A systematic scheme for monitoring waterbird populations at Shengjin Lake, China: methodology and preliminary results

1.   Introduction

Bird ringing has provided undeniable insight into the biology of birds over the years, enabling researchers to gather important data about their behaviour, migration patterns, survival rates, and population dynamics (Bairlein, 2001). As a result, it is widely used in efforts to conserve and protect bird species, and to better understand the complex interactions between birds and their environment (Baillie, 2001; Anderson and Green, 2009). However, the analysis of bird ringing data often comes with some potential sources of error and bias (Thorup et al., 2014). For example, as individuals are no longer identifiable due to ring wear and/or loss, mark-recapture analyses may exaggerate mortality and produce erroneous estimates of survival (Juillet et al., 2011; Allen et al., 2019). As an increasing number of studies use long-term datasets of marked individuals to assess survival, these challenges are becoming more relevant (Reid et al., 2022).

Ring wear and loss rates may differ between and within species based on the habitat they use or the species' life-history traits and behaviour (Harris, 1980; Baylis et al., 2018). Birds that inhabit different environments may experience different levels of exposure to weather, abrasion, or other environmental factors that can affect ring wear and loss rates (Rees et al., 1990; Allen et al., 2019). Of course, such processes can also depend on the ringing method and materials used: the type of ring or band used and the method of attachment can affect the durability of the ring; plastic colour rings typically wear out more quickly than steel or alloy rings (Baylis et al., 2018) and the rate of wear will differ depending on the colour and quality of the plastic ring (Ward, 2000).

Changes in local population size, survival, and dispersal based on ringing data are commonly measured using the capture-mark-recapture method (CMR; White et al., 1982; Lebreton et al., 1992; Lebreton and Pradel, 2002). Most CMR models make the following three assumptions: a) individuals within each group must have the same probability of survival; b) they must be identified (e.g., resighted or recaptured) with the same probability on each occasion independently of their group; and c) markings do not become unreadable or get lost (Lindberg, 2012). When data don’t match those expectations, significant biases can appear in survival predictions (Anderson et al., 1985; Bearhop et al., 2003; McDonald et al., 2003). The only way that ring loss can be quantified – and subsequently accounted for – is when individuals are equipped with multiple marks. Then, recapture events in double-marking studies enable the confirmation of the mark loss directly (Juillet et al., 2011; Laake et al., 2014). Thus, supplemental mark loss probability information can be utilized to adjust a posteriori survival estimations (Reynolds et al., 2009). According to studies aimed at determining ring loss rates in waterbirds, Pink-footed Geese (Anser brachyrhynchus) experience a neck-band loss rate of 3.2% per year (Clausen et al., 2015) whereas neck-band loss in Greylag Geese (Anser anser) was estimated at 3.8% per year (Schreven and Voslamber, 2022). Furthermore, loss of colour-rings was found to be highly variable in swans, estimated at a rate of 2.5% up to 7.1% in Bewick’s Swans (Cygnus columbianus) and Whooper Swans (Cygnus cygnus) respectively (Rees et al., 1990).

The Dalmatian Pelican (Pelecanus crispus) is an iconic waterbird, with a wide distribution ranging from eastern Europe to Mongolia, where it occurs in three distinct flyways: the Black Sea–Mediterranean Flyway, the Central Asian Flyway, and the East Asian Flyway (Catsadorakis et al., 2015). While birds that breed in central and eastern Asia are long-distance migrants overwintering in South Asia and East China, respectively, birds that move within the Black Sea–Mediterranean Flyway overwinter in south-eastern Europe (Crivelli et al., 1991). Dalmatian Pelicans feed almost exclusively on fish and therefore spend most of their life cycle in water (both in freshwater and brackish lagoon ecosystems), except for the breeding season when they breed on the ground on isolated areas such as islands, sand banks, or reedbeds surrounded by water (Crivelli, 1987; Crivelli et al., 1998). The conservation status of the Dalmatian Pelican has improved recently, with the species being downlisted in 2017 from “Vulnerable” to “Near Threatened” (BirdLife International, 2018). Indeed, in the Black Sea–Mediterranean Flyway, implemented conservation measures have resulted in an increase in breeding numbers in SE Europe and Turkey (Catsadorakis et al., 2015). However, the species status is still rather precarious as recently indicated by the avian influenza wave of 2022 that had devastating effects on the species' largest colony and also adjacent ones (Alexandrou et al., 2022). Therefore, continued monitoring and conservation efforts are necessary to ensure the long-term survival of Dalmatian Pelican populations, particularly in regions where they are still facing threats. To this aim, unbiased estimates of population parameters should be obtained to inform management actions.

In this study we use resighting data from a long-term double marking project to estimate the rate of ring loss among different Dalmatian Pelican colonies over time, evaluate any possible factors that could contribute to differential ring loss and assess how it may bias the results of mark-resighting analyses. We expect that since pelican colonies inhabit different environments with varying levels of water salinity, temperature, weather patterns and human activities, the integrity of the rings could be affected thus resulting in variable likelihood of ring loss. Furthermore, technical details of the ringing methodology, such as the accurate ageing of chicks and the use of glue in the overlapping edges of the ring could also affect the rate of ring loss among pelican colonies. Understanding the factors that contribute to differential ring loss among pelican colonies is important for correctly interpreting bird ringing data to estimate demographic parameters without bias, to assess if colour rings are suitable for long-term population studies of waterbirds and to realise if any methodological adjustments are needed to minimize the impact of ring loss in such studies.

2.   Materials and methods

2.1   Study area and ringing data

Birds included in this study originated from three major breeding colonies of the species at the western edge of the species distribution (Fig. 1): Mikri Prespa Lake (40°46′ N, 21°04′ E), Kerkini Lake (41°13′ N, 23°08′ E) and Amvrakikos Gulf (38°58′ N, 20°58′ E). In Prespa and Amvrakikos, birds nest on natural islets (Catsadorakis and Crivelli, 2001) whilst breeding at Kerkini occurs only on artificial nesting structures. All birds included in the analyses were ringed as chicks, 3–10 weeks old, with two yellow plastic rings made of ABS plastic (Haggie Engraving, USA) bearing sequential alphanumeric codes on both tarsi, from 1990 to 2017. Through the course of the study the same colour-rings were used in all colonies. Dalmatian Pelican chicks do not exhibit any sexual dimorphism therefore their sex could not be determined. Ringing conditions varied among colonies: in Prespa and Amvrakikos ringing took place on the species' natural colonies and therefore operations had to be done quickly to avoid disturbance. Therefore, ageing of chicks was not accurate as there was limited time to collect biometric measurements and also overlapping edges of the plastic ring were not glued. In Kerkini, one annual ringing session with many people in artificial platforms visiting the colonies, could cause the chicks to fall in the water and not be able to move up into the platform again. Therefore, the ringing session took place during June when most chicks had left the platform but still did not fly and thus captured with nets, brought to a boat and ringed. This allowed for more careful lower mandible measurements as a proxy to assess age and due to the lower time constraints, every ring was glued.

Figure  1.  Map of our study area. Dots and zoomed insets show the Dalmatian Pelican colonies (Amvrakikos, Kerkini, Prespa).

DownLoad: Full-Size Img PowerPoint

Observations of double-ringed individuals were extracted from a database compiled by the Research Institute of Tour du Valat (project leader: A.J. Crivelli, see also Bounas et al., 2022). To study ring loss, we only kept records of double marked chicks resighted any time of the year, therefore recording individual encounter histories. Ring readings were conducted by one or two observers and each resighting was accompanied with the information that an individual carried one or two rings thus leading to a dataset with 7591 resightings from 429 individuals (out of a total of 530 double-marked) in Prespa between 1990 and 1999, 3007 resightings from 351 individuals (out of 616 double-marked) in Kerkini between 2004 and 2017 and 4251 resightings from 495 individuals (out of 802 double-marked) in Amvrakikos between 1990 and 2005.

2.2   Data analysis

2.2.1   Probability of ring loss

The probability of ring loss, as well as the mean period of retaining a ring, were estimated for each population using multi-state continuous-time hidden Markov models (HMMs) as implemented in the “msm” R package (Jackson, 2011). We chose to use continuous-time HMMs because of their advantages in dealing with irregular datasets (Meira-Machado et al., 2009; Glennie et al., 2023), as in our case ring readings were not conducted at regular time intervals and the time of observation does not necessarily match the times of state transitions (i.e., ring loss). We considered three possible states (two-ring state, one-ring state, and no-ring state). Since no birds were recaptured and ringed after the first double-marking process reverse transitions were not possible; therefore, three unidirectional transitions were considered (two rings → one ring → no ring) as given by the transition intensity (Q) matrix:

A zero value was attributed to all positions where no transition is possible. Then, for each transition identified, a transition intensity q_ij is computed, which represents the probability of moving from a state i to a state j. Transition rates were modelled on an annual basis (a year was considered to start in June, when birds fledged). Age of individual pelicans was entered as a categorical 3-class covariate (juvenile: from fledgling to one year old, immature: from one year old to three years old and adult: breeding birds from 3 years old and above). A comparison of models, with and without covariate, was performed using the Akaike Information Criterion (AIC). Results are reported as mean ± 95 % Confidence Intervals (CIs).

2.2.2   Effects of ring loss on survival estimates

To assess the influence of ring loss on survival estimates, we fitted Cormack-Jolly-Seber (CJS) mark-recapture models in the “marked” R package (Laake et al., 2013). To calculate annual apparent survival (φ) for each different population within ten years of ringing, we compared the survival estimates of double-marked birds (i.e., accounting for ring loss) with survival estimates from models that did not account for ring loss by replacing all encounters in state “one-ring” with “no-ring”, as if birds were ringed with one ring only, thus ignoring the information on the loss of the second ring. The difference in annual survival estimates between the models accounting or not accounting for ring loss, is considered to be the effect of ring loss on apparent survival estimates. All statistical analysis was performed with the software R version 3.0.1 (R Core Team, 2022).

3.   Results

Overall, in all three colonies the majority of birds was encountered twice, representing the ringing occasion and one resighting. One bird has been re-encountered 96 times in Prespa, another one 88 times in Kerkini and a third one 58 times in Amvrakikos (Appendix Fig. S1). In Prespa, 326 out of the 429 double marked individuals (76%) lost one ring through the course of the study whereas in Kerkini and Amvrakikos the percentage of birds that lost a ring was 34% (118 out of 351 individuals) and 64% (317 out of 495 individuals), respectively. Most birds have been encountered in the first and second year after ringing in Prespa and Kerkini whereas the majority of birds were encountered in their third year after ringing in Amvrakikos (Appendix Fig. S2).

The estimated mean sojourn time in the double marked state (the mean period of retaining both rings) based on state transition intensities was 670 days (600–749 days) for Prespa, 3561 days (2972–4267 days) for Kerkini and 1576 days (1410–1763 days) for birds ringed in Amvrakikos. We selected the HMMs with age as covariate for Prespa as they had the lowest AIC whereas for Kerkini and Amvrakikos the inclusion of age did not improve model performance. Ring loss probability showed marked differences among different colonies (Fig. 2). In Prespa, probability of ring loss within the first year was the highest, estimated at 0.42 (0.39–0.46) while after ten years the cumulative probability of losing a ring increased at 0.995 (0.991–0.998). In Amvrakikos probability of loss was lower, estimated at 0.21 (0.19–0.23) within a year from marking, reaching 0.9 (0.88–0.92) ten years after first marking. Finally, in Kerkini, ring loss probability was much lower with a value of 0.1 (0.08–0.12) within the first year of ringing while after ten years ring loss was estimated at 0.64 (0.58–0.71). Considering the age of birds, mean cumulative probability of loss within the five first years of ringing was higher for juvenile birds (0.75) compared to other age classes in Prespa (0.72 and 0.68 for immatures and adults respectively).

Figure  2.  Probability of ring loss in continuous time (years) for Dalmatian Pelicans ringed in Prespa, Kerkini and Amvrakikos. Vertical solid line shows mean sojourn time (average period in which a bird remains in the two-ring state) along with lower and upper confidence intervals (dotted lines).

DownLoad: Full-Size Img PowerPoint

Since ring loss was found to occur in all colonies, annual survival was also overall underestimated when not accounting for ring loss (Fig. 3). Differences were more visible in Prespa, with annual survival probabilities being underestimated by 0.23 in the presence of ring loss (0.54 vs. 0.76 when accounting for ring loss). In Amvrakikos, annual survival was underestimated by 0.1 when not accounting for ring loss (0.74 vs. 0.85) and finally the smallest underestimation was observed in Kerkini (0.05; annual survival of 0.83 vs. 0.88 when accounting for ring loss).

Figure  3.  Annual survival estimates (along with lower and upper confidence intervals) for different colonies when accounting and not accounting for ring loss.

DownLoad: Full-Size Img PowerPoint

4.   Discussion

Plastic rings have been used widely on waterbirds, both marine and freshwater. While they have provided great insight in many studies, ring loss could compromise the estimation of population parameters (e.g., Alisauskas and Lindberg, 2002). Our study provides important insights regarding the extent of ring loss among Dalmatian Pelican colonies and how they may impact demographic estimates of mark-recapture studies. Ring loss and wear have scarcely been directly quantified in pelican species so far, however possible biases that may arise from ring loss have been previously highlighted (Ryder, 1981; Anderson and Anderson, 2005). Compared to previous studies in waterbirds, the rates estimated in this study were substantial. For example, 42% of the ringed Dalmatian Pelicans in Prespa and 21% of birds in Amvrakikos would have lost their rings within a year, whereas the cumulative probability rises to more than 90% after ten years for both colonies. A previous study exploring the population dynamics of a Dalmatian Pelican colony in Amvrakikos from 1985 to 2005 using CMR models, ring loss was indirectly estimated to be around 11% albeit with low model likelihood (Doxa et al., 2010). Regarding other waterbirds, high ring loss rate has been observed in Whooper Swans (26.7% for males) while it seems that species with larger body size show increased ring loss rates (Rees et al., 1990). On the other hand, the rate estimated for birds ringed in Kerkini (10%) is similar to overall rates reported for neckbands and leg rings in geese and swans (Rees et al., 1990; Conn et al., 2004; Juillet et al., 2011).

The fact that we quantified ring loss in different colonies of the same species in continuous time, gives us the opportunity to assess the intra-specific variance in ring loss and possibly distinguish some factors that may be involved. It has been proposed that neckband loss probabilities show a decreasing trend throughout studies implemented over the past decades (Schreven and Voslamber, 2022) thus reflecting improved practices in ringing methodology such as more durable plastic, increased edge overlap, and the use of glue. In our case, indeed ringing of Dalmatian Pelicans at Prespa and Amvrakikos were the first such operations in Greece (started in 1986) whereas ringing at Kerkini started much later, in 2004, when pelicans successfully used for the first time the artificial platforms built in 2002 (Handrinos and Catsadorakis, 2020). However, the considerable variation in ring loss we observed among colonies was probably not a result of advances in ringing methods but rather resulted from a realistic trade-off between time-minimisation and collection of bird measurements during capture operations, rising from the investigator’s disturbance (Carney and Sydeman, 1999). Ringing fledglings in colonial waterbirds can be rather challenging, especially in mixed-species colonies as it might cause disturbance of species with different ecology (Götmark, 1992; Champagnon et al., 2019). Specifically, in Prespa Lake Dalmatian Pelicans breed with Great White Pelicans (Pelecanus onocrotalus) and additionally, different nesting units can be asynchronous during breeding (Hatzilacos, 1986; Hatzilacou, 1992; Catsadorakis and Crivelli, 2001). Therefore, to minimize any negative effects of disturbance on pelicans (Boellstorff et al., 1988), the inaccurate ageing of fledglings might have led in one of both of rings to get easily detached as birds' tarsi were not fully developed. This scenario is further supported by occasional researcher observations of detached rings within the Prespa colony in June–July when all chicks fledged. In fact, we showed that in Prespa, detachment of rings in the nest could be more common as probability of ring loss was higher during the first year. On the other hand, in the other colonies ring loss increased with time, possibly as a result of not using glue in the ring edges as that would prevent coils from opening when the ring got stuck in the environment or when pelicans actively tried to remove them, as they have actually been seen to do so, and therefore rings could detach from the birds easily (Fjetland, 1973; Schreven and Voslamber, 2022). In Kerkini the different ringing conditions, where older birds were ringed and with less time constraints, are reflected on the lower ring loss probabilities estimated.

However, apart from extrinsic factors such as the effect of ringing session conditions, still we detected an increasing loss of rings over the years. Several environmental factors could be implicated such as extreme temperatures, salinity, or ultraviolet radiation (Harris, 1980; Rees et al., 1990; Anderson et al., 2011; Feare, 2011). In general, plastic rings have been found to wear more often than metal ones as they can become brittle through years (Ward, 2000). Available evidence so far show that birds from Amvrakikos show an increased connectivity with the coastal populations in western Balkans even during non-breeding periods (Catsadorakis et al., 2015; Bounas et al., 2022) and can therefore be subject to different pressures than the rest of the colonies. It is possible that increased water salinity and sun exposure in such coastal wetlands could affect the durability of plastic rings at some level in a similar way as indicated by the most pronounced metal ring wear in marine species (Harris, 1980). Furthermore, Prespa Lake is a wetland that freezes almost every winter rendering it highly unsuitable for wintering waterbirds (Catsadorakis, 1997). So, birds from Prespa can experience extremely low temperatures even if for a limited amount of time especially at the early part of the breeding season. Although no direct evidence so far links cold conditions with colour ring loss, it should be noted that ABS plastic although performs well in low temperatures it also changes its behaviour from ductile to brittle (Mura et al., 2018). Additionally, due to the lake freezing and the subsequent decrease in fish availability, birds leave during the winter to find prey elsewhere thus rendering the Prespa population a migratory one (Bounas et al., 2022). Therefore birds could experience a wider variety of threats and risks that can affect both survival and ring loss (Cheng et al., 2019; Sergio et al., 2019). Our approach using irregularly collected resighting data does have an inherent limitation in attributing the ring loss to specific life-stages, as it is practically impossible to pinpoint the precise timing and location of ring losses due to the unpredictable nature of resightings. A different study design with regular time intervals could establish a robust link between increased probability of ring loss and specific life cycle stages.

Based on our estimates, ring loss can be a significant challenge for the assessment of Dalmatian Pelican population dynamics using mark-recapture methods. Our annual survival estimates for Amvrakikos were found to be comparable with the ones calculated using the CMR method (Doxa et al., 2010), albeit directly accounting for ring loss led to higher unbiased survival estimates in our case. We have shown that rates of ring loss may increase over time and can be population specific which can have serious implications especially for long-lived species such as the Dalmatian Pelican. The magnitude of survival underestimation under the effect of ring loss can easily lead to bias on population projections thus lead to misguided management action by shifting conservation resources from other species/populations of concern (Alisauskas and Lindberg, 2002). Accounting for mark loss is therefore needed for obtaining unbiased demographic parameter estimates, whereas implementing robust integrative analysis methods can further improve estimates (Reynolds et al., 2009; Riecke et al., 2019).

Finally, our continuous-time HMM approach can provide some useful information on if plastic colour rings are suitable for long-term population studies of long-lived waterbirds. If trade-offs between sample size of ringed birds and probability of ring loss are further evaluated, some strategies to effectively implement ringing sessions can be shaped; given that appropriate ringing methods are used (aging and gluing of rings) enough number of birds could retain rings long enough to provide useful data for long-term studies of waterbirds even if fewer individuals are ringed.

CRediT authorship contribution statement

Anastasios Bounas: Writing – original draft, Formal analysis, Data curation. Giorgos Catsadorakis: Writing – review & editing, Methodology, Conceptualization. Dionyssia Hatzilacou: Writing – review & editing, Methodology, Data curation. Theodoros Naziridis: Writing – review & editing, Methodology, Data curation. Jocelyn Champagnon: Writing – review & editing, Methodology, Data curation. Alain J. Crivelli: Writing – review & editing, Methodology, Data curation, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

We are thankful to all the people involved in ringing and ring-reading in all Dalmatian Pelican colonies, namely Harris Nikolaou, Myrsini Malakou and Kostas Manikas. Finally, many thanks to members of the public that contributed with reporting their observations.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.avrs.2024.100166.