Audouin’s Gull Colony Itinerancy: Breeding Districts as Units for Monitoring and Conservation

Abstract

We investigated the spatial structure and colony itinerancy of Audouin’s gull (Ichthyaetus audouinii) adult breeders across multiple breeding sites in the central Mediterranean Sea during 25 years of fieldwork. Using cluster analysis of marked individuals from different years and sites, we identified five spatial breeding units of increasing hierarchical scale—Breeding Sites, Colonies, Districts, Regions and Marine Sectors—which reflect biologically meaningful boundaries beyond simple geographic proximity. To determine the most appropriate scale for monitoring local populations, we applied multievent capture–recapture models and examined variation in survival and site fidelity across these units. Audouin’s gulls frequently change their location at the Breeding Site and Colony levels from one year to another, without apparent survival costs. In contrast, dispersal beyond Districts boundaries was found to be rare and associated with reduced survival rates, indicating that breeding Districts represent the most relevant biological unit for identifying local populations. The survival disadvantage observed in individuals leaving their District likely reflects increased extrinsic mortality in unfamiliar environments and the selective dispersal of lower-quality individuals. Within breeding Districts, birds may benefit from local knowledge and social information, supporting demographic stability and higher fitness. Our findings highlight the value of adopting a District-based framework for long-term monitoring and conservation of this endangered species. At this scale, demographic trends such as population growth or decline emerge more clearly than when assessed at the level of singular colonies. This approach can enhance our understanding of population dynamics in other mobile species and support more effective conservation strategies aligned with natural population structure.

1. Introduction

More than 95% of seabird species exhibit colonial breeding behavior [1]. This pattern is consistent across phylogenetically distant groups and significantly higher than in any other avian taxa [2,3], suggesting that colonial behavior evolved due to strong constraints imposed by the marine environment, such as the scarcity of suitable nesting sites, instability of reproductive habitats or extreme climatic conditions. The evolution of this behavior provides clear advantages such as enhancing protection from predators [3,4,5], giving a reference point for juveniles awaiting recruitment [6,7], optimizing sex ratio [5], reducing intraspecific aggression among individuals [8], facilitating social stimulation, synchronizing pair reproduction, promoting familiarity among individuals and influencing partner selection or switching [9]. Additionally, colonies serve as information centers regarding the spatial distribution of suitable nesting sites and food resources, particularly when resources are unpredictably distributed [10,11]. This “collective cognition” [12] is enhanced through the association with the same individuals over time.

Despite its benefits, this evolutionary answer to the constraints of the marine environment has created other drawbacks, such as increased competition for resources during periods of scarcity [13], risks of cuckoldry [14] and inbreeding depression [15,16]. Furthermore, density-dependent effects can influence demographic variables when population size becomes excessively high [17] or falls below a critical threshold, leading to Allee effects [5].

Many seabird colonies have been active in the same locations over entire historical periods, especially in stable and persistent habitats [4]. However, other circumstances, such as habitat instability, may force the whole local population to abandon the colony site permanently or temporarily. This latter case has been often documented, as in South African colonies of various cormorants (Cape cormorant Phalacrocorax capensis, Crowned c. P. coronatus, White-breasted c. P. carbo), gulls (Kelp gull Larus dominicanus, Hartlaub’s gull Chroicocephalus hartlaubii) and terns (Damara tern Sternula balaenarum, Swift tern Thalasseus bergii, Roseate tern Sterna dougallii), with causes ranging from parasite defense to human disturbance or interspecific competition [18]. Species nesting in coastal marshes or sandbanks often exhibit annual variation in breeding sites, as observed in the slender-billed gull (Chroicocephalus genei) [8,19,20]. Similar patterns have been documented for Mediterranean gulls (Ichthyaetus melanocephalus), black skimmers (Rynchops niger), black-billed gulls (Larus bulleri) and sandwich terns (Thalasseus sandvicensis) [21,22,23,24], among other seabird species.

Audouin’s gull (Ichthyaetus audouinii) is known to exhibit highly nomadic breeding behavior, with annual variability in colony sites and breeding pair numbers [25,26,27,28,29,30,31]. Muntaner [27] termed this frequent alternation of colony sites, with abandonment and recolonization in subsequent years, as “itinerancy”. This behavior, however, usually occurs within traditional areas to which the population is faithful [32], typically encompassing several islands or coastal spots [33].

During the breeding season, Audouin’s gulls are primarily distributed within the Mediterranean basin and in Portugal, with most of the population concentrated in the western part of their range (the Iberian Peninsula). In contrast, smaller colonies are scattered across the rest of the Mediterranean [30,34].

This type of distribution, with most individuals concentrated in a few major breeding colonies and the remainder scattered across smaller ones, exposes the species to two main risks. The first is the threat of a sudden catastrophic event affecting the larger sites, potentially leading to a collapse of the overall population. The second is the risk of local extinctions driven by pressures on the smaller colonies, which may result in a progressive contraction of the species’ range [35]. Effective conservation planning should consider the functional role of each colony site at the metapopulation scale to ensure targeted and efficient management actions.

The itinerant breeding behavior of Audouin’s gull complicates the comprehension of this dynamic. The transient nature of nesting sites makes it difficult to protect and monitor local populations, thereby hampering the collection of accurate and comparable demographic data over time. Against this backdrop, the primary aim of our research is to delineate the spatial scale of colony itinerancy to define what constitutes a “local population”. This step is crucial for identifying an appropriate geographic scale to enable effective monitoring and conservation planning, even when colonies are small and dispersed across multiple sites within a single breeding season or over consecutive years.

We investigated which spatial unit—from breeding sites to broader regional areas—is most suitable for demographic data collection and conservation planning. Our analysis was based on three key assumptions:

• The optimal scale is large enough to encompass multiple alternative breeding sites yet small enough to meaningfully represent the local population.
• Individuals in a given area gain an advantage in survival by remaining within rather than leaving the area.
• The emigration from this unit by an experienced breeder is a rare event.

The results aim to fill the gap between ecological understanding and practical conservation, offering a model for managing mobile seabird species like Audouin’s gull.

2. Materials and Methods

2.1. Study System and Data Collection

Audouin’s gull typically prefers rocky cliffs and islands where it finds suitable habitats for nesting and, potentially, an absence of terrestrial predators (Figure 1). Spain has consistently hosted most of the breeding population of Audouin’s gulls in the Mediterranean, primarily concentrated in the Chafarinas Islands and the Ebro Delta, even if, since 2006, a significant decline has been observed [36]. About 31% of the global population has recently colonized some coastal islands of southern Portugal [37]. In the western part of its range, the species often breeds in different habitats like coastal wetlands, including salt marshes, dykes and sand dunes [38]. Roof-top breeding is starting to be observed in Spain [39].

Italy, like Greece, hosts a smaller number of breeding pairs, approximately 5% of the Mediterranean population. Still, due to the high number of colonies and the extent of their range, it represents a metapopulation of high conservation value. Until recently, the Italian breeding population had a relatively stable trend. Since the mid-to-late 2000s, there has been a decrease in the historical part of the range represented by Sardinia and Tuscany, which until 2012 accounted for about 80% of the national population [40]. We focused our research on this historically significant area of the species’ range, encompassing all the known Italian colonies of Sardinia and Tuscan Archipelago, as well as the adjacent French island of Corsica (Figure 2).

Breeding monitoring of this species included data collection on colonies every year from May to July, conducted through boat surveys, land-based observations and drone flights. After adopting the Marine Strategy protocols, the abundance and distribution of this gull are used in Italy as indicators to evaluate the achievement of the target for Good Environmental Status (GES) in the marine waters [41].

A color-banding program for juveniles was implemented in 1998. All individuals have been banded in their natal colony as chicks. Chicks were caught and marked at accessible breeding sites with metal and color rings (white Polymethyl Methacrylate bands with black codes; in Corsica, color was replaced by blue after 2011). Resighting of marked adults by telescope and/or camera was obtained through regular control of all breeding colony sites. Large and easily accessible colonies (e.g., Aspretto in Corsica, Nora and Quartu in Sardinia) received several visits during each breeding season; other sites were checked at least once a year, as a rule, for periods of 1–3 h. Reading success was highly variable and was influenced mainly by vegetation cover at the breeding site.

We analyzed a dataset derived from a database spanning 2002–2023, which contained approximately 3500 ringing records and 13,500 resights. We extracted 6500 events from this database related to readings during the breeding period (from mid-April to mid-July), focusing on approximately 750 adult breeders. Individuals observed one or more times at a given colony during the breeding season were all included as potential breeders, regardless of whether breeding was directly confirmed (e.g., through observation on a nest). While it cannot be entirely ruled out that a few non-breeders or individuals breeding elsewhere were included, several precautions were taken to minimize this possibility. Specifically, individuals observed in more than one colony site during the same season or present in colonies that dissolved before egg-laying were excluded from the dataset.

2.2. Statistical Analysis

Our study aimed to identify the most suitable spatial unit to define and study local Audouin’s gull metapopulations. The first methodological step, therefore, focused on defining and comparing different geographic units. For this reason, we performed a cluster analysis to group different colony sites utilized over the years within the study area based on geographic proximity and similarity in the individual composition of the colonies. Geographic proximity was assessed using the geometric distance between site coordinates, while similarity was evaluated based on shared marked individuals. The entry level for the cluster analysis is represented by the current unit that we used for assessing demographic trends, i.e., the Colony unit, which often corresponds to an island, a small group of islets and/or a section of coastline. Population composition was determined by compiling the list of all marked individuals observed at each site during a breeding season. The clustering algorithm compared the similarity between these lists, with more shared individuals indicating greater similarity between the local populations.

For this analysis, we selected Euclidean distance as the distance measure and Ward’s method as the classification algorithm. Clusters were grouped at varying Euclidean distance thresholds, progressively increasing by a factor of five at each level. These units represent progressively broader geographic scales. Specifically, five spatial units were defined as follows:

• Breeding Site level (BSI): Focuses on the exact locations where the local pairs established their annual breeding sites (e.g., a specific coastal section within an island). This level wasn’t included in the cluster analysis. Still, it was utilized in the subsequent analysis to verify whether information derived from the cluster analysis could also be confirmed at a highly detailed geographic scale. This represented the lowest geographic level.
• Colony level (COL): Groups of geographically close sites representing traditionally recognized Colonies (Euclidean distance = 0).
• District level (DIS): Groups of close Colonies (Euclidean distance = 10).
• Region level (REG): Groups of multiple Districts extending over an area consistent with a regional scale (Euclidean distance = 50).
• Marine Sector (MSE): Aggregation that includes two or more Regions within an extensive macro-area (Euclidean distance = 250). This is the highest geographic level.

In the second step, we compared demographic parameters across the different spatial units identified through cluster analysis. Models were constructed for all five spatial units (BSI, COL, DIS, REG, MSE) to estimate survival, probability of changing breeding unit and philopatry—defined here as the likelihood that a newly recruited breeder reproduces within the same unit where it was born.

To this end, we employed multievent multistate capture–recapture models, applied to data collected between 2002 (the start of breeding by our older cohort) and 2023. These models extend the unistate (Cormack–Jolly–Seber) and multistate frameworks [42] into a multievent framework [43], which incorporates conditional probabilities to connect observable events with underlying biological states.

All models were fitted using E-SURGE version 2.2.3 [44]. To evaluate the fit of the data to the models, we performed a goodness of fit (GOF) test specific to multievent models using U-CARE version 2.3.4 [45]. The GOF test revealed issues related to trap dependence and transience of individuals. To address transience, we adopted an age-structured modeling approach, differentiating the behavior of newly marked individuals. Younger breeders are more likely to disperse after their first breeding attempt and/or exhibit lower apparent survival compared to experienced adults [46]. Trap dependence was modeled following the approach of Pradel and Sanz-Aguilar [47], using two Markovian states: “trap-aware” (A) and “trap-unaware” (U), which describe an individual’s detection history and allow for modeling the behavioral response to previous resightings. A detailed explanation of this implementation is provided in Table S1 in the Supplementary Materials. We defined five mutually exclusive states during the breeding season.

• SFA (Site-Faithful—Aware): Alive observed breeding in its first-ever known or last known site, “trap-aware”.
• FU (Site-Faithful—Unaware): Alive breeding in its first or last known site, “trap-unaware”.
• MA (Mover—Aware): Alive breeding in a different site than the last, “trap-aware”.
• MU (Mover—Unaware): Alive breeding in a different site than the last, “trap-unaware”.
• Dead: Individual dead or permanently emigrated.

Since the “unaware” state corresponds to occasions with no detection, encounter histories consist of a sequence of three possible events for each individual and capture session:

• 0: Not detected.
• 1: Detected at its first site or at its previous known site.
• 2: Detected at a different site than the previous.

Each state may change from one event to another, following the multievent model’s conditional probabilities, structured in five matrices: I, initial state; Φ, apparent survival; Ψ, site change; Γ, awareness change; R, recapture. The structure of each matrix is extensively explained in the Supplementary Information (Table S2). Recapture probability is usually calculated in the recapture matrix, but in our models, accounting for the Markovian states “trap-aware” and “trap-unaware”, the likelihood of a resight is modeled in the awareness change matrix. The models were constructed following an all-combination strategy [48], considering the parameters used to resolve the matrices (constant, dependent on the starting or arrival state, dependent on time or dependent on the individual’s age). These models were subsequently compared using a model selection approach based on the Akaike Information Criterion corrected for small sample sizes (AICc) [49]. For all models, parameters, structures and selection processes; see Supplementary Table S3 in the Supplementary Materials.

3. Results

The cluster analysis revealed spatial relationships among the starting 28 Colonies, grouping them into spatial units of increasing dimensions, as illustrated in Figure 3. Each Colony included multiple nesting sites (Breeding Sites) used across different years. As expected, these Breeding Sites were typically located close to one another (average distance: 2.8 km; Figure 4). A total of 65 Breeding Sites were considered in the analysis.

A clear distinction emerged between two main Marine Sectors: The first includes the Tuscan Archipelago, northern Corsica and the Sardinian sites bordering the north Tyrrhenian Sea (Tyrrhenian Sector); the second encompasses the remaining Sardinian and Corsican sites, primarily oriented towards the western Mediterranean (West-Med Sector). Within both Marine Sectors, five Regions and ten Districts were identified (Figure 5).

The model selection results are summarized in

Table 1. Model selection of capture–recapture analysis. The best model for each spatial unit is shown. Notation: ϕ, apparent survival; ψ, site change; Γ, awareness change; np, number of parameters; dev, deviance; QAICc, Akaike Information Criterion corrected for small sample size; w_AICc, weight of relative likelihood of models; w_ϕ, w_ψ, w_δ, weight of relative likelihood of each parameter; state, difference between states; constant, no difference between states; time, temporal variation; age, two age classes (one year after the first capture or older); +, additive effect of variables.

Unit	Φ	Ψ	Γ	np	Dev	QAICc	w_AIC	W_ϕ	W_ψ	W_δ
BSI	constant	state	state + time	27	8701.631	8756.307	0.690	0.690	1.000	1.000
COL	state + age	state	state + time	29	8399.533	8458.311	0.530	0.530	1.000	1.000
DIS	state	state	state + time	28	7046.145	7102.872	0.789	0.794	0.991	1.000
REG	state	state	state + time	28	7586.539	7643.266	0.782	0.782	1.000	1.000
MSE	state + age	state	state + time	29	7037.907	7096.686	0.534	0.536	0.996	1.000

, highlighting the best-fitting models for each unit. Survival and site change are better modeled by a difference between states (site-faithful individuals vs. movers) for all spatial units, except for breeding site survival, which is modeled as constant.

In the best model for Breeding Site (BSI), no survival differences were found between site-faithful individuals and movers (birds relocating to a different Breeding Site). For the other units, instead, leaving the geographic area where an individual had previously bred can result in a survival disadvantage, which varies depending on the scale of reference (see

Table 2. Survival rates from the best-fit models across the five geographical scales. Except for the BSI category, all other units show a difference between individuals breeding within the same spatial unit (site-faithful, S) and those moving to a different unit (movers, M). The averages of the two groups are significantly different in all cases, as indicated by the F-test.

Unit	S/M	Survival	CI95%	SE	F Test
BSI	S and M	0.854	0.841–0.867	0.006	-
COL	Newly marked	0.800	0.744–0.846	0.026
	S	0.875	0.856–0.891	0.009	F = 4.225, p-value < 0.001
	M	0.840	0.805–0.870	0.017
DIS	S	0.860	0.845–0.873	0.007	F = 24.715, p-value < 0.001
	M	0.769	0.676–0.842	0.042
REG	S	0.862	0.847–0.876	0.008	F = 7.4, p-value < 0.001
	M	0.805	0.751–0.849	0.047
MSE	Newly marked	0.790	0.743–0.831	0.023
	S	0.872	0.855–0.887	0.008	F = 87.042, p-value < 0.001
	M	0.752	0.570–0.874	0.079

): Individuals changing Colonies (COL) experienced slightly lower survival compared to site-faithful individuals (0.84 vs 0.875); those leaving their District (DIS), Region (REG) or Marine Sector (MSE) exhibited significantly lower survival rates (see Table 2). For both Colonies and Marine Sectors, the best-supported models differentiate newly marked individuals, which showed lower survival rates. The effect of transients was considered for all spatial units, although it was generally weak and not always selected in the best models. The probability of awareness change (Γ) was time-dependent in all best-supported models. This variation likely captures differences in field conditions, such as observer effort, visibility or access to specific areas.

The probability of changing breeding (

Table 3. Probability of moving from the previous breeding spatial unit (e.g., DIS₁) to another of the same level (e.g., DIS₂).

Units	Site Change (ψ)	CI95%	SE
BSI1 → BSI2	0.238	0.205–0.276	0.018
COL1 → COL2	0.146	0.121–0.176	0.014
DIS1 → DIS2	0.035	0.024–0.052	0.007
REG1 → REG2	0.054	0.039–0.073	0.009
MSE1 → MSE2	0.034	0.022–0.050	0.007

) sites by moving from one spatial unit to another of the same level (from one Sector to another, from one Region to another, etc.) is higher for BSI and Colonies (23.8% annual probability of changing the specific site, 14.6% chance of changing Colony) and much lower for the others (below 5%).

Philopatry was derived from the initial state probabilities (

Table 4. Return rate of the recruited breeders at the same spatial unit of birth (philopatry).

Unit	% Philopatry	CI95%	SE
BSI	43.0	39.6–46.6	1.8
COL	48.5	44.9–52.0	1.8
DIS	54.0	50.4–57.6	1.8
REG	62.6	59.1–66.0	1.8
MSE	83.3	80.5–85.8	1.4

). At the Breeding Site (BSI) and Colony (COL) levels, less than 50% of individuals returned to their natal site for their first breeding attempt.

4. Discussion

The cluster analysis revealed patterns of association among historically established Audouin’s gull Colonies, primarily driven by similarities in individual composition rather than geographic proximity. Even distant locations were grouped despite geographic proximity being one of the variables included in the model. Therefore, high Euclidean distance values are expected to reflect the separation between local populations. Certain Colonies hosted many individuals and persisted over extended periods (e.g., the Colonies of Nora in southern Sardinia, Aspretto in Corsica and Pianosa in the Tuscan Archipelago). However, due to the species’ itinerant behavior [27,50], the local population relocates annually, shifting to another suitable Breeding Site. Oro and Pradel [51] observed that individual movements between Colonies occur within a metapopulation context—in their case, specifically including all Spanish Colonies—and are distance-dependent. In our study, we introduced an additional perspective, demonstrating that Colonies are not necessarily structured geographically but rather by a spatial reference unit perceived as meaningful by the birds. We observed individuals moving along the western Sardinian coast between Colonies up to 100 km apart, yet within the same District. Conversely, some individuals remain confined to much smaller areas, breeding in sites just a few kilometers apart but which represent different Districts. At Aspretto, a single site effectively seems to function as an entire District, concentrating the whole local population [52,53].

From the District level upward (toward larger units), the probability of individual exchange is very low (never exceeding 5%, see Figure 6). However, Colony-wide displacement may still occur in response to parasite outbreaks [54], or external disturbances, whether anthropogenic or due to predation. A notable example is mammalian predation at the Ebro Delta colony [55,56,57,58,59].

In our study, we cannot assess with certainty whether the movement of a single individual is associated with the relocation of the entire Colony or only a part of it, as our analysis is solely based on marked individuals. What we observe is that, in some cases, multiple individuals from the same Colony are detected breeding elsewhere together in the following season, which may suggest the relocation of a larger part of the local breeding population. The key point is that such movements can range from individual dispersal events to what may represent a coordinated shift of most or all adult individuals of a local colony. While we cannot directly observe colony-wide movements, the simultaneous displacement of several marked individuals supports this possibility. Importantly, these shifts occur almost exclusively within the same District, indicating that even when colonies relocate, they tend to remain within relatively well-defined spatial units. The probability of returning to the same spatial unit as the previous breeding site remains high for Marine Sectors, Regions and Districts (over 94%), but it is lower at the Colony (COL, 85%) and Breeding Site (BSI, 76%) levels. This proves a higher level of turnover or reorganization at finer spatial scales, whereas fidelity to the District remains strong.

Thus, the District spatial unit appears to be defined by a biological boundary encompassing the local population, beyond which only a few individuals disperse. This spatial unit represents the population unit where most individual exchanges occur, ensuring the stability of local populations over time. To interpret philopatry in biological terms, we propose that the District level is the scale at which this behavior becomes ecologically relevant. Additional support for the spatial significance of the District comes from an independent line of evidence: the proportion of individuals returning to their natal site for their first breeding attempt. Notably, this proportion exceeds 50% only at the District level. This threshold is critical because it suggests that more than half of the recruits are successfully returning to breed within the same functional population unit, despite probable changes in the location of breeding sites and colonies. While young breeders may be unable to return to their exact natal Colony if it is no longer in use after some years, the fact that over half remain within the same District indicates that philopatry is preserved at this spatial scale.

When comparing the survival of individuals that remain within the same spatial unit to those that move to a different one (e.g., from one Colony to another, from one District to another, etc.), a disparity begins to emerge at the Colony level, with residents showing an advantage of 3.5 percentage points. This gap increases progressively at broader spatial scales (Figure 7), with the most pronounced difference observed at the District level. Individuals who remain within their District exhibit survival rates approximately 9.1 percentage points higher than those who disperse. This substantial fitness cost appears to constrain local populations from remaining within their District, reinforcing the idea that the breeding District constitutes a biologically meaningful unit for demographic and ecological processes.

Difference in survival between individuals that remain breeding within the same spatial unit and those that move to another locality in a different unit of the same type. Survival rates are consistently higher for individuals who stay, with the disparity always driven by the lower survival rates of movers. No difference is observed in the case of Breeding Sites.

Our findings suggest that the impact of dispersal of Audouin’s gulls depends on the spatial scale at which it occurs. Movements across lower-level units (smaller)—such as between Breeding Sites or neighboring Colonies—do not appear to reduce survival. In contrast, a marked survival gap emerges at broader spatial scales, particularly when individuals move between Districts. This scale-sensitive perspective is crucial for understanding the fitness consequences and behavioral decisions associated with dispersal.

Moving into unfamiliar areas can entail increased mortality risks, primarily due to a lack of local knowledge, such as higher exposure to predation or accidents [60,61]. Nevertheless, such movement-related costs are not always evident: For instance, no significant survival differences were observed in dispersing individuals of the Black-Headed gull (Chroicocephalus ridibundus) [62]. Lower survival among dispersing individuals may result from two non-mutually exclusive mechanisms: (i) an actual increase in mortality due to environmental risks and (ii) the fact that dispersers may be non-randomly selected from individuals with lower intrinsic quality [63]. These two processes likely act additively.

Most insights into colony abandonment mechanisms come from studies in Spain, particularly regarding the once-massive Ebro Delta colony, which exerted a strong attraction on immigrants from smaller, lower-quality colonies [38,55,57,64]. In contrast, in our study system, the population does not appear to be distributed around a central reference site. Instead, it consists of a network of many smaller, more scattered colonies, which may lead to different dispersal dynamics. Dispersing Audouin’s gulls may be attracted by larger colonies due to increased food availability, superior habitat quality and reduced predation pressure [51]. The ability to assess potential fitness benefits—both in terms of physical condition and reproductive success—is crucial for predicting positive outcomes for their reproductive efforts [64]. This assessment often relies on direct information gathered through prospecting, a behavior exhibited by active breeders as they explore alternative nesting sites. Prospecting behavior is common among individuals who subsequently leave their breeding area, especially those who have failed in reproduction. However, it also occurs early in the incubation period, suggesting it is not just a reaction to breeding failure but rather a proactive strategy to enhance future breeding success or prepare for potential site abandonment. This behavior likely evolved as an adaptive response to nesting in ephemeral or unstable environments [65]. Kralj et al. [11] further highlights that prospecting behavior is observed in other Larinae species, such as the Mediterranean gull (Larus melanocephalus) and yellow-legged gull (Larus michahellis), as well as in Sterninae species, like the sandwich tern (Thalasseus sandvicensis) and common tern (Sterna hirundo).

Audouin’s gulls that abandon their colony often do so following reproductive failure [65], frequently due to predation. Less experienced individuals are nearly twice as likely to disperse as established breeders [46]. Lower-quality individuals, particularly those lacking breeding experience, may relocate to better conditions, especially with reduced food availability [66]. These movers may also face higher mortality risks simply because of their inexperience with the new environment. For example, young gulls are more susceptible to longline fishing mortality, which decreases with age and experience [67].

Our data show that reduced survival is limited to movements at broader scales, especially between Districts. Survival differences are minimal at the Colony level and negligible at the Breeding Site level. This result suggests that individuals possess sufficient information to relocate effectively within this geographic range, possibly through the sharing of information within Colonies [64]. However, a lack of shared or personal knowledge may increase costs when dispersal exceeds a certain geographical threshold.

Thus, reduced survival following District-level movements may reflect either higher extrinsic mortality or the lower intrinsic quality of dispersers. In either case, dispersal appears to be disadvantageous at this scale.

A District-Based Approach to Population Conservation

The considerations discussed thus far seem to prove that the breeding District effectively represents the unit encompassing the entire local population and that remaining within its boundaries offers a clear survival advantage. Consequently, the District also emerges as the most appropriate observational scale for studying population dynamics and demographic trends, promoting a targeted approach in the species’ monitoring programs. For instance, examining the population trends of Colonies within selected Districts (Figure 8), based on data collected over two decades, reveals patterns that would be impossible to detect at the individual Colony level. However, at the District level, demographic trends become discernible. In all Districts, there is an initial rapid increase in population size, followed by a subsequent decline of varying magnitude depending on the location. This trend aligns with broader observations indicating a progressive decline in population across all surveyed areas over time [40].

5. Conclusions

Understanding local and global demographic parameters is one of the most critical steps for species conservation [33,68,69]. In this context, a District-based approach to local populations can provide more accurate and meaningful monitoring data, thereby enhancing the effectiveness of global conservation programs for the species.

Applying the District-based approach to other geographic contexts is highly desirable, particularly for highly mobile species like Audouin’s gull. This does not require a complex methodology as it simply requires cluster analysis of marked individual presence data and the evaluation of Euclidean distance, which, within the clustering process, measures the loss of similarity between groups.

Our study did not rely on arbitrarily fixed geographic distances. Instead, it aimed to identify local breeding areas based on the breeders’ perspective, which may vary across different locations. For example, the District unit we defined encompassed territories ranging from a single site to a 100 km long area, underscoring the importance of tailoring monitoring plans to each locality to effectively monitor and study each population unit.