Hotspots in the grid: Avian sensitivity and vulnerability to collision risk from energy infrastructure interactions in Europe and North Africa

sion birds for which sufficient GPS tracking data at high spatial resolution were available. We also map vulnerability hotspots where mitigation at existing EI should be prioritised to reduce collision risks. As tracking data availability improves our method could be applied to more species and areas to help reduce bird- EI conflicts.


| INTRODUC TI ON
The transition to zero carbon energy is essential to avoid runaway climate change (IPCC, 2018). However, the expansion of renewable energy infrastructure (EI) required to achieve this poses a challenge to wildlife conservation due to collision and electrocution risks, particularly for birds and other aerial taxa (Bernardino et al., 2018a;Kiesecker et al., 2019;Marques et al., 2014). European, onshore wind energy capacity is projected to grow from approximately 169 GW in 2018 to between 262 GW and 760 GW by 2050 with enough economically viable wind turbine locations (approximately 3.4 million) for up to 13.4 TW of capacity (Ryberg et al., 2019). Countries in the Middle East and North Africa also have targets to increase the share of electricity supply from onshore wind with Morocco and Tunisia aiming for 100% renewable electricity by 2050 (Timmerberg et al., 2019). Huge investment in the electricity transmission network will accompany this expansion of renewables, with an estimated fivefold increase in transmission capacity required between and 2050(Mckinsey & Company, 2010. However, when poorly designed or situated, wind farms and power lines can result in increased mortality of susceptible birds such as large water birds, gulls, ibis, storks, owls, vultures and other raptors (Janss, 2000;Oppel et al., 2021;Thaxter et al., 2017).
Organisations, such as energy companies, charged with supporting the rollout of renewable energy generation are obliged by national, European legal (2009( /147/EC, 2010 and pan-flyway voluntary (Horns & Şekercioğlu, 2018) frameworks to mitigate risks to birds (Gyimesi & Prinsen, 2015). Methods to evaluate and mitigate these impacts are relatively well understood at project-specific and local scales Serrano et al., 2020). However, such assessments often occur after a development site has already been selected because the initial feasibility studies for energy projects tend to focus on the economic viability of the development over other factors. The scale and pace of new development requires greater integration of high-level assessments of the potential cumulative impact at regional and flyway scales into these feasibility studies to highlight areas where additional EI development is likely to significantly increase the risk to bird populations (Eichhorn et al., 2017;Loss et al., 2019;Thaxter et al., 2019). This is particularly important for migratory bird species who may experience the impact of multiple developments in operation within key migration routes, stopover sites, wintering grounds and breeding sites (Bernardino et al., 2018a;Gove et al., 2013).
Bird sensitivity maps can be developed to illustrate the relative risk associated with EI development for sensitive bird species (Vasilakis et al., 2016;Warwick-Evans et al., 2017). The distribution and behaviour of birds inferred from GPS tracking of individuals can be used to create a spatio-temporal measure of the potential impact of new EI developments, by identifying where and when birds would be most exposed to potential collision risks from EI developments (Ross-Smith et al., 2016;Thaxter et al., 2019;Warwick-Evans et al., 2017). For areas with sufficient tracking data, combining sensitivity maps with other inputs, such as the available wind resources, can help planners optimise new wind farm and power line locations by avoiding high sensitivity areas during the site selection stage of the development process (Kiesecker et al., 2019). This, in turn, can reduce mitigation costs and produce better wildlife outcomes compared with site-based assessments alone (Bradbury et al., 2014;Bright et al., 2008).
Sensitivity mapping is particularly useful for assessing the potential for negative interactions between birds and energy infrastructure at the level of migratory flyways. For example, a wind farm sensitivity map created for the Red Sea flyway estimates the potential collision risks for soaring migratory birds at the flyway scale (BirdLife International, 2015). This tool enables preliminary impact assessment of wind farms by viewing protected areas and raw GPS tracks of susceptible bird species. However, it does not account for all dimensions related with collision risk, such as the height at which birds fly, which in turn may vary depending on landscape, meteorological, seasonal and species-specific factors (Kleyheeg-hartman et al., 2018;Marques et al., 2020). In the terrestrial context, other sensitivity mapping studies largely rely on trait-based analysis in relation to population densities of susceptible bird species (D'Amico et al., 2019;Thaxter et al., 2017).
In this context sensitivity is a measure of potential collision risk identifying areas where the tracked birds could collide if wind turbines or powerlines are present (Thaxter et al., 2019). We calculated this by combining susceptibility traits with GPS location and altitude data for individuals from 27 species, including resident and migratory birds in Europe and Northern Africa, to describe where and when the tracked birds are most sensitive to collision risks from terrestrial EI. Within the areas for which we obtained sufficient high spatial resolution GPS tracking data; this allows us to identify sensitivity hotspots where future onshore EI development should be discouraged. However, our work cannot reveal 'safe' areas where EI development could be encouraged. We then overlay this sensitivity surface onto the density of existing EI to identify vulnerability hotspots where the tracked individuals are most exposed to collision risks due to the presence of wind turbines and powerlines. Similar approaches using GPS tracking data have been applied to assess the impacts of proposed offshore windfarm developments where survey logistics are more challenging (Bradbury et al., 2014;Cleasby et al., 2015;Lees et al., 2016;Ross-Smith et al., 2016;Thaxter et al., 2019). Our work also highlights the spatial variation in GPS tagging effort and data availability which helps identify priority areas for future tracking studies and the need to increase data sharing via online platforms such as Movebank to help fill in the gaps in the existing tracking data where sensitivity assessment is not currently feasible using publicly available GPS tracking data.

| Data acquisition
An overview diagram of the methods is presented in Supporting Information S1, section 1. We sourced bird movement data via the Movebank data repository, a web-based online platform for sharing data from animal tracking studies (Movebank, 2019) used to assess whether the species in these GPS tracking studies were susceptible to mortality associated with EI. This literature search is summarised in Supporting Information S1, section 2. We did not request data from tracking studies with less than five individuals unless multiple individuals of the same species were tracked in other Movebank studies. Data managers were contacted between October 2018 and January 2019 to request access to their datasets with a response deadline of the end of April 2019. Studies using ARGOS Doppler tags (insufficient spatial accuracy; Thomson et al., 2017), captive birds, laboratory-based tests of GPS devices, lacking altitude data or tracking predominantly pelagic species were not included. In total, we obtained permission to use data from 65 suitable GPS tracking studies (Figure 1), representing 27 species and 1,454 individual birds. This included some data hosted on the University of Amsterdam Bird-tracking system database (UvA-BiTS, , offered for inclusion in this analysis by managers of some of the requested Movebank datasets. To our knowledge, all fieldwork associated with the movement datasets included in this study was undertaken with permission from the relevant licensing authority, further details of each dataset are provided in the Data References section of this paper and Supporting Information 1, section 3. The earliest tag deployment within any of the datasets was 2006, while the latest deployment date is 2018; the mean deployment duration was 2.7 ± 1.8 SD years. Infrastructure and terrain data were processed in QGIS and ArcGIS (ESRI, 2019;QGIS Development Team, 2019). We sourced transmission power line data from the open infrastructure project (Garret, 2018;OpenStreetMap, 2018). OpenStreetMap defines transmission lines as >50 kV (OpenStreetMap, 2019a). Flights between 10 and 60 m aboveground were here taken as being within the danger height for transmission power lines ( Figure 2) where birds risk collision (Harker, 2018; F I G U R E 1 (a) The location of the first GPS location of each dataset included in the analysis and the flux of individual birds through each 5 × 5 km grid cell (not controlled for year) for all areas for which we could source GPS tracking data. (b) The density of GPS locations in flight per 5 × 5 km grid cell for all GPS tracking studies included in this study Infante & Peris, 2003). We intersected the data with a fishnet grid consisting of 554,993 individual 5 × 5 km grid cells representing a total area of 13.9 million square kilometres. Power line density is the total length in kilometres per grid cell, normalised onto a 0-100 scale.
Data on the location and size of onshore wind farms were downloaded from windp ower.net (Pierrot, 2018). This dataset contained centroid coordinates and information for 18,681 wind farms within Europe and North Africa. We used this to map the relative density of turbines on a 0-100 scale for each 5 × 5 km grid cell. Density was highest towards the North and West of Europe. From the hub heights and blade lengths in this dataset, we derived a danger height (sometimes known as the rotor swept zone) of 15-135 m for wind turbines (Figure 2), further details in Supporting Information S1 section 4. To our knowledge, all tracking datasets used were collected in line with relevant guidance and licensing requirements of national ethical committees.

| Movement data analysis
Measures of GPS accuracy were not uniformly indicated across all studies. Where the number of satellites was provided, only GPS locations associated with ≥5 visible satellites were included (Morris & Conner, 2017). Duplicate GPS locations were removed. We used the raster package (R version 4.0.5, Hijmans, 2019) to append elevation data from two 30 m horizontal and 5 m vertical accuracy digital surface models (DSMs), STRM-GL1 and STRM-GL1-Ellipsoidal from the OpenTopography Portal (National Science Foundation, 2019;NGA and NASA, 2000) to each GPS location. For the small number of GPS locations at latitudes greater than 60° latitude an ALOS 30 m DSM was used instead (JAXA, 2016). Further details of these DSM surfaces are in Supporting Information S1 section 5. Height above-ground in metres was calculated by subtracting the elevation of the ground from the altitude of the bird. Where bird altitude is in height above ellipsoid, ellipsoidal height of the land surface is used. Where altitude is relative to sea level, orthometric height of the land surface is used (Péron et al., 2020). In some datasets, such as some Lesser Black-backed Gull studies, bespoke correction to obtain orthometric height had already been estimated in the database (Thaxter et al., 2019). GPS locations for each study were classified as breeding or non-breeding (including the migratory period) season by plotting week against latitude, (Supporting Information S1 section 6).
Because instantaneous speed was not available across all datasets we estimated speed in metres per second (m/s) using the time and distance between subsequent GPS locations derived with the anytime and geosphere r packages (Hijmans et al., 2015;Eddelbuettel, 2018).
All GPS locations within the 95% confidence interval for heights relative to ground level and greater than 10 m above-ground or associated with speeds greater than or equal to 1.39 m/s (~5 km/hr) were classified as in flight. This approach accounts for the vertical error given by many GPS devices and excludes locations where the bird is likely to be stationary on the ground. The vertical position error associated with GPS tracking devices is typically in the range of 1.5 m but can be as large as 31 m due to the combined error of the GPS device and the DSM surface (Marques et al., 2020). We categorised each flying GPS location as within each danger height band or not ( Figure 2). Some datasets contained bursts of high frequency GPS measurements (up to 1hz). Because the heights recorded in these bursts are not likely to be independent, the data were filtered to remove this potential source of bias by ensuring a minimum of 1 min between subsequent GPS locations resulting in a total sample size of 18.0 million GPS locations. We then summarised the proportion of GPS locations in flight (6.6 million) observed at each danger height within each grid cell for each species in the dataset.
Due to the nature of data obtained from tracking studies, the distribution of studies and individuals was heterogeneous across the study region (Figure 1), hence we considered cells with more tracking data to have more reliable estimates of the proportion of GPS locations in flight (Péron et al., 2017;Silva et al., 2017). We accounted for this at the species level using the Wilson's score (Reichensdörfer et al., 2017) whereby the lower bound of the Wilson confidence interval (WCI), calculated using the binconf function in the hmisc r package (Harrell, 2018), was used in place of the percentage (Lewis & Sauro, 2006;Lott & Reiter, 2020). Compared to a raw proportion, at low sample sizes, this has the effect of reducing the value assigned to grid cells where uncertainty is higher. For example, if a grid cell contained only three GPS locations, the Wilson score (WS) tended towards zero due to the large WCI around the central point estimate of the proportion of GPS locations at danger height, as sample size increased (n > 50) and the WCI converged towards zero, the WS became comparable to a percentage (Cao, 2018). See further information in Supporting Information S1, section 7.
We weighed this proportion of flying GPS locations to account for the collision susceptibility of different species using a morphobehavioural risk index (MBRI) based on the method utilised in D'Amico F I G U R E 2 Danger height band definitions for energy infrastructure within which birds could be vulnerable to collision. The majority of transmission power lines (66 kV and over) range from 10 to 60 m in height (National Grid, 2014). For current onshore wind turbines, we derive a rotor swept zone ranging from 15 to 135 m above-ground (Pierrot, 2018;Thaxter et al., 2019) et al., 2019 and morphology data provided by Storchová & Hořák, 2018. Wing area and aspect in relation to weight is an important factor in avoidance ability as species with higher wing load are less able to take evasive action (Bevanger, 1998;Janss, 2000;May, 2015). Because wing area values were not available for all species, we simplified the shape of a bird as a rhombus and calculated a simplified area using the wingspan (WS) and body length (BL) in metres using data from Svensson et al., 2016or Storchová & Hořák, 2018. Comparing this with wing area data available for 17 of the 27 species (Hedenström & Strandberg, 1993) using linear regression (R 2 = 0.61, F 1,15 = 23.44, p < 0.001) suggests that it is a good proxy for assessing relative differences between species and an improvement on using wingspan alone (R 2 = 0.46, F 1,15 = 14.38, p < 0.001), further details are provided in Supporting Information S1, section 8. We then estimated a wing load proxy by dividing this area whereas we use flight style to help infer manoeuvrability (May, 2015).
To account for the impact of mortality on the population of each species, this MBRI was then combined with European conservation status (Least Concern = 1, Other categories = 2) to produce a morphobehavioural risk conservation status index (MBRCI) as per Equation 2: MBRCI was then normalised onto a scale between zero and one by calculating the ratio between the MBRCI for each species and the maximum value across all species. MBRCI for each species is detailed in the table in Supporting Information 1, section 8. Sensitivity at the species level for each grid cell was then calculated as the proportion of tracking locations at danger height (quantified by the Wilson Score WS) multiplied by the MBRCI to produce a value between 0 and 1. The final sensitivity across all species is then defined as the maximum sensitivity of any species present in each grid cell. For example, if two species were present and species A was associated with a sensitivity score of 0.2 and species B was associated with a score of 0.4 the sensitivity for species B would be used for that grid cell. Alternative approaches using the raw proportion of flight locations at danger height, the Wilson score proportion or weighting the Wilson score proportion by conservation status did not alter our conclusions significantly and are provided in (Supporting Information 1, section 9).

| Vulnerability to collision for GPS tracked birds
Vulnerability is a measure of how exposed individuals are to the presence of EI in horizontal and vertical space and how sensitive they are to the collision risks posed by this infrastructure (Thaxter et al., 2019). We calculated vulnerability associated with existing infrastructure, for each grid cell, by multiplying the relative density of each EI type (0-100 scale) by the sensitivity value at the relevant height band for each 5x5km grid cell at the species level resulting in a value between 0 and 100. As per Equation 3: A score of zero indicates that either no EI is present or sensitivity is zero whereas vulnerability of 100 would require relative density of EI to equal 100 and sensitivity to equal 1. Combined vulnerability is the sum of vulnerability for each height band. The final vulnerability across all species for each infrastructure is then defined as the maximum value of any species present in each grid cell.

| Defining sensitivity and vulnerability categories
To ensure classification of sensitivity and risk was driven by the data (Gouhier & Pillai, 2020), we defined categories using the 25th, 75th and 97.5th percentiles for all grid cells where the sensitivity (or vulnerability) for a given height band was greater than zero. Grid cells scoring greater than zero but less than the 25th percentile are 'Low', scores between the 25th and 75th percentile are 'Moderate', scores greater than the 75th percentile are 'High' and cells in the top 2.5% of observations are 'Very High'. All other cells are classified as 'Very Low' if there are GPS locations but none at danger height resulting in a score of zero or 'No Data' if data were lacking. We emphasise, therefore, that our method can only identify areas where a high risk of EI exists, but that the absence of a highvulnerability score in our analysis cannot be interpreted as indicative of low impact of EI due to the potential for other bird or bat populations (for which no data were available in our study) to be affected. (2) MBRCI = CI * (WBMR * FS * BV * FL * ND) 5 .
In total, 99,641 of the 554,993 5x5km grid cells in the study area (18%) contained at least one GPS location in flight. Sensitivity to wind turbines was greater than zero in 54.9% (n = 54,703) of these grid cells (Figure 3a). 13.57% (n = 13,516, 337,900 km 2 ) of these cells were classified as high sensitivity, that is, they were in the upper quartile of sensitivity scores (>0.11). There was significant variability in sensitivity between species (ANOVA F 26,59,592 = 432.4, p < 0.001) with Eurasian eagle owl Bubo bubo, whooper swan Cygnus cygnus, eurasian spoonbill Platalea leucorodia, common crane Grus grus and white-fronted goose Anser albifrons exhibiting the greatest sensitivity to wind turbines across the grid cells where data are available for these species (Table 1). Sensitivity to transmission power lines (10-60 m height band) was greater than zero in 37.64% (n = 37,509) of grid cells (Figure 3b). Across Europe and North Africa 9.41% (n = 9375, 234,375 km 2 ) of these cells are classified as high sensitivity, that is, they are in the upper quartile of sensitivity scores (>0.14). Eurasian spoonbill Platalea leucorodia, European eagle owl Bubo bubo, whooper swan Cygnus cygnus, Iberian imperial eagle Aquila adalberti and white stork Ciconia ciconia are the five species which exhibited the greatest sensitivity at the transmission power line danger height band (Table 1).

F I G U R E 3 (a)
Year-round sensitivity to wind turbines across all species (n = 27) and areas for which we could obtain suitable GPS tracking data, (b) year-round sensitivity to transmission power lines across all species using GPS tracking data (n = 27) and areas for which we could obtain suitable GPS tracking data. Sensitivity at the species level for each grid cell was then calculated as the proportion of tracking locations at danger height (quantified by the Wilson score WS) multiplied by the MBRCI to produce a value between 0 and 1. The final sensitivity across all species is then defined as the maximum sensitivity of any species present in each grid cell. Maps for breeding and non-breeding seasons are provided in Supporting Information 2. Basemap from (OpenStreetMap, 2019b) TA B L E 1 Sensitivity and vulnerability across all seasons and grid cells for which sufficient GPS data were obtained: Summarised by species and infrastructure type, sorted according to species name in alphabetical order. Vulnerability hotspots are defined as the upper quartile of the vulnerability scores obtained separately for vulnerability to collision with wind turbines and power lines. Mean combined vulnerability across all grid cells with sufficient data for that species is also described here

| Vulnerability of tracked birds to energy infrastructure risks
We plotted the combined vulnerability score as the sum of vulnerability from wind turbines and power lines present in each grid cell There were marked differences in vulnerability between species, with mean combined vulnerability ranging from 0.042 ± 0.081 SD for western marsh harrier Circus pygargus to 2.14 ± 3.62 SD for white stork (Table 1).

| DISCUSS ION
For areas with sufficient tracking data (currently 18% of the study area), our sensitivity surface identifies sensitivity hotspots associ- Our results also highlighted differences in sensitivity to EI between species and which type of EI poses the most risk to each species (Table 1). It is beyond the scope of this study to provide specific ecological explanations for this observed variation as this is an ongoing topic of research in of itself, however, this is likely a product of ecological and morphological factors such as flight style (flapping vs. soaring), migratory behaviour, habitat preference and how foraging strategy influences flight heights relative to the danger height bands (Bernardino et al., 2018a;Martin & Shaw, 2010;Thaxter et al., 2017).
Despite efforts to obtain as complete coverage of the study region as possible, we acknowledge gaps were present in the available GPS tracking data, particularly within areas such as northern France, northern Spain, Scandinavia, Algeria and Libya. These gaps reflect geographical and seasonal variation in the availability of bird telemetry data (Bouten et al., 2013). As such, our results successfully highlight where sensitivity and vulnerability to collision with EI occurs, but cannot indicate where vulnerability does not occur. Our sample includes only a subset of the most susceptible species, most of which are larger birds with a body mass of 350 g or more, and only a subset of populations of these species, leading to sampling-related bias which is most evident during the breeding season (Supporting Information 2, section 4). These sampling-related biases are a common issue in ecology, and collision risk cannot be inferred for areas where information is not available (Brotons et al., 2004). Despite these limitations, the approach used, based on existing tracking data, accounting for species susceptibility to collision and the proportion of GPS records at danger height, provides a simple way to assess risk in the areas where data are available. As more tracking data become available, this analysis can be updated using the code provided in supplementary material 3 and data from (Movebank, 2019).
This study highlights the benefits of data sharing and we expect data availability to increase significantly in the near future as GPS telemetry becomes more affordable and miniaturisation enables tracking devices to be fitted to smaller bird species (Bouten et al., 2013).
Advances in sensor technology may also soon allow collision mortality to be detected in real-time (O'Donoghue & Rutz, 2016). One priority to aid future research is to help fill these gaps by improving data sharing via platforms such as Movebank or UvA-BiTS, promotion of new bird tracking studies in under-represented areas and taxonomic groups, improved standardisation of biologging datasets and deployment of loggers outwith the breeding season (Sequeira et al., 2021). Other methods to address these data gaps may include the use of GPS data to model the relationship between flight heights and spatio-temporal factors such as weather, time of year, topography and land cover. However, such an analysis is beyond the scope of this paper.
Overlaying sensitivity with the existing wind farms and transmission lines identified a number of vulnerability hotspots where the tracked birds are vulnerable to collision with EI ( Figure 4). While it is beyond the scope of this paper to evaluate the effectiveness of different mitigation options, we suggest that for areas with sufficient GPS tracking data, the vulnerability map can help identify priority areas for mitigation of impacts of EI, to reduce risks to birds. For existing power lines this could include line marking to increase visibility, burying cables or altering routes to avoid high sensitivity areas (Jenkis et al., 2010). For wind turbines, options include repowering with fewer larger turbines (Arnett & May, 2016), marking blades (May et al., 2020), temporary shutdown periods during the peak of the migratory season (de Lucas et al., 2012) which is already a requirement in some countries such as Jordan (Tomé et al., 2017). Another option is to retrofit radar or camera-based systems to monitor bird movements and automatically shut down turbines during periods of high migratory movement (McClure et al., 2018). Future analyses could be improved if official, multi-country, energy network spatial datasets were composed and made available to researchers. This would enable consideration of lower voltage distribution power lines which are under-represented in open-source data and are associated with electrocution, which was not considered in this study, as a major cause of injury and mortality (Garret, 2018;Hernández-Lambraño et al., 2018). In a European study with Northern Bald Ibises 45% of the losses were caused by electrocution (Fritz et al., 2019). As with collision, there are several options such as retrofitting insulators or perches to reduce electrocution risk (Dixon et al., 2019), but the problem could be entirely avoided by constructing safe poles that eliminate electrocution risk in the first place (AEWA, 2012).

| IMPLI C ATI ON S FOR MANAG EMENT AND CONS ERVATION
To our knowledge, this is the first time that an assessment of this kind has been undertaken at the flyway scale across multiple species.
Our methodology provides a readily transferable approach to assess sensitivity and vulnerability for other species and areas as more GPS tracking data become available. The results presented here do not preclude the need for detailed local environmental impact assessment of the potential ecological impacts of EI on birds and other wildlife combined with post-construction monitoring to assess the risks due to disturbance, habitat loss, electrocution as well as collision which was the focus of this paper Gove et al., 2013). However, for areas with sufficient GPS tracking data, our sensitivity maps can inform where new wind farms and power lines should not be constructed and help include consideration of these impacts early in the site selection process for developments.
Moreover, the vulnerability maps can help more effectively target areas for surveys to identify specific locations where mitigation of existing wind farms and power lines should be implemented. In our race to tackle the climate crisis, it is vital that we do not neglect the biodiversity crisis (Vasilakis et al., 2016), sensitivity and vulnerability maps derived from GPS tracking data will be an important tool to help protect wildlife as our energy system transitions to zero carbon.

ACK N OWLED G EM ENTS
We thank all those who have granted us permission to use the data in our analysis and those who assisted in data collection. A full summary of datasets used in the analysis is available in Supporting Information S1, section 3. We are very grateful to all the people who helped to collect these data in the field.

CO N FLI C T O F I NTE R E S T
All authors declare that they have no conflict of interest with the content of this work.

AUTH O R S ' CO NTR I B UTI O N S
J.G.G. is the principal author of the work; A.M.A.F. suggested the initial idea for the paper and provided advice throughout the analysis and writing stages of the work along with P.W.A., J.P.S. and P.R.; All authors contributed critically to data collection, drafting the manuscript and gave final approval. Efforts were made to foster collaboration with individuals from across the entire study region. Our study brought together authors and dataset managers from across Europe and the Middle East; however, there is a lack of representation of authors from African Nations. While this is in part a by-product of the current distribution of available GPS bird telemetry datasets it does highlight a need for greater collaboration and engagement with researchers from across the afro-palaearctic flyways.

DATA AVA I L A B I L I T Y S TAT E M E N T
The raw bird tracking datasets associated with this analysis are available for download on request via https://www.moveb ank.org/cms/ moveb ank-main (Movebank, 2019), many of these datasets relate to sensitive or protected species and therefore permission from the managers of these datasets may be required prior to download. The processed datasets used to produce the sensitivity and vulnerability maps in this paper, including shapefiles of the final sensitivity surfaces and the density of energy infrastructure, are available via the Dryad Digital Repository: https://doi.org/10.5061/dryad.jm63x sjcw (Gauld et al., 2022).