Case Report: Independence day, comparison of methods to estimate the onset of dispersal in large territorial birds

Abstract

Background

In birds, the onset of dispersal is the transition point between the end of the post-fledging dependence period (PFDP) and the beginning of natal dispersal. Identifying this key moment in an individual’s life cycle is crucial for understanding the ecology and demography of a species, but its determination remains a challenge. Some methodologies used to estimate it usually yield biased outcomes, while others are based on adult ranging areas and require previous knowledge on the species’ movement patterns, precluding their use in little-known species. Additionally, the proposed methods have never been compared with data on family breakup timing. Thus, creating a standardized methodology, not based on species-specific information and with a reliable reference, to estimate the onset of dispersal for its use on a wider array of birds is key to optimizing research on rare and/or elusive species.

Methods

We used unique GPS data from parent Chaco eagles (Buteogallus coronatus), a large, territorial, and endangered species, with their corresponding fledglings during the PFDP as a reference against which to evaluate (1) the existing methodologies in the literature; (2) modified methods based on previous ones; and (3) new methods related to changes and overlap values in the ranging areas of young individuals. To quantitatively assess which method best matched the reference values, we calculated the mean deviation and bias on an individual basis, also accounting for the interindividual coefficient of variation.

Results

Most previously published methods assign the onset of dispersal prematurely, probably because they are sensitive to exploratory or foraging movements that young individuals perform prior to dispersal. The modified methods, although more accurate, still provided biased estimates. However, the new methods we propose provided estimates of the onset of dispersal with increased accuracy and low interindividual variability.

Conclusions

The new methods proposed (1) provide the most accurate estimates of the onset of dispersal; (2) are standardized, so they can be used for a wide range of bird species; (3) do not require previous knowledge of adult movement patterns; and (4) are very useful for most bird research studies that tag fledglings with GPS transmitters.

Introduction

Animal dispersal and the movements associated with it, constitute a key process in mobile species that determines their reproductive fitness, leads to gene flow, and thus has significant effects on population dynamics. Consequently, understanding the patterns and mechanisms of dispersal is crucial in addressing issues related to conservation biology, such as the effects of habitat fragmentation and climate change, the spread of pests and diseases, and biological invasions [1,2,3].

In birds, the onset of dispersal (also called emigration or independence) is a transitional point between two fundamental stages of their life cycle [4, 5]: the post-fledging dependence period (hereafter PFDP), or the time elapsed since the fledgling abandons its nest until its natal dispersal or definitive migration [5, 6], and the natal dispersal period, which starts when the young individual moves away from any potential parental care and extends until its first establishment as a reproductive individual [7, 8]. During these periods, inexperienced juveniles may face unknown challenges and risks, which can result in increased mortality [9, 10]. Therefore, the development and duration of both stages are key to determine juvenile survival, the recruitment of breeding individuals, and future reproduction [6, 11, 12]. Additionally, in territorial bird species that maintain relatively stable ranging areas, the correct determination of the extent of these periods aids in correctly identifying the areas used by both adults and fledglings, constituting a significant advantage when planning effective conservation strategies such as protecting these territories [13]. Nevertheless, there is an overall lack of information on the timing, causes, and processes of post-fledging dependence and natal dispersal periods in birds [7].

In large, territorial birds, such as most raptors, studying this topic is even more challenging because their high mobility makes them less predictable, and difficult to follow in the field [12], making them a generally understudied group [14, 15]. This diverse group of birds with a widely recognized role in ecosystems as top predators and scavengers and as flagship species [16], is currently suffering severe declines in global populations [17], a fact that calls for urgent action in studying these topics to develop effective conservation policies [15, 18, 19]. However, determining the onset of dispersal in raptors has traditionally been a matter of discussion [5, 20, 21] because individual variation in dispersal strategies and context-dependent situations make it difficult to decompose movement patterns [22]. The development of GPS satellite devices that allow for the remote tracking of individuals has resulted in a number of approaches for estimating this transitional point [5, 21, 23, 24]. Since these devices are often used in fledglings, mathematical estimations are based on the movements that young birds carry out during the PFDP and try to detect the moment when they change the pattern of movement [5, 21, 23, 24]. However, biased estimations have arisen from GPS-based data, mainly due to the accuracy of the data [23], the type of device [24], and the procedures used to estimate the emigration point, which may confuse excursive behaviours of juveniles with true dispersal [5, 21]. The latter is the most important to address because it is key to developing a standardized methodology to account for the duration of the PFDP and the onset of dispersal. On the one hand, distance-threshold (e.g., distance to nest) and displacement rate-based (e.g., coefficients of variation of distances to nest between consecutive locations) methods yield several inconsistencies and are especially sensitive to predispersal exploratory movements [23, 24]. On the other hand, approaches based on adult ranging areas (e.g., the proportion of locations of the fledgling inside the parental home range, [5]) require previous knowledge of the species’ range and movement patterns, precluding their use in little-known or rare species for which such information is lacking, which are those of conservation concern [14, 17]. Additionally, these studies compare their methods against the best reference method that they consider (e.g., visual examinations of the movement data of the fledgling) [5], but none assess them with real data on family breakup timing (i.e. the moment when the fledgling and the adults abruptly separate in space, which usually coincides with the achievement of independence [5, 6]) by comparing the movements of both adult parents and fledglings. Thus, creating a standardized methodology that is not based on species-specific previous information and provides a reliable reference for estimating the onset of dispersal for its use on a wider array of birds is key to optimizing research on little studied and/or rare species.

The aim of this paper is to test (1) existing methods to estimate the onset of dispersal; (2) modified methods related to previous ones; and (3) new methods based on juvenile ranging areas using unique GPS data from parent Chaco eagles (Buteogallus coronatus) with their corresponding fledglings during the PFDP as a reference against which to evaluate the different methodologies that have been used in the literature and the new ones created in this study.

Methods

Study species

The endangered Chaco eagle is a large raptor of the Neotropical region [25, 26], and a clear example of a little-known species with no data on adult ranging areas or movement patterns [26], being unworkable to use some of the previously developed techniques (i.e. approaches based on adult ranging areas) to estimate the onset of dispersal. Additionally, the Chaco eagle is territorial and lays only one egg per reproductive attempt [26], so parental care after fledgling will be directed towards one individual, with no variability in the timing of the onset of dispersal due to sibling differences characteristic of larger broods [23, 27].

Individual tagging

The study was carried out in La Pampa Province, central Argentina (approx. 37°S, 66°W), along the temperate-arid ecoregions of the Espinal and Monte Desert and the ecotone landscapes between them [28]. We used unique, high temporal and spatial resolution GPS data on the movements of four mother Chaco eagles with their corresponding fledglings during their first year of life. All individuals were tagged with solar-powered GPS satellite transmitters from Microwave Telemetry, Inc. (Columbia, MD, USA). These transmitters were programmed to record a location per hour during the daylight period (i.e. approximately from 8 AM to 9 PM), allowing for continuous monitoring of their locations and movements through the ARGOS satellite system. One mother was captured and tagged at the breeding site using a bow-net trap [29], and the remaining three were female individuals tagged as fledglings that survived until they were recruited as breeding adults [26]. When all these females successfully bred, we tagged their corresponding nestlings a few days before they fledged, so that we had both the mother and its fledgling tagged and providing information on their hourly movements simultaneously.

Reference benchmark for the onset of dispersal

We first determined the individual date of fledging (i.e. day zero of the PFDP) when the distance of the nestling to its corresponding nest exceeded 50 m [30]. We then randomly chose one location (for both adult and fledgling) per day, from the central hours of the day (i.e. 1000 H–1400 H), as the location to measure the daily distance between the mother and the fledgling. This was done to avoid very early (or late) daily locations when the fledgling (and the adult) would presumably roost together near the nest or nearby areas [31], thus underestimating their distance apart. These locations were used to examine the distances of the mother from the fledgling, to visually determine the day when their locations abruptly separated and continued that way during a given time (i.e. the distances between mother and fledgling suddenly increased in a very marked way, and this was not a consequence of excursive behaviour, since the distances between them did not return to previous, low, values, see Fig. 1 for an example of this trend). In all cases, we checked that this sharp increase in the distance between the fledgling and the mother was not a result from increased exploration by females during the final stage of the PFDP. The number of days elapsed since the individual date of fledgling to this moment was taken as the reference length of the PFDP.

Fig. 1

Daily distance (in km) between the four fledgling Chaco eagles (Buteogallus coronatus) and their corresponding mothers (sorted from A to D, and corresponding to the same letters in Fig. 2), all GPS-tagged, during the first months after abandoning the nest (time in days). The vertical lines indicate the estimations from three different methods. Red = Method 22, displacement rate-based (highest C.V. of the distances to nest of the bird of 3 days, based on a previous method—Method 18—Table 1) [28, 29]; grey = reference benchmark, visual (abrupt change in the distances between fledgling and mother, used for comparison); blue = Method 29, ranging-area based, and created in our study (first record of week’s/last week’s ranging areas of the bird ≥ 20, best method according to data, Table 1). For a better visualization of the data, the first 100 days after fledgling have been omitted in the X-axis

Comparison of estimation methods

Starting on day zero for each young eagle (n = 4 fledglings), we collected the data (using all hourly records) of one entire calendar year for each of those fledglings for the following analyses. First, we compared the existing methods in the literature. These methods [23, 24] include distance-threshold and displacement rate-based methods (Table 1). We did not use techniques based on adult ranging areas (i.e. methods 7 and 12, [5]) because they are species-specific, require previous knowledge of adult home ranges (not available for most raptor species) [14], and are highly variable due to habitat characteristics and food availability, among others [32].

Table 1 Description and comparison of the different methods used to estimate the onset of the post-fledgling dependence period in Chaco eagle fledglings tagged with GPS-satellite transmitters in central Argentina

We then modified some of these existing methods by adding more conditions and by standardizing their use. First, since some of these methods are highly sensitive to foraging and predispersal excursions, we required the movement patterns to be maintained for a longer period of time (i.e. the day with the highest coefficient of variation of the distances to the nest, Method 21, instead of the three consecutive records with the highest coefficient of variation of the distances to the nest, Method 18, [23]). Second, since some of them also rely on absolute distances and thus cannot be applied on a regular basis to raptor species of different sizes, development times and/or life-history strategies, we standardized their use (i.e. the highest value of the quotient between the distance to the nest of a given location and the distance to the nest of the previous location, Method 7, instead of the first location over 10 km from the nest, Method 1, [30]).

Last, we built a set of new methods based on fledgling range areas. We calculated weekly ranging areas (95% estimator) of fledglings [33] using autocorrelated kernel density estimations (AKDE) [34] of the Continuous-Time Movement Modeling (ctmm) [35], with the ctmmweb application [36] in R software [37]. This package models the autocorrelation structure of tracking data with the Ornstein–Uhlenbeck Foraging (OUF) or Ornstein–Uhlenbeck (OU) models, and therefore performs more accurate predictions of the home ranges than traditional kernel density estimation (KDE) methods [35, 36], which usually underestimate the ranging areas [34]. With this information, we compared the relative change in the weekly ranging areas of fledglings, and visually determined the onset of dispersal based on abrupt visual changes in the ranging areas (Method 28) or mathematically determined that an individual had started dispersal if there was a significant increase (> = 20 times) in the ranging area from one week to another (Method 29, Fig. 2). In addition, we used the function “overlap” in the ctmm package to analyse the proportion of the area used from one week that matched the area used from the previous one, and determined the onset of dispersal based on the proportions of range overlap during consecutive weeks using different threshold values (Methods 30 and 31, Table 1). Since all these new methods do not yield a day but rather a week, the central (middle) day of the selected week was chosen as the day to account for the end of the PFDP.

Fig. 2

Ratio of the weekly ranging areas (i.e. Method 29, first record of week’s/last week’s ranging areas—RA—of the bird ≥ 20, Table 1) of the four GPS-tagged Chaco eagle (Buteogallus coronatus) fledglings (sorted from A to D, and corresponding to the same letters in Fig. 1) over time (in weeks). The grey, dotted, horizontal line marks the threshold ratio used to determine the onset of dispersal in this method (i.e. ratio ≥ 20). The weeks fulfilling the conditions established in Method 29 are labelled

For method comparison, we calculated the mean deviation from the expected point of emigration (mean deviation: $\sqrt{\Sigma {\left(Residual\right)}^{2}/n}$), and the mean bias $\Sigma Residual$/n (negative values underestimate and positive values overestimate) from the mother–fledgling breakup timing reference data, and the interindividual coefficient of variation C.V [7] (Table 1). These values were used to determine the method that best represented the onset of natal dispersal. All analyses were performed using R [37].

Results

The duration of the PFDP of the four Chaco eagle fledglings evaluated for this study, using the visual assessment of the distances from the four mothers to their corresponding fledglings (reference benchmark), was 245.25 ± 10.56 days (Fig. 1). Using these data on the onset of dispersal for this species, we evaluated a total of 31 methods: 17 distance-threshold methods, 10 displacement rate-based methods and four new methods based on ranging areas of fledglings (Table 1).

All but one of the distance-threshold methods (16 out of 17) underestimated the duration of the PFDP, with a mean deviation from the reference data of 73.90 ± 50.92 days, and an interindividual coefficient of variation of 0.37. From these methods, the ones that overall adjusted best the reference data were Method 14 (i.e. highest value of the daily mean distance to nest/last day’s mean distance to nest), with a negative deviation of 14.75 ± 15.65 days, and Method 17 (i.e. highest value of daily mean distance to nest/ last 5 days’ mean distance to nest), with a positive deviation of 16.00 ± 24.04 days, both standardized and modified from previously published ones (Table 1).

Seven out of 10 of the displacement rate-based methods underestimated the duration of the PFDP (see example of Method 22 in Fig. 1), with a mean deviation from the reference data of 19.85 ± 22.06 days (n = 10), and an interindividual coefficient of variation of 0.15. Among these displacement rate-based methods, the one that adjusted best to the reference data was Method 22 (i.e. the highest coefficient of variation of the distances to the nest of 3 consecutive days), with a negative deviation of 11.00 ± 6.40 days. From the ones that yielded overestimations of the start of dispersal, the ones that adjusted better to the reference data were Method 26 (i.e. week with the maximum distances traversed), with a positive deviation of 17.25 ± 9.88 days, and Method 27 (i.e. week with the maximum rate of change of distances traversed over the previous week), with a positive deviation of 17.25 ± 18.06 days, both of which were modified from previously published ones, and one (Method 27) standardized (Table 1).

Finally, all the new methods based on ranging areas of fledglings yielded an overestimation of the duration of the PFDP, with a positive deviation from the reference data of 7.69 ± 4.49 days and an interindividual coefficient of variation of 0.04. The method that adjusted best to the reference data and the best method of all 31 evaluated, was Method 29 (i.e. first time with an increase in the ranging area of more than 20 times from one week to another), with a positive deviation of 3.25 ± 2.99 days; this method is standardized (Table 1, Figs. 1, 2).

Discussion

This is the first study comparing mathematical and visual methods to estimate the onset of dispersal in large, territorial birds with complete GPS data on the timing of family (i.e. mother and fledgling) breakup. We show that the methods that account for the relative changes in the weekly ranging areas of fledglings are the ones that adjust the best to the reference data. Since these methods are based on ratios (e.g., changes in weekly ranging areas or weekly overlap values), they are not scale- or species-dependent; thus, they are standardized for their use not only in raptors, but also in other large territorial birds. Additionally, they do not require previous knowledge of the species’ movement patterns and ranging areas, which is especially useful for most bird research studies that tag only fledglings (and not adults) with GPS transmitters.

The start of natal dispersal in our population of Chaco eagles took place over a relatively limited period of time, contrary to other studies with species such as the Golden Eagle Aquila chrysaetos [5]. Indeed, the duration of the dependence period in raptors tends to vary considerably among individuals [15, 27, 30]. The mean duration of the PFDP of the four Chaco eagle juveniles analysed (245 days) falls within the observed extension of this period in other resident, large eagles such as the Javan Hawk-eagle Nisaetus bartelsi and the Martial eagle Polemaetus bellicosus [15].

Distance-threshold and displacement rate-based methods assigned the onset of dispersal prematurely and showed high interindividual variability. The available literature has demonstrated that these methods are sensitive to exploratory, pre-migratory and/or foraging movements [5, 24]. These predispersal excursions are a way that young individuals of large and territorial bird species have to prospect the surrounding environment to become ready for dispersal and/or to decide whether to disperse or not [5, 12, 20, 21, 24]. Yet, since young individuals still depend on their parents’ territory for foraging, and/or roosting, this cannot be determined as the onset of dispersal per se. Additionally, and contrary to previous studies with other raptors [27], we show that juvenile Chaco eagles are accompanied, at least by the mother, during their exploratory flights, because the distance between them does not increase dramatically until the onset of dispersal (Fig. 1).

Concerning distance-threshold methods, it is likely that, before the true onset of dispersal, young eagles travel further than the tolerance levels of the threshold values assigned by the models. For instance, although juveniles of some species (for instance, Spanish Imperial eagle Aquila adalberti) remain within a few tens of kilometres of the nest during most pre-emancipation exploratory movements [38], some of them may reach hundreds of kilometres [27]. The election of the threshold value of either 10 or 20 km from the nest remains insufficient in these cases. Also, basing the estimates on distances to nest entails a risk of confusion, because fledglings may eventually abandon the nest area for other reason (e.g., the nest falls or the adults’ breeding home ranges do not match the non-breeding areas), but still stay with their parents. If this is the case and, given that parent raptors are not usually GPS-tagged, the real timing of dispersal may be assigned too soon if the fledgling goes beyond a given distance from the nest.

Regarding displacement rate-based methods, we obtained more accurate estimates, but they still underestimated the extent of the PFDP by some weeks. These methods are created to detect any abrupt change in the movement pattern from the nest, but this highly depends on the previous location points and is not scaled to detect real, significant, dispersing movements. For instance, if a juvenile stays at a short distance from the nest (metres) for several days and, suddenly, goes beyond a couple of kilometres, the method will see it as a significant change (i.e. high coefficient of variation), but in the end, this will not be biologically meaningful to consider this moment as the onset of dispersal.

Besides, the modified methods improved the estimations of its existing counterparts, as we required them to follow a given trend for a longer time or because we standardized the values of the methods to make them more realistic and applicable to other species. For example, instead of choosing the first day with a mean distance to the nest of more than 10 km (Method 10), which is already more demanding than its counterpart (Method 1: first record over 10 km from the nest, [20]), we generated a new, standardized one that chooses the day with the highest value of the quotient between the mean distance of that day divided by the mean distance of the previous day (Method 14). This modified method was the best among the distance-threshold methods and improved the estimations of the onset of dispersal. However, these new methods also yielded inexact estimations, probably because they still confuse exploratory movements with true dispersal.

Finally, the methods that we proposed provided similar estimates on the onset of dispersal with low interindividual variability and only showed a slight delay in the estimates, probably due to the fact that the ranging areas grow from the moment the juvenile disperses, and because the method extends its range to 7 days, it takes some time to capture this change. The success in the estimation ability of these methods can rely on several reasons: (1) since they are weekly estimates based on several location points, the incidence of location points related to exploratory excursions may be smoothed, thus minimizing the effect of these “outliers”; (2) they are not scale-dependent, meaning that we are measuring the relative changes in ranging areas or in the overlap values, so they can be generalized to other bird species of different range area sizes, life-histories, and development and growth patterns; and (3) they can account for unexpected range movements, such as the aforementioned case of the abandonment of the nest area after the breeding season. The creation of these methods is based on the fact that, in some raptors (and in other large birds), the home range of juveniles continuously expands throughout the dependence period, accelerating in the second half of this stage [8, 27, 38], probably due to the gradual acquisition of a better flight capability that allows them to soar and reach greater distances with less effort [38].

Conclusions

Recent technological advances allow us to track birds with unprecedented temporal and spatial resolution to study the spatial patterns of their movements and behaviour [3, 39]. The determination of the onset of dispersal is of particular interest because dispersal constitutes a key transitional point during early development [40], when birds are more vulnerable and prone to mortality [6, 7, 10]. In this way, the standardization of a methodology that allows us to estimate, with accuracy, the onset of dispersal with GPS data is vital when conducting research on the conservation and movement ecology of any species. Our study was the first to assess the precision of different methodologies to estimate the onset of dispersal using a real reference benchmark based on parent–offspring data, thus producing more reliable results. The new methods proposed in this study, based on fledglings’ ranging areas, (1) provide accurate estimates of the onset of dispersal; (2) are standardized, so they can be used for other species; (3) do not require previous knowledge of adult home ranges; and (4) are very useful for most bird research studies that tag fledglings (and not adults) with GPS transmitters. New studies should take into account the outcomes of this study and use the new methods proposed to obtain proper estimates of the timing of dispersal, a key but poorly known moment of the life cycle of birds [15].