in 1989 (6 in springand 26 in fall). Observersoriented the night-visionscope toward the northeast(perpendicular to the main axis of migration), at a slight angle abovehorizontal, so that the upper.edge of the field of viewwas at 30 m agl at the distancewhere we discontinuednight-vision sampling (100 m). To determine the proportion of birds that flew below 30 m agl, the night-visiondata were convertedto flocksper m2, which was multiplied by the area that the vertical radar would have sampledbelow 30 m. Ap- plying the correctionfactor for samplingarea of a vertically-alignedradar to this number allowed us to determine the number of birds that would have been detectedby the radar below 30 m agl (if that waspossible). By dividing this corrected number (x) by x + total corrected number of targetsconcurrently observed with vertical radar, we could estimatethe proportion of birds that flew below 30 m agl during concurrent radar and night-visionsampling.

RESULTS

VisuaL--Flightaltitudes varied among species groups, although the ma- jority of birds in all groups flew below 300 m agl (Fig. 2). In general, ,geese, and cranesflew higher than did raptors,shorebirds and passerines.There wasa great amount of variabilityin flight altitudes among specieswithin each group (Appendix 1). For instance,the bimod- al distributionin altitude of raptors(Fig. 2) largelyreflected the differ- Vol.66, No. 4 Altitudeof Bird Migration in Alaska [595

SWANS GEESE

>300 151-300 31-150 16-30 0-15 r• = 6515 ri = .:'13,898 i i i • • •

DU(;KS RAPTORS >3OO n!=3•58lli ni=•36 n:!.-'- lb15i: :: • • i n•.= 2655: 151-3OO v 31-15O uJ 16-30 0-15 i i , , [ I I I I I i

SANDHILL CRANES SHOREBIRDS

>300 151-300 31-150 • : 16-30 • : .' [ : • :

0-15 n •I 1.9,1I •I I'{i -- •2,7...05I I

PASSERINES >300 151-300 31-15O 16-30 0-15 80 60 40 20 0 20 40 60 80

PERCENT OF INDIVIDUALS FZGU•2. Flightaltitudes of birdsobserved in east-centralAlaska during spring (light shad- ing) andfall (darkshading) 1987-1989. These data include only diurnal, visual obser- vations of birds recorded within 500 m (horizontal distance) of the observer. encebetween low-flying Northern Harriers(see Appendix 1 for scientific names) and high-flyingbuteos. Seasonaland/or annual differencesin flight altitudeswere found for all speciesgroups (2-way ANOVA; Tables 1 and 2). Flightaltitudes of most speciesgroups tended to be higherin fall than springand lowerin 1989 than in other years. 596] B. A. Cooperand R. J. Ritchie j. FieldOrnithol. Autumn 1995

TABLE1. Mean flight altitude (m agl) of flocks"of birds observedvisually in east-central Alaskaduring springand fall 1987-1989.These data includeonly daytimeobservations of flocks within 500 m.

Species Spring Fall Year group Mean SD n Mean SD n 1987 Swans 126 97 60 143 124 48 Geese 170 119 91 353 248 73 Ducks 63 90 139 49 66 12 Raptors 82 110 365 80 140 664 Cranes 173 108 127 224 169 61 Shorebirds 53 62 182 71 145 19 Passefines 17 23 1567 32 69 1001 1988 Swans 138 115 73 204 169 317 Geese 147 ! 13 59 354 256 61 Ducks 56 81 162 108 163 72 Raptors 67 100 309 123 205 609 Cranes 201 129 44 349 258 23 Shorebirds 32 46 192 59 150 18 Passetines 21 30 1362 23 37 1534 1989 Swans 89 61 147 150 164 125 Geese 109 81 83 294 226 120 Ducks 47 55 231 27 45 30 Raptors 52 78 252 63 126 533 Cranes 113 92 43 155 131 25 Shorebirds 30 46 626 12 18 20 Passerines 17 23 2466 27 38 2251

flock wasdefined as two or more birds flying in proximityor anv birds flying singly.

To determine if our visual observationswere biased toward low-flying birds, we compared them with concurrent radar observations,which do not have this bias for observations below 1000 m. When we divided con- current visual and diurnal radar observations into three altitudinal cate- gories (31-150, 151-300 and 301-1000 m agl; no birds were observed

TABLE2. Two-factoranalysis of variance of daytime flight altitudes of flocksa of birds ob- served•,isually in east-centralAlaska during springand fall 1987-1989. Refer to Table 1 for mean (_+SD)flight altitudesfor each seasonand year.

Season Year Season X year Species group df F P df F P df F P

Swans 1 15.02 <0.001 2 8.65 <0.001 2 1.24 0.290 Geese 1 119.57 <0.001 2 5.10 0.006 2 0.16 0.850 Ducks 1 0.24 0.624 2 9.51 <0.001 2 6.96 0.001 Raptors 1 14.37 <0.001 2 12.87 <0.001 2 9.35 <0.001 Cranes 1 19.58 <0.001 2 15.87 <0.001 2 3.06 0.048 Shorebirds 1 1.40 0.237 2 10.47 <0.001 2 3.43 0.033 Passetines 1 142.75 <0.001 2 3.97 0.019 2 22.95 <0.001

a See Table I for definition of flock. Vol.66, No. 4 Altitudeof Bird •Vligration in Alaska [597

T•X•LE3. Mean (+_SD)flight altitudes(m agl) of bird targetsmeasured on radar at night in east-centralAlaska during springand fall 1988-1989.

Spring Fall Year Mean SD n Mean SD n

1988 184 133 211 426 189 1064 1989 146 120 1198 341 199 2835

above 1000 m during concurrent observations),we found that flockshad a significantlyhigher distribution on radar than were recorded visually, both in spring (X2 = 56.5, df = 2, P < 0.001) and fall (X2 = 140.5, df = 2, P< 0.001). In spring,10% of the visualsightings and 14% of the radar observationswere above 300 m agl. In fall, 12% of the visual sightings and 28% of the radar observationswere above300 m agl. These results indicate that lower-flyingbirds were detected better than higher-flying birds when using visualobservation techniques. Radar.--Nocturnal flight altitudeson radar were significantlyhigher in the fall than spring and altitudeswere higher in 1988 than in 1989 (2- wayANOVA; Tables3 and 4; Fig. 3). Night-to-nightvariability in nocturnal flight altitudeswas high for both spring and fall; mean flight altitudes commonlydiffered by 100-200 m on consecutivenights. The night-visiondata indicated that a small percentageof birds flew <30 m agl during nocturnal hours in both spring (0.7% of all targets) and fall (2.4%). In 1989,we alsowere able to obtain information on flight altitudesbetween 1000 and 1900 m agl. Only a small percentageof the targetsobserved in spring 1989 (0.2%) and fall 1989 (0.1%) flew >1000 m agl. We also compared radar observationsof nocturnal flight altitudes among early,mid- and late springperiods (10-25 April, 26 April-10 May, and 11-25 May, respectively)and among early,mid- and late fall (15 Au- gust-5 September,6-21 September,and 22 September-15 October, re- spectively).Flight altitudes were significantlyhigher in early and mid- spring than in late spring (Table 5). In fall, altitudeswere significantly higher in early fall than in mid-fall and late fall. Both thesewithin-season patterns of flight altitudesand day-to-dayvari- abilityin nocturnalflight altitudesapparently were not related to seasonal

T•X•LE4. Two-factor analysisof variance of nocturnal flight altitudes of bird targets mea- sured on radar in east-centralAlaska during spring and fall 1988-1989. Refer to Table 3 for mean (_+SD)flight altitudesfor each seasonand year.

Source df F P

Season 1 1216.2 < 0.001 Year 1 95.3 <0.001 Season X year 1 14.2 <0.001 598] B. A. Cooperand R. j. Ritchie j. FieldOrnithol. Autumn 1995

900-999 800-899 n .• 8371!(1409) n = 10,194 (3899) 700-799 600-699 ! 500-599 400-499 300-399 200-299 100-199 30-99 I I I I 80 60 0 20 40 60 80

PERCENT OF TARGETS FIc•l• 3. Nocturnal flight altitudesof bird targetsdetected by radar (correctedfor sam- pling area) in east-centralAlaska during spring (light shading)and fall (dark shading) 1988-1989. Numbersin parenthesesare the actualnumber of targetsobserved. changesin the magnitude of migration; the correlation between mean nightly flight altitudes and mean nightly migration rates (targets/hour) wasvery low in both spring (r = -0.252, n = 46, P = 0.091) and fall (r = 0.146, n = 82, P = 0.192). Mean altitudesof flight for each hour of the day and night were highly variable and there was no strong near-midnight peak in flight altitudes (Fig. 4). Flight altitudesduring daylight,crepuscular and nocturnal pe- riods alsowere compared.Flight altitudesin springwere similar between daylightand crepuscularperiods and betweendaylight and nocturnal pe- riods, but crepuscularaltitudes were significantlyhigher than nocturnal altitudes (Table 6). In fall, nocturnal altitudes were significantlyhigher than were crepuscularand daylight altitudes,which were similar.

TABLE5. Comparisonsof nocturnalflight altitudes(m agl) of bird targetsobserved by radar during early,mid- and late springand fall periodsin east-centralAlaska during spring and fall 1988-1989.

Period

Early Mid Late Season Mean SD Mean SD Mean SD n H' P

Spring 140 132 157 125 133 107 1409 62.4 •0.001 Fall 449 197 340 189 355 206 3899 121.9 •0.001 Vol. 66, No. 4 Altitude of Bird Migration in Alaska [599

SPRING 1,000 n = 2865 800 ......

600 ......

200 ß ß - ILl 0 0000 0300 0600 0900 1200 1500 1800 2100 I-

FALL • 1,000 800 n=4402 LL. 600

4O0

2OO

I I I I I I I I I I I I I 0000 0300 0600 0900 1200 1500 1800 2100

HOUR FIou•d•4. Mean (_SD) flight altitudesof bird targetsby hour of the day,as measured by radar (correctedfor samlingarea) in east-centralAlaska during spring and fall 1988- 1989. (* -- samplesize -•5)

DISCUSSION Diurnal altitudesofflight.--The flight altitudesof the birdswe studied visuallyin east-centralAlaska generally were lowerthan altitudesreported in the literature.For example,most swans flew --•300m agl,whereas Bell- rose (1976) stated that most Tundra Swansmigrate at 900-1500 m agl. Most of the geesewe studiedflew -•300 m agl; othersreport that geese migratemostly at 100-1110m agl (e.g.,Bellrose 1976, Blokpoel 1974, 600] B. A. Cooperand tL j. Ritchie j. FieldOrnithol. Autumn 1995

T•I•E 6. Comparisonsof flight altitudes (m agl) of bird targetsobserved on radar during daylight,crepuscular and nocturnal periodsin east-centralAlaska during springand fall 1988-1989.

Light condition Daylight Crepuscular Nocturnal Season Mean SD Mean SD Mean SD n H' P

Spring 155 128 168 134 150 122 2865 19.9 (0.001 Fall 319 278 314 227 361 200 4402 50.8 (0.001

Cooch 1955, Meinertzhagen 1955). In addition, most of the ducks that we studied during daylight hours flew --•150 m agl. Seaducksmigrating along the northern coastof Alaska also flew (150 m agl (Johnsonand Richardson 1982). Ducks migrating during the "grand passage"along the Central Flywaygenerally flew at 460-850 m agl during the day and dropped to a minimum of 150 m at night (Bellroseand Sieh 1960). Flight altitudesof raptors alsowere lower in this study (most flew -•150 m agl) than in many others. Flightsbelow 50 m agl are common along some ridges and coastal areas in North America, however (Kerlinger 1989). Migrating raptorsin North America previouslyhave been reported to migrate at 200-1100 m agl (e.g., Kerlinger 1980; Kerlinger and Gauth- reaux 1984, 1985; Kerlinger et al. 1985). The flight altitudesof the other speciesgroups that we studied also were lower than thosein the literature. For example, mostof the Sandhill Cranes we observedflew 30-300 m agl, whereasKessel (1984) reported that most cranesin east-centralAlaska migrated at 300-900 m agl. The shorebirdsthat we observedgenerally flew (150 m agl, but others report flights between -200 m and 4000 m agl (e.g., Kerlinger and Moore 1989, Mascher 1962). Most of the passerinesthat we observed during the day flew below 30 m agl. Similarly, Bingman (1980), Wiedner et al. (1992), and severalof the sourcescited by Kerlinger and Moore (1989) observed diurnal passerinemigration below 50-150 m agl. Others report that pas- serines generally migrate below 1000 m agl but that some have been observedat much greaterheights (e.g., Gauthreaux1972, Kerlinger and Moore 1989, Lack 1960, Meinertzhagen 1955). The primary reasonfor the low diurnal flight altitudesthat we observed visuallymay be related to location: the studyarea is near the northern terminus of migration for many species.Thus, these speciesare not in the middle of a long, non-stoptrip as they passover the studyarea, and many may linger before and after the breeding seasonfor feeding or resting, especiallyif breeding areas are frozen or snow-covered.If birds did stopto use the area, one would expect to observelower flight altitudes as they flew down into, or out of, the area. The extensivewetlands in the Tanana River Valley attract large numbers of migrant waterbirds,and many of the flights that we observedfor those speciesprobably repre- Vol.66, No. 4 Altitudeof Bird Migration in Alaska [601 sentedlow-level flights of birds that had used this area on their way to or from the breeding grounds. It is likely that only a small portion of the differencesin diurnal flight altitudes between this study and others occurred becausesome of the individuals (especiallypasserines) we observedwere local breeders mak- ing low-altitudeforaging flights. Flight directionsindicated that most of our recordswere of migrating birds, even though it was daylight.For all speciesgroups combined, over 90% of the birds that we observedflew toward the West or Northwestin spring,and over 90% flew toward the East or Southeastin fall. ' Another reasonwhy diurnal flight altitudeswere lower for this studyis becauseour visual observationsare biased to low-flying birds, although some of the other visual studieswe cite may also have had this bias. This biasprobably was greatest for the smalleror lessvocal species (i.e., shore- birds, passerines,some of the smallerraptors) and wasgreater in fall than in spring.Other authorsalso report that visualobservers tend to missthe higher flying birds (e.g., Kerlinger et al. 1985; Kerlinger and Gauthreaux 1984, 1985; Meinertzhagen 1955). Comparisonsof our concurrent radar and visual observations did not indicate that this bias in our visual data was great enough to account entirely for the large differencesin flight altitudes between our studyand others, however. Nocturnal altitudesof flight.-•At night, a large proportion (98% in spring and 77% in fall) of the radar targetswere flying •500 m agl. Nocturnal flight altitudesin spring and fall at two other locationsin in- terior Alaska (Fairbanks in 1992 and Gulkana in 1989) also indicated that most birds in the area flew below 500 m agl at night (Cooper, unpubl. data). Likewise, several other radar studies have found that nocturnal migrationusually occurs below 500 m agl (e.g., Bellrose1971; Bruderer and Steidinger1972; Gauthreaux1972, 1978, 1991). Large kills of birds at tall structuresalso indicate that nocturnal migrants fly •500 m on at least some nights (Avery et al. 1980). In contrast,others have found that peak nocturnal densitiesextend over a broad altitudinal range below ap- proximately 2000 m (e.g., Eastwoodand Rider 1965; Nisbet 1963; Rich- ardson1971, 1972). We suspectthat differencesbetween the two groups of studiesare due to differencesin location, speciescomposition of mi- grant birds, and perhapsweather conditions.We do not believe that dif- ferencesbetween our studyand the studiesthat report higher altitudes occurred becausebirds flew above the zone we sampled (i.e., •1000 m agl), becausewe also sampled1000-1800 m agl in 1989 and found that •0.3% of the targetsoccurred in that zone. Further, we do not believe that the differences were due to range limitations of the radars. Our radars (especiallythe FR-1900 model) probablycould not detect someof the smallestbirds that were flying singlyat 1000-1800 m agl; however,the distribution of flight altitudes (Fig. 3) showsthat the number of birds droppedoff to nearlyzero well belowthe minimum rangeof 1000 m agl. Additionally,if large numbersof birdswere migratingat 1000-1800 m agl we should have had more numerous observationsof the large-bodied or 602] B. A. Cooperand P. J. Ritchie j. FieldOrnithol. Autumn 1995 flocked bird targets,which we know were detectable at those altitudes. Thus, we believe that it is extremelyunlikely that a large proportion of birds were flying at 1000-1800 m agl. The relativelylow flight altitudes we observedprobably were due in large part to the location of our study area near the northern terminus of migration. Seasonalpatterns in altitude of flight.--Both radar and visual observa- tions indicated that flight altitudesgenerally were higher in fall than in spring,as hasbeen observedelsewhere (e.g., Bellrose1976, Bellrose and Graber 1963, Blokpoeland Burton 1975, Gauthreaux1978). In contrast, 'otherstudies found that migration occurred atlower altitudes infall than in spring (Eastwoodand Rider 1965, Kessel1984, Lack 1960) and a third group of studiesfound that flight altitudesof both small birds and wa- terfowldid not differ betweenspring and fall (Bellrose1971; Richardson 1971, 1972). Bellrose(1976) noted that, in general,the longer the migratoryflight, the higher the altitude at which it occurs.Perhaps spring flight altitudes in east-centralAlaska were lower becausespring migrants made shorter migration flights than did fall migrants. In spring, migrants to Alaska encounter ice- or snow-coveredhabitat and low food availabilityas they advancenorthward. These barriersare not presentin fall, which may lead to lengthier, higher flights over the area. We observedhigh night-to-nightvariability in nocturnal flight altitudes for both spring and fall. Judging by the low correlation between mean nightly flight altitude and migration rates, the day-to-dayvariability was not related to daily changes in the magnitude of migration. This vari- ability may have been related to changing speciescomposition and changesin vertical structure of the atmosphere,however. Birds tend to migrate at altitudeswhere favorablewinds minimize the costof migration (Bruderer et al. 1995, Gauthreaux 1991, Kerlinger and Moore 1989). As mentioned, we plan to discussthe effects of atmosphericstructure on flight altitude in a different paper. In addition to the high day-to-dayvariability, there were within-season patternsof nocturnal flight altitudesamong early, mid, and late periods of the spring and the fall. These patternswere not related to seasonal changesin the magnitude of migration but probably were related to changingspecies composition. In spring, flight altitudeswere highestin April and early May, when waterfowl were the dominant migrants and lowestin mid-late May, when lower-flyingpasserines were the dominant migrants.The high flight altitudesobserved during late Augustand early September probably were due to Greater White-fronted Geese, which were one of the most numerousfall migrantsand had the highestmean flight altitudesof all speciesin fall (Appendix 1). Diel patternsin altitudeof flight.--The radar data indicated that flight altitudes within a day were highly variable and that there was no near- midnight peak in flight altitudes.Further, birds flew higher at night than during daylight or crepuscularconditions in fall but not in spring. Al- though a few studieshave found that flight altitudesof birds are highest Vol.66, No. 4 Altitudeof Bird Migration in Alaska [603 during the day (Bellroseand Sieh 1960, Gauthreaux1972), most have found that altitudestend to be lowestduring the day (e.g., Bruderer and Steidinger1972, Eastwood and Rider 1965, Gauthreaux1978, Lack 1960). Further, flight altitudes of migrating birds generally are highest near or just before midnight and declineslowly until dawn (e.g.,Able 1970, Blok- poel and Burton 1975, Nisbet 1963, Richardson1971). The lack of a near-midnightpeak in flight altitudesmay be related to the rapidly changingday length in Alaska.During both fall and spring in Alaska,the amount of daylight can change by up to 8 min/d (nearly 1 h/wk), so times of sunriseand sunsetchange dramatically(compared with lower latitudes)over the courseof a springor fall studyperiod. Such changeswould tend to dampen hourly trends in flight altitudes,if alti- tudeswere related to the amount of time elapsedsince sunset. In summary,results of this study indicate that substantialnumbers of birds migrated at low altitudes in east-centralAlaska. The relativelylow altitudeswe observedare a goodexample of the high degreeof geograph- ic variabilityin flight altitudesof birds during migration. Given this vari- ability and the geographicvariability in bird abundance,we recommend that site-specificfield studies,in conjunction with regional assessments (e.g., literature reviews)be conducted to assessthe potential for bird collisions at sites where tall structures are to be constructed. In addition to addressingconservation concerns, there clearly is a need for more studiesof flight altitude in a variety of geographic locations before we will fully understandthis important facet of bird migration.

ACKNOWLEDGMENTS

Fundingfor this studywas provided by the U.S. Air Force.The studywas administered by Metcalf & Eddy/Holmes & Narver, the Alaska Environmental and Information and Data Center, and the StanfordResearch Institute, International. Our radarswere operated under radio stationlicense 901106N-6-1-1B. We thank the numerouspersonnel from ABR who were involvedin this study.Special thanks go to P. W. Banyas,C. L. Cranor, C. B. Johnson,J. G. King, B. E. Lawhead,S. M. Murphy, R.J. Rohleder and J. R. Rose.We thank L. C. Byrne, for help with analyses.W. C. Kappleman(Metcalf & Eddy/Holmes& Narver) providedthor- ough reviewsof the studyand valuableinput into studydesign. We thank H. Blokpoel, R. H. Day, P. Kerlinger,K. Yasukawaand the anonymousreviewers for their valuablecomments.

LITERATURE CITED

ABLE,a. P. 1970. A radar studyof the altitudeof nocturnalpassefine migration. Bird-Band- ing 41:282-290. AL•RST•M,T. 1992. Bird migration. Cambridge Univ. Press,Cambridge, United Kingdom. 420 pp. AVERY,M. L., P. F. SPRINGER,AND N. S. DAILEY.1980. Avian mortality at manmadestructures: an annotatedbibliography (revised).U.S. Fish and Wildlife Service,Biological Services Program, Rep. No. FWS/OBS-80/54. 152 pp. BANKS,R. C. 1979. Human relatedmortality of birds in the United States.U.S. Fish and Wildlife Service,Spec. Sci. Rep.--Wildl. No. 215:1-16. BELLROSE,F. C. 1971. The distributionof nocturnalmigration in the air space. 88:397- 424. ß 1976. Ducks,geese, and swansof North America,2nd ed. StackpoleCo., Harrisburg, Pennsylvania.540 pp. 604] B. A. Cooperand P• J. Ritchie j. riel•iOrnithol. Autumn 1995

--, ANDR. R. G•tBEg. 1963. A radar studyof flight directionsof nocturnal migrants. Proc. XIII Intl. Ornithol. Congr.:362-389. , ANDJ. G. SIEH. 1960. Waterfowlflights in the MississippiFlyway, 1956 and 1957. Wilson Bull. 72:29-59. BINGMAN,V. P. 1980. Inland morning flight behavior of nocturnal passefinemigrants in eastern New York. Auk 97:465-472. BLOga'OEL,H. 1971. The M33C track radar (3-cm) as a tool to studyheight and densityof bird migration.Can. Wildl. Serv.Rep. Ser. 14:77-94ß . 1974. Migration of lessersnow and blue geesein springacross southern Manitoba, Part I: distribution, chronology,direction, numbers, height, and speciesßCan. Wildl. Serv.Rep. 28:1-30. BLOV,rOEL, H., ANDJ. Bugq:ON.1975. Weather and the height of nocturnal migration in east-centralAlberta: a radar study.Bird-Banding 46:311-328. BRUDERER,B., ANDP. STEIDINGER.1972. Methods of quantitativeand qualitativeanalysis of bird migrationwith a trackingradarß Pp. 151-168, in S. E. Galler,K. Schmidt-Koenig, G. J. Jacobs,and R. E. Belleville,eds. Animal orientation and navigation:a symposium. NASA SP262.U.S. GovernmentPrinting Office, Washington,D.C. BRUDERER,B., L. G. UNDERHILL,AND F. LIECHTI. 1995. Altitude choice by night migrantsin a high desertarea predictedby meteorologicalfactors. Ibis 137:44-55. COOCH,F. G. 1995. Observationon the autumn migration of Blue Geese.Wilson Bull. 67: 171-174. Coorsg, B. A., R. H. DAY, R. J. RITCHIE,AND C. L. Cr•qog. 1991. An improved marine radar systemfor studiesof bird migrationßJ. Field Ornithol. 62:367-377. FASTWOOD,E., AND G. C. RIDSg. 1965. Some radar measurements of the altitude of bird flight. Brit. Birds58:393-426. G•gIELSON,I. N., ANDE C. LINCOLN.1959. The birdsof Alaska.Stackpole Co., Harrisburg, Pennsylvania.932 pp. G^uq:Hgr•t•x,S. A., Jg. 1972. Behavioralresponses of migrating birds to daylightand dark- ness:a radar and direct visualstudy. Wilson Bull. 84:136-148. 1978. Migratorybehavior and flight patterns.Pp. 12-26, in M. Avery,ed. Impacts of transmissionlines on birdsin flightßU.S. Fishand Wildlife Service,Office of Biological Services,Washington, D.C. Rep. No. FWS/OBS-78/48. ß 1991. The flight behavior of migrating birds in changingwind fields: radar and visualanalyses. Am. Zool. 31:187-204. IgvINC,L. 1961. The migration of Lapland Longspursto Alaska.Auk 78:327-342. JOHNSON,S. R., ANt) W. J. RICI-t•v•SON.1982. Waterbird migration near the Yukon and Alaskacoast of the BeaufortSea: II. Moult migrationof seaducksin summer.Arctic 35: 291-301. I•UI•INCSg,P. 1980. A trackingradar studyof bird migration.J. Migr. Assoc.N. Am. 2:34-42. 1989. Flight strategiesof migrating .Univ. ChicagoPress, Chicago, Illinoisß 375 pp. , V. P. BINGHAM,AND K. P. ABLE.1985. Comparativeflight behaviorof migratinghawks studiedwith trackingradar during autumn in centralNew York. Can.J. Zool. 63:755- 761. , ANDS. A. G^UTHRrAUX,Jg. 1984. Flight behaviorof Sharp-shinnedHawks during migration, I: over land. Anim. Behav.32:1021-1028. , AND• 1985. Flight behaviorofraptors during springmigration in southTexas studiedwith radar and visualobservations. J. Field Ornithol. 56:394-402. , ANDF. R. Moor•. 1989. Atmosphericstructure and avian migration.Pp. 109-141, in D. M. Power,ed. Current ,Vol. 6. Plenum Press,New York, New York. I•SSSL, B. 1984. Migration of sandhillcranes, canadensis,in east-centralAlaska, with routesthrough Alaska and westernCanadaß Can. Field-Nat.98:279-292. IAcI•, D. 1960. The height of bird migration.Brit. Birds53:5-10ß MASCHEg,J. M., B. Sq:oLq:,AND L. W•LIN. 1962. Migration in spring recorded by radar and by field observationsin Sweden.Ibis 104:205-215. Vol.66, No. 4 Altitudeof Bird Migration in Alaska [605

MEINERTZHAGEN,R. 1955. The speedand altitude of bird flight (with notes on other ani- mals). Ibis 97:81-117. NISBET,I. C. T. 1963. Measurementswith radar of the height of nocturnal migration over Cape Cod, Massachusetts.Bird-Banding 34:57-67. RICI-I•vSON,W.J. 1971. Springmigration and weatherin easternCanada: a radar study. Am. Birds 25:684-690. ß 1972. Autumn migrationand weatherin easternCanada: a radar studyßAm. Birds 26:10-16. ROBBINS,C. S., j. R. S^UER,R. GreENBURG,•ND S. DROEGE.1989. Populationdeclines in North Americanbirds that migrateto the Neotropics.Proc. Natl. Acad. Sci. U.S.A. 86: 7658-7662. SL•a)EN,W. S. L. 1973. A continentalstudy of WhistlingSwans using neck collars. Wildfowl 24:8-14. S•'INDLER, M. A., AND K. E I-L•L. 1991. Local movements and habitat use of Tundra or WhistlingSwans Cygnus columbianus in the Kobuk-SelawikLowlands of northwestAlaska. Wildfowl 42:17-32. STr.•uNS,E. I. 1949. The studyof hawksin flight from a blimp. Wilson Bull. 61:110. TEr•ORGH,J. w. 1989. Where have all the birds gone?Princeton Univ. Press,Princeton, NewJersey. 207 pp. u.s. FISH•D WILDLIFESERVICE. 1990ß Alaska Trumpeter census--1990. U.S. Fishand Wildlife Service,Migratory Bird Management,Juneau, Alaska. 4 pp. W^H•G, C. 1965ßPhysiographic divisions of Alaska.U.S. Geol.Surv. Prof. Pap. 482:1-52. WEGE,M. L., ANDD. G. I•WLING. 1984. Flight speedand directionalresponse to wind by Canada Geese. Auk 101:342-348. WEST,G. C., L.J. PErtON,AND L. IRVING.1968. Analysisof springmigration of Lapland Longspursto Alaska.Auk 85:639-653. WEIR,R. D. 1976. Annotatedbibliography of bird kills at man-madeobstacles: a reviewof the stateof the art and solutions.Can. Wildl. Serv.,Ottawa, Ontario, Canada.85 pp. WIEDNER,D. S., P. KERLINGER,D. A. SIBLEY,P. HOLT, J. HOUGH, AND R. CROSSLEY.1992. Visible flight of neotropicallandbird migrantsat Cape May, New Jersey.Auk 109:500- 510. Received 29 Aug. 1994; accepted 23 Mar. 1995. 606] B. A. Cooperand R.J. Ritchie J. Field Ornithol. Autumn 1995

<•Z• Vol.66, No. 4 Altitudeof Bird Migration in Alaska [607 608] B. A. Cooperand P• J. Ritchie j. FieldOrnithol. Autumn 1995