Work concept and design

We designed a study with three complementary phases. The first was observational, under controlled conditions. We observed the behaviour of birds in relation to six of the most commonly used strain insulator configurations in Spain. We included an analysis of where and how frequently they perched (on which strain insulator and element). The second phase consisted of developing a risk index for evaluating the perchings, based on the combination of the frequency of use of each part of each insulator strain and the hazard of this perching. The third and last phase involved the development of a procedure to evaluate the least risky strain insulators, using the previously obtained risk indices.

To develop the observational phase, we decided to observe perching behaviour in controlled conditions using flying pens. We selected six wildlife rescue centres (WRCs hereafter) in central Spain. At each of these centres, one or more large enough flying pens were used to develop the perch trials (Figure 1). We defined five a priori groups of raptors based on species size (partially after Soldatini et al. Reference Soldatini, Albores-Barajas, Lovato, Andreon, Torricelli, Montemaggiori, Corsa and Georgalas2011 and García, Reference García2013, Reference García2017; Table 1) and aimed to test at least 30 individuals in each group.

Figure 1. General set up of a perching trial in GREFA. An Aquila chrysaetos perching on the metal extension of a 3 U40 strain insulator

Table 1. A priori groups of raptors and species included. Shown is the number of specimens involved in the perch trials (N), the centres where they were perch trialed, the number of complete days for the perch trials (Days), the number of perchings recorded (Perches), (MN), the semi-metacarpal length in centimetres (SL₁), the standard deviation for semi-metacarpal length (SD₁), the semi-wingspan length (SL₂) and the standard deviation for semi-wingspan length (SD₂). If no specimens were measured for a complete species, we assigned it the class average (in bold). *There are two measured specimens for the semi-metacarpal length and four for the wingspan

To build the risk index and the procedure, we used approach of Allan (Reference Allan2006). We defined a hazard and calculated the frequency of use. First, we assigned a hazard to each perch in each element of each insulator strain by each species that varied according to the bird’s dimensions, and the distance to both ends of the strain insulator (Bedrosian et al. Reference Bedrosian, Carlisle, Woodbridge, Dunk, Wallace, Dwyer, Harness, Mojica, Williams and Jones2020). In a second step, we considered the frequency of each kind of perching, compared to the group overall.

To develop a procedure for evaluating the risk indices, we established criteria considering both the electrocution risk index and this index particularised for the most electrocution-prone and endangered group of species in Spain, the great eagles (Aquila genus; Madroño et al. Reference Madroño, González and Atienza2004, Guil et al. Reference Guil, Colomer, Moreno-Opo and Margalida2015). Electrocution is a crucial factor especially for the most threatened great eagles, Spanish Imperial Eagle Aquila adalberti and Bonelli’s Eagle A. fasciata (Real et al. Reference Real, Grande, Mañosa and Sánchez-Zapata2001, González et al. Reference González, Margalida, Manosa, Sánchez, Oria, Molina, Aranda and Prada2007).

Observational phase: Birds and measurements

Birds used in this study were from different WRCs. All of the birds were adults, with the exception of four Spanish Imperial Eagles in their first year. All the birds were considered healthy and fully capable of flight. As electrocution risk is related to bird size (Janss Reference Janss2000), the hazard was designed to consider the differences in raptor dimensions. Two different measures were considered (partially after Dwyer et al. Reference Dwyer, Kratz, Harness and Little2015): spine to metacarpal (hereafter semi-metacarpal) and spine to the 9^th primary remex (hereafter semi-wingspan). Measures from the spine were appraised because the aim of the index was to analyse the risk when the raptor perched on two different elements independently.

Raptors were measured with a flexible tape. We measured only one wing, usually the right wing. We placed each bird on its back, fully extended the wings, and recorded (1) the distance between the location where the flesh ended and the spine (semi-metacarpal) and (2) the distance between the spine and the 9^th primary remex end (semi-wingspan). Measurement results are shown in Table 1.

Observational phase: The perch trials

The trial consisted of the introduction of at least one flying raptor to a flying pen with six different strain insulator configurations for at least two days (48 hours). We counted the number of days per perch trial; initial and final day were only considered as half-days. All flying pens were ample enough to guarantee safe raptor flight considering the minimum standards for wild raptor rehabilitation (Miller Reference Miller2012). The use of different flying pens in a WRC depended on the presence of unrecoverable established raptors: we generally used one or two previously empty flying pens, but the trials were also developed in flying pens where there were unrecoverable established raptors, to minimise disturbances to these specimens. In these cases, we only considered the flying raptors as part of the trials.

Strain insulator configurations were installed on Y-shaped metallic poles except in two flying pens in Fundación para la Investigación en Etología y Biodiversidad (FIEB hereafter) and in “Chaparrillo 1” where the strain insulators were fixed directly on the posts with a V-shaped metallic structure (see Figures 1 and 2). Each model of strain insulator configuration was randomly located in the flying pen and was not repeated within the different flying pens of the same WRC. And all models were available simultaneously. To assess perching behaviour, we used camera traps: at least two BTC-5HDE Browning Trail Cameras in each flying pen, using TimeLapse+ mode, so the camera shot in 10-min intervals and with each detected movement (for more details see Appendix S1 in the online supplementary material).

Figure 2. Strain insulator configurations assayed in this study

Perching selection was based on the number of perches each bird made in each different part of a different strain insulator VS the total number of perches of the bird, in each trial. When there were no distinguishable birds in the same flying pen, we followed a protocol to assign those perchings (Appendix S1). To assess the perching of the birds, we considered that each animal perched at least one time in the flying pen. Perching was assigned to the ground if the bird was not detected on an insulator. We also considered alternative perches (such as trees) as ground perchings.

Observational phase: The strain insulator configurations

We tested the following strain insulator configurations, representing the more frequently used standard models that meet the requirements of Spanish legislation. All these models are different designs fulfilling the same function. And all of these strain insulators are used in electric power distribution lines (up to 66 kV):

• • CAON-C3670. Anti-perching composite insulator model C3670EBAV_AR (total length of 123,5 cm)

• • 7 U40 glass. 7 glass insulators type U40 (total length of 100,3 cm)

• • Composite-CS40. Composite insulator type CS40 and metal extension with anti-perching sheet APA 16-590 (total length of 131,8 cm)

• • 3 U40 glass. 3 glass insulators type U40 and metal extension with anti-perching sheet APA 16-590 (total length of 117,8 cm)

• • PECA-1000. Anti-perching composite insulator model PECA-1000 (total length of 135,5 cm)

• • 2 U70 glass. 2 glass insulators type U70 and metal extension with anti-perching sheet APA 16-590 (total length of 109,4cm)

Risk index phase

As a first step, for building hazard, we characterised the risk of each perching as a combination of the bird size and the exact location on the strain insulator. We defined five levels of hazard (after Allan Reference Allan2006; see Figure 3 and a full diagram in Figure S1):

• • Extreme hazard: if the bird perched on an element where the distance between each extreme of the strain insulator is smaller than double the semi-metacarpal length.

• • Very high hazard: if the bird perched on an element where the distance between the centre of the strain insulator to one extreme of the insulator is smaller than the semi-metacarpal length and the distance to the other extreme of the insulator is smaller than the semi-wingspan.

• • High hazard: if the bird perched on an element where the distance between each extreme of the strain insulator is smaller than double the semi-wingspan length or if the bird perched on an element where the distance between one extreme is smaller than the semi-metacarpal length and the other is smaller than the semi-wingspan+15 cm (according to INSHT 2011).

• • Moderate hazard: if the bird perched on an element where the distance between each extreme of the strain insulator is smaller than double the semi-wingspan length+15 cm.

• • Low hazard: the rest of the cases.

We estimated the distances between the perching site and the extremes of the strain insulator. For most of the perches, we considered the centre of the element, with three exceptions:

• • Insulators: We distinguished this in thirds, one proximal (nearer to the strain clamp), one central and one distal. In each case, we considered the bird to be perched on the centre of each third. The exception is the insulator composed of two U70 glass insulators, for which we considered each insulator independently.

• • Wire: If the bird perched on the wire, we considered it to be perched on the strain clamp.

• • If the bird perched on two elements simultaneously, each length was calculated independently.

For calculating the frequency of use (FoU hereafter), we considered the number of perchings for a group on a certain insulator and hazard level and divided it with the global number of perchings of the group. This is:

$$ FoU=\frac{x_{ij}}{\sum_{i=1}^n{\sum}_{j=1}^n{x}_{ij}} $$

Where:

• • FoU is the frequency of use of a certain insulator and a hazard level

• • x_ij is the number of perchings of a certain group in the insulator j with hazard level i

Figure 3. Extreme hazard of a vulture on a CAON-C3670 strain insulator. MC is used for metacarpal position. Original vulture picture from Eprdox at Wikimedia Commons.

Statistical analysis

We analysed different factors that influenced the use of each strain insulator, to assess how results were conditioned by the experimental design. We considered the number of perchings on each strain insulator, per species and perch trial as the dependent variable, and as independent variables:

• • Type of WRC: whether or not there were alternative perches

• • WRC: nested in the type of WRC, we considered:

  • ◦ With alternative perches: Chaparrillo, CRAS, and Sierra de Fuentes

  • ◦ Without alternative perches: CERI, FIEB, and GREFA

• • Group, with five levels: vultures, great eagles, eagle owls, medium raptors, and small raptors

• • Species, nested in group

• • Strain insulator, with seven levels: one per insulator and the ground

• • Number of days each specimen spent in each perch trial (cofactor)

• • The total number of different images taken, multiplied by the number of birds of the same species in each perch trial (offset)

We used a negative binomial generalised linear model, with a log-link function. We also calculated deviance using the modEvA package (Barbosa et al. Reference Barbosa, Real, Muñoz and Brown2013) in R 3.4. (R Core Team 2017).

Procedure for prioritising strain insulator use

For general prioritising, we conceived a step-by-step procedure for prioritising the use of the different strain insulators. This is necessary because there are variations in the risk index between different raptor species. When a power pole is retrofitted, prioritisation of strain insulators should consider all raptor species, but attend to conservation priorities. First, we prioritised the absence of extreme hazard perchings, following Allan (Reference Allan2006). We then considered great eagles as the most sensitive group, both due to their susceptibility (Guil et al. Reference Guil, Colomer, Moreno-Opo and Margalida2015) and their conservation status in Spain (Madroño et al. Reference Madroño, González and Atienza2004. Thus, we developed the following procedure:

1. 1. Absence of extreme hazard perchings for the target group; in our case, the great eagles

2. 2. Lower FoU for very high hazard and for the target group (great eagles), if differences for a couple of strain insulators in step 1 are less than 5%

3. 3. Lower FoU for extreme hazard for all the groups, if differences in step 2 are less than 5%

4. 4. Lower FoU for very high final hazard for all the groups, if differences in step 3 are less than 5%

While these criteria can be changed for other purposes, when prioritising per species or group we conceived a similar procedure but with the specific data of each species or group:

1. 1. Absence of extreme hazard perchings (for the species or group).

2. 2. Lower FoU for very high hazard (for the species or group).

3. 3. Lower FoU overall for all the groups.