Extreme Fattening by Sedge Warblers, Acrocephalus Schoenobaenus, Is Not Triggered by Food Availability Alone

ANIMAL BEHAVIOUR, 2007, 74, 471e479 doi:10.1016/j.anbehav.2006.11.030

Extreme fattening by sedge warblers, Acrocephalus schoenobaenus, is not triggered by food availability alone

NICHOLAS J. BAYLY*† *The Wetland Trust, U.K. ySchool of Life Sciences, University of Sussex, U.K. (Received 23 May 2006; initial acceptance 10 August 2006; ﬁnal acceptance 6 November 2006; published online 22 August 2007; MS. number: 8976)

The strategies adopted by birds during migration are expected to be strongly influenced by resource availability and the efficiency with which these resources are used. The migratory strategy of the sedge warbler in northwest Europe has been linked to superabundances of plum-reed aphids Hyalopterus pruni that enable birds to accumulate extensive fat reserves and make nonstop flights to sub-Saharan Africa. Food availability was therefore expected to be the main determinant of whether sedge warblers accumulated extensive reserves or not. In this study, birds were provided with an unlimited supplementary food source, but only 10 out of 24 birds accumulated large reserves. In view of this, temporal and geographical cues reflecting seasonal variations in aphid abundance are considered, in addition to food availability, to determine whether extensive fuelling occurs. Optimality models of resource use by migrants predict two modes of fuelling behaviour, as seen here, if strategies have evolved to divide a journey into an optimal number of stages or if the perception of resources at future sites varies between birds or in time/space. If sedge warbler perception of future resources reflects the temporal and geographical variation in aphid abundance in northwest Europe, then the distribution of fuelling behaviours observed here is expected. Despite some birds accumulating large fuel reserves, they did not show a high degree of sensitivity to wind conditions when initiating migratory flights.

Keywords: Acrocephalus schoenobaenus; migration; optimal fuelling behaviour; sedge warbler; stopover behaviour

Migration is an energetically demanding process and While sedge warblers stopover for long periods and consequently migratory strategies are expected to be accumulate enormous fuel reserves (80e120% of their influenced by the ability to utilize various food supplies lean body mass) in some reedbeds in northwest Europe, and the distribution of those supplies both in space and the proportion doing so can be relatively small. Indeed, time (Alerstam 1990). The contrast between two migratory when aphid abundance is low, up to 85% of birds may birds, the reed warbler Acrocephalus scirpaceus and sedge remain at a site for 2 days or less, accumulating only small warbler illustrates this principle. Sedge warblers perform reserves (Bibby et al. 1976). This suggests that the spatial a remarkable long-haul migration from northwest Europe distribution of resources, i.e. reed aphids, influences fuel- to sub-Saharan Africa by feeding extensively on super- ling behaviour and that short stays and consequent rapid abundant food supplies, specifically reed aphids (e.g. Hya- movements between sites are a searching behaviour adop- lopterus pruni), while the reed warbler must stop and ted by birds looking for a resource rich reedbed (Bensch refuel two or three times to complete the same journey. & Nielsen 1999; Wernham et al. 2002). However, even This difference arises because reed warblers are less effi- where aphid infestations occur not all birds remain to cient at preying on reed aphids despite considerable eco- fuel extensively (Bibby & Green 1981), suggesting that logical and morphological similarity to the sedge warbler other factors influence which behaviour is adopted. (Bibby & Green 1981). Optimality models have been used to examine how migrants should respond to resource availability under various selection pressures during stopovers/fuelling Correspondence and present address: N. J. Bayly, 17 Walton Street, periods (cf. Alerstam & Hedenstro¨m 1998). The stopover Oxford, Oxfordshire OX1 2HQ, U.K. (email: [email protected]). behaviour of several long-distance migrants is consistent

with time-minimization (Bayly 2006) and given the res- tricted temporal availability of reed aphids, whose abundance peaks in early August in the U.K. and is in rapid decline by late August (Bibby & Green 1981), the same is likely to be true of sedge warblers. While time-minimizing models predict that fuel loads will increase with resource availability/fuel deposition rate (FDR), stopover durations are expected to decrease, not increase as is true of sedge warblers (Alerstam & Lindstro¨m 1990). However, varying the assumptions of the model can change the outcome: for example, if migratory distance is considered finite then a stepped function arises because the journey is best divided into equal steps, rendering the same fuel load optimal over a range of FDRs (Weber & Houston 1997). In this case, fuelling duration increases in steps with FDR. Al- ternatively, fuel loads and fuelling durations are predicted to increase if decreasing resources are expected at future sites (Weber & Houston 1997). In addition to resources, Figure 1. Sedge warbler using remote weighing set-up. The perch is attached to the pan of the balance, which is reading 15.99 g. The weather is expected to influence stopovers (Weber et al. Perspex box around the bowl of mealworms is to encourage birds 1998) and given the extremely large reserves laid down to access the food from the perch. by sedge warblers, they are expected to be highly sensitive to wind conditions and to delay departure until conditions are favourable (Liechti & Bruderer 1998). To examine how resources influence fuelling behaviour, previous studies (e.g. Fransson 1998; Bayly 2006)andisex- free-living sedge warblers were attracted to an unlimited pected to give comparable results. In 2001 and 2002 birds supply of food at a reedbed in the U.K. and their body mass discovered the feeders by chance, while in 2003 and 2005, was recorded daily. The departure mass of birds utilizing the sedge warbler song was played close to unattended feeders food source was expected to reflect extensive fuelling if during the morning to increase the rate of discovery. abundant resources were the cue for this strategy. The departure loads and FDRs of experimental birds were compared to time-minimizing models to determine whether birds were responding to resources in an optimal Body Mass Recordings way. In conjunction with weather data, the sensitivity of departure decisions to wind conditions was also tested. Body mass, which reflects the mass of stored fuel (Redfern et al. 2004), was recorded using an electronic balance inserted into the feeder box (Ohaus Scout II, Ohaus Corpo- ration, Pine Brook, NJ 07058, U.S.A.; Fig. 1) and a video METHODS camera (Sony CCD-TR748E Hi8, Sony Corporation) accord- Study Site ing to the methods described in Bayly (2006). Remote weighing avoids the potentially confounding effects on The Pannel Valley Reserve (PVR) is located in southeast body mass and foraging behaviour associated with catching England (50540N, 0410E) and consisted of a 23 ha mosaic of and handling birds (e.g. Gosler 2001). Phragmites reedbed, sallow Salix scrub and open water, Recordings were made twice daily between 0700 and 1.7 km from the English Channel. Daily mist-netting at 1000 hours and 1700 and 2030 hours, with filming PVR between July and October has resulted in a mean sessions lasting ca. 45e90 min. Where possible, evening annual capture of 7965 sedge warblers (mean of 1995e2004). recordings were made in the hour preceding sunset. Videotapes were analysed on a colour television and the identity, time of visit (to the nearest minute), body mass Food Supplementation Experiments (to the nearest 0.01 g if possible, otherwise to 0.1 g) and the number of mealworms consumed on each visit were Sedge warblers were attracted to supplemental food during recorded. Some individuals could be identified either 1Auguste16 September 2001, 12 Julye30 September 2002, because they had a unique combination of colour rings 6Auguste2 September 2003 and 26 Julye12 September or because of a combination of the following features: 2005. Food was presented in a feeder consisting of a wooden the presence/absence of a metal ring, small plumage box (25 40 15 cm) attached to a stake (0.7e1.5 m high), differences, and by marked differences (>1.5 g) in body driven into the ground within the reedbed (Fig. 1). A plastic mass (e.g. of 24 birds: colour rings: two juveniles; metal bowl (10 4cm),kepttoppedupwithatleast50meal- ring: 10 juveniles, one adult; no ring: 10 juveniles, one worms, Tenebrio molitor, from dawn till dusk, was inserted adult). Individuals that could not be confidently identified into a purpose-made hole on the box top. Feeders some- were excluded from the analyses. Licences for the trap- times had to be emptied overnight to avoid consumption ping, ringing and colour ringing of birds were obtained by small rodents. This set up is similar to that used in from the British Trust for Ornithology. BAYLY: SEDGE WARBLER MIGRATION STRATEGIES 473

Fuel Deposition Rates, Departure Loads which it was the number of days since the start of fuelling and Intake Rates (see above). On arrival at a feeder, several individuals were >1 g heavier than the mean mass of sedge warblers cap- Departure loads and three measures of FDR were tured in mist nets at PVR in 2005 (mean ¼ 10.8 g, calculated using the following equations where LBM N ¼ 4682), suggesting that fuelling had commenced prior stands for lean body mass (N.B. values could not be to finding a feeder. To take this into account, an extended calculated for all birds): duration was calculated as an average of two extrapolated overall FDR ¼ (mass change between first and last durations: these involved estimating how many days evenings/days fuelling)/LBM; a bird may already have fuelled for if: (1) its FDR was the initial FDR ¼ (mass change between 1st and 3rd same as when at a feeder; (2) its FDR was that of ‘natural’ evenings/days fuelling)/LBM; birds (5.3% LBM/day) and it had begun fuelling at a mass hourly FDR ¼ (evening mass morning mass)/hours in of 11 g. ‘Natural’ FDR was the mean for retrapped birds at between. PVR in 2005 that gained mass and whose FDR was in the Departure load ¼ (body mass on last evening LBM)/ top 25% (chosen to exclude nonfuelling birds). LBM Evening body mass was defined as the mean of all mass recordings in the 30 min prior to a bird’s last feed of the Optimal Migration Models day. If recordings did not include this 30-min period, evening body masses were extrapolated using a linear re- Predictions for a time-minimizing model were generated gression (a linear model provided the best fit to regular using the information in Hedenstro¨m & Alerstam (1997).A mass recordings from an individual over the course of constant FDR between stopovers, an infinite migratory a day). Starting body mass was that recorded on the first distance, and variation in FDR being interpreted as ‘global’ evening of a bird’s attendance at a feeder, except in two variation, were assumed. The assumption that there is cases where birds did not initially gain mass (instead, a time and energy cost to initiating a stopover is central to the start of fuelling was defined as the day before two con- the formulation of the model. The model predicts the opti- secutive days with gains of >1 g/day). LBM was estimated mal number of days to spend at a stopover site (written as from a regression model of body mass on wing length t*), for a given FDR (k), and varies with the values entered using sedge warblers with a fat score of zero trapped at for the stopover cost in time (t0) and energy ( f0). Optimal PVR in 2005 (regression: LBM ¼ 0.706 þ 0.144wing; departure loads ( f *) were calculated by multiplying t*by 2 the FDR (k). An additional time-minimizing model in F1,708 ¼ 138, P < 0.001, adjusted R ¼ 16.2). To examine behaviour during fuelling phases, daily mass gains and in- which migratory distance is finite and FDR is expected to take rates (number of mealworms consumed per hour) decrease along the migratory route is illustrated in Fig. 2. were calculated in addition to hourly FDRs (see above) This model’s predictions were not generated but modified for each day that a bird was present. from Fig. 3 of Weber & Houston (1997). Accurate estimation of departure load depends on a bird departing on the evening after which it no longer Weather Data attended a feeder. Sedge warblers made a mean of 8.9 visits (SD ¼ 3.7, min ¼ 3.2, max ¼ 16) during 2 h of film- The Meteorological Office provided hourly recordings of ing on their last day. Given this rate of visitation, there wind strength/direction, precipitation and air pressure at is a very low probability of a bird being present and not sea level from Herstmonceux (50890N, 00320E, altitude being recorded the following day (P ¼ 0.006, if visitation 52 m), approximately 25 km from PVR, for the years rate is modelled as a normal distribution). To determine 2001e2003. For 2005, data from a weather station at whether birds underwent ‘extensive’ fuelling, it is neces- PVR were used. Birds were expected to depart 2 h after sary to define ‘extensive’. Here, a bird capable of complet- sunset (A˚ kesson et al. 2001) and conditions closest to ing a nonstop flight from PVR to the Sahel of West Africa this time were used. Assuming a preferred migratory direc- (approximately 3800e4200 km), which requires a body tion of 225(approximate mean direction of sedge warbler mass > 19.75 g (equivalent departure load ¼ 99%), is con- movements from the U.K.; Wernham et al. 2002) wind di- sidered to have fuelled extensively and is termed EX. A rection and strength were converted into a wind compo- nonextensively fuelling bird is termed NEX. Flight ranges nent using Fransson’s (1998) methods. Positive wind were estimated using Flight 1.15 (Pennycuick & Battley components relate to tailwinds, negative to headwinds 2003) using the following assumptions/values: still and zeros to crosswinds. air; 85% of any mass increase above LBM consisted of To examine departures in relation to wind, chi-square fat; wing span ¼ 0.183 m (mean of 12 measurements tests used expected values based on the frequency of from juveniles at PVR); aspect ratio ¼ 5.27 (mean of five different wind types/components. Two frequency distri- birds at PVR); otherwise default values were used. butions were generated for this purpose: (1) including all days on which birds were present at feeders, on the assumption that birds could leave at any time; (2) Estimating Fuelling Durations including only those days when individual birds were considered ‘ready’ to depart. The criteria for determining ‘Raw’ fuelling duration was calculated as the number of ‘readiness’ differed between groups: NEX: all days a bird days a bird was present at a feeder, except in two cases for was present plus the day after departure; EX: all days after 474 ANIMAL BEHAVIOUR, 74,3

load: t21 ¼ 10.05, P < 0.001; fuelling duration: t21 ¼ 9.67, 140 P < 0.001; Fig. 3). Overall FDRs were generally lower in EX birds but not significantly (t test: t ¼1.49, P ¼ C 19 120 B 0.152). Conversely, initial FDRs were higher in EX birds (t test: t19 ¼ 3.71, P < 0.001; Fig. 2). No apparent difference in the date when fuelling commenced was found between 100 NEX and EX birds (t test: t19 ¼ 1.24, P ¼ 0.230). The extended stays and initial nonfuelling behaviour of one adult 80 and one juvenile were indicative of premigratory fuelling while all other birds could have been on stopover. Notably A this adult bird was observed feeding the juvenile and while 60 the juvenile fuelled extensively the adult did not. A binary logistic regression including initial FDR and Departure load (% LBM) 40 ‘start date’ found that higher initial FDRs and later ‘start dates’ increased the likelihood of extensive fuelling, 20 although only the former was significant (logistic regression: P ¼ 0.032 and 0.061, respectively). When initial FDR is plotted against start date, only one EX individual over- 0 laps in the two-dimensional spaces occupied by EX and 0 5 10 15 20 NEX birds (see Fig. 4). This bird may be better classified as Initial FDR (% LBM/day) NEX, as its departure load only just qualifies as EX (99.9% Figure 2. Departure load (fuel load expressed as a percentage of LBM) and examination of Fig. 6 (bird fuelling between 17 LBM) against initial fuel deposition rate (% of LBM/day) for 22 sedge and 24 August 2003) suggests that its fuel load may have C warblers at PVR ( ).Therelationshipiscomparedtothreetime- been elevated through delayed departure associated with minimising models where (A) migratory distance is infinite and the adverse winds between 18 and 23 August. A second logistic stopover cost in terms of energy ¼ 3% of LBM and the cost in days ¼ 1; (B) as model A except that the cost in days ¼ 4; (C) migra- regression including the terms initial FDR, ‘start date’ and tory distance is finite and there is an expectation of decreasing future initial FDR*‘start date’, in which the bird is reclassified, can- FDRs. Importantly, models A and B also reflect the difference not converge and fails to generate accurate P values because between an expectation of increasing or decreasing FDR along the the model is a ‘perfect predictor’ (Field 2000) of whether migratory route, respectively. The boundary between EX and NEX a bird fuels extensively (Pearson goodness-of-fit test: birds is indicated by a dotted line. P ¼ 1.0). Using the same reclassification, the initial FDRs of EX birds were found to decrease with date (regression: 2 and including day 7 since fuelling began plus the day after F1,7 ¼ 5.83, P ¼ 0.046, adjusted R ¼ 37.6%). departure (no EX birds departed before day 8). Using these periods of readiness, a probability of occurrence was assigned to the conditions on each day based on their Optimization of Fuelling Behaviour frequency during an individual’s period of readiness. These probabilities were summed across birds to give the Energy/time-minimizing models predict that departure final expected frequencies. load will increase smoothly with FDR if migratory distance

160 Statistical Analyses 140 EX and NEX birds were compared using two-tailed 120 t tests, while a binary logistic regression was used to inves- tigate the factors inﬂuencing fuelling behaviour. Predic- 100 tions from optimality models regarding the relationship between FDR and departure load/fuelling duration were 80 tested using regression models. A nested ANOVA was used to examine day-to-day variation in fuelling variables 60 Departure load within the population and individuals. In no cases were 40 the assumptions of the parametric tests described above broken (Field 2000). 20

0 0 246810 12 14 16 RESULTS Extended fuelling duration During four autumn migrations, 22 juvenile and two adult Figure 3. Departure load (fuel load expressed as a percentage of sedge warblers attended feeders at PVR, 10 of which (nine LBM) against extended fuelling durations (see Methods) in days juveniles and one adult) fuelled extensively (Fig. 3; depar- for 22 sedge warblers attending supplementary food at PVR (C). ture load > 99%). EX birds had larger departure loads and The solid line indicates the departure load above which birds are longer fuelling durations than NEX birds (t tests: departure considered to have fuelled extensively. BAYLY: SEDGE WARBLER MIGRATION STRATEGIES 475

25 fuelling durations in relation to initial FDR suggests that the durations of NEX and EX birds form two separate steps (in a similar fashion to the relationship between departure load and initial FDR in Fig. 2). 20

Variation in Fuelling Rate Between Days 15 One assumption of Hedenstro¨m & Alerstam’s (1997) time-minimizing model is that FDR is constant through- 10 out a fuelling period. At the population level, both hourly FDR and daily mass gain decreased as departure ap- proached (Fig. 5; regression: hourly FDR: F1,55 ¼ 16.33,

Initial fuel deposition rate P < 0.001, adjusted R2 ¼ 21.5%; KruskaleWallis test: daily 5 mass gain: H5 ¼ 17.1, P ¼ 0.004), a relationship whose sig- nificance held when individual birds were nested within ‘day relative to departure’ (ANOVA: hourly FDR: 0 F10,21 ¼ 7.18, P < 0.001; daily mass gain: F7,16 ¼ 5.11, 05 Aug 10 Aug 15 Aug 20 Aug 25 Aug 30 Aug 04 Sep P ¼ 0.003). Intake rate (mealworms/hour) showed no Date at start of fuelling trend with day relative to departure (Fig. 5; nested Figure 4. Initial fuel deposition rate against date at the start of ANOVA: intake rate: F17,41 ¼ 0.67, P ¼ 0.812). fuelling for 22 sedge warblers at PVR. EX (C)andNEX(,)birds occupy unique spaces on this plot except for one EX bird (-), whose departure load may have been elevated as a result of delayed departure because of adverse wind conditions (see Fig. 6;bird Departure Decisions and Weather fuelling between 17 and 24 August 2003). Of the 24 sedge warblers, eight departed with headwinds, one with a crosswind and 15 with tailwinds. Departures in relation to wind component, which re- is modelled as infinite or in a stepped function if flects both wind direction and strength, can be seen in migratory distance is finite (Fig. 2); although energy-min- Fig. 6. On no occasion did birds depart on nights with imizing models give rise to a shallower slope. Departure rain or when rain had fallen in the 6 h prior to depar- load was not significantly related to ‘overall’ FDR (regres- ture. Air pressure was higher on nights with departures 2 sion: F1,19 ¼ 0.15, P ¼ 0.705, adjusted R ¼ 0.0%) but did relative to nights without (t test: t40 ¼ 2.30, P ¼ 0.027). increase significantly with ‘initial’ FDR (regression: Relative to the frequency of head/cross winds and tail- 2 F1,19 ¼ 35.2, P < 0.001, adjusted R ¼ 63.1%; see Fig. 2). winds whilst birds were present at feeders (see Methods, In Fig. 2, neither infinite distance model fits the whole distribution A), birds departed with tailwinds more often data set while a stepped finite model provides a better than would be expected by chance (chi-square test: 2 qualitative fit. A finite model predicts a constant departure c1 ¼ 4:92, P ¼ 0.027). However, if birds are treated indi- load across a range of FDRs. To test this, EX and NEX birds vidually and are considered ready to depart only on cer- were considered representative of two steps and were tain days (see Methods, distribution B), conditions at analysed separately. Departure loads of NEX birds did departure were not significantly different from those ex- 2 not increase significantly with ‘initial’ FDR (regression: pected by chance (chi-square test: c1 ¼ 0:555, P ¼ 2 F1,8 ¼ 2.02, P ¼ 0.193, adjusted R ¼ 10.1%) when an indi- 0.456); although the greatest difference between ob- vidual with markedly lower FDR was excluded (see Fig. 2). served and expected values was for EX birds departing For EX birds no relationship was found in relation to in tailwinds. ‘overall’ FDR (regression: F1,8 ¼ 0.00, P ¼ 0.975, adjusted That sedge warblers do not appear to actively choose R2 ¼ 0.0%) but a positive relationship was just significant tailwinds is surprising, however, this may be because with ‘initial’ FDR (regression: F1,8 ¼ 5.39, P ¼ 0.049, light cross/headwinds have little impact on flight range. adjusted R2 ¼ 32.9%). Indeed, six birds departed in cross/headwinds of Extended fuelling durations were significantly longer 3.6 km/h and only three birds departed in stronger than raw fuelling durations (paired t test: t23 ¼ 4.76, winds. To examine whether birds were avoiding stronger P < 0.001, mean difference ¼ 1.52 days) as would be ex- headwinds, the expected frequency of wind components pected, and are used here to examine model predictions on nights when birds were ‘ready’ to depart was calcu- relating to durations. When migratory distance is treated lated (see Methods). However, the observed pattern of as infinite, fuelling durations are expected to decrease wind components on departure nights could still have 2 with increasing FDR, while a finite model predicts fuelling occurred by chance (chi-square test: c3 ¼ 4:0, P ¼ durations will increase in steps with FDR but decrease 0.261). Some of the departures in headwinds by EX birds within steps. Extended fuelling durations decreased non- might be explained by extensive delays forcing birds to significantly with ‘overall’ FDR (regression: F1,19 ¼ 3.27, depart whatever the conditions. This may be true of P ¼ 0.086) and increased significantly with initial FDR (re- the bird that began fuelling on 16 August 2005 (see gression: F1,19 ¼ 11.61, P ¼ 0.003). Qualitative analysis of Fig. 6). 476 ANIMAL BEHAVIOUR, 74,3

DISCUSSION

0.22 Dual Fuelling Behaviours (a) Given access to the same resources, migrants might be 0.2 expected to adopt the same fuelling behaviour and therefore migratory strategy, assuming there is little variation in 0.18 the ability to use those resources. In this study, sedge warblers had access to a clumped, superabundant food 0.16 supply, which was believed to be the cue for extensive fuelling, however, only 10 out of 24 birds did so. Indeed, it was evident that two distinct fuelling behaviours had been 0.14 adopted. One set of birds (NEX) remained for a mean of

Hourly FDR (g/h) 3.7 days and accumulated a limited amount of fuel while 0.12 the second set (EX) fuelled for a mean of 11.6 days and gained sufficient fuel for a nonstop flight from the U.K. to 0.1 sub-Saharan Africa (Fig. 3). The ability to utilize the avail- able resources did differ between groups, as reflected by significantly higher initial FDRs in EX birds, however, initial 0.08 FDRs were not mutually exclusive between groups, show- 2.5 (b) ing considerable overlap (Fig. 2). Food availability is therefore unlikely to be the sole determinant of whether birds fuelled extensively. Indeed, the threshold FDR above which 2 extensive fuelling occurred (>13% LBM/day; Fig. 3)israrelyat- tained under natural conditions: max FDR ¼ 12% LBM/day in Bibby et al. (1976). 1.5 Temporal and geographical cues are plausible additional determinants of fuelling behaviour, given that the temporal abundance of aphids varies on a geographical scale. Phragmites reeds mature and senesce earlier in the year as 1 mean summer temperature increases, giving rise to

Daily mass gain (g) a northesouth cline within Western Europe and the tim- ing of reed aphid abundance reflects this cline (Bibby & 0.5 Green 1981). The optimal region for seeking an abundance of aphids will therefore depend on time of year: early migrating birds might be expected to use more 0 southerly sites (which have the added benefit of reducing 10 fuel load requirements) while later migrants should use in- (c) creasingly northerly locations (e.g. PVR). The results presented here partly agree with this prediction, as mean 9 ‘start’ date for EX birds was later than NEX birds but not significantly so (a larger sample size may clarify this). This nonsignificant result might be explained if exten- 8 sive fuelling at PVR is favourable for early migrants when a threshold FDR is attained. Given the temporal and 7 geographical distribution of aphids, early migrants at PVR would be expected to stop briefly before heading south to where aphid abundance is peaking, unless 6 a threshold FDR is exceeded, cancelling out the cost of a greater migratory distance. This threshold is expected to Intake rate (mealworms/h) decrease with date in line with a decline in aphid 5 abundance at sites south of PVR. Initial FDR did decrease with ‘start date’ in EX birds if one individual was 4 reclassified (see Results and Fig. 4 for reasoning), poten- −5 −4 −3 −2 −1 0 tially reflecting this decreasing threshold. Indeed, using Day relative to departure the same reclassification, a logistic regression including Figure 5. Variation in the population mean SE of (a) hourly FDR, the interaction between initial FDR and start date was (b) daily mass gain and (c) intake rate (of mealworms) in relation a ‘perfect predictor’ of whether birds fuelled extensively. to the day of departure for experimental sedge warblers at PVR. Considered alongside the distribution of NEX and EX The departure day ¼ 0. Note: sample sizes vary between days. birds in Fig. 4, these results suggest that in early August all birds stop briefly at PVR, as southerly sites are optimal BAYLY: SEDGE WARBLER MIGRATION STRATEGIES 477

15 20 25 (a) (b) 16 15 10 20 14 10 5 15 12 5 Body mass (g)

0 Wind component 10 10 0

−5 8 –5 5 Body mass (g) 21 Aug 24 Aug 27 Aug 30 Aug 08 Aug 12 Aug 16 Aug 20 Aug 24 Aug 28 Aug 01 Sep 2002 2003

20 25 (c) Wind component 15 20

10 15 5

10 0

−5 5 05 Aug 10 Aug 15 Aug 20 Aug 25 Aug 30 Aug 04 Sep 09 Sep 2005 Figure 6. Fuelling curves for 22 sedge warblers attending supplementary food sources at PVR ((a) 2002; (b) 2003; (c) 2005) plotted as body mass (right axis) against date and time (solid lines, symbol colour varies to allow separation of individuals; a lone bird from 2001 and one bird from 2002 are not included). Wind component (left axis) is illustrated (dotted line) on the same date scale to allow the assessment of delays caused by adverse wind conditions. Positive components refer to tailwinds, negative to headwinds and zero values to crosswinds (zero is indicated by a straight black line). Wind components represent the conditions 2 h after sunset and those with white squares indicate nights with departures. for fuelling, while in mid-August, the threshold for switch- Time-minimizing models predict a positive relationship ing behaviours is lower and birds with high initial FDRs re- between FDR and departure load (Alerstam & Lindstro¨m main to fuel extensively. By late August, the threshold 1990). This was not true of sedge warblers at PVR when drops further and most birds fuel extensively. This is sup- overall FDR was considered but was for initial FDR. This ported by Bibby et al.’s (1976) finding that later migrating departure from the models, when considering overall sedge warblers in the U.K. have larger reserves. Sedge war- FDR, arises from the violation of the underlying model as- bler fuelling behaviour may therefore have evolved in a re- sumption that FDR remains constant throughout the fuel- markable way to respond to the temporal and ling period (Hedenstro¨m & Alerstam 1997; Weber & geographical distribution of aphid abundance. Houston 1997). This assumption is broken by EX birds, in particular, whose FDR gradually decreases in the later stages of fuelling (see Figs 5 and 6). Rising metabolic de- Optimization of Fuelling Behaviour mands associated with the maintenance and transport of an increasing fuel load are most probably the cause of The brief period in which aphids are superabundant in this decrease (Klaassen & Lindstro¨m 1996). Reed warblers northwest European reedbeds (Bibby & Green 1981) sug- seem to compensate for similar rising energy demands by gests that time-minimization should be a key selective increasing their intake rate, which may otherwise be re- pressure on autumn migrating sedge warblers. Further- duced to minimize foraging-intensity dependent preda- more, sedge warblers breeding in northwest Europe moult tion risk (Bayly 2006). There is no evidence that sedge in the Sahel region of West Africa (Morel & Morel 1992; warblers do the same (Fig. 5), implying that sedge warblers Wernham et al. 2002), where invertebrate abundance is are already foraging at maximum rates; as expected of determined by the erratic wet season between June and a strongly time-selected migrant (Alerstam & Lindstro¨m October. Therefore, early arrival in this region may be ad- 1990). vantageous in order to finish moulting before the onset of It is apparent that a time-minimizing model in which the dry season (Bensch et al. 1991). migratory distance is considered finite, giving rise to 478 ANIMAL BEHAVIOUR, 74,3

a stepped function, provides a good qualitative fit to the warblers (taken from Bayly 2006) attracted to mealworms data on sedge warblers (Fig. 2). Nevertheless, both finite at PVR highlights a number of differences (see Fig. 7). For and infinite distance models can give rise to the observed example, FDRs are higher in sedge warblers, which may be duality in fuelling behaviour. In a finite model, the opti- achieved by reducing predator awareness and increasing mal solution is to divide the journey into equal stages feeding rates and/or through physiological adaptations where each step in the resulting stepped function repre- such as increased gut size. It also shows the fuel loads of sents a decrease in the number of stages (Weber & Houston reed warblers falling half way between those of EX and 1997; Fig. 2). If the expectation of FDR at future sites is NEX birds. If NEX birds have an expectation of increasing lowered, these steps become shorter and the switch be- resources and EX birds a decreasing expectation, then reed tween steps occurs at lower FDRs. For early migrants at warblers may well have a constant expectation. Finally, PVR, aphid abundance and therefore FDR are expected this comparison further highlights that food availability to be higher at future sites, while the opposite is true of and therefore FDR is not the only determinant of fuelling later birds. The finite model therefore predicts that few behaviour; there are clearly hardwired behaviours that early migrants will fuel extensively while later birds are give rise to quite different outcomes under the same con- more likely to, as the step occurs at a lower FDR. The ditions: even within ecologically and morphologically data presented here are in agreement with this prediction similar species. and there is even an overlap in the FDRs of EX and NEX Other species of long-distance migrants utilize clumped, birds as would be expected (Fig. 2). When distance is con- superabundant resources (e.g. garden warblers, Sylvia borin sidered infinite (smooth function) two models are needed, and fruit) but do not accumulate extensive fuel reserves. the first resulting from an expectation of increasing re- Evidently, the close association between sedge warblers sources ahead (NEX birds) and the second from an expec- and aphid abundances has made this strategy optimal in tation of decreasing resources (EX birds). These two this species. Ultimately, sedge warblers adopt a high-risk models predict that EX and NEX birds will have very dif- strategy that is dependent on a highly variable food ferent departure loads but departure load should still in- source, both in time and space, and presumably expose crease with FDR within groups (Fig. 2). There is some themselves to a high level of predation risk by fuelling evidence for these increases but the case is not clear-cut. at maximum intensities and compromising their escape Regardless of which model sedge warblers conform to, ability by grossly increasing their body mass (Gosler they point to the evolution of an innate perception of fu- et al. 1995). The success of sedge warbler migration is ture conditions as being a highly plausible mechanism for likely to be as sensitive to the effects of climate change explaining the observed duality in fuelling behaviour. on aphid abundance and phenology, as their overwinter survival is to erratic rainfall in the Sahel (Peach et al. 1991). Departure Decisions and Weather

While sedge warblers appeared to avoid departing in rain or high winds, their departures were not as closely related to tailwinds as might be expected (Liechti & Bru- 150 derer 1998). This may be due to light head/crosswinds being considered suitable for departure, which is not 125 unexpected given that a light wind will have little negative impact on potential flight range; especially for NEX birds which are likely to be making relatively short flights. 100 Alternatively, birds may be finding favourable winds at higher altitudes, which can differ from those recorded at ground level (Schaub et al. 2004). However, for EX birds 75 that must manage their fuel effectively to reach sub-

Saharan Africa, this result is surprising. The greatest con- Departure load tributors to the chi-square analyses suggest that EX birds 50 were more likely to chose tailwinds and avoid strong headwinds but not signiﬁcantly so. This apparent insensitivity 25 to wind conditions may have resulted from a small sample size, combined with some individuals departing regardless of conditions due to long delays (Weber et al. 1998). How- 0 ever, it may also be argued that migrants making long 0510 15 20 ﬂights have reduced sensitivity to initial conditions. Fuel deposition rate Figure 7. Comparison of the relationship between the fuel deposition rate and departure load of food supplemented sedge and reed Sedge Warblers versus Reed Warblers warblers at PVR. Reed warblers (B;fromBayly 2006), regression line: departure load ¼ 35.1 þ 3.98 FDR. Sedge warblers (EX: -; A comparison of the relationship between FDR and NEX: ¤), regression line for NEX birds: departure load ¼ 35.9 þ departure load for sedge (data presented here) and reed 1.74 FDR. BAYLY: SEDGE WARBLER MIGRATION STRATEGIES 479

Acknowledgments Fransson, T. 1998. Patterns of migratory fuelling in whitethroats Syl- via communis in relation to departure. Journal of Avian Biology, 29, The University of Sussex and the Wetland Trust funded 569e573. this study. I thank all the staff at the Wetland Trust, as well Gosler, A. 2001. The effects of trapping on the perception, as Anna Gray, Naomi Scuffil and Manu Shirwiwa for their and trade-off, of risks in the great tit Parus major. Ardea, 89, assistance with fieldwork. The Meteorological office pro- 75e84. vided crucial weather data and David Harper supervised Gosler, A. G., Greenwood, J. & Perrins, C. 1995. Predation risk and the project in its early stages and commented on the the cost of being fat. Nature, 377, 621e623. manuscript. I am grateful to two anonymous referees for Hedenstro¨m, A. & Alerstam, T. 1997. Optimum fuel loads in migra- providing valuable feedback on the manuscript. tory birds: distinguishing between time and energy minimization. Journal of Theoretical Biology, 189, 227e234. Klaassen, M. & Lindstro¨m, A. 1996. Departure fuel loads in time- References minimising migrating birds can be explained by the energy costs of being heavy. Journal of Theoretical Biology, 183,29e34. ˚ Akesson, S., Walinder, G., Karlsson, L. & Ehnbom, S. 2001. Reed Liechti, F. & Bruderer, B. 1998. The relevance of wind for optimal warbler orientation: initiation of nocturnal migratory flights in migration theory. Journal of Avian Biology, 29, 561e568. relation to visibility of celestial cues at dusk. Animal Behaviour, Morel, G. & Morel, M. 1992. Habitat use by Palaearctic migrant e 61, 181 189. passerine birds in West Africa. Ibis, Supplement, 134,83e88. Alerstam, T. 1990. Bird Migration. Cambridge: Cambridge Univer- Peach, W. J., Baillie, S. R. & Underhill, L. 1991. Survival of British sity Press. sedge warblers Acrocephalus schoenobaenus in relation to west Alerstam, T. & Hedenstro¨m, A. 1998. The development of bird mig- African rainfall. Ibis, 113, 300e305. e ration theory. Journal of Avian Biology, 29,342 369. Pennycuick, C. J. & Battley, P. F. 2003. Burning the engine: Alerstam, T. & Lindstro¨m, A. 1990. Optimal bird migration: the rel- a time-marching computation of fat and protein consumption in ative importance of time, energy and safety. In: Bird Migration: the a 5420 km flight by great knots (Calidris tenuirostris). Oikos, e Physiology and Ecophysiology (Ed. by E. Gwinner), pp. 331 351. 103, 323e332. Berlin, Heidelberg: Springer. Redfern, C., Topp, V. & Jones, P. 2004. Fat and pectoral muscle in Bayly, N. J. 2006. Optimality in avian migratory fuelling behaviour: migrating sedge warblers Acrocephalus schoenobaenus. Ringing e a study of a trans-Saharan migrant. Animal Behaviour, 71, 173 182. and Migration, 22,24e34. Bensch, S. & Nielsen, B. 1999. Autumn migration speed of juvenile Schaub, M., Liechti, F. & Jenni, L. 2004. Departure of migrating reed and sedge warblers in relation to date and fat loads. Condor, European robins, Erithacus rubecula, from a stopover site in relation e 101, 153 156. to wind and rain. Animal Behaviour, 67, 229e237. Bensch, S., Hasselquist, D., Hedenstro¨m, A. & Ottosson, U. 1991. Weber, T. & Houston, A. 1997. A general model for time- Rapid moult among palaearctic passerines in West Africa: an adap- minimising avian migration. Journal of Theoretical Biology, 185, e tation to the oncoming dry season? Ibis, 133,47 52. 447e458. Bibby, C. J. & Green, R. E. 1981. Autumn migration strategies of Weber, T., Alerstam, T. & Hedenstro¨m, A. 1998. Stopover e reed and sedge warblers. Ornis Scandinavica, 12,1 12. decisions under wind influence. Journal of Avian Biology, 29, Bibby, C. J., Green, R. E., Pepler, G. R. M. & Pepler, P. A. 1976. Sedge 552e560. e warbler migration and reed aphids. British Birds, 69, 384 399. Wernham, C. V., Toms, M. P., Marchant, J. H., Clark, J. A., Field, A. 2000. Discovering statistics using SPSS for Windows: ad- Siriwardena, G. M., Baillie, S. R. (Eds). 2002. The Migration Atlas: vanced techniques for beginners. Introducing Statistical Methods. Movements of the Birds of Britain and Ireland. London: T. & A. D. Thousand Oaks, California: Sage Publications. Poyser.