Tagging and Tracking of Seabirds on the West Coast of Scotland

4. Results

4.1 GPS tag deployment and retrieval rates

The numbers of GPS tags deployed, retrieved, containing usable data, and trips obtained are shown in Table 2 for both species in each of the three study years.

¹ The remaining bird was recaptured, but the tag had detached.

4.2 Impacts of GPS tagging

Within the limits of data availability (see methods) there were no signs of adverse impacts of tagging operations on either species. The size and weights of the tags are within the limits of conventional GPS tags used successfully on these species for several years, using identical attachment methods, without signs of adverse impacts. Over the course of the three years of fieldwork, the overall recapture rates were extremely high (93% for Leach’s and 84% for European Storm-petrel) and all birds appeared fully healthy and had maintained a good weight when recaptured. The inability to recapture several birds in 2024 resulted from field teams having to evacuate St Kilda and Treshnish prematurely. Comparison of nest attendance patterns and nest survival rates of tagged and control birds also indicated no quantifiable impacts of tagging as follows: we found no difference in the duration of foraging trips of tagged Leach’s Storm-petrels compared to the foraging trip directly prior to tagging for the same individual (Wilcoxon signed-rank test N = 7, P = 0.09), though there was a trend towards shorter trips when tagged. This may be due to trips generally becoming shorter as breeding progresses, and a mixed-effects model found that trip duration decreased with date (X²₂ = 22.69 P < 0.0001), but found no additive effect of tagging on trip duration (X²₂ = 0.0039 P=0.95), and no interaction effect of date and tagging (X²₂ = 2.24 P=0.33). For European Storm-petrels, daily nest survival rates were slightly (but not statistically significantly) higher at 23 nests where tagging occurred, compared with 52 control nests (0.991 + 005 SE and 0.986 + 0.0038 SE respectively, X²= 0.5296 P=0.47).

On the basis of the data available, we found no evidence to suggest that the birds’ foraging behaviour or breeding success was affected by the tagging procedures.

4.3 Individual specialisation

4.3.1 Leach’s Storm-petrel

There were similar levels of overlap in locations of core area usage during both incubation and chick-rearing stages among pairs of trips drawn at random from within or between individual Leach’s Storm-petrels, indicating an absence of individual specialisation (D = 0.3741 P=0.35 and D=0.1702 P=0.053 for incubation and chick-rearing stages respectively, Figures 4 and 5). We therefore considered that trips made by the same individual were independent, and calculated summary statistics directly from the raw trip-level data.

4.3.2 European Storm-petrel

For the small number (n=5) of European Storm-petrels tracked during the incubation stage (in 2021 only) only a single foraging trip was obtained, so there was no need, or possibility, to compare within- and between-individual space use. Among chick-rearing birds, we found significantly greater overlap in core area usage among trips by the same individual than among individuals (Figure 6, D = 0.45, P <0.001). We therefore considered that trips made by the same individual were not independent, and derived summary statistics from mixed effects models that included individual identity as a random term.

4.4 Representativeness of colony-level foraging distributions

Bootstrap resampling of the samples of tracked Leach’s and European Storm-petrels indicated that the samples achieved acceptably high representativeness of the wider space use relating to their respective colonies as a whole (representative values of 89.3% and 81.5% respectively, Figures 7 and 8). The sample of tracking data collected by this project can therefore be considered to provide a good representation of space use by Leach’s Storm-petrels breeding on Hirta, St Kilda and European Storm-petrels breeding on Lunga, Treshnish Isles.

4.5 Trip metrics and marine distributions

Summary trip statistics for both Leach’s and European Storm-petrel are presented in Table 3. For Leach’s Storm-petrel only the overall mean and standard deviation figures are derived from model estimates because within each breeding stage category there was no evidence of individual specialisation. As such individual trips can be regarded as independent. For Leach’s Storm-petrel it was necessary to exclude trips 02 - 07 for nest 44 because these were associated with a failed breeding attempt and therefore not considered to be representative. Five further trips where incubating birds had departed the colony very soon after arriving were also excluded because these were thought to be linked to failed re-capture attempts and therefore the birds were unlikely to be exhibiting natural behaviour.

For European Storm-petrel the trip analysis suggests potential for foraging specialisation within individuals. As such, to avoid potential bias due to the presence of multiple trips for some individuals at the chick rearing stage, the mean and standard deviation values presented for trip duration, total distance and maximum range are derived from the output of a linear fixed effects model with individual ID used as a random term to account for the variation in numbers of trips between individuals. This was not necessary for trips for incubating European Storm-petrels because we don’t have any cases of multiple trips being recorded by the same individual and can therefore be considered independent.

For European Storm-petrel, only 2021 data were used to test of the effect of breeding stage on trip metrics because no incubating birds were tracked in 2023 or 2024. As such it was necessary to produce separate models to understand the effect of year and breeding stage. Summary statistics for individual trips of Leach’s Storm-petrels are presented in Table 4 - 9 and mapped in Figures 9 - 17. Summary statistics for individual trips of European Storm-petrels are presented in Tables 11 - 13 and mapped in Figures 24 - 28.

4.5.1 Leach’s Storm-petrel trip metrics

Mean trip duration across all birds and years was 52.2 ± 20.2 hours. This was found to differ significantly between breeding stages (Kruskal-Wallis X²₂ = 26.95, df = 1, P <0.001) with incubating birds spending longer at sea (66.5 ± 19.9 hours) than chick rearing birds (43.8 ± 15.1 hours). A difference between years was also confirmed (Kruskal-Wallis X²₂ = 9.16, df = 2, P = 0.002) with birds spending more time at sea in 2021 (62.0 ± 14.6 hours) compared to 2023 (49.0 ± 17.9 hours) or 2024 (51.5 ± 24.4 hours). This indicates that foraging behaviour was more similar in 2023 and 2024 than in 2021.

The mean maximum range (distance travelled from the colony) across all birds was 281.2 ± 103.6 km. This varied significantly with breeding stage (Kruskal-Wallis X²₂ = 23.03, df = 1, P < 0.001). Incubating birds travelled further from the colony (347.2 ± 106.7 km) compared to birds attending chicks (243.7 ± 81.4km). The effect of year was not significant (Kruskal-Wallis X²₂ = 1.42, df = 2, P = 0.49).

Total trip distance averaged across all birds tracked from St Kilda was 848.4 ± 332.8 km. As with maximum distance and total distance incubating birds (1084.9 ± 333.9 km) travelled significantly further than those attending chicks (708.5 ± 242.5 km) (Kruskal Wallis X²₂ = 27.10, df = 1, P=0.002). No significant difference between years was found (Kruskal Wallis X²₂ = 3.62, df = 2, P=0.16).

Figures 18 - 20 present the movements of all Leach’s Storm-petrels tracked from St Kilda in 2021, 2023 and 2024.

4.5.1.1 Leach’s Storm-petrel Utilisation Distributions

The utilisation distributions (UDs) of Leach’s Storm-petrels tracked from Hirta (Figure 21) highlighted the core foraging and home range of the tracked birds. These UDs were calculated for complete and ‘almost complete’ trips only and exclude any trips associated with possible disturbance at the nest (n = 5) or failed breeding attempts (n = 6). The core usage range (50% KDE) includes an area to the North of St Kilda (approximately centred on -59°N 8°W,) and an area to the West of the colony likely reflecting predominant commuting routes to and from the colony (Figure 21A). As expected, birds are more likely to be present close to the colony at night than during the day (Figure 21B and C). The utilisation distributions appear to be broadly consistent between years (Figure 21D - F). This agrees with the analysis of the trip statistics in which no significant effect of year on maximum distance from the colony or total travel distance was found. In contrast to this, incubating birds appeared to target different areas (Figure 21G) compared with those attending chicks (Figure 21H).

4.5.1.2 Leach’s Storm-petrel foraging habitat selection

The top-selected model for Leach’s Storm-petrel breeding on Hirta, St Kilda included non-linear smoothers for log-transformed water depth and SST (Table 10). The non-linear smoother for SST also incorporated a two-way interaction with time of day (day vs. night). Additionally, linear terms for seabed slope and surface chlorophyll concentration were included in the model.

Of these retained covariates, water depth was highly correlated with both distance from the colony and distance from the coast. However, models including either distance from the colony or distance from the coast showed errors related to model convergence, which led to our preference for selecting models that included log-transformed water depth instead. Habitat usage of Leach’s Storm-petrels declined as water depth increased from relatively shallow depths of approximately 25 m to 150 m. Beyond this point, habitat usage began to increase from 200 m to 1000 m, with a slight peak in usage around 1000 m (Figure 22).

During the day, we found that habitat usage was positively associated with SST, with a slight peak in usage occurring when SST was in the range of ~12.5°C to 13°C (Figure 23). In contrast, at night, there was little evidence that SST influenced habitat usage, though there was slight evidence of reduced usage of areas with higher SSTs (> 14.75°C, Figure 23). Additionally, model coefficients suggested that seabed slope was negatively associated with habitat usage, while surface chlorophyll was positively associated with habitat usage (Table 10).

Table 10: Results for gam SSFs for European and Leach’s Storm-petrel. Tables shows outputs for best performing model (on the basis of model AIC scores) for each species. Covariates were included as either on-linear smooths (s) or as parametric coefficients (β). When fitted as parametric coefficients covariates were standardized prior to modelling. Parametric coefficients are reported together with associated 95% confidence intervals (CI). For non-linear smooths we report associated χ² and p values. N = 4173 observations for European Storm-petrel; N = 6105 observations for Leach’s Storm-petrel. Each observation was paired with N = 50 available points.

European Storm-petrel

Leach’s Storm-petrel

4.5.2 European Storm-petrel trip metrics

The modelled mean trip duration across all birds and years was 36.9 ± 17.3 hours. As with total trip distance and maximum range, this was found to differ significantly between breeding stages in 2021 (GLMM Estimate = 35.9, SE = 8.0, P = 0.001). Incubating birds spent significantly longer at sea (61.3 ± 24.9 hours) than those provisioning for chicks (26.1 ± 19.6 hours). However this result is derived from only 3 complete trips for incubating birds and as such should be taken as indicative only. There were small, but significant differences among years in the duration of trips of chick-rearing birds: trip durations were shorter in 2021 (26.1 ± 20.0 hours) than 2023 (38.8 ± 4.0 hours, GLMM Estimate = 12.7, SE = 5.8, P = 0.039) and 2024 (40.8 ± 5.2 hours, Estimate = 14.7, SE = 6.8, P = 0.038).

The modelled mean maximum range (accounting for multiple non-independent trips for some individuals) across all trips was 116.6 ± 53.0 km. In 2021, when birds were tracked both during both breeding stages, birds travelled significantly further from the colony during the incubation stage (153.9 ± 25.1 km) compared to birds attending chicks (89.6 ± 69.9km, GLMM Estimate = 50.1, SE = 15.9, P = 0.007). There was a significant difference in the foraging range of chick-rearing birds in 2021 (89.7 ± 69.9km) compared with 2023 (149.3 ± 73.6 km, GLMM Estimate = 59.6, SE = 25.7 P = 0.033).

The modelled mean total trip distance of birds tracked from Treshnish was 418.6 ± 178.0 km. In 2021 birds travelled significantly further during the incubation stage than those attending chicks (602.2 ± 253.1 and 303.3 ± 204.7 km respectively, GLMM Estimate = 291.2, SE = 92.3, P = 0.007). However, this should be interpreted with caution as only three complete trips were obtained from incubating birds in 2021 compared to 18 complete trips for birds attending chicks. There were significant differences among years in the total distance travelled on trips by chick-rearing birds: in 2023 (481.5 ± 222.1 km) the total was considerably greater (GLMM Estimate = 178.2, SE = 77.54, P = 0.035) than in 2021 (303.3 ± 211.1 km) and 2024 (441.4 ± 426.5 km).

Table 11: Summary of complete trips by individual for European Storm-petrels tracked during the 2021 field season. There were two further partial trips, for which trip metrics could not be reliably estimated.

Incubating Birds

Birds Attending Chicks

Figures 29 - 31 present the movements of all European Storm-petrels tracked from Lunga, Treshnish Isles in 2021, 2023 and 2024

4.5.2.1 European Storm-Petrel Utilisation Distributions

The utilisation distributions (UDs) of European Storm-petrel in Figure 32 highlight the core foraging and home range of the tracked birds. The 25% UD highlights the importance of an area to the north of Coll and Tiree, particularly during the daytime (Figure 32B). As expected, the night-time UD was concentrated around the colony (Figure 32C). The UDs for 2021 (Figure 32D) and 2024 (Figure 32F) were broadly similar with however in 2023 (Figure 32E) the core range (50% UD) was located further to the south and the 2023 home range (95% UD) extended further north and south of the colony than in 2021 and 2024. This reflects the significantly larger variation in maximum distance from the colony observed in 2023 (57.4 km – 483.5 km) compared to the other years (Table 3). In 2021, a small number (n = 5) of birds were tracked during the incubation stage of the breeding cycle. The difference in UDs supports the results of the trip analysis which indicated a within-year difference between the foraging strategy of incubating birds (Figure 32G) and those who were attending chicks (Figure 32H).

4.5.2.2 European Storm-petrel foraging habitat selection

Based on the forward stepwise model selection approach, the final model for European storm-petrels on Lunga incorporated non-linear smoothers for log-transformed distance from the colony, seabed slope, log-transformed NPP, and the chlorophyll gradient (Table 10). Among these covariates, we also included an interaction between distance from the colony and the time of observation (day vs. night).

We found a clear negative association between European Storm-petrel habitat use and log-transformed distance from the colony during daylight hours, with usage decreasing sharply, particularly after birds were approximately 50 km from the colony (Figure 33). During the night, habitat use also tended to decline with increasing distance from the colony; however, there was evidence of a slight peak in habitat use at approximately 100 km from the colony (Figure 33). It is important to note that the ratio of daytime to night-time observations was 70/30, which introduces a higher level of uncertainty in estimates of behaviour during the night.

We found a positive relationship between seabed slope and habitat usage, with habitat use increasing as seabed slope rose up to a value of approximately 1.5. Beyond this point, the relationship between seabed slope and habitat usage levelled off (Figure 34). For context, the 10%, 50%, and 90% quantiles for seabed slope across the observed and available locations in the European Storm-petrel dataset were 0.12, 0.45, and 1.52, respectively. This suggests that birds tended to use areas with steeper slopes within their observed foraging range.

Habitat usage was highest in areas characterized by low levels of NPP, then levelled off for NPP values between 10 and 25 mg/m³/day (Figure 35). Beyond this point, there was weak evidence suggesting a further decline in habitat usage as NPP increased. However, the estimated 95% confidence intervals were quite wide for higher NPP values which were rarely observed in the dataset. Additionally, habitat usage also declined as the magnitude of chlorophyll gradient increased (Figure 36).

4.6 Overlap with Offshore Wind Farm lease areas

4.6.1 Leach’s Storm-petrel

Leach’s Storm-petrels tracked from the breeding colony on St Kilda were observed to consistently use an area to the NNW of St Kilda, remote from current offshore windfarm development areas. Therefore, it is not expected that OWFs included in the current leasing round will impact birds from this colony. The data provide a useful base line for potential future assessments of the impact of marine activities such as shipping and other human activities.

4.6.2 European Storm-petrel

Across all years a total of 19 out of 38 tracked European Storm-petrels accessed leased OWF areas (Figures 37 - 39). The OWF sites accessed were: Machair Wind (Machair Wind Ltd), Talisk Offshore Wind Project (Magnora Offshore Wind N3), Malin (Malin Sea Wind Limited), Lease Area 18 - Northern Ireland (No further Information available) and the Clogerhead OWF lease area off the east coast of the Republic of Ireland. For the 19 birds accessing OWF areas, the proportion of time per trip at sea within OWF sites ranged from 0.03 – 35.2%. Across all tracked birds this equates to a mean of 2.1 ± 5.7% of time at sea occurring within OWF areas. The OWF area leased to Machair Wind Ltd located to the south-west of Mull falls within the core range (50% KDE) of birds from Lunga (Figure 32). Eighteen of the 19 birds which interacted with OWFs passed through this area Figures 37 – 39). The other bird which interacted with an OWF area passed through the Talisk Offshore Wind Project area near North Rona. One of the birds tracked in 2023 which travelled south to the Irish Sea passed through an OWF area in Irish waters in addition to the Machair Wind site.

4.7 GLS tracking

Device effects

Three of 20 Leach’s Storm-petrels that had been fitted with GLS tags in 2021 were recaptured in 2023, compared with 11/20 control birds recaptured. Two of the GLS-tagged birds had suffered leg injuries that resulted in the loss of the ring bearing the tag. The other individual retained the GLS tag, which was removed. Close inspection of its leg revealed no sign of injury. One of the control birds recaptured in 2023 had also suffered a similar leg injury, on the unringed leg, which had occurred since it was ringed in 2021.

It is highly likely the injuries observed, and low recapture rates, were caused by the GLS tag deployment, possibly exacerbated by the inability to retrieve tags in 2022. In light of these negative impacts, no further GLS tag deployments were made on Leach’s Storm-petrels in 2023.

None of the 20 European Storm-petrels fitted with GLS tags in 2021 were retrapped, in 2023 compared with 7 of 22 controls re-encountered in 2023 (including one bird found freshly depredated by a Short-eared Owl Asio flammeus). It is highly likely that the low recapture rates were caused by the tag deployment, possibly exacerbated by the inability to retrieve tags in 2022. No further GLS tag deployments were made on European Storm-petrels in 2023.

In light of these low recapture rates, and likely impacts of ring-mounted GLS tags, Mark Bolton consulted widely with other researchers using the same, or similar, methods to fit GLS tags to a range of Storm-petrel species elsewhere in Europe (Norway, Spain, Iceland, Azores) and Canada. As a result, it became apparent that whilst some studies had been successful, achieving high return rates and no signs of injury, others have encountered similar problems of low (or zero) recapture rates and birds with leg injuries. The group of researchers agreed to supply data on ring size and design, fitting method, and recapture rates of tagged birds and controls. The information was compiled into a database, comprising 57 studies, which indicated that the likely cause of the problems encountered by some workers were due to small differences in ring size, leading to ring fit being slightly too tight on occasion. Agreement has been reached on best-practice ring design, size and fitting method, and the need for adequate controls in future GLS tagging studies, to assess potential impacts. It is expected that as a result of these changes, it will be possible to resume GLS tagging of Storm-petrels in UK.

Movements of Leach’s Storm-petrel outside the breeding season

Data obtained from the single GLS tag retrieved from a Leach’s Storm-petrel in 2023, tagged during incubation in July 2021, indicate that it continued to undertake several further incubation shifts (shown by whole days of light shading of the tag as the bird remained in its nest burrow, Figure 40) and commenced migration in mid-September (shown by the abrupt divergence of recorded light intensities from the local light cycle), suggesting it bred successfully. The locations show a southward migration to the west of Ireland, rather than an alternative route through the Irish Sea. Thereafter it travelled to waters off Mauritania (October) and Benguela (December) followed by a trans-Atlantic westward movement to waters of the central Atlantic between January and March (Figure 41). The tag battery unfortunately did not last for the whole deployment, meaning that no data was collected on the bird’s return migration.

4.8 HPAI impact assessment

Eight blood samples from Leach’s Storm-petrels and one from European Storm-petrels were adequate for analysis (by the Animal and Plant Health Agency), and all tested negative for antibodies to Avian Influenza virus.