Scottish Marine and Freshwater Science Volume 6 Number 10: At-Sea Turnover of Breeding Seabirds - Final Report to Marine Scotland Science

5. Discussion

5.1. Summary of Results

We have found evidence for four key general patterns in relation to turnover, which hold across a range of species and footprints:

(1) turnover decreases as site fidelity increases;

(2) turnover decreases as the spatial scale of site fidelity becomes finer;

(3) turnover is typically much higher for "commuting" behaviour than for "foraging" or "resting at sea" behaviours;

(4) variation in turnover between surveys (as represented by the 95% confidence intervals shown in Section 4) is general very substantial. Variation appears to be much lower for puffin than for the remaining three species, but it is unclear if this is a real effect or a statistical artefact that results from the sparseness of available GPS data on puffin.

These four key broad trends are in line with the patterns that we would have expected to find, but there are also more subtle differences between individual species and footprints.

In general, kittiwake and razorbill had higher levels of turnover than did guillemots or puffins. This is true for both foraging and resting at sea. For all wind farm footprints, kittiwake and razorbill had estimates of turnover between approximately 100 and 150 with a site fidelity level of zero, in comparison to guillemot and puffin that had estimates between approximately 60 and 100. The differences in turnover levels among species may in part be due to differences in their foraging ranges and behavioural activity patterns.

Kittiwakes have larger foraging ranges compared to the other three species meaning a lower proportion of their total population is likely to be present in any given survey location within their foraging range, resulting in higher estimates of turnover compared to species with a smaller foraging range. The remaining three species have similar foraging ranges but differ in their time activity budgets.

Puffins spend less time at the colony than the other species because they are a burrow nesting species, which may in part explain the lower estimates for turnover in this species. Birds spending a greater proportion of their overall time budget attending a nest at the colony will have higher turnover estimates because they will be less likely to be observed at sea during a particular snapshot survey.

Finally, the higher overall estimate of turnover for razorbills compared to guillemots is likely due to differences in the proportion of time each species spends in commuting flight and foraging. The empirical time activity budgets used in this project (from Thaxter et al. 2013) show that razorbills spend relatively more time flying (approximately two hours flight per day) than do guillemots (approximately one hour flight per day). Razorbills also spend comparatively less time foraging (approximately four hours foraging per day) than do guillemots (approximately five hours foraging per day). These combined activity patterns mean that, for any snapshot survey, razorbills are more likely to be engaged in flight than are guillemots. Birds in flight will be less likely to be observed in a particular time window for a specific survey area because they only spend a short amount of time passing through each grid cell. Therefore, birds tending to spend a greater proportion of their time commuting are less likely to be observed within a particular snapshot survey and so resulting estimates of turnover will be greater.

Within a species, there was variation in estimates of turnover between wind farm footprints. Guillemots displayed the lowest variation in turnover estimates between the different footprints, with estimates of turnover for all wind farm footprints ranging between 70 and 90 when site fidelity was zero, for both foraging and resting at sea. Razorbills also exhibited relatively low variation in turnover estimates between wind farm footprints. When site fidelity was zero, estimates of turnover for foraging razorbills ranged between approximately 110 and 140 for all wind farm footprints, and estimates of turnover for razorbills resting at sea ranged from 80 to 120. Kittiwakes displayed a similar pattern in relation to variation amongst wind farm footprints as seen for razorbills, although overall turnover estimates for kittiwake were slightly higher than those estimated for razorbills, for both foraging and resting at sea. These differences in estimates of turnover between windfarms for each species are likely due to a combination of the relative foraging densities within each wind farm footprint (determining the likelihood of birds being present during a snapshot survey), and the relative location of the wind farm footprint in relation to all the colonies (determining in part the density of foraging birds, but also the number of birds commuting over the area to and from each colony).

Puffins had the lowest overall estimates of turnover for both foraging and resting at sea of the four species, but they did have noticeably higher estimates of turnover for foraging birds at the NnG and Bravo wind farm footprints in comparison to the other wind farm footprints. This may be due to the more patchy foraging distribution of puffins in relation to the other three species, although this patchiness may be an arbitrary artefact of the small sample size in the GPS tracking dataset for this species.

5.2. Assumptions

The results that we have obtained depend on a number of key assumptions, which are discussed below.

5.2.1. Definitions

Most importantly, the results presented in this document are entirely contingent upon the definitions of turnover and site fidelity that we have used. The definition of turnover refers to the number of birds using a particular area ( e.g. a wind farm footprint) during the entire breeding season, relative to the number of birds using it during an idealised snapshot census survey in which the entire population is recorded at a specific point in time. The idealised snapshot census assumes that all birds in the survey area in the survey time window are observed - we make this assumption not because it is true in practice (real survey data do not provide a snapshot census), but because it allows us to disentangle the effects of turnover from those of other features of survey data (non-detection, spatial sampling).

We define site fidelity relative to birds making an independent selection of sites on different days, not relative to a completely random selection of sites. In our simulations, all birds - regardless of site fidelity, have a tendency to return to areas of high bird density (estimated from the GPS tracking data for each species). This means that some birds will select the same foraging site multiple times during the breeding season even in the absence of site fidelity. Therefore, in this project, a site fidelity value of zero does not necessarily mean birds never return to the same foraging site over the breeding season. It is important to bear this in mind when looking at the results.

5.2.2. Scenarios

The results are also explicitly contingent upon the levels and spatial scales of site fidelity that have been assumed. As such, we have presented results under a range of different scenarios for these values to show how estimates of turnover change when the level of site fidelity is increased, or when the spatial scale over which site fidelity is assumed to operate increases. Estimating 'true' site fidelity values, or spatial scales of fidelity, is beyond scope of this work, but could be possible for some species where GPS deployments have been of sufficient duration to obtain multiple trips per individual. Of the species considered here, this is likely to be possible in the auk species, where a proportion of deployments are of sufficient duration, though may be more challenging in kittiwakes where deployments are typically shorter and, in the majority of cases, only one to two foraging trips are recorded per bird. We have implemented site fidelity using a particular methodology ( Appendix A), and while we believe this to be a reasonable approximation of the way in which foraging seabirds display site fidelity, the lack of precise empirical data for these processes means simulation output cannot be validated against observed patterns for these species.

5.2.3. Data Quality

The methodology assumes that the estimated bird density maps that we are using (and which were constructed through statistical modelling of GPS tracking data) provide an accurate representation of the spatial distribution of foraging birds. Our results suggest that bird density is not a key factor influencing turnover, evidenced by obtaining similar turnover estimates for the different wind farm footprints even though they have substantially different bird densities. As such the results presented here may be fairly robust to the failure of this assumption.

The method also assumes the empirical activity budgets we have used are an accurate representation of the behaviour of each species. These budgets were derived from a subset of the population from each colony from a small number of years or, in one case (puffins), a single year, corresponding to an assumption that activity budgets from this subset of the population represent those for the whole population. Turnover estimates are directly related to the percentage of time spent on each activity, so if time budgets are systematically wrong this will affect the turnover estimates for each behaviour.

5.2.4. Other Assumptions

The methodology that we have used to derive turnover values also depends upon a number of other specific assumptions:

1. Our method assumes there is complete site fidelity within a day. In contrast, if individual birds went to different foraging locations within the same day the estimate of turnover would be different. If foraging locations selected by an individual are far apart (for instance some are in the footprint and some are not) we might expect this effect to be more important than if they are close together, but in general it is difficult to know how important this assumption is likely to be. In addition, this assumption constrains the estimates of turnover to behave in a certain way in relation to the time activity budget. It forces the estimates of turnover to change in a particular direction (to decrease) as the time birds spend performing activities at sea increases. This is because by assuming birds only forage in one location per day the numerator of the turnover equation (number of birds that perform behaviour B within area A at any point during the entire breeding season) is constrained to not change in relation to the proportion of the four behavioural categories in the time activity budget, whereas the denominator (Mean across time-points of the number of birds that perform behaviour B visiting area A at each time-point t) will necessarily get larger as birds spend more time performing activities out at sea. As a consequence, estimates of turnover have to decrease as birds spend proportionately more time commuting, foraging or resting out at sea. The same directional relationship between time spent at sea and turnover would result from a simulation where birds selected multiple foraging locations each day (the more time birds spend away from the colony the more likely they are to be counted in an at-sea survey), but it is likely it would be less strong than that resulting from the formulation used in this project.

2. The method also assumes site fidelity is defined in terms of grid cells (used to estimate bird densities), not points or foraging patches. The results may be somewhat sensitive to the precise resolution and alignment of the grid used, but we expect that this effect is likely to be small.

3. Our implementation of site fidelity does not depend on foraging success, meaning that whether or not a bird is successful at a foraging location has no influence on their subsequent fidelity to that location. We suspect that this is not a critical assumption because we calculate turnover as a population-level quantity that will tend to average out individual-level effects.

4. The method also assumes that the order at which foraging sites are returned to is random ( i.e. they are just as likely to return to the same foraging location on the following day as compared to any other day in the breeding season. We do not believe this assumption will have a significant effect on estimates of turnover because such effects are likely to be averaged out in the calculation of turnover at the population level.

5. Finally, we assume that birds fly in a straight line from the colony to the foraging locations.

5.3. Conclusions

The turnover values that we have presented could, in principle, provide a basis for scaling the abundance estimates of breeding individuals obtained during bird surveys of a particular area (such as a wind farm footprint) up to estimates of the number of breeding birds that are using that area during the entire breeding season. There are three key reasons why considerable caution needs to be taken in trying to do this, however:

1. The results that we have presented are contingent upon particular scenarios regarding the level and spatial scale of site fidelity. They provide a guide to assess how the level of turnover changes with site fidelity behaviours and patterns, and with the spatial scale of wind farm footprints, but they cannot provide specific estimates of turnover until further data on both the level and spatial scale of site fidelity of these species become available.

2. The literature review we conducted highlighted the considerable variability in seabird foraging ranges and foraging trip characteristics both within and between species, and within and between years. These parameters, in any one population and in any one year, will be influenced by food availability and distribution as well as stage of the breeding cycle. In addition, foraging behaviour in some species can be affected directly by human activities (gannets, for example, are known to follow fishing boats and feed on discards). Similarly, activity budgets and foraging site fidelity are likely to be affected by factors such as environmental conditions, predictability of prey distribution and population density-dependence, and can, therefore, vary among colonies and years. The variation in all these parameters may translate into among-population and inter-annual differences in turnover of individuals at sea that must be considered when assessing the potential impacts of offshore renewable energy developments on breeding seabirds.

3. The turnover values that we show here represent the value that a "snapshot" census of the complete population of birds within the footprint at a particular instant in time would need to be scaled up by in order to gain an estimate for the total population of birds that use the footprint at any point during the breeding season. In reality, current methods for surveying seabirds cannot achieve a complete census of all birds within an area the size of most wind farm footprints. At-sea surveys will, therefore, generally be a sample, rather than a complete census, and will typically take place over a longer time period rather than at an instantaneous snapshot. In order to scale actual survey data ( e.g. at-sea surveys) up to the total population it is, therefore, also necessary to use statistical adjustments to account for factors other than turnover: non-detection, for example, and the spatial up-scaling involved in translating transect counts (or other sample counts) up to an estimate of the total population within the area (Thomas et al. 2010). In addition, at sea survey estimates cannot distinguish between breeding and non-breeding individuals, nor assign birds to specific colonies. An additional step is required to adjust the at sea estimate by the proportion of non-breeding birds and to assign remaining birds to the appropriate colony or population of interest.

This project provides estimates of turnover for four species in the Forth-Tay region. However, the turnover estimates are contingent upon assumptions regarding the level and spatial scale of site fidelity. Empirically estimating site fidelity from tracking data was beyond the scope of the work within this project. Furthermore, available tracking data are generally from short-term deployments that would likely constrain our ability to reliably estimate these site fidelity parameters. In future work it may be possible to estimate these parameters for guillemots and razorbills, where birds may carry a logger for several days. However, deployment durations are very short for kittiwakes (12-36 hours, typically), and puffins are limited by the small sample size of GPS data sets. Analysis of appropriate existing data and detailed tracking studies involving long-term deployments would be required to better understand how foraging site fidelity operates in each species, and how it may be influenced by environmental conditions and seasonality. This more detailed understanding would allow for a more realistic capture of site fidelity processes within models, thereby facilitating a more accurate depiction of how turnover varies between species and survey methods.

This project has nonetheless provided an important first step in quantifying turnover in relation to wind farm footprints, and in understanding the ecological factors that influence turnover. It provides a basis for identifying knowledge gaps that will benefit from further data collection, and the results contribute to informing assessments of the potential impacts of development projects. The project, therefore, has significant strategic relevance for site characterisation and monitoring in Scotland and beyond. Turnover is clearly only one factor that will need to be considered when assessing the risks to seabird populations from offshore developments. A related task will involve quantifying the fate of birds that lie within the development footprint, and it is important to note that these two questions cannot meaningfully be considered in isolation because they are fundamentally linked: higher levels of turnover imply that larger number of birds are using an area during the breeding season, but also imply that the impact of a development on any individual bird is likely to be lower (because the bird is present in the area for less time than if turnover were low). There will be a potentially complicated trade-off between these two processes, and further work is needed in order to understand the precise nature of this trade-off - i.e. to understand whether higher levels of turnover lead, all else being equal, to higher or lower estimates of development-related mortality.