Extreme moulting site fidelity of the Saimaa ringed seal

Site fidelity is commonly observed in pinnipeds and has direct consequences for individual space use and population dynamics. Here, we used photo-identification recapture data to quantify site fidelity of the endangered Saimaa ringed seal (Pusa hispida saimensis) over four successive moulting seasons. We identified 337 seals based on their permanent fur patterns, and 192 of them were observed during at least 2 years. Over the study period, the median number of terrestrial haul-out sites used by an individual seal was four, and nearly 50% of the seals reused them over the years. Although eight seals performed movements (up to 48 km) between the different water basins of Lake Saimaa, most of the studied seals remained in the same water basin over the years. The median distance between successive moulting sites used by an individual seal was 643 m. While these distances were similar within years in both sexes, the distances between years were longer in females, suggesting post-nursing related behaviour. The extreme site fidelity of the Saimaa ringed seal has important implications for its conservation, especially in the land use management of the Lake Saimaa shoreline.


Introduction
Site fidelity is defined as the tendency of animals to return to previously occupied locations in one or more stages of their life history (Switzer 1993). It has been documented for many migratory and non-migratory species (Greenwood 1980). Site fidelity is particularly common in terrestrial habitats, but it also occurs among many aquatic species, such as highly mobile pinnipeds. It may vary across multiple spatial and temporal scales, including very fine-scaled fidelity to breeding and moulting sites, daily to seasonal foraging areas or relatively coarse-scale annual and seasonal migration. By returning to familiar locations, individuals may benefit from more efficient exploitation of resources and improved reproductive success or survival (e.g., Campbell et al. 2008;Kelly et al. 2010;Oksanen et al. 2014;Cordes and Thompson 2015;Koivuniemi et al. 2016). Site fidelity can also have population-level effects, for example, on metapopulation dynamics (Matthiopoulos et al. 2005), social structure (Wolf and Trillmich 2007) and gene flow (Valtonen et al. 2012. While phocids are primarily aquatic, critical parts of their biological cycle-breeding, nursing and moulting-are connected to ice or terrestrial substrates. Therefore, understanding phocid site fidelity patterns in the use of these habitats, which are often threatened by anthropogenic activities, is important for their conservation. Traditionally, various aspects of pinniped spatial ecology have been studied with telemetry (e.g., McConnell et al. 1992;Bjorge et al. 2002;Cunningham et al. 2009;Oksanen et al. 2014) or methods based on individual identification, such as flipper tagging (Pomeroy et al. 2000), painting (Griben et al. 1984), branding (Walker et al. 2010), genetics ) and, more recently, photo-identification (Photo-ID) (Karlsson et al. 2005;Graham et al. 2011;Hastings et al. 2012;Koivuniemi et al. 2016;Sayer et al. 2019). Photo-ID is a non-invasive method, which relies on individual recognition via permanent natural markings, such as scars or pelage patterns. This approach can be applied to a wide range of studies, for instance, estimation of population size (Hiby et al. 2007; Koivuniemi et al. 2019) and reproductive success (Thompson and Wheeler 2008). It is also a powerful tool in the assessment of the degree of site fidelity (Cordes and Thompson 2015;Koivuniemi et al. 2016).
The Saimaa ringed seal (Pusa hispida saimensis, Appendix Fig. A1) is an endemic subspecies of ringed seal that is landlocked in Lake Saimaa, a freshwater lake in Eastern Finland (Fig. 1). Its current estimated population size is ~ 400 individuals (Metsähallitus 2020). Primary threats to the Saimaa ringed seal include incidental by-catch mortality, habitat loss and climate change (Sipilä 2003;Auttila 2015;Jounela et al. 2019;Kunnasranta et al. 2021). Compared to its marine relatives (Härkönen et al. 2008;Kelly et al. 2010;Oksanen et al. 2015), Saimaa ringed seal's relatively small home range and its high site fidelity (e.g., Koskela et al. 2002;Koivuniemi et al. 2016;Niemi et al. 2012Niemi et al. , 2019 are unparalleled in any other pinniped species and reflect the unique life-history of this endangered seal. During the last 9,500 years of being landlocked , Saimaa ringed seals have adopted various strategies to cope with their lacustrine habitat. While arctic ringed seals are ice obligates (Kelly 2001) and depend on sea ice for breeding, resting and moulting, in Lake Saimaa, they give birth in snow lairs located along shorelines (Sipilä 1990) instead of marine ice pressure ridges (Smith and Stirling 1975). Moreover, the ice cover melts in early spring. Consequently, Saimaa ringed seals moult on terrestrial platforms, such as flat rocks on the shores of islands. They also use these for resting at night after moulting Niemi et al. 2013a). Terrestrial habitat is, therefore, crucial for the Saimaa ringed seal. The subspecies is strictly protected through Annex II of the European Union Habitats Directive, in which article 12 especially frames the protection of its breeding and resting sites (Council directive 1992). Yet, anthropogenic land use is increasingly degrading and fragmenting ringed seal habitat in Lake Saimaa (Liukkonen et al. 2017). Therefore, understanding Saimaa ringed seal terrestrial habitat use, including site fidelity patterns, is essential for the design of conservation measures to ensure adequate protection of terrestrial moulting and resting habitat.
Ringed seals have distinctive and permanent fur patterns, which enables identification of individuals through their lifetime (Koivuniemi et al. 2016). In this study, our primary objectives were to investigate the moulting behaviour of the Saimaa ringed seal using the Photo-ID method, with the focus on estimating the degree of moulting site fidelity. For the first time, we include the entire distribution range of the Saimaa ringed seal to characterize their site fidelity at the population scale. We aim to quantify the movement patterns during the moulting seasons and the spatial and temporal factors affecting them. Finally, we discuss how these findings should be considered in the conservation of the Saimaa ringed seal.

Methods
Lake Saimaa (61° 05´ to 62° 36´ N, 27° 15´ to 30° 00´ E) is the largest lake of Finland extending over approximately 4,400 km 2 . The labyrinthine lake has a mean depth of 12 m and comprises 13,700 islands (Kuusisto 1999) and ten different water basins (Fig. 1). We collected Photo-ID data from 1 April through 30 June, spanning the seals' annual moulting season, between 2016 and 2019. Our Photo-ID team (average of 17 photographers per year) covered the entire distribution range of the Saimaa ringed seal, which includes some 70% of the lake (Niemi et al. 2012) (see Supplementary Table S1 for details). We conducted surveys on outboard motor boats (< 6 m) by photographing the seals using digital single-lens reflex cameras with telephoto lenses (up to 300 mm). GPS coordinates of each seal location were recorded for every sighting. In addition, we set up game camera traps (average of 53 locations per year; Scout Guard 550VB and 560 K-8, UoVision UV785 superb full HD 12 MP) at some moulting sites mainly in Pihlajavesi (PV) water basin ( Fig. 1 area 6). Game cameras were set to motion trigger activation and to take two pictures over a 0.5-2 min time span (see details in Koivuniemi et al. 2016), with the exception of 2019 when they were set to time lapse, two pictures every 10 min.
We stored the image data into a Wildbook (https:// www. wildb ook. org/ doku. php) based catalogue (https:// norpp agall eria. wwf. fi/) managed since 2010. We identified individual seals by visually matching their unique ring-shaped fur patterns to images in the catalogue. We also determined the sex from ventral photographs when available. We identified nursing females by the presence of a pup in photographs taken immediately prior to and during the Photo-ID data collection period previously described. In addition, the data set was supplemented by images, only if associated with precise GPS coordinates, collected by the members of the public, who directly submitted them to the Norppagalleria platform.
For the site fidelity analyses, we only included individuals that had been photographed during at least two of the 4 years. For each individual seal, we merged moulting site locations that were within < 50 m of each other to account for a potential error of the GPS device and observer assessment of seal location. However, if the locations were situated clearly on different islands, different shorelines or bays of islands within this 50 m range, they were treated as distinct sites. For each seal, we calculated the number of years that it was observed, the number of overall sightings and the number of sightings per year. Furthermore, we calculated the number of different moulting sites used per seal as well as the number of sites used within each year. In addition, we determined the total number of seals observed per year, the portion of individuals reusing the same moulting sites over the years and the number of seals sharing moulting sites.
As common measures of site fidelity are difficult to apply in a restricted lake system, we assessed, in this study, site fidelity by calculating the distance between each successive moulting site (within and between years) for each individual seal, using the R package geosphere (Hijmans 2019). We tested the effect of sex (Sex), water basin (Water_Basin), year (Year, between or within) and nursing (Pup, absence or presence) on the distance between successive moulting sites (logarithm of distance, log_Distance) using the R package glmmTMB (Brooks et al. 2017). We accounted for individual variation (and pseudoreplication) in distances by adding a random intercept of seal ID (1|Seal_ID) in the models. We used Akaike's Information Criterion (AIC) values to select the variables that best explained the variation in the distances, based on maximum likelihood estimation (REML = FALSE). Individuals of unknown sex were excluded from the models as including them would not have been biologically meaningful. We ran models with Gaussian, Gamma and Tweedie distributed errors on untransformed data to find the best fitting distribution. In addition, we ran a Gaussian error model on log-transformed distances after converting zero distances (46 cases, where the seal used the same site in successive years, resulting in a distance between successive moulting sites of 0 m) to 10 m to normalize the error residuals. We chose the model with the best fitting error distribution by inspecting the error residual plots drawn with the function 'check_model' from the R package performance (Lüdecke et al. 2020).
Because of the low number of seals observed in other water basins, only the data from the main distribution areas-Haukivesi ( Fig. 1 area 2) and Pihlajavesi ( Fig. 1 area 6)-were included to further investigate the combined influence of sex and water basin (as an interaction of the two) on the distance between moulting sites. We ran a Gaussian error model on the previously described log-transformed distances and verified the fit by inspecting the error residual plots. Finally, the effect of nursing was examined using only females from all water basins.

Results
A total of 337 individual seals were identified from 2164 sightings (36% from game cameras) containing sufficient data for identification. We were able to determine the sex of 102 females and 69 males, but the sex of 166 seals remained undetermined. In total, 192 seals (76 females, 55 males and 61 unknown) were sighted during at least two of the 4 years. As expected, the median number of seals observed per year was the highest in Pihlajavesi (61) and Haukivesi (34) water basins (Table S2), corresponding to the current abundance estimates (Metsähallitus 2020). We identified 788 different moulting sites, the majority of which were located in the Pihlajavesi (44%) and Haukivesi (29%) water basins (Table S3). The total number of sites observed per year was relatively consistent over the study period (2016 = 233; 2017 = 193; 2018 = 181; 2019 = 181). Overall, 53 moulting sites were used by multiple seals, mainly within the same year (87% of the cases). This involves a total of 73 seals (32 females, 24 males and 17 unknown), but does not systematically imply that more than one seal was observed simultaneously at a given site. The median number of moulting sites used by an individual was four (range 1-13), while the median number per year was one (range 1-7). Almost half of the seals (49.5%) used at least one (range 1-3) specific site during more than 1 year.
The median distance between the successive moulting sites used by individual seals (N = 192) was 643 m (range 0-47,550 m) (Table S4, Fig. A2). The median distance within years was shorter (494 m) than the median distance between years (982 m). In the mixed models that accounted for the random effect of the individual seal, the Gaussian distribution (and log-transformed response) had the best fit to the data out of all the candidate models with no observed significant lack of fit in the error residuals. The best GLMM fitted to the data set of successive distances from males and females from all water basins included the factors Water_ Basin, Sex and Year as well as the Sex*Year interaction. Of the fixed effects, Water_Basin had the greatest influence on the successive distance, with medians ranging from 109 m in Kolovesi ( Fig. 1 area 5) to 23,103 m in Haukivesi/Pihlajavesi ( Fig. 1 movements between area 2 and 6) ( Table A1, Fig. 2A). Eight individuals (five females and three males) moved between water basins, resulting in the longest median distances. When omitting these individuals, the maximum median drops to 2,058 m in Joutenvesi ( Fig. 1 area 3). The Sex*Year interaction was also significant; while the distances within years were relatively similar in both sexes (females median: 445 m; males median: 594 m), the distances between years were longer in females than in males, with median distances of 126 m and 842 m, respectively (Fig. 2B).
The best fitting model including only Haukivesi and Pihlajavesi data comprised the factors Water_Basin and Year and their interactions with Sex and accounted for the random effect of the individual seal. Distances in Haukivesi were longer than in Pihlajavesi, with a median of 880 m compared to a median of 473 m (Table A2). Females moved shorter distances between their moulting sites within years compared to between years (median of 416 m and 1,350 m, respectively), while male distances were, in general, more constant (median of 585 m and 626 m, respectively) (Fig. 2C). Both males and females moved longer distances between their moulting sites in Haukivesi (median of 917 m and 829 m, respectively) compared to Pihlajavesi (median of 486 m and 450 m, respectively). However, the Sex*Water_ Basin interaction was not statistically significant.
The best fitting model including only females from all the water basins comprised the factors Water_Basin, Year and Pup. Distances were shorter within years (445 m) compared to between years (1,126 m) (Table A3). Water basins had an effect on the distances between moulting sites, with medians ranging from 104 m in Kolovesi ( Fig. 1 area 5) to 14,676 m in Pihlajavesi/Puruvesi (Fig. 1  movements between area 6 and 7). However, nursing did not have an effect on the distances between moulting sites. Abbreviations separated by a slash correspond to movements between the two water basins. The numbers above each water basin correspond to the number of individuals within each water basin. B Comparison between male and female for the whole Lake Saimaa. The "Between" series represents the dis-tances between successive moulting sites used in different years. The "Within" series represent the distances between successive moulting sites used within the same year. The numbers above each sex categories correspond to the number of individuals within each sex group. C Comparison between male and female for the main distribution area of the Saimaa ringed seal. Abbreviation of the water basins: HV = Haukivesi, PV = Pihlajavesi. The numbers above each sex categories correspond to the number of individuals within each sex group for each water basin. The figure was drawn using the R package ggplot2 (Wickham 2016)

Discussion
Over the study period of 4 years, individual seals used a median of four moulting sites, and nearly half of the individuals reused at least one site. Some seals even used the exact same rocks during several successive years. In addition, the median distance between successive moulting sites was just over half a kilometre. Our study, which incorporates data from the whole distribution area, not only confirms earlier findings on the site fidelity of the Saimaa ringed seal Valtonen et al. 2012;Koivuniemi et al. 2016) but further defines the extreme extent of such behaviour. Moreover, our results bring additional evidence to the site fidelity behaviour of the ringed seal in general (McLaren 1958;Smith and Hammill 1981;Härkönen et al. 2008). Kelly et al. (2010) reported an analogous degree of high site fidelity in Arctic ringed seals, where several individuals returned to the vicinity of their breeding sites (3-54 km) in the following year after moving up to 1,800 km. However, in the case of Saimaa ringed seals, the specificity of their habitat and the scale in which such behaviour is observed suggest extreme fidelity to moulting sites.
It is important to consider that the definition of moulting site fidelity and the scale of home range applied to marine populations (Kelly et al. 2010) may not be applicable in the case of the Saimaa ringed seal due to the restricted yet highly complex lake habitat. In fact, a similar spatial range, which in a more open marine environment would typically represent only one site, may cover five or more different moulting sites at different islands in the case of the Saimaa ringed seal (Fig. 3). Therefore, in this study, we estimate site fidelity by quantifying the distance between moulting sites, which provides a fine-scale measure of fidelity, more suitable for the lake system. We found that the distances between moulting sites were 400 m less in Pihlajavesi than in Haukivesi. This may be attributable to the distinctive landscapes of these two basins, with the labyrinthine archipelago of Pihlajavesi offering a higher density of potential haulout locations, thus underlining the effect of a restricted but complex habitat on reducing the moulting range (i.e., higher site fidelity). Because of the restricted lake system, sense of scale Fig. 3 Example of four Saimaa ringed seals using multiple moulting sites located at different islands within the restricted yet highly complex lake habitat. Individuals Phs125 and Phs149 are males; individu-als Phs164 and Phs174 are females. White areas are land. This map represents a small area of Pihlajavesi (PV) water basin is primordial when studying the behaviour of the Saimaa ringed seal. For instance, the average post-moult home range of Baltic ringed seals (P. h. botnica) is almost twice as big as the surface area of Lake Saimaa (Oksanen et al. 2015), while it is only 90 km 2 for Saimaa ringed seals (Niemi et al. 2012). Our results further emphasize that Saimaa ringed seal behaviour plays out at a fine scale, especially when considering their moulting range. This is accentuated by the fact that nearly half of the seals in this study were re-using the exact same locations throughout the years. We suggest that this extremely fine-scale site fidelity could be the adaptational result from the highly labyrinthine and limited lake habitat.
Our study revealed that the females' distances between successive moulting sites were larger between years (1,126 m) compared to within years (445 m), while no similar effect was observed for males. This could be explained by the reproductive status of females. Females' breeding and moulting areas are not necessarily the same, and females may leave their weaned pups and breeding sites to moult in different locations. Saimaa ringed seal pups tend to stay in their nursing area shortly after weaning (Niemi et al. 2013b), which may induce females to leave the site. Moreover, Reder et al. (2003) reported female dispersion and behavioural changes following lactation in high Arctic harbour seals (Phoca vitulina vitulina). Therefore, we could have observed nursing females in their breeding sites before they moved to their moulting locations, leading to larger distances between years. However, the low number of confirmed reproductive status of females in our study may hinder our ability to make definitive conclusions. Female ringed seals nurse their offspring in snow lairs (Sipilä 1990) that are difficult to monitor before the moulting season. Thus, a substantial number of them might have been misidentified as non-nursing in our study. Besides, females may not give birth every year (Sipilä 2003), which makes the reproductive status designation even more challenging. Finally, we may lack post-weaning observations of identified nursing females and thus underestimate the movements within years that would be expected. Further research combining Photo-ID and genetics, based on placentas for instance, could help to better understand the effect of nursing on female moulting site fidelity.
Our study shows that Saimaa ringed seals exhibit restricted moulting range, as the majority of the seals had their interannual moulting sites situated within a few kilometres. We observed only a few movements across water basins, with only eight individuals undertaking trips of up to 50 km. Such sporadic trans-basin movements of adults (see also Niemi et al. 2012) are an interesting phenomenon, because they are more typical of juveniles (Niemi et al. 2013b). Most of these movements took place over two moulting seasons, which may reflect a recolonisation of previously occupied habitats. One example is a female that travelled approximately 30 km from Pihlajavesi to Puruvesi and subsequently gave birth there in 2018, which was the first pup for that region for several decades (Metsähallitus 2020).
Unlike many other pinnipeds that gather in colonies for breeding or moulting, ringed seals are generally considered solitary (McLaren 1958). Exceptions to this include some low latitude ringed seals from Lake Ladoga (P. h. ladogensis) in Russia (Sipilä et al. 1996;Kunnasranta et al. 1996Kunnasranta et al. , 2001, and to some extent from the Baltic Sea (Härkönen et al. 1998), that are occasionally found hauling out in gregarious groups. Although our study reveals that Saimaa ringed seals are usually solitary, almost 40% of the seals observed for at least 2 years were using the same moulting sites with one or two of their conspecifics, suggesting that some associative behaviour might also exist within Lake Saimaa.
The Photo-ID method, which was used to collect data in this study, was originally implemented for population abundance estimation , as the traditional lair census method (see Sipilä 2003) is hampered by mild winters. However, our study has shown that it has further applications in the research on other aspects of seal ecology and is a well-suited method for a complex and unique environment, such as Lake Saimaa. Telemetry is a powerful tool for the monitoring of individuals at a very fine spatiotemporal scale and has revealed information on the home range and habitat use of Saimaa ringed seals (Hyvärinen et al. 1995;Koskela et al. 2002;Kunnasranta et al. 2002;Niemi et al. 2012Niemi et al. , 2013aNiemi et al. , 2019. However, the applicability of such a method during the moulting season and on a larger scale is financially, technically and logistically challenging. For instance, the tag will detach as the seal renews its fur annually. Thus, the Photo-ID method is the most appropriate tool, up to date, to study seals during the moulting season in their whole distribution area. Seal sightings are also reported year-round by the public through a citizen science programme, opening perspectives to use this technique in a larger time frame. Nonetheless, manual matching of thousands of images is time and labour consuming, and an automatic identification tool that takes variations in seal posture into consideration is urgently needed. Some preliminary developments with automatic matching approaches have already been carried out (Zhelezniakov et al. 2015;Chehrsimin et al. 2017;Nepovinnykh et al. 2020) but are still in progress.
Understanding site fidelity patterns of the Saimaa ringed seal, occurring in both breeding (Valtonen et al. 2012;Niemi et al. 2019) and moulting seasons Koivuniemi et al. 2016; this study), is of paramount importance for the effective implementation of conservation actions. High site fidelity and restricted ranging patterns makes marine mammals more prone to population declines due to local anthropogenic threats, such as habitat degradation and human-caused mortalities (e.g., Rojas-Bracho et al. 2006;Campbell et al. 2008;Atkins et al. 2016). These risks are even more acute for the Saimaa ringed seal because of its restricted lake habitat and high proximity to humans. While it is commonly accepted that ringed seals are most vulnerable to disturbance during the breeding season (Sipilä 2003), we would also like to highlight their sensitivity during the moulting season, which is considered as an energetically demanding period (Paterson et al. 2012). Niemi et al. (2013a) also reported that Saimaa ringed seals rest in the same terrestrial platforms outside of the moulting season, which further highlights their importance as a key habitat. Our study not only emphasizes the high conservation value of the current haul-out sites in the core distribution area (see Fig. 1) but further underlines the need to extend the protection to all suitable habitats and safeguard future recolonisation of previously used areas. Indeed, the effective conservation measures implemented for decades have resulted in the Saimaa ringed seal population growth, which will logically expand its range from the current core areas to more peripheral areas of the lake. However, in the context of climate change, together with low genetic diversity ) and limited suitable lake habitat (Liukkonen et al. 2017), the high site fidelity makes this subspecies especially vulnerable to changes in its environment.
This study, together with earlier findings, shows that Saimaa ringed seals exhibit strong site-fidelity and relatively limited mobility across water basins. Such affiliation to their breeding and moulting areas is already affected by habitat degradation and disturbance resulting from growing human activities. High site fidelity of this subspecies is a major biological factor that should be considered in the conservation strategies that integrate anthropogenic activities, particularly in land use planning, for safeguarding key habitats of the Saimaa ringed seal.

Fig. A1
Photograph of a female Saimaa ringed seal (Phs229) at its moulting site in Pihlajavesi water basin. The highly labyrinthine water basin of Pihlajavesi is home to nearly half of the Saimaa ringed seal population. From mid April to mid June, the seals are moulting on terrestrial platforms, the only period when they are easily observable. Despite the thousands of islands in Lake Saimaa, the seals present restricted moulting range and often reuse the exact same locations over years. Photographer: Vincent Biard    Availability of data and materials The data sets generated and analysed during the current study are not publicly available due to the protection by law of the location of the haulout sites of the Saimaa ringed seal but are available from corresponding authors on reasonable request.

Conflicts of interest
The authors have no conflicts of interest to declare that are relevant to the content of this article. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.