Research articlesEffect of kelp gull harassment on southern right whale calf survival: a long-term capture–recapture analysis
Abstract
Kelp gulls (Larus dominicanus) commonly feed on the skin and blubber of surfacing southern right whales (SRW, Eubalaena australis) in the near shore waters of Península Valdés (PV), Argentina. Mothers and especially calves respond to gull attacks by changing their swimming speeds, resting postures and overall behaviour. Gull-inflicted wounds per calf have increased markedly since the mid-1990s. Unusually high mortality of young calves occurred locally after 2003, and increasing evidence points to gull harassment as a factor contributing to the excess deaths. After leaving PV, calves undertake a long migration with their mothers to summer feeding areas; their health during this strenuous exertion is likely to affect their probabilities of first-year survival. To explore the effects of gull-inflicted wounds on calf survival, we analysed 44 capture–recapture observations between 1974 and 2017, for 597 whales photo-identified in their years of birth between 1974 and 2011. We found a marked decrease in first-year survival associated with an increase in wound severity over time. Our analysis supports recent studies indicating that gull harassment at PV may impact SRW population dynamics.
1. Background
Southwest Atlantic southern right whales (SRWs, Eubalaena australis) migrate every winter to raise their calves along the coasts of Argentina, Brazil and Uruguay [1–5]. The breeding population that gathers at Península Valdés (PV), Argentina, has been studied closely since 1971 [6]. At this site, kelp gulls (Larus dominicanus) feed on the skin and blubber of SRWs as they surface, creating wounds of various sizes (figure 1a) and primarily attacking mother–calf pairs, which interrupts lactation and affects the whales' behaviour [8]. This harassment was first reported at Golfo San José (figure 1b) in the 1970s [9] and described as a parasitic interaction in the 1980s [10]. By the 1990s, it had spread to the adjacent Golfo Nuevo (figure 1b) where it rapidly increased during the 2000s [11,12]. The percentage of mothers and calves with lesions caused by gulls increased from 2% in the 1970s to 99% in the 2000s. Initially, calves were rarely attacked by gulls but, since the mid-1990s, calves have become the main targets of attacks and their average wound severity has increased [7].
Figure 1. (a) Images 1 to 3 show the sequence of a gull attack: 1, gull landing on the whale's back, 2, skin gouging, and 3, feeding on the whale's skin and/or blubber. Image 4 shows an open gull-inflicted lesion as a result of several attacks. (b) Map of the study area: Península Valdés, Argentina. (c) Lesion sizes on the back of SRW calves: extra-small (XS), small (S), medium (M), large (L), extra-large (XL), double XL (XXL). The lesion index used in the current study is represented by the equivalent number of XS lesions provided by [7] and represents the area of the whale's back affected by gull lesions. Photos by Macarena Agrelo (a1), Rodrigo A. Martínez Calatalán (a2–a4) and Fredrik Christiansen (c).
At PV, whales spend a significant portion (at least 24%) of daylight hours fleeing from gull-induced disturbance [11], which has been shown to affect their physiology and overall health [7,13–17]. Physiological stress from injuries and an increase in energy demand resulting from gull harassment could be contributing to calf deaths in this population [8,13,14]. Unexplained local high mortality occurred at PV between 2003 and 2013; of 672 dead whales, 91% were calves less than three months old [18,19]. A recent study based on long-term behavioural observations shows a positive relationship between gull harassment and the number of dead calves registered at PV each year [20]. First-year survival probabilities of individual SRW exposed as calves to different severities of gull wounding has not been estimated. In an attempt to connect gull-attack behaviour to SRW population dynamics, we used capture–recapture methods to test the hypothesis that wounding decreases calf survival.
2. Methods
(a) Study area and database
Photo-identification aerial surveys were conducted along the shoreline of PV (figure 1b). Whales inhabit PV from April to December [21,22]. Individuals without calves stay a mean of 52 days (range 8–145), while mothers with calves stay longer (77 days, range 15–170) [21]. In the 1970s the area was surveyed repeatedly within each calving season, but since the 1980s it has been surveyed once a year in September or October, close to the peak of whale abundance [21]. We followed aerial survey procedures and methodology previously reported [6,21,23]. SRWs are individually identified from photographs of their callosity patterns and dorsal pigmentation markings [6]. The reference catalogue up to 2017 includes 3777 photo-identified individuals, of which 773 were identified in their year of birth. The total number of calves recorded during aerial surveys is much higher than this, but only identifiable individuals—those with a developed callosity pattern and/or a distinct skin pigmentation pattern—can be added to the catalogue. Individual sightings were pooled into annual sampling occasions to create a presence–absence matrix of individual yearly sightings.
(b) Variation of gull-inflicted lesions among years
To investigate gull-attack effects, we used the data provided by [7] of the area of gull-inflicted lesions (hereafter referred to as a lesion index) on calves born between 1974 and 2011. The lesion index represents the number of extra-small sized lesions that, when summed, is equivalent to the total wounded area—considering that each extra-small lesion represents 0.13% of the individual's back area (see [7] for details). Data included the lesion indices of 740 individuals, either photo-identified calves (n = 192) or unidentified calves with known mothers (n = 548). The lesion index was calculated from aerial survey pictures obtained during the peak of whale abundance (September and early October), during which gull attack rates are also highest [24]. Wounding severities estimated for calves photographed in aerial surveys from the 1980s onwards are considered to be representative for that particular year because the area of a calf's back carrying lesions tends to reach its maximum by October [7].
The years 1991, 1992, 1994, 1997, 1998 and 2001 were excluded because of a lack of enough information about gull wounding in those years. We used the lesion index estimated for a calf in its year of birth, and did not include information about gull-inflicted lesions present in subsequent years when it was photographed as juvenile or adult. We fitted a generalized linear model (GLM) of the lesion index (a count) as a function of the year of birth with a negative binomial error structure, to allow for overdispersion, and log link function [25]. Predicted values from this model were later used as a temporal covariate (hereafter referred to as the lesion index covariate) in the capture–recapture analysis. All analyses were performed in R with packages stats and MASS [26,27].
(c) Modelling calf survival: the effect of gull-inflicted lesions
We used a subset of the data comprising the encounter histories from 1974 to 2017 of 597 whales identified at PV in their year of birth between 1974 and 2011. We used the encounter histories up to 2017 so that individuals that entered the dataset in recent years (in 2011 or just before that year) had a chance to return to PV and be recaptured. To investigate the influence of gull-inflicted lesions on calf survival, we used Cormack–Jolly–Seber (CJS) mark–recapture models. First, goodness-of-fit (GOF) tests were performed to assess the quality of fit of CJS models. GOF tests indicated a lack of fit of the CJS model resulting from a difference in recapture probability between newly and previously captured individuals (test 3.SR: = 149.71, = 37, < 0.001). This lack of fit is often attributed to transient individuals (captured only once) and is conventionally accommodated by modelling two time-since-marking classes for survival probability (first year after marking; all subsequent years). In our dataset all individuals were marked in their year of birth, so implementing this formulation provided an age class model for first year (calf) survival and age 1 + year (non-calf) survival. There was no indication of overdispersion in the dataset ( = 0.95).
Recapture probability was modelled as constant over time, or as a function of: the year (t) to test for time-dependent effects; a temporal trend (T), as a continuous integer variable to test whether the recapture rate increased or decreased over time; and a period, defined as either 1974 to 1995, when the main gull attack target was the mothers, or 1996 to 2011, when the main target switched to calves [7].
Survival probability was modelled as constant for calves and non-calves, or with only calf survival varying with t, T, period, and lesion index covariate. Models with additive effects between the lesion index covariate and period for calves were also fitted. Model selection was based on Akaike's information criterion (AIC) [28], as a measure of the support from the data for each model among the set of models considered. If more than one model had support, a model average was constructed based on the models' AIC weights. We used the R [29] package RMark [30] to build models in software MARK [31], and package R2ucare [31] to perform GOF tests. Additionally, we estimated mean calf survival for each period by using delta methods to estimate standard errors [32].
3. Results
Of all calves (identified: n = 192 and unidentified, n = 548), 483 (65.3%) had gull-inflicted lesions. Of 192 identified calves, individuals with no lesions (n = 77) were all identified prior to 1995 after which all calves showed one or more lesions. Most identified calves (77.4%, n = 89) with gull-inflicted lesions were not seen again at PV. By contrast, less than half (44.2%, n = 34) of calves without lesions were not seen again.
The area of gull-inflicted lesions on a calf's back varied with year of birth (z = 28.55; p < 0.001). Mean calf lesion index was 1.72 (range 0–28) between 1974 and 1995, increasing to 17.0 (range 0–147) between 1996 and 2011 (figure 2a). These values represent an increase in the average injured back area from 0.2% (range 0–3.6%) to 2.2% (range 0–19.1%).
Figure 2. (a) Lesion index (area of lesions on the whale's back) per calf from 1974 to 2011 fitted by the GLM model. Points indicate observed values per calf. Data obtained from [7] (b) SRW calf survival probability. (c) Recapture probability for SRWs identified in their year of birth. (d) Relationship between calf survival probability and the lesion index per calf. Estimates of (b), (c) and (d) are shown with 95% CI (error bars). Shadows indicate the period when the main target of gull attacks were mothers (from 1974 to 1995, blue) and calves (from 1996 to 2011, white).
(a) Calf survival decreases with increasing gull-inflicted lesions
Of the 24 candidate models considered, the best model included calf survival probability as a function of the lesion index covariate (β = −0.09, 95% CI −0.06 – −0.13), allowed a time-varying recapture probability, and was well supported by the data (81% of the AIC weight, table 1). Other models with some support included those with an additive effect between period and lesion index, and a trend in calf survival (AIC of 3.81 and 4.86; 12% and 7% support, respectively).
| survival probability | recapture probability | k | AICc | DeltaAICc | AICc weight |
|---|---|---|---|---|---|
| (a + c:lesion) | (t) | 46 | 3204.16 | 0 | 0.81 |
| (a + c:period + c:lesion) | (t) | 48 | 3207.98 | 3.81 | 0.12 |
| (a + c:T) | (t) | 46 | 3209.02 | 4.86 | 0.07 |
| (a + c:period) | (t) | 47 | 3220.59 | 16.43 | 0 |
| (a + c:t) | (t) | 83 | 3226.33 | 22.17 | 0 |
| (ca) | (t) | 45 | 3236.81 | 32.64 | 0 |
| (a + c:lesion) | (T) | 5 | 3262.44 | 58.28 | 0 |
| (a + c:period + c:lesion) | (T) | 7 | 3266.25 | 62.09 | 0 |
| (a + c:T) | (T) | 5 | 3266.46 | 62.29 | 0 |
| (a + c:t) | (T) | 42 | 3273.51 | 69.35 | 0 |
| (a + c:period) | (T) | 6 | 3276.55 | 72.38 | 0 |
| (a + c:lesion) | (period) | 5 | 3286.05 | 81.89 | 0 |
| (ca) | (T) | 4 | 3288.47 | 84.3 | 0 |
| (a + c:T) | (period) | 5 | 3288.6 | 84.43 | 0 |
| (a + c:period + c:lesion) | (period) | 7 | 3289.08 | 84.92 | 0 |
| (a + c:t) | (period) | 42 | 3295.94 | 91.78 | 0 |
| (a + c:T) | (.) | 4 | 3301.8 | 97.64 | 0 |
| (ca) | (period) | 4 | 3303.07 | 98.91 | 0 |
| (a + c:lesion) | (.) | 4 | 3303.26 | 99.1 | 0 |
| (a + c:period + c:lesion) | (.) | 6 | 3305.5 | 101.33 | 0 |
| (a + c:period) | (period) | 6 | 3305.52 | 101.36 | 0 |
| (a + c:t) | (.) | 41 | 3311.99 | 107.83 | 0 |
| (a + c:period) | (.) | 5 | 3325.69 | 121.52 | 0 |
| (ca) | (.) | 3 | 3376.61 | 172.45 | 0 |
Following model averaging, estimated apparent calf survival showed a marked decrease after 1995, even though the recapture probability remained low but stable since the 1980s (figure 2b,c). Results showed a clear relationship between calf survival and lesion index. Calf survival decreased from 0.659 (95% CI: 0.570–0.737) for calves without lesions to nearly zero (0.026, 95% CI: 0.007–0.093) for calves with a lesion index of 45 (figure 2d), which was close to the mean number of lesions per calf registered in 2011 (46.92 ± 0.08). Between 1974 and 1995—the period when mothers were the main targets of gull attacks—mean calf survival was 0.622 (95% CI: 0.346–0.898), while between 1996 and 2011—when calves were the main targets—it dropped markedly to 0.291 (95% CI: 0.198–0.394) (table 2). After surviving the first year, mean non-calf survival was estimated to be 0.959 (95% CI: 0.944–0.970).
| n = 597 | 1974–1995 | 1996–2011 |
|---|---|---|
| mean calf survival | 0.62 (0.35–0.90) | 0.29 (0.19–0.39) |
| gull attack main target | mothers | calves |
| identified calves | 281 | 316 |
| recaptures | 133 | 49 |
| percentage of recaptures | 47.3 | 15.5 |
| mean lesion index/calf | 1.72 [0–28] | 17 [0–147] |
4. Discussion
Our results provide evidence that gull harassment has a negative impact on the survival of SRW calves born at PV, Argentina. Most calves showed a relatively lower lesion index between the 1970s and 1990s than in the 2000s. When SRW mothers were the target of gull attacks, calf survival remained stable. Individual calf survival probabilities varied as a function of their wounding severity; when the lesion index increased, apparent calf survival probability decreased, and calves that suffered greatly elevated gull harassment were unlikely to be resighted in the PV area. These findings are consistent with recent research about the increasing local mortality—based on carcass recovery—of calves at PV that has followed an increase in gull attack frequency and pressure over the last two decades [20]. In addition, mortality of calves less than three months old reaches its maximum at PV in September [18,20], which is also the time of highest gull attack rates [24]. Thus, most calves identified during aerial surveys in Sep–Oct are likely to survive at least until leaving PV to migrate to the feeding grounds.
Recapture probabilities of SRWs identified in their year of birth at PV appear to be lower since the 1980s, when the frequency of aerial surveys was reduced to just once per year, during the peak of whale abundance. However, if the lower calf survival probability was only a result of a drop in recapture probability, a marked decrease in calf survival would be expected from the 1980s, instead of from the mid-1990s as estimated, when calves became the main targets of gull attacks. Ongoing studies are incorporating new techniques that may provide important information about the life histories of the whales that visit PV.
In particular, two new sources of images have recently been developed to photograph individuals for later identification: citizen science photos taken during whale-watching trips and UAV (unmanned aerial vehicle) drone surveys [8,33–35]. By contrast to single annual aerial survey data, these additional sources of data cover most of the whale season and have contributed to expansion of the database. The analysis of photos taken by citizen scientists during whale-watching tours throughout the calving season from 2003 to 2007 added 105 new individuals and new sightings of 45 previously known individuals to the reference catalogue [33]. Drone surveys add around 300–400 whales per year to the catalogue. Thus, future analyses are expected to show higher rates of recapture.
The calf survival probabilities estimated here must be considered with caution, especially since the mid-2000s. Without additional information, it is not possible to distinguish between death and permanent emigration in estimates of survival probability [36]. If whales abandon PV and emigrate permanently to other areas, such as southern Brazil, calf survival estimated in this study will be underestimates of true survival. However, a recent comparison of the photo-ID catalogues for Argentina and Brazil, between 1971 and 2017, documented just 124 individuals seen in both calving grounds; in particular, only approximately 3% of whales in the Argentine catalogue were seen off Brazil [37].
In the present study, of 773 individuals identified as calves at PV, 553 have not been recaptured and only six have been seen off Brazil but not at PV. Efforts are underway to estimate movement rates between both breeding grounds, which may help us better understand the effect of gull harassment, calf mortality and density-dependence processes [38]. For example, a shift in the population distribution along the Argentine coast may be a response to increased density. Mother–calf pairs have continued wintering at PV, while other age groups have expanded their distribution range [39]. Golfo San Matías, 300 km to the north of PV, has been recolonized by solitary individuals and mating groups since 2013 [5]. Catalogue comparisons with other areas in Argentina are underway or planned. Even during periods of a constant low recapture probability, our results showed that calf survival decreased over time at PV together with increased levels of gull-inflicted lesions. Previous studies have suggested that gull harassment is a local stressor that may reduce calf survival [8,11,13,14]. The endocrine response of calves to gull harassment has been analysed using glucocorticoids and thyroid hormone levels. Despite no post-mortem evidence of malnutrition [40], high glucocorticoid levels suggested that calves with severe gull lesions suffered elevated physiological stress before death [13]. Calves increase their respiration rates during attacks and gulls focus their attacks on previously wounded calves, enlarging the lesions [8]. Our results provide further evidence that gull attacks are contributing to calf mortality. Whether calves abandoned their breeding area or actually died during their first year, our analysis suggests that gull harassment may affect future adult recruitment, female reproductive success, and consequently local population growth [41].
In the light of the high calf mortality recorded in some years at PV [18,42] and the conservation challenges the population faces due to climate change [43], our results strongly suggest a need to include gull harassment in measures of habitat quality used by wildlife managers and government officials. Effective reduction of anthropogenic food subsidies may help to control kelp gull population growth [44]. Our results add detail to an emerging picture in which the southwest Atlantic SRW population, although continuing to grow, is increasingly burdened by a number of stressors whose combined effects could threaten its future viability.
Data accessibility
All data needed to reproduce the analyses, including the R code, are available from the Dryad Digital Repository: https://doi.org/10.5061/dryad.c59zw3rb3 [45].
Authors' contributions
M.A.: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, software, validation, visualization, writing—original draft, writing—review and editing; C.F.M.: conceptualization, data curation, investigation, methodology, resources, supervision, writing—review and editing; F.G.D.-J.: conceptualization, formal analysis, methodology, supervision, writing—original draft, writing—review and editing; V.J.R.: data curation, funding acquisition, investigation, methodology, project administration, resources, writing—review and editing; M.S.: conceptualization, data curation, funding acquisition, investigation, methodology, project administration, resources, supervision, writing—review and editing; P.S.H.: formal analysis, methodology, supervision, writing—review and editing; S.N.I.: conceptualization, formal analysis, methodology, supervision, writing—review and editing; F.O.V.: data curation, methodology, writing—review and editing; J.S.: data curation, methodology, resources, writing—review and editing; P.C.S.-L.: conceptualization, funding acquisition, investigation, methodology, supervision, writing—original draft, writing—review and editing.
All authors gave final approval for publication and agreed to be held accountable for the work performed therein.
Conflict of interest declaration
We declare we have no competing interests.
Funding
We thank the Coordination of Superior Level Staff Improvement (CAPES) from Brazil for providing a doctoral scholarship and grant CAPES-PRINT (grant no. 88887.370641/2019-00) to M.A. The National Council for Technological and Scientific Development (CNPq) from Brazil provided research grants to F.G.D.-J. (grant no. 308867/2019-0), and P.C.S.-L. (grant no. 305573/2013-6). Aerial surveys and curation of the database since 1971 were funded by numerous donors through Ocean Alliance and Instituto de Conservación de Ballenas.
Acknowledgements
We thank John Atkinson, Roger Payne and many researchers and photographers for their work during aerial surveys since 1971. Instituto de Conservación de Ballenas, Ocean Alliance, Armada Argentina and Prefectura Naval Argentina provided essential support. Research permits were issued annually by the Dirección de Fauna y Flora Silvestre and the Subsecretaría de Turismo y Áreas Protegidas of Chubut Province, Argentina. We thank Rodrigo A. Martínez Calatalán and Fredrik Christiansen for photographs. Finally, we thank the Editor and anonymous reviewers for their constructive comments that greatly improved the manuscript.
