Times and partners are a-changin’: relationships between declining food abundance, breeding success, and divorce in a monogamous seabird species

Study site and field work

The fieldwork was conducted on Bonaventure Island (48°30′ N, 64°09′ W) located in the Ile-Bonaventure-et-du-Rocher-Percé National Park in the Gulf of St. Lawrence (GSL, Quebec, Canada) where a long-term study started in 2008 (Fig. 1). This colony holds approximately 50,000 pairs of breeding gannets as well as several thousand immatures. Annually, this colony is monitored for partnership status and breeding success from incubation to second part of the chick-rearing period (from May to September). Gannets were caught from 108 to 184 nests per year using a noose-pole within four plots in the peripheral section of the colony (the first 108 nests were monitored through 2019 and we added nests over the years). All bird capture and handling methods were approved by the Animal Care Committee (ACC) of the Université du Québec à Rimouski (CPA-49-12-102, CPA-65-16-177), and complied with the guidelines of the Canadian Council on Animal Care (CCAC). Field experiments were approved by Canadian Wildlife Service-Environment and Climate Change Canada (permit numbers SC25, RE-27) and by Societe des establissements de plein air du Quebec (permit numbers PNIBRP-2008-001 to PNIBRP-2019-001). Birds were marked with a U.S. Fish and Wildlife Service steel ring and an alphanumeric coded and colored plastic band (permit number 10704). Because gannets regurgitate regularly during handling, we were able to study the diet annually throughout the breeding season. We were therefore able to identify the main prey items and establish their proportion in their diet (see Guillemette et al. (2018)). Occurrences of individually banded gannets within the colony were established visually twice daily over a period from the end of May (before hatching) to the end of August (or September), from 2008 to 2019. We identified established partnerships in the set of monitored nests coupled with an intensive visual survey of the study area and surrounding sectors to ensure that we resighted all live breeding birds. Because gannets have a strong breeding philopatry (Nelson, 2002), the absence of a partner within the colony was considered as a death or a skipping reproductive behavior.

Sex determination

Four thoracic covert feathers were plucked during capture and stored in plastic bags at −20 °C until subsequent analysis. DNA was isolated from these feathers. Quills were cut into 2- to 5-mm-long pieces and submerged in 50 µL of a solution of QuickExtract™ DNA Extraction Solution (Lucigen, Middleton, WI, USA) in a 96-wells plate. Each plate was incubated in a G-Storm GS4 thermal cycler during 5 h at 65 °C for DNA extraction, and 10 min at 95 °C for ending the process. One set of primers was used for the amplification of CHD gene (Chromo Helicase DNA-binding gene): P2 (5′-TCTGCATCGCTAAATCCTTT-3′) and P8 (5′-CTCCCAAGGATGAGRAAYTG-3′) described by Griffiths et al. (1998) for sex determination in birds. The amplification was carried out in a total reaction volume of 25 µL containing 1× ThermoPol buffer (New England Biolabs, Ipswich, MA, USA), 0.2 mM of each dNTP, 0.2 µM of each primer (P2 and P8), 0.05 units of Taq polymerase, 5 uL of DNA (100 ng), and HPLC grade water for completing volume. PCR was performed in a G-Storm GS4 thermal cycler. An initial denaturing step at 94 °C for 1 min 30 s was followed by 40 cycles of 48 °C for 45 s, 72 °C for 45 s and 94 °C for 30 s. A final run of 48 °C for 1 min and 72 °C for 5 min completed the program. PCR products were separated by electrophoresis for 3 h at 95 V in a 3% agarose gel stained with 1× GelRed Nucleic Acid Gel Stain (Biotium, San Francisco, CA, USA). One µL of the DNA amplified solution was diluted in 4 µL of HPLC grade water and mixed with 1 µL of 6× gel loading buffer composed of 0.03% Bromophenol Blue, 0.03% Xylene Cyanol FF, 60 mM EDTA, pH 7.6 and 60% glycerol in HPLC grade water (Griffiths et al., 1998; Redman et al., 2002).

Mackerel and herring biomass

Quantification of diet, breeding success and divorce

We monitored diet of gannets from 2013 to 2019 by analyzing the content of the regurgitation they make when handled or disturbed. The proportions of occurrence of each prey species in the regurgitated food was calculated yearly and, from these proportions, we calculated the Shannon H-index as an index of diet diversity (where higher values signifying a more diverse diet, and inversely): $\mathrm{H}=-{\sum }_{i=1}^s(Pi\ast \ln Pi)$ where Pi = proportion of a species i relative to the total number of species present in diet and s = number of species encountered. We added data from 2004, 2005 and 2009 reported in Guillemette et al. (2018). Given the high rate of mate changing within the monitored nests (108 to 184), breeding success and partnership status were studied in a total of 809 birds for a total of 704 pairs from 2008 to 2019 (Table 1). We estimated annual breeding success of the colony as the number of fledged chicks divided by the number of nests with eggs. A divorce is said to have occurred when two birds that bred together in year t − 1 were alive and present in the colony at year t, but not breeding together. Mate retention occurred when both partners were together at year t − 1 and t, and mate loss occurred when one of the two partners was absent in our study plot (Coulson, 1966; Ens, Safriel & Harris, 1993; Choudhury, 1995). Divorce rate is the number of divorced birds divided by the total number of birds alive.

Statistical analyses

Prey biomass in marine ecosystem and breeding success

Two generalized additive models were nearly similar in explaining the relationship between northern gannet breeding success and the biomass of its main prey (Table 3). The model containing the smoothed function of fall herring and mackerel had a greater proportion of the variance explained (R² = 0.70) and deviance explained (DEV = 77%), with a similar AICc to the model containing only the smoothed function of mackerel biomass, but the smoothed function of fall herring biomass was not significant in the model (P = 0.09). Therefore, the most parsimonious model with the greatest explanatory power was the model containing only the smoothed function of mackerel biomass with a positive relationship (Fig. 2A, ${\chi }_{-2.3}^2$ = 1,068, P = 0.048, AICc = 165.9, delta = 0.00, weight = 0.54).

Diet of gannets and breeding success

The breeding success of gannets increased with the proportion of mackerel and herring in their diet. Years when the diet was dominated by these two species and therefore less diverse were beneficial for gannet reproduction (Figs. 2B–2D). For example, in 2005, the diet was dominated by mackerel (70%) and herring (28%) with a low Shannon H-index of 0.66 and a breeding success of 68%. In contrast, in 2019, the diet was very diversified (Shannon H-index = 1.73), with only 37% mackerel and a wide variety of species (e.g., sand lance, redfish, capelin, squid, cod, etc.).

Breeding success and divorce rate

The breeding success influenced the divorce rate with a lag of 1 year (Table 4, F_1,14 = 9.67, P = 0.008), but not the inverse relationship, i.e. the divorce rate did not influence the breeding success at population level. These results suggest that changes in breeding success can be used to predict variation in divorce rate in gannets the subsequent year (Fig. 3). The most parsimonious model obtained for divorce rate (DivR) is presented in this equation (where ‘BrS’ is breeding success and ‘t − 1’ is the year before divorce): DivR_t = 0.41***−0.57* BrS_t−1 (***:P < 0.001, *: P < 0.05). In this equation, the significant coefficient −0.57 captures the cross-correlation between BrS and DivR 1 year later (P = 0.01). This model explains 89% variance in divorce rate (Adjusted R² = 0.89, F_2,8 = 43.38, P < 0.0001).

Annual differences in divorce rate

The proportion of gannets that divorced after breeding failure was almost three times greater overall than after breeding success (28% vs. 11%, Fig. 4). However, the probability that an individual divorce more after breeding failure than breeding success is highly variable across years according to best parsimonious model selected including the interactions between year t and breeding success at year t − 1 (χ²₁₆ = 47.1, P < 0.001, AICc = 850, delta = 0.00, weight = 0.974). Between 2010 and 2016, individuals experiencing breeding failure did not appear to be more likely to divorce the year after failure than individuals experiencing breeding success. However, the difference was significant between 2017 and 2019 (P < 0.01).

Partnership status and breeding success

Breeding success in gannets was 0.32 ± 0.05 chick.yr⁻¹ on average between 2009 and 2019 (Table 2). At the individual-level, the model with partnership status, time and interactions provided a significantly better fit to the data than the model with only partnership status, time or intercept alone (GLMM, χ²₄ = 28.1, P < 0.0001) (AIC = 2777, delta = 0.00, weight = 1.00). According to the area under the curve (AUC) that measures performance of the model (where AUC = 1 is perfect), the discrimination power of the full model is very good (AUC score = 0.83). Breeding success values and interannual trends were different according to the partnership status and time (Fig. 5). At year t, when partnership status is determined for an individual, breeding success was significantly different between partnership categories. Individuals that retained their previous mate presented higher breeding success (0.35 ± 0.02 chick.yr⁻¹), followed by divorced birds (0.24 ± 0.03 chick.yr⁻¹) and individual that lost their mate (0.16 ± 0.04 chick.yr⁻¹) the previous year (t − 1). There was no significant difference between ‘divorced’ and ‘lost’ categories at year t (P = 0.24). At year t − 1, retained birds had a two-fold higher breeding success than divorced birds (P < 0.0001). In each partnership status categories, pairwise comparison indicated different trends. Gannets that retained their partner decreased their breeding success at year t (P = 0.03) and at t + 1 (P = 0.002). At the opposite, gannets that divorced at year t doubled their breeding success after mate change from t − 1 to t + 1 (P = 0.04). One year after the partnership status determination, breeding success was similar between all individuals (P > 0.97).

When do gannets divorce?

Our results are consistent with our first hypothesis: divorce is a result of breeding failure and prey decrease. As predicted, divorce is more often observed in northern gannets 1 year after breeding failure while breeding failure is induced by mackerel decrease in the Gulf of St. Lawrence. Guillemette et al. (2018) showed that breeding success of gannets between 1979 and 2014 starts to decline below a threshold of mackerel spawning-stock (mackerels over 3 years old) biomass of 132,300 t (or 97,370 t when considering biomass corrected for size of fish (<35 cm) available to the gannets). During the period in which we studied the relationship between breeding success and divorce rate (2009 to 2019), even if we used the total stock biomass of mackerel (including mackerel of 0 to 10 years old), the biomass was quite below this threshold on average (67,336 ± 8,623 t, except for 2009: 141,243 t). These results illustrate that mackerel abundance was at its lowest during our study period. With the analysis of the diet during the same last years, the observed relationship between proportion of this prey and breeding success was similar. We also observed such a relationship with Atlantic herring in the diet, but it is almost three times less important than mackerel. The decrease in the presence of mackerel in the diet also results in an increase in diet heterogeneity, with a negative impact on breeding success. It further supports the importance of mackerel in the diet of gannets during the chick-rearing period (Garthe et al., 2007; Guillemette et al., 2018) as an extrinsic modulator of breeding performance of gannets in the Gulf of St. Lawrence.

Consequently, under conditions of low availability of their preferred prey, both parents increase their time spent foraging (Guillemette et al., 2018) and leave their offspring temporarily unattended. Normally, seabirds divide parental care between sexes: one parent leaves for foraging and the other waits for its return by protecting its chick (Schreiber & Burger, 2002). However, when food is scarce, foraging trips are longer and further away from the colony, inducing the attendant partner to leave the nest to ensure its own survival, leaving a younger chick or the chick earlier (Regehr & Montevecchi, 1997). The unattended chick becomes vulnerable to the assault of adults from neighboring sites (Ashbrook et al., 2008), attacked by adults from nearby sites (Ashbrook et al., 2008), or assaulted by non-breeders attempting to usurp sites (Porter, Anderson & Ferree, 2004; Hamer et al., 2007), or killed by predators (Oro & Furness, 2002), which turns out very often into breeding failure. As observed in other monogamous bird species, a poor breeding performance recorded at a specific nest site with a specific partner influences negatively the occurrences of pair reunion at this site (Ens, Choudhury & Black, 1996; Bried & Jouventin, 2002; Dubois & Cézilly, 2002 but see Choudhury, 1995; Taborsky & Taborsky, 1999).

However, according to a review of 93 species, seabirds exhibit generally very high fidelity to their nesting site and their partner, 75% and 82% on average, respectively (Schreiber & Burger, 2002). In this review, it is indicated that northern gannets have very high nest and mate fidelity (90% and 83.5% respectively), compared to other species. However, divorce rate estimates reported for two closely related species vary between 40–43% for Australasian gannet (Morus serrator) (Ismar et al., 2010), and 45% for blue-faced booby (Sula dactylatra) (Kepler, 1969), which are twice as high as the results obtained for northern gannets in our study. The only divorce rate reported in literature for northern gannets (17% in Cézilly, Dubois & Pagel (2000) from Nelson (2002) at Bass Rock Island, Scotland, between 1961 and 1976) is probably an overestimate of divorce rate because the methodology described included the loss rate. According to our study, the mate change rate (including loss and divorce rates) was on average 30 ± 14% (where loss rate was 7 ± 4% (range: 4–13%) and divorce rate, 22 ± 12%). Considering that loss rate is dependent of adult survival rate (less influenced by food conditions during breeding period) and assuming that the loss rate measured in our study may be used as a fair estimate, we may expect that the more exact divorce rate reported in Cézilly, Dubois & Pagel (2000) should be between 4% and 13% (mean ± SD: 9 ± 4%) between 1961 and 1976. Breeding success recorded during the sixties and the seventies was very high and ranged between 73% and 85% at Bass Rock colony. In this context, divorce rate recorded in a period of poor food conditions at Bonaventure Island (between 2009 and 2019) would be around 2.6-fold higher than the rate of divorce observed during better environmental conditions. During the seventies, in the Gulf of St. Lawrence, mackerel biomass was widely higher than now (mean ± SD: 235,313 ± 75,950 t, Smith et al., 2020). Thus, we could infer that divorce rate was lower during this decade (and after, during the eighties and the nineties).

Breeding habitat quality and environmental predictability have been proposed for some time as factors influencing temporal variability in divorce rate. As an example, European blackbirds (Turdus merula) increase their divorce rate when they are in low quality nesting sites (Desrochers & Magrath, 1993). Recently, a study on a long-lived seabird species (black-browed albatross, Thalassarche melanophris) has demonstrated for the first time empirical evidence that the prevalence of divorce can be directly modulated by environmental temporal variability via sea surface temperature variations (Ventura et al., 2021). In lower quality years, with warmer sea surface temperature anomalies, the probability of switching mate increased in albatross populations. The underlying mechanism proposed to explain the link between sea surface temperature and divorce rate is through a bottom-up process of reduced food availability during warming periods (Behrenfeld et al., 2006), causing a subsequent reduction in breeding success. Thus, our results support the proposed explanatory hypothesis directly linking food abundance, breeding success and divorce rate.

However, our study does not allow us to identify the temporal mechanism of pair formation and mate change at fine scale (timing and proximal causes of divorce) but we explore here few aspects guided by our results. Timing of the formation of old and new pairs, or their break-up, is different between migratory and resident species and between species with continuous and part-time partnership (Ens, Choudhury & Black, 1996). Northern gannet is a migrant species but it is unknown if individuals migrate in pairs or if they reunite on their winter areas as waterfowl species do (Robertson & Cooke, 1999). Pair bonding behavior is unknown during winter for gannets. In two studies reporting migration destination for both sexes in four couples (Fifield et al., 2014; Pelletier et al., 2020), partners wintering in the same area were reported in only one couple (Pelletier et al., 2020). Sex differences in migration patterns are common in seabirds (Phillips et al., 2009; Bosman et al., 2012), and males are generally wintering closer to the colony than females (Phillips et al., 2017). The same is observed for northern gannets nesting in Europe (Deakin et al., 2019), but both sexes are observed throughout the species’ winter range of gannets nesting in North America (Fifield et al., 2014; Pelletier et al., 2020). Because no difference is detected in the arrival date at colony between males and females (Pelletier et al., 2020), it suggests that gannets nesting on Bonaventure Island reunite, form new pairs or break-up from their partners essentially on the breeding ground. We thus hypothesize that individual’s decision process to divorce probably occurs at the colony at the beginning of breeding season.

Why do gannets divorce?

At the individual level, gannets brooding a chick until fledging tend to stay with the same partner while gannets that have failed breeding are more likely to divorce. In a context of food depletion of the preferred prey as observed between 2009 and 2019 in the Gulf of St. Lawrence, individuals that change partners seems to optimize their fitness (1 year after the divorce). It supports the hypothesis that divorce may be an adaptive strategy as it would not happen by chance. Divorce in gannets may be a form of adaptive mate choice, triggered by low breeding success (Dubois & Cézilly, 2002) and by the potential to improve their breeding success with a new partner (Choudhury, 1995; Black, 1996). During a period of low food abundance, gannets that divorce seem to benefit from this change since they increase their breeding success the subsequent years with the new partner. Our results reflect the “win-stay/lose-switch” theory (Switzer, 1993; Naves, Yves Monnat & Cam, 2006; Piper, 2011), which hypothesizes that the decision to repeat a breeding event on the same territory or with the same partner will depend on past performances. Other studies support the theory demonstrating that previous breeding failure predicts infidelity (e.g., Bai & Severinghaus, 2012; López-López, 2016), but it is not always the case. For example, prior reproductive success was not predictive of divorce in Australasian gannet (Morus serrator) (Ismar et al., 2010). These conflicting results reveal the complexity of the mechanisms underlying mate fidelity and suggest that this behavioral modification induces variable physiological costs depending on individual quality and ability to respond to stressors.

According to literature reviews written on divorce in monogamous birds (Choudhury, 1995; Culina, Radersma & Sheldon, 2015), our results suggest that divorce is a short-term adaptive strategy to counteract breeding failure occurring when birds are faced with reduced food supplies. Therefore, divorce should be viewed as a reproductive strategy that maximize individual fitness (e.g., Coulson, 1966; Ens, Safriel & Harris, 1993). According to various hypotheses reviewed by Choudhury (1995), it is difficult and hazardous to determine which is the best hypothesis that explains the divorce in a specific bird species. For instance, gannets that divorce for improving reproductive success may both change partners because their combined qualities result in reduced fitness (incompatibility hypothesis, Coulson & Thomas, 1980), or one member of a pair changes to improve its reproductive success by obtaining a better-quality mate (better option hypothesis, Ens, Safriel & Harris, 1993), or divorce may arise from errors made in the original choice of a mate (errors of mate choice hypothesis, Johnston & Ryder, 1987). These hypotheses are not mutually exclusive and only an experimental study in which the conditions and quality of the partners are controlled would reveal the one that would best explain the divorce in northern gannets.

Our study showed that divorce can be seen as an example of behavioral flexibility to counteract the low productivity observed during a period of limited food resources. Indeed, temporal fluctuations in colony productivity were directly related to Atlantic mackerel abundance and inversely related to divorce rate. At the individual level, gannets that change partners do so following a reproductive failure and there is an increase in reproductive success 1 year following the divorce. In the context of rapid environmental changes (accelerated by anthropogenic pressures), behavioral flexibility would be important because opportunities for dispersal and adaptation are often limited for seabirds. These behavioral responses can therefore lead to the avoidance of ecological traps in which the demographic parameters of the animals (as birth, death, and migration rates) could be altered. Gannets that divorce could improve their individual fitness by increasing their subsequent breeding success. However, breeding success is only one component of fitness and divorce could also affect survival (Nicolai et al., 2012; Culina et al., 2013) and probably decrease lifespan. In a future study, the cost and benefits of mate choice in northern gannets should be explored at physiological and behavioral levels to understand and explain the potential short- and long-term consequences of an individual’s mating decision process on stress regulation and individual health status.