Available online at ww

w.sciencedirect.com

ANIMAL BEHAVIOUR, 2008, 76, 1133e1138doi:10.1016/j.anbehav.2008.05.024

ARTICLES

Sustained bird flights occur at temperatures far beyond

expected limits

HEIKO SCHMALJOHANN, BRUNO BRUDERER & FELIX LIECHTI

Swiss Ornithological Institute

(Received 18 February 2008; initial acceptance 8 April 2008;

final acceptance 25 May 2008; published online 21 July 2008; MS. number: D-08-00106)

Migratory birds deposit fat and protein before passing ecological barriers and must economize these duringsuch crossings. Birds crossing the western Sahara during autumn face a trade-off between cold and humidair along with head winds at high altitudes versus warm and dry air along with tail winds at low altitudes.Since water loss rate increases with temperature, migrants should avoid warm and dry air to save water andhence fly at high altitudes. By quantifying nocturnal songbird migration across the western Sahara withradar, we found that more than 60% of the songbirds migrated below 1000 m above ground level.Thus, the majority of songbirds performed sustained migratory flights in much warmer and drier condi-tions than predicted (weighted means: 30 �C; relative humidity: 27%; water vapour density: 7.8 g/m3).Based on the metabolically available water from fat and protein catabolism, we estimated the maximumpossible overall water loss rate of a flying model bird, the garden warbler, Sylvia borin, for the entire Saharacrossing at 0.29 g/h. This is considerably lower than water loss rates for the same model bird passing ourstudy site, 0.62 g/h at 30 �C, based on applied calculations of physiological studies. Our results clearlyshow that migrating songbirds can fly at much higher temperatures, and have considerably lower waterloss rates, than predicted. This new insight based on observations under natural conditions will have sub-stantial impact on the development of new physiological models for birds and other animals withrestricted access to water.

� 2008 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.

Keywords: bird; flight; garden warbler; humidity; migration

CorrespoAvian RGermanand F. L6, 6204

0003e3

; Sahara; Sylvia borin; temperature; water loss

Migratory birds crossing large ecological barriers, such asdeserts or oceans, have to prepare for extended periods offasting. Therefore, they accumulate large amounts of fatand increase their flight muscles in advance (Jenni &Jenni-Eiermann 1998). The amounts of energy and waterconsumed during such passages depend on the size ofthe barrier, but are highly influenced by wind (Liechti2006), temperature and humidity (Torre-Bueno 1978; Bie-sel & Nachtigall 1987; Carmi et al. 1992; Klaassen et al.1999; Kvist 2001; Klaassen 2004; Engel 2005). Since theseparameters change with time and space, bird migrants are

ndence and present address: H. Schmaljohann, Institute ofesearch, An der Vogelwarte 21, 26386 Wilhelmshaven,y (email: heiko.schmaljohann@ifv.terramare.de). B. Brudereriechti are at the Swiss Ornithological Institute, LuzernerstrasseSempach, Switzerland.

1133472/08/$34.00/0 � 2008 The Association for the Stu

thought to adjust their energy expenditure and water lossby choosing the right time and altitude for their flight(Carmi et al. 1992; Bruderer et al. 1995; Klaassen et al.1999; Klaassen 2004; Liechti 2006). Most songbirds mi-grate nocturnally across the Sahara (Schmaljohann et al.2007), thus avoiding the highest diurnal temperaturesand reducing water loss (Klaassen 2004). In autumn, fa-vourable winds (trade winds) coincide with high noctur-nal temperatures and low humidity at low altitudes,whereas low temperatures occur with unfavourable winds(antitrades) and high humidity at high altitudes. Conse-quently, birds face a trade-off between minimizing energyexpenditure and water loss. Physiological models (Carmiet al. 1992; Klaassen et al. 1999) and several studies(Tucker 1968; Torre-Bueno 1978; Biesel & Nachtigall1987; Wolf & Walsberg 1996; Giladi & Pinshow 1999; Tie-leman et al. 1999, 2002b; Williams & Tieleman 2000;

dy of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.

Wind support (m/s)

−2 0 2 4 6

0−200

600−800

1200−1400

1800−2000

2400−2600

3000−3200

3600−3800

Alt

itu

de

(m)

0 5 15 25 35

Temperature (°C)

28 34 40 46 52 58

Relative humidity (%)

Head wind Tail wind

Figure 1. Altitudinal profiles of wind support (as tail wind compo-

nent in grey bars), air temperature (unbroken line) and relative hu-midity (broken line) at Ouadane above ground level around

midnight (Nnights ¼ 62). Means are shown � SEM. The tail wind

component was defined as cos (wind direction �main migration di-

rection) � wind speed. The main migration direction was defined asthe mean heading direction of tracked songbirds (177�, r ¼ 0:764,

N ¼ 1524), which we simplified to 180�. Meteorological data were

provided by nightly radiosonde measurements.

ANIMAL BEHAVIOUR, 76, 41134

Kvist 2001; Michaeli & Pinshow 2001; Engel 2005) haveshown that, in general, water loss increases with tempera-ture. According to these results, immense water loss wouldprevent birds from flying for longer than 2e4 h at hightemperatures (Biesel & Nachtigall 1987; Giladi & Pinshow1999; Ward et al. 1999; Michaeli & Pinshow 2001; Engel2005). In addition, in most experiments birds refrainedfrom flying at high temperatures, that is, in most casesabove 25 �C (Torre-Bueno 1976; Biesel & Nachtigall1987; Ward et al. 1999; Engel 2005). Based on this currentknowledge, we would predict that (1) birds do not performsustained migratory flights in air layers of such high tem-peratures along with low relative humidity and (2) theywould cross the Sahara at high altitudes in autumn.

We quantified the altitudinal distribution of songbirdmigration by radar during autumn migration (Schmaljo-hann et al. 2007). We then chose the garden warbler, Syl-via borin, as a model songbird and calculated its water lossrate at the meteorological conditions encountered whenactually crossing the Sahara based on our current physio-logical knowledge (Torre-Bueno 1976, 1978; Biesel &Nachtigall 1987; Carmi et al. 1992; Adams et al. 1997;Giladi & Pinshow 1999; Ward et al. 1999; Kvist 2001;Michaeli & Pinshow 2001; Engel 2005). We comparedthis rate with the maximum possible water loss rate basedon the metabolically available water of the model song-bird given different fuel loads.

METHODS

Radar Study

Bird migration was quantified by a fixed-beam radarmethod (Schmaljohann et al. 2008) in Mauritania close tothe oasis Ouadane, about 500 km east of the Atlantic coast(20�560N, 11�350W, 420 m above sea level; 23 August to 24October 2003). Songbirds passing our radar site had alreadycovered about 1500e1700 km of the 2000 km Sahara cross-ing. We carried out radar measurements (N ¼ 484) everyhour on the hour at both low and high elevation angles(11� and 79�) to survey high and low altitudes with nearlythe same effort. The recording time per measurement was4 min and the detection range for songbirds was restrictedto 7500 m (Schmaljohann et al. 2008). The radar beamwas directed towards the west (270�), perpendicular to themain migration direction (Fig. 1). We considered onlyfixed-beam measurements from 2100 to 0100 hours GMTand thereby excluded the large number of songbirds depart-ing after sunset; the altitudinal distribution of songbirds didnot change after 0100 hours (unpublished data). The echosignature of each target crossing the beam was registered;this enabled us to distinguish between birds (Nechoes ¼2307) and insects (Nechoes ¼ 13 357; Schmaljohann et al.2008). To quantify songbird passage of nocturnal migrants,we selected only identified songbirds (Nechoes ¼ 1172) ac-cording to their echo signature, that is, wing beat pattern,and we also excluded hirundines and corvids because theyare not typical nocturnal songbird migrants. Migration in-tensity is given as migration traffic rate in birds/km per hper night and 200 m height intervals (Schmaljohann et al.2008). Launching sounding balloons around midnight

every night provided altitudinal profiles of air pressure,temperature, relative humidity and wind up to 6000 mabove ground level (Bruderer 1994).

Model Songbird

We chose the garden warbler, a well-studied songbird ofthe PalaearcticeAfrican migration system (Bairlein 1991),as a model songbird to estimate water loss rates. Model birdswere provided with hypothetical fuel loads of 75, 100 and125% of their lean body mass (15 g, Bairlein 1991) consist-ing of 95% fat and 5% protein (Jenni & Jenni-Eiermann1998) prior to the Sahara crossing. The highest load is aboutthe upper limit of recorded fuel loads in songbirds andwould provide sufficient energy to fly far more than2000 km without refuelling (Dierschke et al. 2005). Modelbirds were assumed to cross the 2000 km of desert with anintermittent flight strategy, that is, by nocturnal flightsand diurnal rests but without feeding or drinking (Schmal-johann et al. 2007). Thus, fuel loads would decrease withongoing migration across the Sahara. We assumed theshortest possible time to cross the Sahara was 4 nights of10 h flights, because the average ground speed of singlytracked songbirds at our study site was about 50 km/h.Hence, for the 2000 km crossing, a songbird would needto fly for 40 h. This comprises 4 nights combined with threedaytime rests of 14 h each (Schmaljohann et al. 2007).

To obtain the model bird’s body mass when passing ourstudy site, we used two methods. (1) Model birds wereassumed to lose 1% of their body mass during a 1 h flight,which is a realistic figure for small songbirds (Delingat

SCHMALJOHANN ET AL.: BIRDS MIGRATE AT HIGH TEMPERATURES 1135

et al. 2008 and references within) and 0.5% duringrest (Meijer et al. 1994; J. Delingat, V. Dierschke, H. Schmal-johann & F. Bairlein, unpublished data). Consequently, themodel songbirds’ body mass declined in respect of their fuelload from 26.25 (75%), 30 (100%) and 33.75 (125%) g to23.74, 27.13 and 30.52 g during the first 10 h of flightand, after the first 14 h of rest, to 22.13, 25.29 and28.45 g, respectively. We assumed that model birds passingour study site were within their third migratory night, andthus had a body mass of about 17.7, 20.2 and 22.7 g depend-ing on the initial fuel load. After the Sahara crossing, themodel birds’ body mass would have declined to 13.26,15.15 and 17.05 g. Since the body mass of the model birdwith the lowest fuel load, that is, 75%, would be lowerthan its lean body mass, we assumed that a fuel load of atleast 75% is required to cross the Sahara successfully. We,therefore, did not consider birds with a lower fuel load inour model. (2) The mean body mass � SD of garden war-blers mist netted close to the radar site during the same pe-riod as the radar study was 18.4 � 1.9 g (N ¼ 627,unpublished data), which is within the body mass estima-tions based on the assumptions above.

Water Loss Rates: Current Knowledge

We calculated the water loss rates for the model birds giventhe meteorological conditions encountered during theirtrans-Sahara flight based on current knowledge of waterloss rates in flight (Tucker 1968; Hudson & Bernstein 1981;Biesel & Nachtigall 1987; Adams et al. 1997; Giladi & Pin-show 1999; Kvist 2001; Michaeli & Pinshow 2001; Engel2005). Body mass-adjusted water loss rates for flight werecomputed for the temperature range 10e30 �C in steps of5 �C (linear least-squares regression with adjusted R2 values:10 �C: slope¼ 1.75, F1,3 ¼ 23.52, P ¼ 0.017, R2 ¼ 0.85,N ¼ 5; 15 �C: slope ¼ 1.40, F1,4 ¼ 9.24, P ¼ 0.038,R2 ¼ 0.70, N ¼ 6; 20 �C: slope ¼ 1.20, F1,5 ¼ 20.39,P ¼ 0.006, R2 ¼ 0.76, N ¼ 7; 25 �C: slope ¼ 1.02,F1,7 ¼ 62.52, P < 0.001, R2 ¼ 0.88, N ¼ 9; 30 �C:slope ¼ 0.98, F1,3 ¼ 21.30,P ¼ 0.019,R2 ¼ 0.84, N ¼ 5;Tiele-man et al. 2002a). We applied these dependencies to ourmodel songbirds with different body masses (17.7, 20.2and 22.7 g) when crossing our study site to predict bodymass-adjusted water loss rates for the different temperatures.

Water Loss Rates: Metabolic Water

To estimate water loss rates conservatively, we assumedthat (1) catabolizing 1 g of fat liberates 1.1 g of water,while 1 g of protein liberates 0.82 g of water (Jenni &Jenni-Eiermann 1998). The amount of available water istherefore 1.09 times higher than the fuel load (75%:11.25 g; 100%: 15 g; 125%: 18.75 g) resulting in ‘water re-serves’ of 12.3, 16.4 and 20.4 g depending on fuel load. (2)We assumed that birds do not lose water with faeces, al-though faecal water loss could represent up to 10% of totalwater loss (Giladi & Pinshow 1999). (3) Songbirds were as-sumed to need the shortest possible time to cross the Sa-hara, that is, 4 nights of 10 h flights (equalling a flightdistance of 2000 km with a ground speed of 50 km/h),

and three daytime rests of 14 h (Schmaljohann et al.2007). (4) Metabolically produced water from fat and pro-tein catabolism was assumed to be used entirely to balancethe bird’s water budget.

Since measuring water loss rates during flight is morecomplex and hence more imprecise than at rest, weestimated first the total water loss of the model birdswith their different fuel loads (75, 100 and 125%) for thewhole rest phase based on published water loss rates at rest(Wolf & Walsberg 1996; Tieleman et al. 1999, 2002b; Wil-liams & Tieleman 2000). To estimate water loss rates con-servatively, we assumed that (1) model songbirds wouldencounter on average a ground temperature of 35 �C dur-ing three diurnal rests of 14 h each. At our study site themean daytime ground temperature � SD was 35 � 6.2 �C(Ndays ¼ 63) measured 2 m above ground, in shade, be-tween 0600 and 1900 hours GMT, a mean � SD of12 � 0.8 times per day (Nmeasurements ¼ 733). (2) We alsoassumed that model songbirds resting in the desert wouldshow similar water loss rates to desert species, althoughspecies from mesic habitats, such as our trans-Sahara mi-grants, have generally higher water loss rates than speciesadapted to arid conditions (Tieleman et al. 2002a, b).Among three desert songbird species (Wolf & Walsberg1996; Tieleman et al. 1999, 2002b; Williams & Tieleman2000) the water loss rate at 35 �C was related to bodymass (Tieleman et al. 2002a) as log water loss rate(g/h) ¼ �4.06 þ 0.56 log body mass (g) (linear least-squares regression: F1,3 ¼ 11.34, P ¼ 0.043, R2 ¼ 0.72,N ¼ 5). Our model songbirds had body mass-adjusted waterloss rates at 35 �C of 0.099, 0.108 and 0.116 g of water/h ac-cording to their body mass. During their three 14 h rests at35 �C, the model songbirds would consequently lose in to-tal 4.55, 4.91 and 5.24 g of water, respectively. These are,according to our assumptions, conservative estimates. Di-viding the ‘remaining water’ (7.65e15.16 g) by the 4 nightswith 10 h flights (in total 40 h) yields the maximum aver-age water loss rates during flight for the entire Sahara cross-ing. Because we have divided the ‘remaining water’ just bythe time necessary to cross the Sahara, the water loss rateduring flight is dependent only on time but not on temper-ature and is a rough overall average estimation for waterloss rates in flight across the Sahara.

We used the statistical software package R (R Develop-ment Core Team 2006) with adjusted R2 values through-out the analyses.

RESULTS

At low altitudes, tail winds prevailed, whereas at highaltitudes, above 2000 m above ground level, head windsdominated (Fig. 1). Temperature decreased considerablywith altitude from an average of 31.8 �C (95% confidenceinterval (CI) 30.9e32.7 �C) in the lowest 200 m to about3.5 �C (95% CI 3.2e3.9 �C) at 4000 m, whereas relativehumidity increased (Fig. 1).

At night, the majority of songbirds (64%, 95% CI 57e71%) chose low altitudes for migration, that is, below1000 m (Fig. 2). More than 60% of the nocturnal songbirdmigration occurred at temperatures above 25 �C (weighted

0−200

400−600

800−1000

1200−1400

1600−1800

2000−2200

2400−2600

2800−3000

3200−3400

3600−3800

Alt

itu

de

(m)

0 10 20 30 40

Migration traffic rate (songbirds/km per h)

Figure 2. Altitudinal distribution of songbird migration as migration

traffic rates above ground level at night (2100e0100 hours;

Nnights ¼ 62). Seasonal means are given � SEM.

10 15 20 25 30 35

Temperature (°C)

0.25

0.5

0.75

1

1.25

1.5

Wat

er l

oss

rate

du

rin

g fl

igh

t (g

/h)

Water surplus

Water deficiency

Predicted waterloss rate

Maximum possiblewater loss rate

Figure 4. Flight water loss rates of a model songbird, a garden war-

bler, Sylvia borin, with an initial fuel load of 100% of its lean body

mass (15 g) prior to the Sahara crossing. Body mass when passing

the study site, Ouadane, was estimated to be 20.2 g and was usedfor the water loss rate calculations. The unbroken line indicates pre-

dicted water loss rates in flight of the model songbird according to

free-flight and wind tunnel studies (Tucker 1968; Hudson & Bern-

stein 1981; Biesel & Nachtigall 1987; Adams et al. 1997; Giladi &Pinshow 1999; Kvist 2001; Michaeli & Pinshow 2001; Engel 2005).

The broken line shows maximum possible average water loss rate

of the model songbird in flight based on the metabolically produced

water (for further details see Methods).

ANIMAL BEHAVIOUR, 76, 41136

average means � SD: temperature: 30 � 2.3 �C; relative hu-midity: 27 � 5.1%; water vapour density: 7.8 � 0.65 g/m3;Nmeasurements ¼ 178) and more than 20% above 30 �C(weighted average means � SD: temperature: 32 � 1.1 �C;relative humidity: 22 � 2.8%; water vapour density:7.2 � 0.59 g/m3; Nmeasurements ¼ 79; Fig. 3).

We calculated first the corresponding water loss rates inthese meteorological conditions for our model songbird,the garden warbler, based on free-flight and wind tunnelstudies (see Methods). At ambient flight temperatures of25 �C water loss rates would increase with fuel load ofthe model songbirds (estimated body mass when crossingour study site: 17.7, 20.2 and 22.7 g) from 0.47 to 0.60 g/hand at 30 �C from 0.54 to 0.69 g/h, respectively. Accordingto these results, model songbirds that started their Sahara

Nocturnal flight temperature (°C)0 5 10 15 20 25 30 35

Cu

mu

lati

ve f

req

uen

cy o

f m

igra

tion

in

ten

sity

0

0.2

0.4

0.6

0.8

1

Figure 3. Cumulative frequency distribution of nocturnal songbird

migration (2100e0100 hours; Nnights ¼ 62) in relation to ambient

nocturnal flight temperature around midnight. Arrows specifyproportion of nocturnal songbird migration that occurred below

the indicated temperature.

crossing with an initial fuel load of 100% should fly attemperatures below 21 �C to prevent dehydration whencrossing the Sahara (Fig. 4). However, we could detectsongbird migration even up to 36 �C.

Maximum possible water loss rates of our modelsongbird based on metabolic water production variedbetween 0.19 and 0.38 g/h depending on fuel load (initialfuel load of 75%: 95% CI 0.142e0.223; 100%: 95% CI0.229e0.323; 125% fuel load: 95% CI 0.315e0.379;Fig. 4). These mass-dependent relationships are indepen-dent of ambient temperature (see Methods).

DISCUSSION

Songbirds adjusted their flights to low altitudes, wherethey encountered high temperatures and low relativehumidity, but benefited from tail wind (Figs 1, 2). Thesesustained migratory flights under such conditions are con-trary to our current physiological knowledge about waterbudgets of flying birds (Torre-Bueno 1976, 1978; Biesel &Nachtigall 1987; Carmi et al. 1992; Adams et al. 1997;Giladi & Pinshow 1999; Ward et al. 1999; Kvist 2001;Michaeli & Pinshow 2001; Engel 2005). However, little isknown about the actual rate of water loss in such condi-tions, since birds would not fly in such warm and dry airin wind tunnels, nor were free-flight studies carried outin similar conditions.

SCHMALJOHANN ET AL.: BIRDS MIGRATE AT HIGH TEMPERATURES 1137

Our calculations of water loss rates in flight based oncurrent physiological knowledge revealed very highvalues, twice as high as the rate of 0.29 g/h estimated onthe basis of metabolically available water at 25 and30 �C and five times higher at 35 �C, respectively. Hence,birds would run out of water well before they had crossedthe Sahara, if water loss is as high as predicted by our cur-rent physiological knowledge (Fig. 4).

This discrepancy might be caused by overestimatingwater loss rates in free-flight and wind tunnel studies(Tucker 1968; Hudson & Bernstein 1981; Biesel & Nachti-gall 1987; Adams et al. 1997; Giladi & Pinshow 1999; Kvist2001; Michaeli & Pinshow 2001; Engel 2005). One reasonmight be that the general assumption that expired air issaturated (Michaeli & Pinshow 2001; Engel 2005) maybe invalid. If so, as in ostriches, Struthio camelus (Witherset al. 1981), respiratory water loss would decrease by atleast a third (Engel 2005). Furthermore, carrying instru-ments (equipment) for recording expired air temperaturein flight (Michaeli & Pinshow 2001) and flying in the sta-tionary environment of the small flight chamber ina wind tunnel (about 1 � 1 � 1 m3, Tucker 1968; Torre-Bueno 1976; Biesel & Nachtigall 1987; Ward et al. 1999;Kvist 2001; Engel 2005) might be more costly than freeflight (Gessaman et al. 1991; Liechti & Bruderer 2002). Be-sides the general difficulties of measuring water loss dur-ing flight, there are further possible sources of error. Nophysiological experiments have been published in whichbirds performed sustained flight at temperatures wellabove 25 �C. Furthermore, the bird species that havebeen investigated (Tucker 1968; Biesel & Nachtigall1987; Adams et al. 1997; Giladi & Pinshow 1999; Kvist2001; Michaeli & Pinshow 2001; Engel 2005) were eithernot songbirds or not typical trans-Sahara migrants andtheir body weights were mostly well above the averageof trans-Sahara songbird migrants. Since it is unknownwhether rates of water loss depend linearly on tempera-ture at high values, and below the body mass consideredin previous studies (Biesel & Nachtigall 1987; Adamset al. 1997; Giladi & Pinshow 1999; Kvist 2001; Michaeli& Pinshow 2001; Engel 2005), linearly extrapolated waterloss rates of songbirds may not be realistic. Furthermore,our model songbirds were not allowed to dehydrate. De-hydrating birds, using water reserves not used by ourmodel songbirds, can consequently lose more water thanthe latter. However, water gain by dehydration does notseem to be large enough to explain the large difference be-tween the water loss rates predicted by flight studies andby metabolic water reserves.

If these possible sources of error explain the discrepancywe found in rates of water loss, then birds do not need anyspecial physiological mechanisms to migrate in suchwarm and dry conditions. However, if data from free-flight and wind tunnel studies are correct, then songbirdscrossing the Sahara must reduce their water loss consid-erably by some kind of adaptation. At least in respect ofenergy requirements, carrying large amounts of fuel is lessexpensive than formerly thought (Kvist et al. 2001). Thismight also apply for water loss. Our estimated rates of wa-ter loss based on metabolically available water were two tothree times higher than the water loss rates at rest. There is

a similar (four times) increase in the energy requirementfrom resting to flight metabolism (Wikelski et al. 2003).

In the eastern Sahara during autumn, predicted waterloss rates of bird migrants, contrary to our results, wereassumed to increase with flight altitude (Klaassen & Bie-bach 2000). However, the meteorological conditions inthe eastern Sahara differ significantly from those in thewestern Sahara (Fig. 1), with low nocturnal temperatures(<25 �C) and high relative humidity close to the ground(>50%, decreasing with altitude). Because of this altitudi-nal distribution of meteorological conditions, it is not pos-sible to refute or confirm dehydration risk as a threat fortrans-Sahara migrants (Klaassen 2004).

Songbird migrants select altitudes with the best windsupport regardless of the high ambient temperature andlow relative humidity. Despite the uncertainties in esti-mating water loss rates in flying birds, our results showthat migrants tolerate much higher temperatures thanpredicted and must have considerably lower water lossrates than previously assumed. This will have a significantinfluence on future physiological models in birds andbeyond.

Acknowledgments

We are most grateful to all those who helped to gather thedata in Mauritania. Special thanks to Thomas Steuri whoengineered the hardware and the software together withErich Bachler. We thank the main supporters of ourproject: the Swiss Army and Oerlikon-Contraves AG forthe radar equipment; for important financial support wethank the Swiss National Science Foundation (Project No.31-65349), the Foundations Volkart, Vontobel, MAVA forNature Protection, Ernst Gohner, Felis, and Syngenta, thecompanies Bank Sarasin & Co, Helvetia Patria Insurancesand F. Hoffmann-La Roche AG, the European ScienceFoundation, BirdLife Switzerland and BirdLife Interna-tional. The Swiss Ornithological Institute was the strong-hold of the whole project throughout. Comments fromSusi Jenni-Eiermann, Lukas Jenni, Marcel Klaassen, IreneTieleman and two anonymous referees significantlyimproved the manuscript.

References

Adams, N. J., Pinshow, B. & Gannes, L. Z. 1997. Water influx and

efflux in free-flying pigeons. Journal of Comparative Physiology B,167, 444e450.

Bairlein, F. 1991. Body mass of garden warblers (Sylvia borin) on mi-gration: a review of field data. Vogelwarte, 36, 48e61.

Biesel, W. & Nachtigall, W. 1987. Pigeon flight in a wind tunnel. IV.

Thermoregulation and water homeostasis. Journal of ComparativePhysiology B, 157, 117e128.

Bruderer, B. 1994. Nocturnal bird migration in the Negev (Israel):a tracking radar study. Ostrich, 65, 204e212.

Bruderer, B., Underhill, L. G. & Liechti, F. 1995. Altitude choice ofnight migrants in a desert area predicted by meteorological fac-

tors. Ibis, 137, 44e55.

Carmi, N., Pinshow, B., Porter, W. P. & Jaeger, J. 1992. Water and

energy limitations on flight duration in small migrating birds. Auk,

109, 268e276.

ANIMAL BEHAVIOUR, 76, 41138

Delingat, J., Bairlein, F. & Hedenstrom, A. 2008. Obligatory barrier

crossing and adaptive fuel management in migratory birds:

the case of the Atlantic crossing in Northern Wheatears (Oenantheoenanthe). Behavioral Ecology and Sociobiology, 62, 1069e1078.

Dierschke, V., Mendel, B. & Schmaljohann, H. 2005. Differentialtiming of spring migration in northern wheatears Oenanthe

oenanthe: hurried males or weak females? Behavioral Ecology and

Sociobiology, 57, 470e480.

Engel, S. 2005. Racing the wind. Water economy and energy expen-

diture in avian endurance flight. Ph.D. thesis, University of

Groningen.

Gessaman, J. A., Workman, G. A. & Fuller, M. P. 1991. Flight per-

formance, energetics and water turnover of tippler pigeons withharness and dorsal load. Condor, 93, 546e554.

Giladi, I. & Pinshow, B. 1999. Evaporative and excretory water lossduring free flight in pigeons. Journal of Comparative Physiology B,

169, 311e318.

Hudson, D. M. & Bernstein, M. H. 1981. Temperature regulation

and heat balance in flying white-necked ravens, Corvus cryptoleu-

cus. Journal of Experimental Biology, 90, 267e281.

Jenni, L. & Jenni-Eiermann, S. 1998. Fuel supply and metabolic con-

straints in migrating birds. Journal of Avian Biology, 29, 521e528.

Klaassen, M. 2004. May dehydration risk govern long-distance mi-

gratory behaviour? Journal of Avian Biology, 35, 4e6.

Klaassen, M. & Biebach, H. 2000. Flight altitude of trans-Sahara mi-

grants in autumn: a comparison of radar observations with predic-

tions from meteorological conditions and water and energybalance models. Journal of Avian Biology, 31, 47e55.

Klaassen, M., Kvist, A.& Lindstrom, A. 1999. How body water and fuelstores affect long distance flight in migrating birds. In: Proceedings of

the 22nd International Ornithological Congress (Ed. by

N.Adam&R.Slotow),pp.1450e1467.Durban:BirdLifeSouthAfrica.

Kvist, A. 2001. Fuel and fly: adaptations to endurance exercise in mi-

grating birds. Ph.D. thesis, Lund University.

Kvist, A., Lindstrom, A., Green, M., Piersma, T. & Visser, G. H.2001. Carrying large fuel loads during sustained bird flight is

cheaper than expected. Nature, 413, 730e732.

Liechti, F. 2006. Birds: blowin’ by the wind? Journal of Ornithology,

147, 202e211. doi:10.1007/s10336-006-0061-9.

Liechti, F. & Bruderer, L. 2002. Wingbeat frequency of barn swal-

lows and house martins: a comparison between free flight and

wind tunnel experiments. Journal of Experimental Biology, 205,2461e2467.

Meijer, T., Mohring, F. J. & Trillmich, F. J. 1994. Annual and dailyvariation in body mass and fat of starlings Sturnus vulgaris. Journal

of Avian Biology, 25, 98e104.

Michaeli, G. & Pinshow, B. 2001. Respiratory water loss in free-

flying pigeons. Journal of Experimental Biology, 204, 3803e3814.

R Development Core Team. 2006. R: a Language and Environment

for Statistical Computing. Vienna: R Foundation for Statistical Com-

puting. <http://www.R-project.org>.

Schmaljohann, H., Liechti, F. & Bruderer, B. 2007. Songbird mi-

gration across the Sahara: the non-stop hypothesis rejected!. Pro-ceedings of the Royal Society of London, Series B, 274, 735e739.

doi:10.1098/rspb.2006.0011.

Schmaljohann, H., Liechti, F., Bachler, E., Steuri, T. & Bruderer, B.2008. Quantification of bird migration by radar: a detection prob-

ability problem. Ibis, 150, 342e355. doi:10.1111/j.1474-

919x.2007.00797.

Tieleman, B. I., Williams, J. B., Michaeli, G. & Pinshow, B. 1999.

The role of the nasal passages in the water economy of crestedlarks and desert larks. Physiological and Biochemical Zoology, 72,

219e226.

Tieleman, B. I., Williams, J. B. & Bloomer, P. 2002a. Adaptation

of metabolism and evaporative water loss along an aridity gradi-

ent. Proceedings of the Royal Society of London, Series B, 270,

207e214.

Tieleman, B. I., Williams, J. B. & Buschur, M. E. 2002b. Physiolog-

ical adjustments to arid and mesic environments in larks (Alaudi-dae). Physiological and Biochemical Zoology, 75, 305e313.

Torre-Bueno, J. R. 1976. Temperature regulation and heat dissi-pation during flight in birds. Journal of Experimental Biology, 65,

471e482.

Torre-Bueno, J. R. 1978. Evaporative cooling and water balance

during flight in birds. Journal of Experimental Biology, 75,

231e236.

Tucker, V. A. 1968. Respiratory exchange and evaporative water loss

in the flying budgerigar. Journal of Experimental Biology, 48,

67e87.

Ward, S., Rayner, J. M. V., Moller, U., Jackson, D. M., Nachtigall,W. & Speakman, J. R. 1999. Heat transfer from starlings Sturnusvulgaris during flight. Journal of Experimental Biology, 202, 1589e

1602.

Wikelski, M., Tarlow, E. M., Raim, A., Diehl, R. H., Larkin, R. P. &Visser, G. H. 2003. Costs of migration on free-flying songbirds.

Nature, 423, 704.

Williams, J. B. & Tieleman, B. I. 2000. Flexibility in basal metabolic

rate and evaporative water loss among hoopoe larks exposed to

different environmental temperatures. Journal of Experimental Biol-ogy, 203, 3153e3159.

Withers, P. C., Siegfried, W. R. & Louw, G. N. 1981. Desert ostrichexhales unsaturated air. South African Journal of Science, 77, 569e

570.

Wolf, B. O. & Walsberg, G. E. 1996. Respiratory and cutaneous

evaporative water loss at high environmental temperatures in

a small bird. Journal of Experimental Biology, 199, 451e457.