Prospecting movements during the transit phase of immature eagles are driven by age, sex and season

Abstract

Background

Dispersal includes three phases: emigration, transit, and immigration. The transit phase, which involves all movements between departure and arrival, is the least understood phase of dispersal. During the transit phase, individuals prospect their environment to gather information about potential breeding sites, thus enhancing their future reproductive success and survival. Studies have revealed a wide inter-individual variability in prospecting behaviours which may result from complex interactions between external and internal factors affecting the costs and benefits of prospecting. Age, sex, and season are expected to strongly influence prospecting behaviours, yet their effects are far from established.

Methodology

We investigated how age, sex, and season interact and influence prospecting movements throughout the transit phase. We analysed telemetry data from 106 immature Golden eagles (Aquila chrysaetos), whose natal dispersal involves a transit phase lasting several years. Using a trajectory segmentation method, we identified the areas sequentially prospected by each individual and we assessed the size, duration of use, and spacing between these areas to uncover spatio-temporal variations in prospecting behaviours.

Results

We confirmed our predictions, revealing strong influences of age, sex, and season, as well as their interactions, on prospecting movements. First, age had a significant effect on prospecting behaviours: individuals displayed a progressive spatial concentration of prospecting, consistent with patterns observed in colonial species. Second, seasonal variations were detected, with peaks of prospection in spring and autumn, likely resulting from the constraints imposed by territorial adult reproduction and weather-related flight conditions. Third we found sexual differences in movement patterns, with females prospecting over a larger spatial range than males, in line with the female-biased dispersal existing in most bird species. The level of inter-sexual differences and seasonal variations in prospecting behaviours differed depending on the age of the individuals.

Conclusions

Our work strongly supports that individuals adjust their prospecting behaviour in response to interacting intrinsic and extrinsic factors, in order to reduce prospecting movement costs while maximising the information gathered to inform their immigration decision.

Background

Dispersal refers to any movement from the birth site to a reproductive site (natal dispersal), or between two breeding sites (breeding dispersal) [1]. Dispersal plays a key role in many ecological and evolutionary processes: it affects population dynamics and genetics [2, 3], it structures communities and ecosystems [4], and it impacts local adaptation and speciation [5]. Dispersal is selected because it reduces competition with kin, grants access to higher quality habitats, and increase individuals fitness in a context of environmental and demographic stochasticity [6].

Dispersal includes three phases: (i) emigration, which is the departure from the birth or reproductive site; (ii) transit, which includes all movements between departure and settlement; and (iii) immigration, when individuals settle in a new breeding site [7]. According to the informed dispersal theory [8], in active dispersers such as birds, individuals prospect their environment before immigrating in order to acquire personal and/or social information on potential breeding sites [8, 9]. Prospecting individuals can then select the best breeding sites among the several prospected ones, thereby enhancing their future reproductive success and survival [8, 9]. However, prospecting implies significant costs for individuals, such as energy costs of performing long distance movements and increased mortality risk in unknown environments [10]. Depending on the life history traits of each species, individuals can prospect either before emigration or afterward, during the transit phase [11, 12]. In long-lived species with delayed sexual maturity, the transit phase of natal dispersal can last several years and is commonly characterised by large-scale prospecting movements [11, 13]. This period is critical for individuals because it is associated with a high mortality rate, and also determines their immigration site which impact their future reproductive [9, 14]. Despite its considerable impact on survival and reproduction, the dispersal transit phase has received less attention than emigration and immigration phases [11]. This is due to the fact that most studies on dispersal relied on genetic and on capture-mark-recapture methods which do not allow the study of the transit phase [15,16,17], or on tracking devices that have limitations in monitoring individuals over extended periods with high spatial and temporal resolution, particularly for species with small body sizes. Improvements in tracking technologies now provides opportunities to study this critical phase of dispersal [9, 18, 19].

During the transit phase, prospecting individuals display specific patterns of movements. They move further, straighter and faster [11, 20], and have larger home ranges than individuals already settled on breeding sites [21, 22]. However, studies have revealed a wide inter-individual variability in these movement behaviours during the transit phase [23, 24]. For instance, in mammals such as the Roe deer (Capreolus capreolus), dispersers have several alternative prospecting tactics characterized by various timing, amplitude and duration of movements [25]. These various prospecting behaviours have a substantial impact on the dispersal process by affecting the distance at which individuals disperse [26]. The intraspecific variability in prospecting behaviour may result from complex interactions between external factors (e.g., physical and social environment), and intrinsic factors (e.g., personality, age, sex), that affects the costs and benefits of prospecting movements [24, 27]. Yet, empirical studies investigating which factors influence prospection pattern and how they interplay during the transit phase remain scarce [24].

Prospecting movement patterns are first expected to depend on the age of individuals [24]. Studies on long-lived species, such as colonial seabirds, suggested that during the pre-breeding period, younger individuals may benefit more from prospecting on a large spatial range than older individuals that have already identified areas representing potential future breeding sites [27, 28]. During the transit phase of natal dispersal, old pre-breeders are therefore expected to move less while displaying a more intensive prospecting to a few selected areas compared to young ones (Hypothesis 1). In territorial species (that constitute the bulk of bird species), only few research programs have successfully tracked a significant number of dispersing individuals over several years of their transit phase. This is partly due to technical limitations as a large part of territorial birds are passerines, whose small body size makes long-term tracking by telemetry particularly challenging. As a result, most studies have examined only the first year of dispersal and could not test the effect of age on prospecting behaviours [13, 28]. The few studies that have succeeded in tracking dispersal of territorial species for more than one year have shown contrasting results. Studies found that in the Spanish imperial eagle (Aquila adalberti) and the Red kite (Milvus milvus), old pre-breeders move less than young ones suggesting that aging individuals tend to spatially focus their prospecting efforts [29, 30], while in the Bearded vulture (Gypaetus barbatus) and the Golden eagle (Aquila chrysaetos), the opposite was shown [22, 31]. However, these studies were carried out on small sample sizes of individuals within each age class, which makes the results difficult to interpret, and could explain the lack of consistency between studies.

Second, intra-annual variations in prospecting behaviours may arise due to both the breeding phenology of adults and seasonal variations in weather conditions [32, 33]. In colonial birds and in passerines, a peak in prospecting activity has been observed during the chick-rearing period [32, 34]. Prospecting during the breeding period would be beneficial because individuals could assess the reproductive success of ongoing breeding individuals, which is a good indicator of habitat quality for future immigration [32]. The seasonal fluctuations in weather conditions may also influence prospecting behaviours, especially for soaring birds [24, 33]. Prospecting raptors fly over greater daily distances when solar radiation is high and the wind is favourable [33], because these conditions generate updrafts that reduce their cost of flight [35, 36]. In temperate regions, thermal updrafts are weak in winter, and flight conditions are more favourable between spring and autumn which may facilitate prospecting movements during this period [37]. We therefore expect intra-annual variations in prospecting behaviours during the transit phase, with more extensive movements during the breeding season and when flight conditions are favourable (Hypothesis 2).

According to Hypotheses 1 and 2, dispersers are expected to undertake fewer long-distance movements as they age, with such movements being synchronized with seasons that provide favourable flight conditions. Since long-distance travel is more energetically demanding and its efficiency depends on favourable seasonal conditions (e.g., thermal updrafts), aging dispersers that cover shorter distances may benefit less from this synchronisation. Consequently, we may expect an interaction between age and season to shape prospecting behaviours, with seasonal variations in prospecting declining during the later stages of the transit phase (Hypothesis 3). In previous studies, the effect of season on prospecting behaviour was tested only by mixing all age classes, making it impossible to detect an interaction between the effect of age and season on prospecting behaviours [31, 33].

Third, sex may affect prospecting patterns [38]. In birds, as females tend to disperse further than males [39], they should also prospect at a larger spatial range than males do (Hypothesis 4) [24, 38]. Studies on long-lived territorial species that have investigated intersexual differences in prospecting during transit have shown contrasting results. For instance, in territorial raptor, Ferrer et al. [40] has shown that female Spanish imperial eagles prospect significantly further than males during transit, whilst in Bonelli eagles a congeneric species with a relatively similar ecology, Cadahía et al. [13] found no sexual difference. This lack of consensus may be due to the interplay between age and sex, resulting in sexual differences fluctuating throughout the transit period that last several years in these species [41]. In long-lived bird species where competition for breeding sites is strong, males, which often access and defend territories, would be constrained to delay their first breeding more than females and would therefore tend to settle at an older age [42, 43]. In such species, the transit phase of the natal dispersal should be shorter in females, and we can expect that they spatially focus their prospecting on a few areas more quickly than males (Hypothesis 5).

Here we aimed at investigating how age, season, and sex influence prospection behaviours, as well as the age-season and age-sex interactions during the transit phase of dispersal. This research focused on the Golden eagle, a long-lived territorial raptor, whose first reproduction typically occurs between the ages of four and seven [44]. Natal dispersal in this species includes a transit phase lasting between one and six years [44, 45]. We studied prospecting behaviours during the three first years of transit of 106 GPS-tagged immatures individuals (61 males and 44 females) from three regions from France. According to Hypothesis 1, we predicted a spatial concentration of prospecting over the three first years of transit. In our study population, egg-laying starts on average at 20 March, hatching occurs on average at 01 May, fledgling happens on average at 20 July, and the flight conditions are favourable between spring and late autumn [46]. In line with Hypothesis 2 and 3, we therefore expected to find more extensive prospecting during the spring (breeding season) and reduced in winter (unfavourable flight condition), with these seasonal variations reducing throughout the transit phase as individuals age. Female Golden eagles disperse significantly further than males, and as in many long-lived birds of prey, they tend to settle earlier than males [43, 45, 47, 48]. In accordance with Hypotheses 4 and 5, we thus assumed that females will prospect more extensively but will spatially concentrate their prospecting efforts quicker than males. We used a trajectory segmentation method to identify, for each individual, the areas prospected sequentially during their transit phase. By assessing the size, the duration of use, and the spacing between those areas, we could accurately detect spatio-temporal variations in prospecting pattern. Through this analysis, we could examine whether the variations in prospecting behaviours during the transit phase of the Golden eagle were in line with the five predictions aforementioned.

Methods

Study species and area

The Golden eagle is a raptor with Holarctic distribution, living in habitats ranging from arctic tundra to subtropical deserts. Individuals form monogamous pairs and build nests on cliffs or in tree canopies [44]. Pairs defend a territory whose size ranges between 20 and 200 km², and generally rear one, rarely two, fledglings each year [44]. The departure from the natal territory (emigration phase) of juvenile individuals usually occurs between autumn and spring of their first year of life [46, 49]. The transit phase can last between one and six years, and ends when the individual settles in a new territory (immigration phase) [45]. During this period, individuals prospect their environment in search of a vacant or already occupied territory [50]. Most of the time, individuals do not reproduce immediately after immigration, and the age of first reproduction generally ranges from four to seven years old [44]. After immigrating, the individuals are very faithful to their territory, and breeding dispersal is very rare [44].

In France, Golden eagles are sedentary and nest in mountain ranges of the Alps, Pre-Alps, Massif Central, Pyrenees and Corsica [51]. We focused on three study areas (hereafter referred to as “regions”; see map Appendix S1: Figure S1), the ‘Alps’ (central point 6.5°E 44.7°N, WGS84), the ‘Pre-Alps’ (central point 5.3°E 44.4 °N, WGS84) and the ‘Massif central’ (central point 3.8°E 44.3°N, WGS84). The Alps are characterized by an alpine climate, with snow cover that extends for up to six months at high altitudes. The Massif Central and the Pre-Alps have a continental climate, with frequent heatwaves in summer in the southern part [52].

Bird tagging and GPS filtering

Between 2016 and 2022, 128 juveniles were captured and tagged at nest between 48 and 58 days old (Appendix S1: TABLE S1). Tags were fitted using a Teflon harness, attached as back-pack (X-strap thoracic harness) [53]. Six models of solar-powered GPS tags from four manufacturers (E-obs, Ecotone, Microwave Telemetry, and Ornitela) were deployed (Appendix S1: TABLE S1). Including colour rings, the material weighted between 74 and 105 g, corresponding to 1.8–3% of an adult’s body weight, in accordance with the recommendation of Kays [54]. Feather samples were collected during tagging, in order to determine the sex of each individual, using the method of Fridolfsson and Ellegren [55].

The GPS transmitters were programmed to record 3D positions at intervals of 5–15 min during the day, with a higher frequency of 1-minute intervals in sunny weather when the battery was fully charged. The GPS was programmed to stop recording at night and start tracking at the sunrise time of the study area centroid. Data were transmitted daily via GSM network, except for the first Ecotone and Skua models, which used UHF transmissions. Data were stored on the www.movebank.org online database in three movebank studies (see Appendix S1: TABLE S1). All the data were processed and analysed using R ver.4.2.0 (www.r-project.org). Data were first extracted using the ‘move’ package [56], and were then cleaned to remove inaccurate locations following the methods from Gupte et al. (2022). We applied a spatial filter to remove all obvious erroneous locations (too far south i.e. Latitude < 40.995°; and too far west i.e. Longitude < -1.740°, coordinate system WGS84). We filtered on data quality attributes: duplicated timestamps were deleted, and we retained locations with a horizontal dilution of precision (Hdop) < 10 or a satellite number > 3 [57,58,59]. We removed unrealistic movements by filtering speed above 400 kph, which is a threshold slightly above the maximum speed recorded for Golden eagles [44, 57]. We resampled the whole dataset to 1-day intervals by keeping only the first location of each day, i.e. the place where individuals had stayed for the night. This provided an overview of the transit trajectory without the daily routine movements, which were not relevant for this study. For each individual, we determined the emigration date using the method #7 proposed by Weston et al. [49], i.e. the first day further than 9 km from the birthplace, with no return within 6 km (see “Date emigr” in Appendix S1: TABLE S1). All the locations before the emigration dates were filtered, so that only the transit trajectories were retained. Between 2016 and 2022, 106 of the 128 tagged individuals emigrated. We then assessed for each individual whether territorial settlement, marking the end of dispersal, occurred during the study period. We applied the methods of Whitfield et al. [60], based on an algorithm specifically designed for Golden eagles which determine whether individuals have settled territorially and, if so, establish the settlement date. The algorithm considered that an individual settled if for at least 30 consecutive nights, the nightly resting distance between each night and 10 nights earlier was less than 10 km, and the eagle remained within 15 km of the centre of its implicit territory at least 250 days after fledging. We identified 7 individuals which have settled territorially during the time interval studied, and for these individuals we removed the data after their settlement date. We then confirmed the territorial settlements, by routinely visualising the individual movements to ensure that they did not leave the territory nor undertake long distance extra-territorial movements. In addition, we conducted field verifications of new settlements thanks to a network of observers, who checked for the presence of reproductive behaviour of the tagged individuals such as nest building, mating or chick rearing. The tracking durations after the emigration dates depended on the death of individuals, on tag failure or loss, on the settlement dates for the individuals concerned, and on the data cutoff date (2023-09-15) for individuals still tracked (mean = 739 days, range = 42–2471 days, Appendix S1: TABLE S1, Appendix S1: Figure S2).

Segmentation and metrics computation

Prospecting involves a sequential visit of patches that could become potential breeding areas [24]. Describing how these visits are structured in space and time provides a picture of a species’ prospecting patterns. To prospect, immature Golden eagles move sequentially into several geographical areas (hereafter Activity Areas, ‘AAs’), corresponding to patches of prospection [50]. We used a path segmentation method (package segclust2d [61]) to split trajectories of each individual into a series of successive AAs. This method consists of a bivariate segmentation on the longitudinal and latitudinal coordinates that detects changes in the mean and variance of X(t) and Y(t), where X(t) and Y(t) correspond to the longitudinal and latitudinal time series (Fig. 1.B.1). A change in the mean of X(t) and Y(t) indicates a geographical shift of the AA locations, while a change in variance relates a shift in the AA sizes (Fig. 1.B; Patin et al., 2020). We used a Lavielle threshold of 0.2 for the segmentation, making the algorithm very sensitive to changes in the mean and variance of X(t) and Y(t). The minimum duration of the segments was set to the minimum value required to warrant a reliable estimation of the parameters (5 days in our case), in order to give a high level of flexibility to the segmentation, since we had no prior assumption on the duration of the Activity Areas [61]. This allowed detection of AA shifts with high spatial and temporal resolution [61, 62]. Since we kept only the first location of each day in our trajectories, an AA identified by segclust2d corresponded to a sequence of locations where an immature eagle spent its nights.

Fig. 1

Methodological framework to assess the variations in prospecting behaviours during the transit phase of 106 dispersing Golden eagles. (A) Trajectory of one individual during the transit phase of dispersal. Points represent the first location of each day, and lines connect two successive locations. Longitude and latitude are Lambert RGF93 coordinates. (B) Representation of the path segmentation from segclust2d. B.1) Longitude and latitude time series are split into Activity Areas (AAs). The colored lines and rectangles illustrate the different AAs determined by segclust2d. Rectangle widths are equal to the variance values, and the lines are the mean values. Segclust2d detect an AA shift when the mean or the variance of the time series vary significantly. B.2) Map of the AAs. Each color indicate one AA. (C) Illustration of the three prospecting metrics. AA-size is the mean distance between the AA centroid and each locations. AA-duration is the number of days the individual stays on the AA. AA-spacing is equal to the spacing between the two centroids of successive AA

We constructed a time variable representing the progress of the transit phase for all individuals. This ‘Time’ variable corresponded to the number of days elapsed since the first of June of the year of tagging. We limited our analysis period to the time when we had at least 15 males and 15 females simultaneously tracked, to ensure a total sample size of 30, allowing the central limit theorem to apply while maintaining a representative balance between sexes. This interval started on the tenth of December of the first tracking year and ended on the tenth of September of the fourth tracking year (see Appendix S1: Figure S3 for further detail).

In order to characterise how the AAs were spatially and temporally structured, three metrics were computed and attributed to each date of the ’Time’ variable (Fig. 1.C). The first metric was a proxy of the Activity Area sizes (hereafter named AA-size), and is equal to the mean Euclidian distance between all the locations of the AA and its centroid. The value of the AA-size was equivalent to the radius of a temporary home-range, and was expressed in kilometres. The second metric was a proxy of the spacing between successive Activity Areas (hereafter named AA-spacing). It was estimated as the distance between the centroid of an AA and the centroid of the next AA. The value of the AA-spacing was a distance expressed in kilometres. The third metric referred to the time spent on the Activity Areas (hereafter named AA-duration), which was calculated by counting the number of days spent at each AA (Fig. 1.C). It was expressed in number of days. This procedure was repeated for each individual (Fig. 1).

Statistical analyses

We fitted generalized additive mixed models [63] (hereafter GAMM) for each metric instead of LMM because we expected seasonal variations in the metrics. To investigate intra-annual variations in prospecting behaviour, we chose to use time as a continuous variable rather than a categorical seasonal factor, as this allowed finer detection of temporal patterns that may not have aligned with human-defined seasons. Models were fitted with the ‘mgcv’ package 1.9.0 [64], using thin plate regression splines as smoother [65]. A random effect was included to take into account the inter-individual variation in prospecting behaviours, allowing the model to fit distinct intercepts for each individual. Sex was included in interaction with the time variable within the smoother term. Sex was also added as fixed effect to test for mean differences in metrics between sexes. In this way, the GAMMs fitted mean metric values over time, and accounted for both intra- and inter-annual variations for each sex separately. In order to test whether regions affected prospecting patterns, the region of origin of the individuals was added as a fixed effect in the GAMMs to test for differences in means between regions for each metric. We also added a regional effect interacting with time to test whether age and season affected prospecting behaviours in the same way across the three regions. For choosing the adequate basic dimension (k) for GAMMs, we took the smallest value of k for which there was a small probability (p < 0.05) to have missed pattern left in the residuals, following the method described by Wood [66]. For each date of this studied interval (Appendix S1: Figure S3), we computed for each sex separately the population mean of each metrics in order to check the match between the mean values of the metrics and the fit of the models (Appendix S1: Figure S4).

The variations of the population mean of metrics were analysed as time-series to seek for seasonality signal. The autocorrelation functions for each metric time series were computed using the acf function from the R base, in order to test for seasonal oscillations [67]. These steps were performed for males and females separately, for the three metrics. For each metrics, we computed the difference in means between the female and the male for each date of the ‘Time’ variable. This provided an overview of the sexual differences in prospection pattern throughout the transit period. The 95% confidence intervals of the sexual differences were computed (± 1.96*Standard Error).

Results

Result of segmentation and metrics computation

We found an average number of 18.53 AAs ± SE 1.05 (range = 3–62) per individual, through the entire tracking period. AA-sizes measured on average 13.37 ± SE 0.33 km (range 5 m to 133.7 km). The mean AA-spacing was 40.12 ± SE 0.87 km (range 107 m − 339 km). The AA-duration lasted on average 38.15 ± SE 1.13 days (range 5 days to 429 days). The population mean of the three metrics varied strongly throughout the transit period (Appendix S1: Figure S4), and were well fitted by the GAMMs (see r ² in Table 1, and Appendix S1: Figure S4).

Table 1 Table of generalized additive mixed models modelling the effect of sex, regions, and time on prospecting metrics during the transit phase of dispersing Golden eagles. The three metrics are the size of the Activity Areas (AA-size), the spacing between successive Activity Areas (AA-spacing), and the duration of the Activity Areas (AA-duration). K is the number of basis functions to use for each smooth term

Effect of age on prospecting behaviours

The AA-spacing decreased significantly over the transit phase for both sexes (Fig. 2). Between the first and the last day of the studied interval, AA-spacing decreased from 52 ± 1 km to 28 ± 1 km for females, and from 38 ± 1 km to 28 ± 1 km for males (Fig. 2). AA-duration significantly increased throughout transit: from 40 ± 1 days to 169 ± 1 days for females, and from 55 ± 1 days to 105 ± 1 days for males (Fig. 2). Thus, as they aged, dispersing eagles prospected among areas closer from each other’s, and remained longer in those areas.

Fig. 2

Variations of prospecting metrics during the transit phase of dispersing Golden eagles. The three metrics characterise the Activity Areas (AAs) successively prospected by the individuals. The three metrics are the size of the Activity Areas (AA-size), the spacing between successive Activity Areas (AA-spacing), the duration of the Activity Areas (AA-duration). Mean ± 95% confidence intervals from the Generalized Additive Mixed Model are represented. The top x-axis represents the time variable corresponding to the number of days since the 01 June of the year of tagging. The bottom x-axis represents the corresponding date over the transit period

Seasonal variations in prospecting behaviours

There were strong oscillatory variations in the mean AA-size (Appendix S1: Figure S4), which were well fitted by the GAMM (Fig. 2, Appendix S1: Figure S4). The auto-correlogram showed a significant temporal autocorrelation at 365.25 days for males and females (Appendix S1: Figure S5), which indicates a statistically detectable seasonality signal for this metric. These oscillations revealed seasonal variations in the AA-size, with peaks in spring and autumn (Fig. 2). For females only, a positive autocorrelation with a lag of half a year was significant, showing the existence of two significant peaks in the mean AA-sizes every year, in autumn and in spring (Appendix S1: Figure S5). These two peaks were also apparent for males (Fig. 2), but not sharp enough to be detectable on the auto-correlogram (Appendix S1: Figure S5). The amplitude of the seasonal oscillations of AA-size decreased throughout the transit period, especially for females, for whom the spring peaks were significantly smaller in the second and third years of dispersal compared to the first year (Fig. 2). In males, the prospecting trough in summer were marked in the first year, and were attenuated in the second and third year of the transit period. This result revealed an interaction between the effect of season and age on AA-size, with seasonal variations in AA-size less pronounced as individual aged.

No seasonal variations were noticeable for AA-spacing and AA-duration (Fig. 2). The auto-correlogram revealed no positive temporal autocorrelation every year, confirming the lack of a seasonal signal for those metrics (Appendix S1: Figure S5). However, AA-spacing reached a peak during the first spring, which was not well fitted by the GAMM (Fig. 2, Appendix S1: Figure S4).

Sex differences in prospecting behaviours

AA-size and AA-spacing were greater in females than in males during the transit phase (AA-size: 18.7 ± 1.4 km for females and 15.4 ± 3.4 km for males; AA-spacing: 45.9 ± 3.6 km for females and 34.6 ± 4.7 km for males). The GAMMs showed that these differences were significant for AA-spacing but not for AA-size (Table 1).

For all three metrics, sexual differences were not constant throughout the transit phase (Fig. 3). For AA-spacing and AA-size, sexual differences were particularly pronounced and significant during the first spring of the first year, with females prospecting on larger and more spaced areas than males. Subsequently, these differences reduced and became no longer significant (Fig. 3). In females, the AA-spacing decreased by 24 km in 3 years of tracking, compared with 10 km in 3 years in males (Fig. 2), showing that females spatially restricted their prospecting faster than males. At the early stages of transit, AA-duration were slightly shorter in females than in males, but by the third year of dispersal, AA-duration became significantly longer in females, which show that females were more spatially restricted at the end of the tracking period (Figs. 2 and 3).

Fig. 3

Variations in sexual differences of the prospecting metrics during the transit phase of dispersing Golden eagles. Sexual difference (female – male) in the size of the Activity Areas (AA-size), the spacing between successive Activity Areas (AA-spacing), duration of the Activity Areas (AA-duration). Difference values (black line) ± confidence interval 95% (in grey shading) are represented. In order to smooth out the variations of sexual differences, we represented the moving average of the differences over 10 days (Black line)

Regional differences in prospecting behaviours

The regional fixed effects were not significant for any of the metrics, showing no significant difference in mean between regions for the three metrics (Table 1). The Region-Time interaction splines significantly improved the model fits, revealing an interregional variability in prospecting behaviours. However, prospecting patterns over the transit phase were similar for all three regions, as shown in Appendix S1: Figure S6, demonstrating that the effect of age and season was consistent across regions.

Discussion

Effect of age on prospecting behaviours

We observed significant variations in prospecting behaviours throughout the transit phase. As individuals aged, they prospected in increasingly closer geographical areas and spent more time in these areas, demonstrating a spatial narrowing of prospection movements as individuals get closer to the immigration phase. These results are consistent with the few studies that have examined the effect of age on prospecting during the pre-breeding period [68,69,70]. For instance, Wolfson et al. [70], showed that older non-breeding Sandhill Cranes (Antigone canadensis) tend to revisit the same areas more often and stay longer than younger individuals. This supports Hypothesis 1, which suggests that in long-lived species, the benefit of large-scale prospecting diminishes throughout the pre-breeding period because individuals already identified areas that could represent potential breeding sites of good quality [69, 71]. Focusing prospecting efforts on a few areas may allow older individuals to familiarise themselves with potential future breeding sites and gather precise information on a small spatial scale. In contrast, younger individuals would collect information over a larger spatial scale through long-distance prospecting movements [8, 24]. Our finding suggests that during the transit phase, individual adjust their movement strategies from large-scale random prospecting to more memory-guided strategies, such as systematic searches, to maximize encounters with high-quality areas while avoiding previously explored sites of lower quality [72]. In territorial species, such as Golden eagles, this would involve identifying existing nesting sites or favourable areas for building new ones, finding food-rich spots, or queuing for good quality territory (sensu Heg et al., [73]). The spatial prospecting concentration might also concern areas that do not necessarily represent future breeding sites. Such areas may correspond to strategic habitats where food resources are abundant and neighbouring potential breeding sites. With age, individuals would acquire a better knowledge of these areas, and using them while waiting for an opportunity to settle in a breeding site, would improve their competitiveness. To further understand the role of age-related prospecting concentration, recursion analyses could be used to determine whether the areas prospected intensively at the end of the transit phase match the settlement areas of individuals [74]. In addition, emerging movement analysis approaches based on network theory could offer a more detailed characterization of changes in prospecting behaviours, particularly by comparing observed movement patterns with theoretical movement models such as Brownian motion or Lévy walks [75].

Our analysis revealed that immature Golden eagles gradually concentrated their prospecting movements rather than undergoing an abrupt transition. During the transit phase, we observed no clear-cut shift between an exploratory phase and a subsequent phase of more restricted movements during the transit phase. In contrast, previous studies on the transit phase have described a clear transition from exploratory to routine movements, leading to a spatial concentration of prospecting [11, 20]. For example, in the Eagle owl (Bubo bubo), the transit phase is divided into two stages: an initial exploratory phase characterized by rapid, straight-line movements over great distances, followed by a second phase with more tortuous movements confined to non-breeding settlement areas [11]. It has been suggested that these two similar stages could occur in Golden eagles and long-lived raptors in general [44, 76]. Our findings thus challenge this view by highlighting a more gradual spatial concentration of prospecting throughout the transit phase.

Effect of season on prospecting behaviours

A notable seasonal variation in AA-size was observed in both sexes, and the areas prospected were larger during spring and autumn. These findings corroborate previous studies on the same species, indicating that monthly home ranges and daily distances are greater during these seasons, suggesting that this pattern is a general strategy of the species rather than specific to our population [31, 33]. A first explanation for this pattern could be weather-related. The weather is most favourable for the formation of thermal updrafts between March and October, in contrast to winter when the solar radiations and temperatures are low [37]. Such unfavourable flight conditions may explain the significant decrease of AA size in winter (Fig. 2). However, during summer, strong thermal updrafts should allow immature birds to cover greater distances, as observed in Griffon vultures (Gyps fulvus) within our study area [77], which contrasts with our findings. This may suggest that thermal updraft is not the only weather factor driving prospecting behaviour. Wind also considerably increases the flight capacity of birds, especially for long-distance movements [78, 79]. In our study area, the prevailing winds are expected to be stronger in autumn and spring, which could explain the prospecting peaks during these periods. To further investigate which weather factor primarily drives the prospecting behaviours, it would be relevant in a future study to correlate daily weather data, such as wind or solar radiation with prospecting movements. This would allow to better assess what season is optimal for prospecting flights, and may provide stronger support for the Hypothesis 2, which postulates that dispersers synchronise long-distance prospecting movements with seasons offering optimal flight conditions [36, 80]. Another explanation discussed in the literature is that seasonal variations in movements during the transit phase could be attributed to fluctuations in food abundance [81,82,83]. The benefit of changing of Activity Area could depend on the food supply; it would be advantageous to stay in an area with abundant food, and advantageous to leave the area when food supplies decrease [81]. In the White-backed vulture (Gyps africanus), it has been shown that the home ranges of dispersing immatures are smaller during the winter dry season when ungulate mortality is highest, i.e. when food is most available [83]. Hence, a greater food abundance in summer in our study areas may enables immature eagles to stay and prospect in smaller areas during this period, explaining the summer trough in AA-sizes. The lack of data about food supply variations in our study areas makes it difficult to determine to what extent food resource drive the prospection behaviours of the dispersing eagles. However, it would be interesting for future research to investigate whether the Activity Areas overlap between individuals, in which case we could assume that these areas are used for foraging. Similarly, a habitat selection analysis would inform about the nature of the Activity Areas prospected: open environments favourable to hunting or cliffs favourable to nesting.

The seasonal variations in prospecting spatial range that we observed could be also linked to neighbouring adult breeding birds. In some territorial passerines, juveniles prospect on a larger spatial range during the breeding season to visit more adult territories [84, 85]. In line with this, the AA-size peaks that we observed between March and June would suggest that young eagles prospected in larger areas to visit a greater number of territories during this period, coinciding with the laying, hatching and chick-rearing season of the Golden eagle in our study areas [46]. This would be consistent with Hypothesis 2, stating that the optimal timing for prospecting is when information on the quality of the territories is the most available, i.e. during the chick-rearing period [32]. However, in the Spanish imperial eagle, it has been demonstrated that juveniles engage in fewer incursions into adult territories during the breeding season, likely because breeding adults are more aggressive towards prospectors during these periods [40]. The peaks of AA-size in spring could then also be explained by the exclusion of immatures from breeding territories that are then constrained to prospect at the periphery of territories and in more distant vacant areas during the breeding season [40, 82]. To determine whether the spring peaks in prospecting range result from a constraint of exclusion by breeding adults, or by a decision by immatures to prospect when habitat quality information is available, future research should further investigate how dispersers visit territories and nests of breeding adults across different seasons.

Effect of sex on prospecting behaviours

The AA-size and AA-spacing were in average greater in females than in males, showing that females prospected over larger spatial range than males. In the Golden eagle, Murphy et al. [86] showed that females disperse twice further than males. Our results are thus consistent with the Hypothesis 4, whereby females dispersing further would benefit from prospecting over a larger geographical area than males. In contrast, males incur higher dispersal costs as they must defend resources and territories to attract females, and thus prospecting in more familiar areas would reduce their dispersal costs [1, 38]. Therefore, female-bias in dispersal distances observed during the immigration phase of birds is preceded by a female-bias in prospecting during the transit phase, highlighting the link between these two phases of dispersal [9].

Interaction between age, season and sex

We found an interaction between the effect of age and season on prospecting behaviour. Seasonal variations in AA-sizes were less pronounced from the second year of the transit phase onwards, especially in females. This is consistent with Hypothesis 3, which suggests that as long-distance movements become less frequent in the later stages of transit, eagles benefit less from synchronizing their movements with favourable seasons. This may be because when individuals fly over short distances, the energetic benefits of timing movements with favourable conditions, like thermal updrafts, are reduced. The intensity of seasonal variations in prospecting differed between males and females, likely reflecting differences in their tendency to perform long-distance movements. Because synchronizing long-distance prospecting with optimal flight periods helps reduce the high energetic costs of transit, females, which prospected over larger spatial ranges, should exhibit stronger seasonal variation in their movements than males, which prospected smaller areas.

The sexual differences in prospecting behaviour varied according to the age of individuals, reflecting a strong interaction between the effect of sex and age on prospecting behaviour. During the first spring of the transit phase, females prospected on a larger spatial range than males, but from the second year onwards, the sexual differences in AA-size and AA-spacing were not anymore significant. Furthermore, the AA-duration were greater in females from the third year onwards, showing that females prospected for longer time in each area and concentrated their prospecting more than males at the end of transit phase. This is consistent with the result of a study in Scotland where female Golden eagles settle in average 60 days earlier than males, even though these differences were not significant (895 ± 110 days for females and 957 ± 111 days for males; 80). Those evidence support the Hypothesis 5, stating that in species whose females tend to immigrate and reproduce earlier than males, females should also concentrate their prospecting earlier than males [42]. Furthermore, the interaction between age and sex on prospecting behaviour likely explains the previous conflicting results on sexual differences in prospecting, and should be addressed in future studies examining the effect of sex on prospecting behaviour.

Conclusions

Our study revealed a strong influence of age on prospecting behaviours during the transit phase of long-lived species. Individuals showed a progressive spatial concentration of prospecting, suggesting that they gradually refined their prospecting strategy based on information gathered in the early stages of the transit phase. Prospecting behaviours also varied strongly with sex, with females prospecting on a larger spatial range than males in line with the female-biased dispersal widely observed in birds. The effects of sex and age observed in our study support the theory of informed dispersal, as they suggest that individuals adjust their behaviour at each stage of dispersal based on information previously acquired. As for many species, seasonal variations in prospecting were detected, which may result from the constraints and opportunities imposed by adult reproductions and seasonal variations in flight conditions. More notably, we detected significant interactions between these factors: the effect of sex and season depended strongly on the age of individuals. Individuals seem to adjust their prospecting behaviour in response to interacting intrinsic and extrinsic factors, in order to reduce prospecting movement costs while maximising the information gathered to inform their immigration decision. Complementary work to this study would involve establishing a link between the prospecting strategies observed during the transit phase and the immigration phase of individuals, although this would require long-term data. Such an approach would provide insights into how early dispersal decisions influence settlement decisions and future reproductive success. Furthermore, it would be particularly relevant to explore the relationship between the movement patterns described here and social information-gathering behaviours during transit. Specifically, investigating nest-visiting behaviours throughout the transit phase could help determine whether these movements primarily serve a social information-gathering function or rather reflect personal information collecting based on the physical characteristics of the landscape.