Effect of weather conditions on the spring migration of ... · Woodcock Scolopax rusticola ﬁtted with Argos satellite tags in the British Isles and France (2012–17). First, we

Effect of weather conditions on the spring migrationof Eurasian Woodcock and consequences for

breedingK�EVIN LE REST,1* ANDREW HOODLESS,2 CHRISTOPHER HEWARD,2 JEAN-LOUIS CAZENAVE3 &

YVES FERRAND1

1Office National de la Chasse et de la Faune Sauvage, 8 boulevard Albert Einstein, Nantes, 44300, France2Game & Wildlife Conservation Trust, Burgate Manor, Fordingbridge, Hampshire SP6 1EF, UK3Club National des B�ecassiers, 105 rue Louis Pergaud, Villeneuve, Champniers, 16430, France

Migration is a critical period of time with fitness consequences for birds. The develop-ment of tracking technologies now allows researchers to examine how different aspectsof bird migration affect population dynamics. Weather conditions experienced duringmigration are expected to influence movements and, subsequently, the timing of arrivaland the energetic costs involved. We analysed satellite-tracking data from 68 EurasianWoodcock Scolopax rusticola fitted with Argos satellite tags in the British Isles and France(2012–17). First, we evaluated the effect of weather conditions (temperature, humidity,wind speed and direction, atmospheric stability and visibility) on migration movementsof individuals. Then we investigated the consequences for breeding success (age ratio)and brood precocity (early-brood ratio) population-level indices while accounting for cli-matic variables on the breeding grounds. Air temperature, wind and relative humiditywere the main variables related to migration movements, with high temperatures andnorthward winds greatly increasing the probability of onward flights, whereas a trendtowards greater humidity over 4 days decreased the probability of movement. Breedingsuccess was mostly affected by climatic variables on the breeding grounds. The propor-tion of juveniles in autumn was negatively correlated with temperature in May, but posi-tively correlated with precipitation in June and July. Brood precocity was poorlyexplained by the covariates used in this study. Our data for the Eurasian Woodcock indi-cate that, although weather conditions during spring migration affect migration move-ments, they do not have a major influence on subsequent breeding success.

Keywords: age ratio, Argos transmitter, breeding ground conditions, carry-over effects, early-brood ratio.

The most common way for birds to cope with sea-sonal changes in weather and feeding conditions isto make flights between separate wintering andbreeding grounds, a mechanism which has a stronggenetic basis (Berthold et al. 1990, Berthold &Helbig 1992, Pulido et al. 1996, Pulido 2007,Liedvogel et al. 2011, Liedvogel & Lundberg2014). The expression of this behaviour in a popu-lation is related to the fitness of migrating individ-uals vs. resident ones, e.g. by breeding and/or

wintering in habitat with higher resources, lowerpredation risk and/or less competition (Alerstamet al. 2003, McKinnon et al. 2010). Undertakingmigration, however, carries potential costs. Itencompasses an important part of the annual cycle(up to 40–60% for some long-distance migrants,Newton 2011) and comprises a period duringwhich birds cross unknown areas with uncertainfood resources, weather conditions and predationrisk. Conditions experienced during migration maythus have major effects on the timing andenergetic costs of the journey (Jenni & Schaub2003), as well as on individual survival (Klaassen

*Corresponding author.Email: [email protected]

© 2018 British Ornithologists’ Union

Ibis (2018) doi: 10.1111/ibi.12657

http://orcid.org/0000-0003-1592-8402

http://orcid.org/0000-0003-1592-8402

http://orcid.org/0000-0003-1592-8402

http://orcid.org/0000-0001-7286-6776

http://orcid.org/0000-0001-7286-6776

http://orcid.org/0000-0001-7286-6776

mailto:

et al. 2014). For instance, food abundance andavailability at stopover sites are known to beimportant variables influencing the duration ofstopovers for refuelling (Newton 2006). Moreover,the weather conditions such as wind strength anddirection (Liechti 2006, Shamoun-Baranes et al.2017) experienced during the trip strongly influ-ence migration progress, depending on the flightstrategy of the species, i.e. soaring or flappingflight, fuel deposit capacity and altitude of flight(Alerstam 2011).

Pre-breeding migration has profound conse-quences for the breeding season. Indeed, it setsimportant characteristics for breeding, such as thedate of arrival at the breeding ground and bodycondition on arrival. The latter determines thecapacity of the bird to invest energy directly inbreeding activity (Sandberg & Moore 1996) andthe former constrains the timing of breeding, e.g.breeding success may depend on the relationshipbetween hatching date and seasonal availability offood for chicks (Both et al. 2005, Møller et al.2008, Saino et al. 2011). The energetic cost ofmigration is especially important for species usingstored resources for breeding, i.e. capital breeders(Stephens et al. 2009), but the timing of migrationhas consequences for breeding whatever the breed-ing strategies of the species (Both & Visser 2001,Newton 2006).

The study of bird migration has been revolu-tionized by the miniaturization of geolocationtechnologies, which now enable the tracking ofmost birds throughout their travels. The largeamount of geolocation data recorded has revealedunexpected migration routes and behaviours (Bur-ger & Shaffer 2008, Klaassen et al. 2014,Weimerskirch et al. 2015). This technology alsooffers new opportunities to study the mechanismsand factors driving migration strategy at the indi-vidual level. One of the main challenges lies inlinking these geolocation data to relevant informa-tion on the factors expected to influence birdmovements at relevant spatial and temporal scales(e.g. meteorology, Shamoun-Baranes et al. 2010).Remote sensing data appear to be the best solu-tion, because it is almost impossible to realizefield surveys at each bird position (Gordo 2007).Worldwide data are now easily available frompublic datasets, e.g. land cover classification, vege-tation indices and climate datasets, which offervaluable information on parameters affecting birdmigration.

This study focuses on the effect of weather con-ditions on the pre-breeding migration of a crypticbird, the Eurasian Woodcock Scolopax rusticola(hereafter Woodcock). It is a widespread speciesliving in woodland habitats. It has a broad diet,composed mainly of lumbricids, myriapoda andarthropods. The spring migration of the Woodcockalternates between flights of several hundred kilo-metres at night and stopovers for one or severaldays (https://Woodcockwatch.com; http://www.becassesmigration.fr; Crespo et al. 2016). Foodabundance and availability play an important rolein habitat selection during winter (Duriez et al.2005) but are not expected to influence migrationmovements because most of the habitats crossedduring migration are suitable for Woodcock (Cre-spo et al. 2016). However, weather conditions areexpected to have a large influence on stopoverduration during Woodcock migration and, conse-quently, on the duration of the complete springmigration.

To test this, we linked satellite tracking datafrom 68 individuals with remote sensing datasetsavailable daily at a global scale. The effect of14 climatic variables expected to affect birdmigration, related to wind, temperature, visibil-ity, humidity and atmospheric conditions(Richardson 1990a, Shamoun-Baranes et al.2017), and when relevant their trends over afew days, were first tested to explain the migra-tion movements of Woodcock at the individuallevel. We then investigated the consequences ofweather conditions experienced during migrationon the breeding success (proportion of juve-niles) and brood precocity (proportion of juve-niles having completed their post-juvenilemoult) during the subsequent breeding seasonat the population level. However, because theenergetic needs of breeding are thought to begained mainly from local resources, especiallyfor an income breeder such as Woodcock (seeStephens et al. 2009), the timing and success ofbreeding will also depend on climatic variableson the breeding grounds (Guzm�an & Arroyo2015). Temperature and precipitation on thebreeding grounds were thus calculated andincluded in the analyses. We discuss the impli-cations of our findings for Woodcock populationdynamics and, more generally, for our under-standing of carry-over effects of spring migrationconditions on the reproductive success of migra-tory birds.


2 K. Le Rest et al.

https://Woodcockwatch.com

http://www.becassesmigration.fr

http://www.becassesmigration.fr

METHODS

Migration movements

From 2012 to 2017, 87 Woodcock were fittedwith Argos satellite tags (solar PPT MicrowaveTelemetry, Inc., Columbia, MD, USA; 9.5 g) inFebruary–March in Western Europe (63 in the Bri-tish Isles in 2012–17 and 24 in France in 2015–16). Tags were attached to the birds using a leg-loop harness (Rappole & Tipton 1991) composedof 1.6-mm-diameter, UV-resistant, marine-graderubber cord (EPDM cord; Polymax Ltd, Bordon,UK) passed through biomedical tubing (Silastictubing; Cole Palmer, St Neots, UK). The tagschedule alternated between a 10-h ON period(Argos messages transmitted every 60 s) and a 48-h OFF period (no messages transmitted). Some ofthe 87 tags failed prematurely and some othersfailed to transmit any signal during spring migra-tion. Moreover, several Woodcock were resident inthe UK (no migration data) and some others died(hunted or unknown causes). Consequently, only68 tags from the 87 used provided meaningfuldata for studying spring migration.

The spring tracking data were converted to abinary variable indicating whether the bird hadcontinued its migration (1) or not (0) betweentwo ON periods (first 10 h ON + 48 h OFF +next 10 h ON, i.e. 68 h elapsed time). A buffer of50 km was defined to assess whether the distancetravelled between the two ON periods corre-sponded to local movement or a migration flight.The threshold of 50 km was chosen according tothe home-range of Woodcock during the breedingperiod, when movements were typically < 50 km(mean � sd = 14 � 23 km, n = 12). Such binarydata thus described the probability of migrationmovement where at least two accurate locationswere transmitted during a 68-h period. In total,729 spring locations from 68 individuals were used(Fig. 1). Spring migration movements in morethan 1 year were recorded for 31 individuals, withsome birds yielding data for up to five consecutiveyears.

Weather variables were extracted from theNCEP Reanalysis (Kalnay et al. 1996) 2.5° 9 2.5°gridded data provided by the NOAA/OAR/ESRLPSD (Boulder, CO, USA) website (http://www.esrl.noaa.gov/psd/). Fourteen weather variablesrelated to wind, temperature, humidity, visibilityand atmospheric conditions (see Table S1 in

Appendix S1 for details) and, when relevant, theirtrends over 4 days were extracted. The accuracy ofthese variables was not available, which may haveintroduced additional uncertainty into the analysis(see Baker et al. 2017). All weather variables wereavailable four times daily, but only information at18 h (UTC) was extracted because Woodcockmigration movements started at dusk. Moreover, allaltitude-dependent weather variables wereextracted near the ground surface because GPS log-gers fitted on Woodcocks showed that they mainlymigrate at low altitude (A. Hoodless unpubl. data).Weather variables were averaged over 3 days (timet, t + 1 and t + 2) to fit the tracking data time reso-lution (time t corresponded to the first 10-h ONperiod; time t + 1 and t + 2 to the 48-h OFF per-iod) and their trends were assessed over 4 days(adding the previous day, t � 1).

Age and early-brood ratios

The effect of weather conditions experienced dur-ing spring migration on subsequent breeding ide-ally would be investigated at the individual level.However, tracking data did not provide reliableinformation about breeding success. Frequentmovements of females during the breeding periodindicated incomplete incubation, but this wasinsufficient to assess breeding success confidently.However, the proportion of juveniles (hereafterage ratio) and the proportion of juveniles hav-ing finished their post-fledging moult before migra-tion (hereafter early-brood ratio) were availableeach year from French and Russian monitoringschemes.

Hunting of Woodcock wintering in WesternEurope is popular, especially on the winteringgrounds and during autumn migration. In France,a non-profit organization of Woodcock hunters(Club National des B�ecassiers) collected wings ofhunted Woodcock in order to determine the ageand progress of moult (8000–10 000 wings col-lected each year). At the same time, a ringing pro-gramme was managed by a governmentalorganization (Office National de la Chasse et de laFaune Sauvage – ONCFS) and hunter organiza-tions (F�ed�eration Nationale des Chasseurs – FNC,F�ed�erations D�epartementales des Chasseurs – FDC).Thousands of birds (5000–7000) were ringed inFrance each year through this programme and theage and moult of individuals were assessed. Tensto a few hundred Woodcock were also caught in


Weather conditions and Woodcock migration 3

http://www.esrl.noaa.gov/psd/

http://www.esrl.noaa.gov/psd/

autumn near their breeding grounds in CentralRussia, thanks to a partnership between ONCFSand a Russian organization (BirdsRussia).

Age (juvenile or adult) can be assessed easilyfor Eurasian Woodcock using plumage characteris-tics (Clausager 1973, Ferrand & Gossmann 2009),and age ratio (proportion of juveniles) provided anindex of breeding success at the population level.The age ratio of birds caught in Central Russia inearly autumn was especially valuable because itgave a measure of breeding success at the start ofautumn migration. However, the number of indi-viduals caught was low in comparison with thethousands of birds collected each year in France.The moult stage of juveniles in autumn/wintergave important information on their hatchingdate because early-hatched juveniles (earlybroods) moulted all their secondary coverts,whereas late-hatched juveniles (late broods)showed some juvenile coverts (autumn migrationstops moult of the wing feathers, Ferrand & Goss-mann 2009). The proportion of early-brood juve-niles therefore provided information on breedingphenology and its possible consequences forbreeding success, e.g. higher mortality rate inchicks reared later.

To link such population-level indices withspring migration weather, the tracking data loca-tions and dates were used as a sample to representWoodcock flyway movements. Important weathervariables found to affect migration movement were

extracted from 1996 to 2017 for each location anddate. As in the migration movement analysis,weather variables were averaged over 3 days (timet, t + 1 and t + 2) and their trends assessed over4 days (adding t � 1). All values were then aver-aged yearly, which resulted in a unique set ofweather conditions each year and enabled a correl-ative analysis with the breeding indices. We con-sidered three scenarios of change in migrationphenology from 1996 to 2017: (1) no change; (2)a linear shift of ¼ day per year (based on meanspring passage at Helgoland; H€uppop & H€uppop2003); and (3) a non-linear shift driven by springprecocity. Scenario (2) was evaluated by retrospec-tively shifting migration dates by 1 day every4 years because weather conditions were onlyextracted in the evening. Scenario (3) wasachieved using air temperature data extracted forscenario (1) to give an index of rise in spring tem-perature at the flyway scale. An annual delay oradvancement in spring temperatures was then cal-culated. This variable was scaled, as the overalltrend from 1996 to 2017 corresponded to a gen-eral shift of ¼ day per year, as in scenario (2). Thedifference in temperature values was then con-verted into a number of days, which allowed anon-linear shift of dates to be considered.

Age and early-brood ratios were also expectedto be influenced by weather on the breedinggrounds. If so, failing to account for these vari-ables may lead to fallacious results due to

Figure 1. The 729 locations from 68 individuals where spring migration movement of Woodcock could be assessed in a 68-helapsed time.


4 K. Le Rest et al.

spurious correlations between weather duringspring migration and the breeding period.Monthly mean temperature and precipitation inMay, June and July from 1996 to 2017 weretherefore extracted from known breeding loca-tions of 63 tracked Woodcocks. Total precipita-tion from January to April was extracted as anindex of pre-breeding soil–water conditions eachyear.

Statistical analyses

The effect of weather conditions on springmigration movements was examined using gener-alized linear mixed models, with a binomialerror term and a logit link function. Weathervariables were added as fixed effects and theindividual bird was specified as a random effect.Candidate covariates (fixed effects) were chosenaccording to prior knowledge or their expectedeffect on migration movements. It was not possi-ble to consider all subsets of these candidatecovariates (2k possibilities with linear effects).Covariate selection was thus done in severalsteps starting from a model including only tem-perature and wind effects (strong support fromprior knowledge). In total, about 150 subsets ofcovariates were considered (weather conditionsand individual-based variables, e.g. age andbreeding location) while avoiding collinearitybetween covariates (cross correlation < 0.7, seeDormann et al. 2013). All subsets of covariateswere then classified according to Aka€ıke’s infor-mation criterion (AIC) and we retained the bestmodel and alternative models having DAIC < 2,i.e. those having ‘substantial support’ accordingto Burnham and Anderson (2002). Row coeffi-cients of the models and their uncertainty areprovided in Appendix S2 (see Table S2) andaveraged parameters were calculated from thissubset of models (Symonds & Moussali 2011).Model selection was performed on a subsampleof 668 observations for which there were nomissing values for any of the covariates. How-ever, coefficients were estimated from the maxi-mum number of observations with no missingvalue in the set of covariates considered (from668 to 729 depending on the covariates consid-ered). Model residuals were checked for spatialcorrelation.

The effects of climatic variables extracted onthe breeding grounds, and the weather conditions

during migration on age and early-brood ratios,were investigated using linear models with normalerrors. Model residuals were visually checked fornormality. A linear model accounted for overdis-persion due to the independent variance parame-ter r of the error term, whereas the use of abinomial model would imply that variance was afunction of the mean, which ignores overdisper-sion. An alternative would be to use quasi-bino-mial models, but AIC could not be strictlyevaluated with such an approach. Note that theresults remained equivalent using either quasi-binomial or linear models. The source of data(birds ringed in Russia, ringed in France or huntedin France) was used as a factor to account for thedifferences between these sources, ensuring thatcovariates explained relative changes. Resultsremained unchanged whether data source wasconsidered as a random or a fixed effect. Parame-ters were estimated by weighted least squares toaccount for the differences in the number of indi-viduals ni used to calculate age and early-broodratios.

Candidate climatic variables were thoseselected in the migration movement analysis(eight variables from the best model selected)and conditions at the breeding grounds (sevenvariables, see above). These covariates could notall be considered in the same model because thenumber of covariates considered was high rela-tive to the number of observations, and becauseof collinearity between some of them. The analy-sis was thus carried out considering two mainpotential drivers of variation in observed indexvalues: (1) the climatic variables extracted onthe breeding grounds and (2) the weather condi-tions experienced during migration. The secondpotential driver of variation was investigated bytesting the three scenarios detailed above on thechanges in migration phenology since 1996, i.e.(scenario 1) no change in migration dates,(scenario 2) a shift of ¼ day per year and(scenario 3) an annual change in migration datesdepending on air temperature. Covariate selectionwas done independently for each of these foursets of covariates (see Tables S3 and S7, inAppendices S3 and S4, respectively). Model selec-tion was carried out using a corrected AIC (AICc)with n = 23 (number of studied years). AICc wascompared between the different scenarios, whichallowed assessment of the main cause of variationin age ratio and early-brood ratio.



RESULTS

Migration movements

Variable selection based on AIC resulted in reten-tion of 15 models with DAIC < 2. The modelshad from 10 to 15 parameters (Table 1,Appendix S2). Air surface temperature, winds(especially northward wind), atmospheric stabilityand relative humidity trend were the mainweather variables associated with migration move-ments (Table 1, Fig. 2). These variables were pre-sent in all of the 15 competing models, and theircoefficients were very similar across them (seeAppendix S2). Air pressure and relative humiditywere selected in most of the alternative models,but the estimated coefficients showed high uncer-tainties. Neither cloud cover nor cloud ceilingheight (visibility) were selected in the models. Thesame was true for vertical wind speed and soil

wetness. Precipitation rate had a skewed distribu-tion, which prevented a good evaluation of itseffect on migration movements.

Migration movements more frequently occurredin the presence of positive air surface temperatures(> 3 °C), stable atmospheric conditions (positivebest four-layer lifted index) and favourable winds(especially northward winds, Fig. 2). A quadraticeffect was fitted to atmospheric conditions, suchthat the highest values resulted in a decrease inmovement probability, but the number of observa-tions with such extreme conditions was small and,hence, the uncertainty was high. Conversely,migration movements occurred less frequently inthe presence of an increasing trend in relativehumidity. Movements were most frequent in thepresence of a slightly decreasing or stable relativehumidity trend. Uncertainty was very high forconditions with a strongly decreasing relativehumidity trend (Fig. 2). Bird breeding origin (lon-gitude and/or latitude of breeding area, linearmigration distance; Table 1) also showed weakpositive correlations with probability of migrationmovement. Longitude of the breeding area (Fig. 2)was the variable which best highlighted this corre-lation, with reduced uncertainty in most of themodels considered (Table 1, Appendix S2).Selected interactions (lifted index 9 air pressuretrend and eastward 9 northward wind) enhancedAIC values, but coefficient estimates showed highuncertainty.


The proportion of young among late autumnand winter birds collected in France was lowerthan that of Russian birds caught in earlyautumn. Moreover, age ratio differed betweenringed and hunted individuals in France, butboth showed synchrony over time (Fig. 3a,r = 0.79, n = 21 years). All the models consid-ered accounted for these differences throughinclusion of the factor ‘type of data’. Variableselection showed that the lowest AIC wasachieved by a model accounting for the climaticvariables during the breeding period (Table 2,Table S3 in Appendix S3). The alternative bestmodel, accounting only for weather conditionsduring spring migration, was that of the non-shifted migration hypothesis, but it performedvery poorly (DAIC > 30, Table S4). The mostimportant variables correlating with age ratio

Table 1. Averaged parameters from the best set of models(15 models with DAIC < 2, see Table S2 in Appendix S2) onthe probability of migration movement.

Covariates(scaled)

Averagedcoefficients

Averagedstandarderrors t-values

Intercept 0.60 0.15 3.95Longitude ofbreeding area

0.23 0.11 2.03

Air temperature 0.74 0.13 5.73Eastward wind(u-component)

0.15 0.09 1.66

Northward wind(v-component)

0.50 0.10 4.90

Air pressure 0.15 0.10 1.46Air pressure trend 0.14 0.10 1.40Best four-layerlifted index

0.23 0.12 1.90

(Best four-layerlifted index)²

�0.25 0.07 �3.55

Relative humidity 0.15 0.12 1.27Relative humiditytrend

�0.14 0.09 �1.53

(Relative humiditytrend)²

�0.16 0.06 �2.71

(Eastward wind9 Northward wind)

0.13 0.09 1.55

Best four-layer liftedindex 9 Air pressuretrend

0.14 0.10 1.44

Latitude of breedingarea

�0.12 0.11 �1.03

Migration distance 0.17 0.12 1.42


6 K. Le Rest et al.

were temperature in May and precipitation inJuly. The former was negatively related to theproportion of juveniles, whereas the latter was

positively related and stabilized in the presenceof high precipitation (Table 2, Fig. 4). Age ratiowas also higher in the presence of high

Figure 2. Effect of air temperature (°C), atmospheric stability (best four-layer lifted index), wind speed (m/s), relative humidity trendand longitude of breeding area (decimal degree) on migration movements. The grey line with shaded area represent the mean andits 95% confidence interval. The dashed horizontal line corresponds to a probability of migration movement of 0.5.



temperature and precipitation in June. Overall,climatic variables on the breeding groundsexplained 65% of the observed variation in ageratio (calculated from the deviance ratio of themodel including only the factor ‘source ofdata’). In contrast, the models accounting forweather conditions during migration onlyexplained 10% (non-linear shifted migration),15% (linear shifted migration) and 37% (non-shifted migration) of the variation in age ratio(Tables S3_1 to S3_3 in Appendix S3). Moreimportantly, the main weather variables corre-lated with age ratio were not the ones thatwere strongly related to migration movements.In particular, coefficients estimated from airpressure and air pressure trend exhibited highuncertainties (Table 1).

Early-brood ratio was quite similar betweenhunted and ringed individuals in France (r = 0.48,

n = 21 years). Values in the Russian data wereslightly higher (Fig. 3b) but between-year variationwas high, so selection based on AIC favouredmodels without the factor ‘type of data’(Table S4_1 in Appendix S4). The lowest AIC inthe models examining early-brood ratio wasachieved by a model accounting for weather con-ditions under the hypothesis of non-shifted migra-tion. Air pressure and relative humidity showednegative correlations with early-brood ratio(Table 3), but their effects on migration move-ments had high uncertainty (Table 1,Appendix S2). The models evaluated under thealternative hypotheses of linear and non-linearshifted migration phenology performed poorly(Appendix S4). Of the climatic variables extractedat the breeding grounds, only temperature in Maywas selected. Early-brood ratio was higher in thepresence of higher temperatures in May.

Figure 3. Trends in (a) age ratio and (b) early-brood ratio from 1996 to 2017. Each source of data is drawn using a different colour(see key).


8 K. Le Rest et al.

DISCUSSION

Migration movements

Spring migration movements of Eurasian Wood-cock were clearly associated with particularweather conditions. Air surface temperatureshowed the strongest positive correlation, butwind, atmospheric stability and relative humiditytrend were also highly related to migration move-ment probability (Fig. 2). Many studies of birdmigration have shown that temperature has amajor role in migration schedules (e.g. Marra et al.2005, H€uppop & Winkel 2006, Gordo 2007,Tøttrup et al. 2010). Woodcock in particularavoided moving when temperatures were below ornear 0 °C, and migration movement frequencyreached 50% above 4 °C and 80% above 11 °C.Temperature may have an indirect effect onmigration movements, i.e. increasing food abun-dance and availability at new sites along the migra-tion route, which consequently stimulates birds tocontinue their travel. This would be analogous tothe green wave hypothesis for goose migration (Siet al. 2015, and references therein).

Wind is also known to be an important factorinfluencing migration movements (Liechti 2006,Sinelschikova et al. 2007). Biologging technologiespermitting the tracking of birds throughout theirmigration travels have recently highlighted theimportance of this variable at the individual level(Conklin & Battley 2011, Safi et al. 2013). In ourstudy, wind direction and strength were strongly

correlated with spring migration movements ofWoodcock (Fig. 2), with individuals logically tak-ing advantage of tailwinds. Intriguingly, the effectof northward wind was much stronger than theeffect of eastward wind, although there was noapparent advantage to Woodcock in heading northfaster. Wind speed and direction may not onlyfacilitate migration flights but may also providesome information on conditions at the followingstopover sites. In particular, northward winds maygive important information about temperaturetrend (and snowmelt) at a broad scale, whereaseastward winds are associated with rainy weatherand atmospheric instability (which may generatesnow when reaching cold areas). Another explana-tion might be related to the structure of winds inthe different layers of the atmosphere. Indeed,birds may adapt their flight altitude by a compro-mise between wind support and air temperature(Kemp et al. 2013). We need reliable data on theflight altitude of tracked Woodcock to investigatethis hypothesis.

Atmospheric conditions have not often beenshown to influence non-soaring bird migration(Richardson 1990b). Yet, in theory, stable atmo-spheric conditions could also be advantageous forflapping-flight migration by reducing turbulence inthe air and the risk of encountering storms, thusencouraging the bird to migrate. Atmospheric sta-bility was evaluated here by the best four-layerlifted-index and was the third most important vari-able (after temperature and wind) explainingWoodcock spring migration movements (based onestimated coefficients, Table 1). Moreover, asexpected, Woodcock migration movements werelower when relative humidity increased. We wereunable to find any study in which this variable haspreviously been related to migration movements.Both atmospheric stability and relative humiditytrend should thus be given greater consideration infuture studies modelling bird migration movements.

Migration movements of Woodcock stilloccurred when weather conditions were not opti-mal. Indeed, the movement probability was rarelylower than 0.2 for each single effect (Fig. 2). Thus,it seems that only a combination of unfavourableelements could drastically reduce migration move-ment probability towards zero and, conversely, justa few favourable factors could encourage somebirds to continue spring migration. A large amount(88%) of the observed variance (evaluated fromthe null model with only the individuals

Table 2. Estimated parameters of the best model fitting ageratio to the climatic variables at the breeding grounds (seeAppendix S3 for other models, all having DAIC ≫ 2).

Covariates (scaled)Estimatedcoefficients

Estimatedstandarderrors t-values

Intercept 59.27 0.83 71.41Source: ringedin Russia

16.94 3.40 4.98

Source: huntedin France

7.35 0.73 10.07

Temperature May �2.50 0.41 �6.10Precipitation July 2.89 0.40 7.22Precipitation June 1.69 0.41 4.12Temperature June 0.75 0.45 1.67(Temperature June)² 1.57 0.37 4.24(Precipitation July)² �1.54 0.46 �3.35R²: 0.80 AICc: 397.2



considered as a random effect) remained unex-plained by the best model. A part of this unex-plained variance may be related to the accuracy ofthe variable used in the model. In particular, somevariables may have spatial variation in their

accuracy according to the density of weather sta-tions, which may create additional variance oreven bias in analysis (Baker et al. 2017). Anotherpart of this unexplained variance results from vari-ables and processes unaccounted for in the models.For instance, energy expenditure and body condi-tion could affect the decision of Woodcock to stopor fly on at the individual level throughout themigration. Variables related to visibility (cloudcover and cloud-ceiling height) were not shown toaffect Woodcock migration movements.

We found no difference in migration move-ments between juvenile (second calendar year)and adult Woodcock (third or more calendaryear). However, Woodcock breeding further easthad a higher rate of migration movements thanbirds breeding closer to the wintering grounds(Table 1). This is mainly related to the migrationdistance the birds have to cover, as the longitude

Figure 4. Effect of temperature (°C) in May and June and precipitation (mm) in June and July on age ratio. The grey line withshaded area represent the mean and its 95% confidence interval.

Table 3. Estimated parameters of the best model fitting early-brood ratio to the weather conditions under the hypothesis ofnon-shifted migration phenology (see Appendix S4 for othermodels, all having DAIC > 2).

Covariates(scaled)

Estimatedcoefficients

Estimatedstandarderrors t values

Intercept 63.89 0.47 135.94Air pressure �1.98 0.45 �4.4Relative humidity �1.56 0.47 �3.32(Air pressure)² �0.97 0.30 �3.23R²: 0.33 AICc: 379.01


10 K. Le Rest et al.

of the breeding area is highly correlated with thedistance between the wintering and breeding areas(r = 0.97, n = 82 spring travels, DAIC of the alter-native model = 1.1, see Appendix S2). Woodcockbreeding further east and undertaking longermigrations thus compensate for their long-distancetravel by increasing their rate of migration move-ment. Migration will have a greater impact on thefitness of these individuals, and they may need tobe more efficient at finding resources at stopoverareas and/or their breeding grounds might beexpected to be of higher quality.


Age ratio was higher among hunted Woodcockthan those captured for ringing, but both trendsacross years were very similar. The higher age ratioamong hunted birds is likely to be due to differen-tial distribution of young birds according to hunt-ing pressure, such that young birds reachingwintering areas for the first time are more likely tosettle in hunted areas by accident than are adults,which almost always returned to the same winter-ing location (Fadat 1991). Once accounting for thefactor ‘source of data’, annual age ratio was mostlyrelated to climatic variables on the breedinggrounds. Conversely, weather conditions duringspring migration were unlikely to influence the pro-portion of juveniles. Temperature in May and Juneand precipitation in June and July showed strongcorrelations with age ratio. In May, most femaleWoodcock on the Russian breeding grounds areincubating, so a higher temperature might be con-strued as good for breeding. However, temperatureand precipitation were positively correlated in May(r = 0.43, n = 23 years) and heavy precipitationmight reduce hatching success. The negative corre-lation between age ratio and temperature in Maysuggests that unstable weather in May (high tem-peratures and precipitation) was detrimental tohatching success. Conversely, in June, most youngwould have hatched and would benefit from hightemperatures and precipitation, which was consis-tent with the correlations we found. In July, dryweather will probably drive earthworms deeper intothe soil and make it harder for Woodcock to probe,which could prevent them from feeding on soilinvertebrates (Peach et al. 2004, Hoodless & Hirons2007). Low rainfall in July could particularly affectWoodcock chicks, most of which will be full grownby July but still relatively na€ıve foragers, and could

thus increase post-fledging mortality. Guzm�an andArroyo (2015) also showed a negative effect of tem-perature in July and a positive effect of precipitationon relative Woodcock abundance in Spain duringthe following winter, which would be partly relatedto the breeding success that summer. However,they also found a positive effect of precipitation inMay, but did not provide interpretation of thisresult.

We evaluated breeding success using the ageratio, which only gave the proportion of youngWoodcock in the population a few months afterthe breeding period. The limitations in the inter-pretation of results on age ratio must be borne inmind because, although breeding productivity willhave a strong influence on age ratio values, condi-tions during autumn migration may affect adultsand juveniles differently, e.g. differential mortalityand/or migration routes. The abundance of Wood-cock in Western Europe in a given winter is likelyto depend strongly on weather conditions (espe-cially frost) in northern and eastern countries,which push Woodcock towards southern areassuch as Spain (P�eron et al. 2011). Unfortunately,this was not taken into account by Guzm�an andArroyo (2015), and hence the relative importanceof breeding ground conditions and autumn/winterweather (and their correlations) on the abundanceof wintering Woodcock still need to be evaluated.

In contrast to age ratio, the early-brood ratio waspoorly related to the covariates studied. The weathervariables during spring migration that werecorrelated with early-brood ratio were not those cor-related with migration movements. Such inconsisten-cies were probably due to multicollinearity betweenthe covariates measured during spring migration andother important variables unaccounted for in ourmodels. Of the variables extracted on the breedinggrounds, only temperature in May may have a weakpositive influence on early-brood ratio. This is plausi-ble, as high temperatures in May advance spring phe-nology and hence hatching date (Hoodless &Coulson 1998). However, because temperature inMay was negatively correlated with age ratio, itmight produce antagonistic effects on breeding suc-cess and breeding precocity.

Implications for population dynamics

In a recent review, Shamoun-Baranes et al. (2017)provided an overview of the effect of atmosphericconditions on bird migration at the individual level



and gave some insights into the consequences ofthese conditions at the population level. Theauthors highlighted that the feedback of weatherconditions experienced during migration on popu-lation dynamics has not yet been sufficiently inves-tigated. We have shown that the weatherconditions experienced during spring migrationhad little influence on breeding success (age ratioafter the breeding period) or breeding phenology(early-brood ratio) of Woodcock. Breeding successwas mainly correlated with climatic variables onthe breeding grounds, which could influence bothhatching success and the survival of juveniles. Pop-ulation dynamics of Eurasian Woodcock winteringin Western Europe are thus more likely to berelated to the conditions on the breeding groundsthan to conditions experienced during the springmigration, which is consistent with the incomebreeding strategy of this species (Stephens et al.2009). Although many studies have examinedcarry-over or delayed effects of previous conditionson breeding success, most have also concluded thatconditions on the breeding grounds were the mostimportant (Legagneux et al. 2012, Ockendon et al.2013, but see Finch et al. 2014).

Although spring migration weather conditionswere not shown to affect breeding success of Eura-sian Woodcock wintering in Western Europe at thepopulation level, they could affect the fitness ofbirds at the individual or sub-population level.Indeed, indices used here reflected what was hap-pening on average, but may have hidden differencesbetween individuals or between geographicalregions. Breeding success should ideally be evalu-ated at the individual level to assess further its linkwith spring migration conditions. In particular, birdsbreeding further from their wintering areas (east ofthe Ural Mountains) would theoretically be themost impacted by weather conditions during migra-tion. These birds represented a small proportion ofthe population studied here, probably because ofevolutionary processes exerting selection pressureon migration (Bell 2000, Liedvogel et al. 2011, Sha-moun-Baranes et al. 2017). However, the propor-tion of these long-distance migrants is higher thanpreviously thought (see Bauthian et al. 2007). Itwould be interesting to evaluate whether theiroccurrence in winter in Western Europe has chan-ged over recent decades and, if so, why.

GWCT funded the study in Britain and Ireland, thanksmainly to private donations for tags and funds raised

by the Shooting Times Woodcock Club. ONCFS andCNB financed the tags in France. CNB financed thetags thanks to a membership fee as well as privatedonations. We would like to thank Bruno Meunier(President of CNB) and all the members of CNB fortheir input since the first stages of this project. OwenWilliams (The Woodcock Network) and Luke Harman(University College Cork) helped to fit tags in Walesand Ireland. Pierre Launay (CNB), Damien Coreau andFranc�ois Gossmann (ONCFS), as well as many ringersfrom R�eseau B�ecasse (ONCFS/FNC/FDC) helped tocatch Woodcock and fit tags in France. One anony-mous reviewer, the associate editor, James Pearce-Hig-gins, and the editor, Dan Chamberlain, providedvaluable comments and corrections which greatlyimproved the manuscript.

REFERENCES

Alerstam, T. 2011. Optimal bird migration revisited. J.Ornithol. 152(S1): 5–23.

Alerstam, T., Hedenstrom, A. & Akesson, S. 2003. Long-distance migration: evolution and determinants. Oikos 103:247–260.

Baker, D.J., Hartley, A.J., Pearce-Higgins, J.W., Jones,R.G. & Willis, S.G. 2017. Neglected issues in using weatherand climate information in ecology and biogeography.Divers. Distrib. 23: 329–340.

Bauthian, I., Gossmann, F., Ferrand, Y. & Julliard, R. 2007.Quantifying the origin of Woodcock wintering in France. J.Wildl. Manage. 71: 701–705.

Bell, C.P. 2000. Process in the evolution of bird migration andpattern in avian ecogeography. J. Avian. Biol. 31: 258–265.

Berthold, P. & Helbig, A.J. 1992. The genetics of birdmigration: stimulus, timing, and direction. Ibis 134(S1): 35–40.

Berthold, P., Wiltschko, W., Miltenberger, H. & Querner, U.1990. Genetic transmission of migratory behavior into anonmigratory bird population. Experientia 46: 107–108.

Both, C. & Visser, M.E. 2001. Adjustment to climate changeis constrained by arrival date in a long distance migrant bird.Nature 411: 296–298.

Both, C., Bijlsma, R.G. & Visser, M.E. 2005. Climatic effectson timing of spring migration and breeding in a long-distance migrant, the Pied Flycatcher Ficedula hypoleuca. J.Avian. Biol. 36: 368–373.

Burger, A.E. & Shaffer, S.A. 2008. Application of trackingand data-logging technology in research and conservation ofseabirds. Auk 125: 253–264.

Burnham, K.P. & Anderson, D.R. 2002. Model Selection andMultimodel Inference: A Practical Information-theoreticalApproach, 2nd edn. New York: Springer.

Clausager, I. 1973. Age and sex determination of theWoodcock, Scolopax rusticola. Danish Rev. Game Biol. 8:1–18.

Conklin, J.R. & Battley, P.F. 2011. Impacts of wind onindividual migration schedules of New Zealand Bar-tailedGodwits. Behav. Ecol. 22: 854–861.

Crespo, A., Rodrigues, M., Telletxea, I., Ib�a~nez, R., D�ıez, F.,Tobar, J.F. & Arizaga, J. 2016. No habitat selection during



spring migration at a meso-scale range across mosaiclandscapes: a case study with the Woodcock (Scolopaxrusticola). PLoS One 11: e0149790.

Dormann, C.F., Elith, J., Bacher, S., Buchmann, C., Carl,G., Carr�e, G., Garc�ıa Marqu�ez, J.R., Gruber, B.,Lafourcade, B., Leit~ao, P.J., M€unkem€uller, T., McClean,C., Osborne, P.E., Reineking, B., Schr€oder, B., Skidmore,A.K., Zurell, D. & Launtenbach, S. 2013. Collinearity: areview of methods to deal with it and a simulation studyevaluating their performance. Ecography 36: 27–46.

Duriez, O., Ferrand, Y., Binet, F., Corda, E., Gossmann, F.& Fritz, H. 2005. Habitat selection of the EurasianWoodcock in winter in relation to earthworms availability.Biol. Cons. 122: 479–490.

Fadat, C. 1991. Age-ratio des tableaux de chasse deb�ecasses (Scolopax rusticola). Signification biologique etutilisation pour la bonne gestion des populationsb�ecassi�eres. Bulletin Mensuel de l’ONC – Num�ero Sp�ecialScientifique et Technique, 141–172.

Ferrand, Y. & Gossmann, F. 2009. Ageing and sexing series5: ageing and sexing the Eurasian Woodcock Scolopaxrusticola. Wader Study Group Bull. 116: 75–79.

Finch, T., Pearce-Higgins, J.W., Leech, D.I. & Evans, K.L.2014. Carry-over effects from passage regions are moreimportant than breeding climate in determining the breedingphenology and performance of three avian migrants ofconservation concern. Biodivers. Conserv. 23: 2427–2444.

Gordo, O. 2007. Why are bird migration dates shifting? Areview of weather and climate effects on avian migratoryphenology. Clim. Res. 35: 37–58.

Guzm�an, J.L. & Arroyo, B. 2015. Predicting winterabundance of Woodcock Scolopax rusticola using weatherdata: implications for hunting management. Eur. J. Wildl.Res. 61: 467–474.

Hoodless, A.N. & Coulson, J.C. 1998. Breeding biology ofthe Woodcock Scolopax rusticola, in Britain. Bird Study 45:195–204.

Hoodless, A.N. & Hirons, G.J.M. 2007. Habitat selection andforaging behaviour of breeding Eurasian WoodcockScolopax rusticola: a comparison between contrastinglandscapes. Ibis 149: 234–249.

H€uppop, O. & H€uppop, K. 2003. North Atlantic Oscillationand timing of spring migration in birds. Proc. R. Soc. Lond.B Biol. Sci. 270: 233–240.

H€uppop, O. & Winkel, W. 2006. Climate change and timing ofspring migration in the long-distance migrant Ficedulahypoleuca in central Europe: the role of spatially differenttemperature changes along migration routes. J. Ornithol.147: 326–343.

Jenni, L. & Schaub, M. 2003. Behavioural and physiologicalreactions to environmental variation in bird migration: areview. In Berthold, P., Gwinner, E. & Sonnenschein, E.(eds) Avian Migration: 155–171. Berlin: Springer.

Kalnay, E., Kanamitsu, M., Kistler, R., Collins, W., Deaven,D., Gandin, L., Iredell, M., Saha, S., White, G., Woollen,J., Zhu, Y., Chelliah, M., Ebisuzaki, W., Higgins, W.,Janowiak, J., Mo, K.C., Ropelewski, C., Wang, J.,Leetmaa, A., Reynolds, B., Jenne, R. & Joseph, D. 1996.The NCEP/NCAR 40-year reanalysis project. Bull. Am.Meteor. Soc. 77: 437–470.

Kemp, M.U., Shamoun-Baranes, J., Dokter, A.M., vanLoon, E. & Bouten, W. 2013. The influence of weather on

the flight altitude of nocturnal migrants in mid-latitudes. Ibis155: 734–749.

Klaassen, R.H.G., Hake, M., Strandberg, R., Koks, B.,Trierweiler, C., Exo, K.M., Bairlein, F. & Alerstam, T.2014. When and where does mortality occur in migratorybirds? Direct evidence from long-term satellite tracking ofraptors. J. Anim. Ecol. 83: 176–184.

Legagneux, P., Fast, P.L.F., Gauthier, G. & Bety, J. 2012.Manipulating individual state during migration providesevidence for carry-over effects modulated by environmentalconditions. Proc. R. Soc. B Biol. Sci. 279: 876–883.

Liechti, F. 2006. Birds: Blowin’ by the wind? J. Ornithol. 147:202–211.

Liedvogel, M. & Lundberg, M. 2014. The genetics ofmigration. In Hansson, L.-A. & �Akesson, S. (eds) AnimalMovement Across Scales: 219–231. New York: OxfordUniversity Press.

Liedvogel, M., �Akesson, S. & Bensch, S. 2011. The geneticsof migration on the move. Trends Ecol. Evol. 26: 561–569.

Marra, P.P., Francis, C.M., Mulvihill, R.S. & Moore, F.R.2005. The influence of climate on the timing and rate ofspring bird migration. Oecologia 142: 307–315.

McKinnon, L., Smith, P.A., Nol, E., Martin, J.L., Doyle, F.I.,Abraham, K.F., Gilchrist, H.G., Morrison, R.I.G. & Bety, J.2010. Lower predation risk for migratory birds at highlatitudes. Science 327: 326–327.

Møller, A.P., Rubolini, D. & Lehikoinen, E. 2008.Populations of migratory bird species that did not show aphenological response to climate change are declining.Proc. Natl Acad. Sci. USA 105: 16195–16200.

Newton, I. 2006. Can conditions experienced during migrationlimit the population levels of birds? J. Ornithol. 147: 146–166.

Newton, I. 2011. Migration within the annual cycle: species,sex and age differences. J. Ornithol. 152: 169–185.

Ockendon, N., Leech, D. & Pearce-Higgins, J.W. 2013.Climatic effects on breeding grounds are more importantdrivers of breeding phenology in migrant birds thancarryover effects from wintering grounds. Biol. Let. 9:20130669.

Peach, W.J., Denny, M., Cotton, P.A., Hill, I.F., Gruar, D.,Barrit, D., Impey, A. & Mallard, J. 2004. Habitat selectionby Song Thrushes in stable and declining farmlandpopulations. J. Appl. Ecol. 41: 275–293.

P�eron, G., Ferrand, Y., Gossmann, F., Bastat, C.,Guenezan, M. & Gimenez, O. 2011. Nonparametric spatialregression of survival probability: visualization of populationsinks in Eurasian Woodcock. Ecology 92: 1672–1679.

Pulido, F. 2007. Phenotypic changes in spring arrival:evolution, phenotypic plasticity, effects of weather andcondition. Clim. Res. 35: 5–23.

Pulido, F., Berthold, P. & van Noordwijk, A.J. 1996.Frequency of migrants and migratory activity are geneticallycorrelated in a bird population: evolutionary implications.Proc. Natl Acad. Sci. USA 93: 14642–14647.

Rappole, J.H. & Tipton, A.R. 1991. New harness design forattachment of radio transmitters to small passerines. J. FieldOrn. 62: 335–337.

Richardson, W.J. 1990a. Timing of bird migration in relationto weather: updated review. In Gwinner, E. (ed.) BirdMigration: 78–101. Berlin: Springer.

Richardson, W.J. 1990b. Wind and orientation of migratingbirds: a review. Experientia 46: 416–425.



Safi, K., Kranstauber, B., Weinzierl, R., Griffin, L., Rees,E.C., Cabot, D., Cruz, S., Proa~no, C., Takekawa, J.Y.,Newman, S.H., Waldenstr€om, J., Bengtsson, D., Kays, R.,Wikelski, M. & Bohrer, G. 2013. Flying with the wind: scaledependency of speed and direction measurements inmodelling wind support in avian flight. Move. Ecol. 2013: 1–4.

Saino, N., Ambrosini, R., Rubolini, D., von Hardenberg, J.,Provenzale, A., H€uppop, K., H€uppop, O., Lehikoinen, A.,Lehikoinen, E., Rainio, K., Romano, M. & Sokolov, L.2011. Climate warming, ecological mismatch at arrival andpopulation decline in migratory birds. Proc. R. Soc. Lond. BBiol. Sci. 278: 835–842.

Sandberg, R. & Moore, F.R. 1996. Fat stores and arrival onthe breeding grounds: reproductive consequences forpasserine migrants. Oikos 77: 577–581.

Shamoun-Baranes, J., Bouten, W. & van Loon, E. 2010.Integrating meteorology into research on migration. Integr.Comp. Biol. 50: 280–292.

Shamoun-Baranes, J., Liechti, F. & Vansteelant, W.M.G. 2017.Atmospheric conditions create freeways, detours and tailbacksfor migrating birds. J. Comp. Physiol. A. 203: 509–529.

Si, Y., Xin, Q., de Boer, W.F., Gong, P., Ydenberg, R.C. &Prins, H.H.T. 2015. Do Arctic breeding geese track or overtakea green wave during spring migration? Sci. Rep. 5: 8749.

Sinelschikova, A., Kosarev, V., Panov, I. & Baushev, A.2007. The influence of wind conditions in Europe on theadvance in timing of the spring migration of the SongThrush (Turdus philomelos) in the south-east Baltic region.Int. J. Biometeorol. 51: 431–440.

Stephens, P.A., Boyd, I.L., McNamara, J.M. & Houston, A.I.2009. Capital breeding and income breeding: their meaning,measurement, and worth. Ecology 90: 2057–2067.

Symonds, M.R.E. & Moussali, A. 2011. A brief guide tomodel selection, multimodel inference and model averaging

in behavioural ecology using Akaike’s information criterion.Behav. Ecol. Sociobiol. 65: 13–21.

Tøttrup, A.P., Rainio, K., Coppack, T., Lehikoinen, E.,Rahbek, C. & Thorup, K. 2010. Local temperature fine-tunes the timing of spring migration in birds. Integr. Comp.Biol. 50: 293–304.

Weimerskirch, H., Delord, K., Guitteaud, A., Phillips, R.A. &Pinet, P. 2015. Extreme variation in migration strategiesbetween and within wandering albatross populations duringtheir sabbatical year, and their fitness consequences. Sci.Rep. 5: 8853.

Received 11 September 2017;revision accepted 7 August 2018.

Associate Editor: James Pearce-Higgins.

SUPPORTING INFORMATION

Additional supporting information may be foundonline in the Supporting Information section atthe end of the article.

Appendix S1. Details on the weather variablesused to study migration movements.

Appendix S2. Covariate selection, parameterestimates and model averaging for migration move-ment probability.

Appendix S3. Covariate selection and para-meter estimates for age ratio.

Appendix S4. Covariate selection and para-meter estimates for early-brood ratio.



Documents

Effect of weather conditions on the spring migration of ... · Woodcock Scolopax rusticola ﬁtted with Argos satellite tags in the British Isles and France (2012–17). First, we