Effect of tower base painting on willow ptarmigan collision rates with wind turbines

Abstract Birds colliding with turbine rotor blades is a well‐known negative consequence of wind‐power plants. However, there has been far less attention to the risk of birds colliding with the turbine towers, and how to mitigate this risk. Based on data from the Smøla wind‐power plant in Central Norway, it seems highly likely that willow ptarmigan (the only gallinaceous species found on the island) is prone to collide with turbine towers. By employing a BACI‐approach, we tested if painting the lower parts of turbine towers black would reduce the collision risk. Overall, there was a 48% reduction in the number of recorded ptarmigan carcasses per search at painted turbines relative to neighboring control (unpainted) ones, with significant variation both within and between years. Using contrast painting to the turbine towers resulted in significantly reduced number of ptarmigan carcasses found, emphasizing the effectiveness of such a relatively simple mitigation measure.

the other hand, there are few reports of birds colliding with turbine towers.
Studies of grouse in relation to wind-power plants are addressing both mortality due to collision and displacement due to disturbance (e.g., Hovick, Elmore, Dahlgren, Fuhlendorf, & Engle, 2014;Pruett, Patten, & Wolfe, 2009;Winder, Gregory, McNew, & Sandercock, 2015;Winder et al., 2014a;Winder et al., 2014b;Zeiler & Grunschachner-Berger, 2009). Considering collisions, carcasses of willow ptarmigan (Lagopus lagopus) at the Smøla windpower plant are often found only a few meters from the tower base, showing signs of direct impact with a "wall" rather than cuts and fractures usual for hits by turbine blades (Figure 1). In one case, fresh blood smear and feathers was also observed on the tower base where a fresh ptarmigan carcass was found (Bevanger et al., 2010). Galliformes typically fly relatively low above ground; 97% (138 of 142 flights recorded) of willow ptarmigan that were flushed on Smøla, Norway showed a flight height lower than 15 m (Pedersen, 2017). Both data from autopsy and flight height indicate that grouse are more prone to collide with the turbine tower bases than the rotor blades. In support of this, several black grouse (Tetrao tetrix) that were found immediately under turbines in an area in Styria, Austria, presumably died because of collision with tower bases and not the rotor blades, even though the cause of death was never observed directly (Zeiler & Grunschachner-Berger, 2009). Corpses were analyzed by veterinarians, who concluded that injuries were consistent with the birds flying into a hard surface. Furthermore, collision between a willow ptarmigan and the tower base has been confirmed by actual observation once in Sweden (Falkdalen, Lindahl, & Nygård, 2013;Pedersen, 2017), and twice in Scotland (Coppes et al., 2020). In Sweden, it was observed that one individual, part of a group of 10 birds, crashed directly into the tower base at 2.7 m height above ground (25 September 2011, at 07:05 a.m.). The rest of the group passed the tower on both sides. At the time, there was no precipitation or wind, but overcast weather (Falkdalen et al., 2013).
Willow ptarmigan has, due to low population size, been protected from hunting at Smøla from 2005 (Farstad pers. comm.). It is a popular small game species, frequently hunted in Fennoscandia.
Due to reduction in population size in most of this species' circumpolar distribution, several restrictions have been introduced on hunting in later years to reduce hunting mortality (Pedersen & Karlsen, 2007;Pedersen & Storaas, 2013;Sandercock, Nilsen, Brøseth, & Pedersen, 2011). The number of planned wind-power plants is rapidly increasing, not only in coastal areas in Norway, but also in alpine and subalpine areas in Scandinavia. Therefore, any additional negative effects on willow ptarmigan associated with wind turbines are important to assess.
Documenting mortality due to collisions with wind turbines is clearly important, but the next step would be to find solutions to reduce the risk of collisions. It is pivotal to understand why and how birds are killed in order to adopt proper mitigation measures (de Lucas, Ferrer, Bechard, & Muñoz, 2012;Martin, 2011;Wang et al., 2015), and then use this information to select the most efficient tools to reduce bird mortality (Dai, Bergot, Liang, Xiang, & Huang, 2015;Marques et al., 2014;May, Reitan, Bevanger, Lorentsen, & Nygård, 2015).
Norway's largest wind-power plant at the time was constructed in the period 2002-2005 on the northwestern part of Smøla, an island off the coast of central Norway, consisting of 20 2.0 MW and 48 2.3 MW wind turbines distributed within an 18 km 2 area (Bevanger et al., 2010;Follestad, Flagstad, Nygård, Reitan, & Schulze, 2007;May, Nygård, Dahl, & Bevanger, 2013). After having first documented the effects of wind turbines on birds within the BirdWind research project -2011Bevanger et al., 2016), the main aim of the INTACT (INnovative Tools to reduce Avian Collisions with wind Turbines) project (2013-2017) was to advance one step and experimentally test various mitigation measures at the Smøla windpower plant. One of these measures was painting of tower bases to increase the contrast of the turbine base against the background, thus making them more visible and easier to avoid for low-flying ptarmigan (May, 2017).

| Study location
Smøla consists of a large main island together with about 5,500 smaller islands, islets, and skerries and is located off the coast of Møre and Romsdal County, central Norway (63°24′N, 08°00′E).
The terrain is flat with the highest point only 64 m above sea level.
Habitats are dominated by moors of heather (Calluna vulgaris), marshlands and low rocky outcrops (May et al., 2013).

| Search regime
Searches for dead birds below turbines started in August 2006. Dogs trained to find carcasses and remains (e.g., feathers) of birds were used, as this has been shown to increase search efficiency (Paula et al., 2011). An area of ca 120 m radius from turbines was searched.

| Statistical analyses
We tested for any effects of tower base painting on ptarmigan mortality rates (i.e., carcasses found), before and after painting following a Before-After-Control-Impact (BACI) approach. performed at each turbine. In the analyses, the number of recorded carcasses at ten control turbines (Turbine number: 27,29,32,34,36,38,43,45,48,51) was compared with the painted turbines (Turbine number: 26,30,31,33,35,37,44,46,49,52) before and after painting, while taking into account search effort by including an offset term. To control for any potential effects of turbines and either year or season, random effects were included in a generalized linear mixed-effects model using the glmer function of the lme4 library with a Poisson distribution (Bates, Machler, Bolker, & Walker, 2015) in the statistical software program R 3.3.2 (R Core Team, 2016). To control for potential overdispersion in the data, we also included an observation-level random effect (Harrison, 2014). The distribution of turbine distances of ptarmigan carcasses across time and treatment was tested with the Levene's test for homogeneity, using the leveneTest function in the car library (Fox & Weisberg, 2019). Using a similar model structure as for the mortality rates, effects of painting on the average (log-transformed) distance where ptarmigan carcasses were found were tested using linear mixed-effects using the lmer function of the lme4 library. When regressing distance (log-transformed, maximum cutoff distance of 120 m) against species using linear regression, and considering all turbines for the whole study period, there were significant species-specific differences at what distance from the turbine base carcasses were found (F = 3.111, p = .046). Ptarmigan were found significantly closer to turbines compared to eagles (t = −2.265, p = .024). The variation among species falling within the grouping "other species" (other than eagles or ptarmigan) led to a nonsignificant effect (t = −1.413, p = .16). For ptarmigan, 23.9% of all carcasses were found within 10 m of the turbine tower, against 0.0% and 9.8% for white-tailed eagles and other species, respectively ( Figure 3).

| RE SULTS
When considering the 10 control turbines, there were 11 ptarmigan carcasses found in the period before painting (1,400 searches) and 19 carcasses found in the period after painting (505 searches).
Same numbers for the 10 painted turbines were 25 carcasses in the period before painting (1,023 searches) and 14 carcasses in the period after painting (523 searches).
The generalized linear model disclosed that the yearly number of carcasses (across seasons) after painting was reduced at the painted turbines (z = −2.884, p = .004, Table 1). Overall, there was a 48.2% (95% confidence interval: 44.2%-52.0%) reduction in the annual number of recorded carcasses per search at painted turbines relative to unpainted ones. This effect was mainly due to the relatively large increase in fatalities per search at the control turbines (before: 0.005, after: 0.030), but lack of such a large increase at painted turbines (before: 0.019, after: 0.023) (Figure 4a). We found no effect of birds having a higher probability of collision at the neighboring control turbines due to painting. This was tested by comparing control turbines to other untreated turbines before-after "treatment" within the wind-power plant (z = 0.283, p = .78). The average distance of ptarmigan carcasses from the turbine base increased significantly at the painted turbines after painting (F = 6.535, p = .014). After painting, no ptarmigan carcasses were found within 30 m of the painted turbines ( Figure 5). However, the number of carcasses fluctuated considerably between years (Figure 6a).

| D ISCUSS I ON
Carcasses of willow ptarmigan were the most frequently observed of all species close to wind turbines at the Smøla wind-power plant.
Ptarmigan carcasses were usually found closer to the turbines relative to other species. Together with the general findings that (a) these TA B L E 1 Model estimates testing the effect of painting on the yearly (upper table) and seasonal (middle table) rate of ptarmigan carcasses found at the Smøla wind-power plant using a Before-After-Control-Impact (BACI) design. The lower table provides the effect of painting on distances from turbines where carcasses were found. The models controlled for search effort using an offset term carcasses often had injuries that comply with impact from hitting the tower base instead of the turbine blades (Bevanger et al., 2010;Pedersen, 2017), (b) ptarmigan carcasses were found closer to turbines than other species (Figure 3), (c) number of carcasses found close to turbines were lower after painting than before (Figure 3), and (d) overall number of carcasses found declined after painting of the tower bases, strongly indicates that turbine tower base collisions are an important cause of death for this species at the wind-power plant on Smøla. However, importantly, it is also probable that some ptarmigan collide with the turbine blades, as there is also a peak in the distribution of distance from turbine of carcass at 40-60 m ( Figure 3). In addition, predation by raptors such as golden eagles (Aquila chrysaetos) and gyrfalcons (Falco rusticolus) is an important source of mortality among ptarmigan on Smøla, especially during winter (Bevanger et al., 2010;Brøseth, Nilsen, & Pedersen, 2012).
Hence, predation was likely the source of mortality for an unknown proportion of the ptarmigan carcasses found. In a telemetry study on ptarmigan at the Smøla wind-power plant (Bevanger et al., 2010), collisions with wind turbines accounted for 35.7% of the mortalities, while predation accounted for 42.9%-57.1% (N = 28). However, there is no reason to assume that there should be difference in predation pressure at painted versus unpainted turbine towers.
Our results also show that there was a variation in collision risk both among years and seasons Martin, Perez-Bacalu, Onrubia, Lucas, & Ferrer, 2018), which has also been found in studies on other bird species colliding with man-made objects (e.g., Avery, Springer, & Dailey, 1978;Barrios & Rodriguez, 2004;Bevanger & Brøseth, 2000). Annual variation in collision risk in F I G U R E 4 Effect plots testing the effect of painting on the yearly (a) and seasonal (b) rate of ptarmigan carcasses found at the Smøla wind-power plant using a Before-After-Control-Impact (BACI) design. Panel (c) provides the effect of painting on distances from turbines where carcasses were found. Control = unpainted turbine towers, Impact = painted turbine towers, Before = period before painting, After = period after painting. Estimates are controlled for search effort using an offset term ptarmigan may be influenced by environmental factors such as variation in weather conditions. Spells of weather regimes resulting in fog and rain, for instance, may result in poor visibility and hence a greater danger of colliding with stationary objects (Bevanger, 1994;. Potentially, fluctuating population sizes could also lead to variation in collision risk, as it has previously been found that the ptarmigan population size on Smøla may vary significantly from year to year (Bevanger et al., 2010). A higher winter mortality can cause reduction in number of breeding pairs and hence chick production and vice versa (Pedersen, 1984). Also, the chick production can vary greatly from year to year due to weather conditions and other environmental factors (Kvasnes, Pedersen, Storaas, & Nilsen, 2014). In years with larger ptarmigan population sizes, the probability of finding birds colliding with turbines could increase simply due to more individuals using the air space in the area. However, there is no indication that the population size on Smøla has declined after the construction of the wind-power plant.
Furthermore, both the density of ptarmigan and chick production inside the power plant has not been significantly different from outside (Bevanger et al., 2010).
Seasonal variation in collision risk may also be influenced by weather since amount and nature (i.e., rain or snow) of precipitation, wind speed, etc. may be connected to season. In addition, variation in light conditions and species-specific behavioral patterns may explain seasonal variation in collision risk (e.g., Bevanger, 1994). Bevanger (1995) found a peak of collisions with power lines for black grouse in the autumn (September to October), while most collisions in willow ptarmigan occurred in winter and early spring (November to March). This fits well with the results from the Smøla wind-power plant, where collisions most frequently occurred in winter and spring ( Figure 5). This is also supported by earlier findings of the same willow ptarmigan population by Brøseth et al. (2012).
The effect of painting of the turbine tower bases was most pronounced in spring and autumn. The lack of effect of painting during winter could be due to generally poor light conditions, making tower bases hard to observe no matter their appearance (Pedersen, 2017 (Pedersen & Karlsen, 2007).
Failure to detect carcasses may obviously influence estimation of collision frequencies. At the control turbines, there was an increase in the number of carcasses found after painting than before ( Figure 4). This result is difficult to explain, but it could be due to variation in search efficiency. In the present study, the search regime was constant throughout the whole search period, and variation in search intensity was controlled for in the analyses. In addition, Another possible reason for the increase in number of carcasses found at the control turbines before and after painting could be that ptarmigan showed anticipatory evasion (cf. May et al., 2015) of the painted turbines by changing flight paths, and were therefore more likely to collide with the unpainted turbines. However, we found no indications of ptarmigan being "forced into" neighboring turbines due to an evasive response to the painted rotor blades. This possibility, however, merits further investigations focusing on ptarmigan movements within the study area.
Scavenger removal of collision victims may obviously influence estimation of collision frequencies (e.g., Loss et al., 2015). The intro- port. We would also like to thank Ole Reitan for collecting data on bird carcasses, and personnel at the Smøla wind-power plant for their logistical support during the experiment. We want to thank two anonymous reviewers for their insightful and constructive comments that significantly improved the quality of the manuscript.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study have been deposited