Abstract
The aim of this workpackage is to evaluate apportioning methods identified during WP1 by utilising the datasets with the most potential from WP2 to determine their consistency and assess their strengths/weaknesses. WP1 and WP2 demonstrated, however, that the set of methods that have already been used for apportioning in a UK context is very limited, whilst identifying situations in which the data are available to extend these methods or develop new methods. In consultation with the Project Steering Group, WP3 therefore focussed on developing or extending more advanced methods for apportioning in two key situations in which this was identified, based on WP1 and WP2, to be both feasible and a high priority, and then on evaluating these new or extended methods against existing methods.
In the context of breeding season apportioning, the SNH/NatureScot Apportioning Tool is the only apportioning tool that is currently in use for most species – it is a simple method that is easily implemented but makes strong assumptions and does not utilize all of the available data. GPS summer tracking data were previously used to develop spatial distribution maps (Wakefield et al., 2017) and thereby apportioning estimates (Butler et al., 2020) for four species, providing an alternative to the SNH/NatureScot Tool. Extensive GPS tracking data, although not available for all species, are available for species beyond those considered previously (WP2). Accordingly, we here extend the spatial distribution mapping approach of Wakefield et al. (2017) to include a new species, lesser black-backed gull, in order for these maps to provide an alternative apportioning method to the SNH/NatureScot Tool for this species.
Breeding season apportioning is aided by the strong central place foraging constraint, which necessitates a strong relationship between the spatial distribution of birds and the distance from the colony. Non-breeding season apportioning is more challenging, and the only approach that is currently used in practice for this in the UK is BDMPS, an approach that considers broad spatial regions. We develop an alternative approach, using GLS winter tracking data of guillemot and razorbill (Buckingham et al., 2022) to develop maps that can be used for apportioning. Unlike previous apportioning methods, this approach also provides a quantification of uncertainty – the locational uncertainty in GLS data is potentially large, and we account for this uncertainty explicitly.
We evaluate the new summer GPS-based lesser black-backed gull maps against the NatureScot tool, and the new winter GLS-based maps for guillemot and razorbill against BDMPS. In both cases, the comparison is against the spatial distribution that underpins the methods, not the apportioning percentages i.e., it is an evaluation of the underlying assumptions of the methods rather than a comparison of the actual apportioning results. This is the fundamental analytical progress that is required in order to develop new apportioning estimates, and the most useful, and easily interpretable, way to compare the key differences between the two approaches. As such, we were able to fulfil two main objectives: to broaden estimation of underpinning distributions used in apportioning calculations to new species and seasons, and to compare methods of generating colony-specific distributions underpinning apportioning calculations.
The next work package (WP4) is tasked with developing a tool to calculate apportioning percentages in a suitable format, and will involve developing the code for the apportioning tool based on the methods presented here, including the development of a user interface. As well as the existing methods 6 (SNH/NatureScot and BDMPS), and the two new methods described here, it will also include a simple extension of the NatureScot method, in which the rate of decay with distance is estimated from published foraging ranges rather than being fixed to be equal to minus two. WP4 will also include a comparison of the apportioning values obtained using different methods, including comparisons of methods for all species for which at least two methods are now available – in contrast, the evaluations here are comparing the colony-specific spatial distribution maps underpinning the methods, and have focused specifically on the two situations in which new or extended methods have been developed. The reason the apportioning calculations are undertaken in WP4, and not in this report, is because it has emerged that for each of the two new/extended methods there is an extra, non-modelling, step required in order to carry the modelling through to apportioning, and it has also become clear that a comparison of the final apportioning results can most usefully and defensibly be undertaken once the methods have been implemented within the tool. For the lesser black-backed gull analysis the additional, non-modelling, step that will be required in order to translate the spatial distribution maps described here into apportioning estimates will be an adjustment for proportion of time spent foraging on land: this was not required for the four species considered in Butler et al. (2020), but is required here because of the substantial proportion of time that lesser black-backed gulls can spend foraging on land, and because this proportion can vary substantially between colonies. For the auk GLS analysis, there are areas of the UK coast that do not lie close to any of the tracked colonies used in creating the GLS maps, and in these situations, we propose that the apportioning method will default to using BDMPS, so that the new apportioning method will use a hybrid of GLS-based maps (where feasible and defensible) and BDMPS (in other situations).
The two main sections of this report focussed on the methods and results associated with developing the two main new analyses outlined above – an analysis of summer distributions of lesser black-backed gulls (Section 2) and an analysis of winter distributions of guillemot and razorbill (Section 3) – and on evaluating these against existing approaches (SNH/NatureScot apportioning and BDMPS, respectively). Within each section a more detailed description of how the results of these analyses will be used in apportioning within WP4 is provided.