Avian-power line interactions in the Gobi Desert of Mongolia: are mitigation actions effective?

Electrocution and collisions on power lines are among the leading causes of non-natural mortality for birds. Power lines are exponentially increasing, particularly in developing countries, but mitigation strategies to prevent bird mortality are questionable. Mongolia combines a recently increased power line network, an abundant raptor population, a dangerous crossarm configuration and a habitat with no natural perches, producing many bird-power line interactions. Our aim is to assess the bird mortality caused by power lines in the Gobi Desert of Mongolia, to determine the factors increasing the risk of bird electrocution, and to evaluate the effectiveness of used retrofitting measures. In July 2019 we covered 132.9 km of 15 kV power lines checking 1092 poles. We also conducted bird transects to record raptor and corvid richness and abundance, to assess species vulnerability to electrocution. We recorded 76 electrocuted birds of 7 species. Electrocution rate was 6.96 birds/100 poles. The most affected species were Common Raven (Corvus corax) and Upland Buzzard (Buteo hemilasius), highlighting the electrocution of 5 endangered Saker Falcons (Falco cherrug). By contrast, we only recorded 8 individuals of 5 species colliding with wires, the most affected being Pallas’s Sandgrouse (Syrrhaptes paradoxus). About 76.1% of sampled poles had some mitigation measure. Of these, 96.6% were brush perch deflectors and 3.4% rotating-mirrors perch deterrents. We found differences in electrocution rates among crossarm configurations, with the strain insulator with one jumper being the most lethal. Additionally, we found no correlation between bird abundance and electrocution rates, suggesting that some species are more sensitive to electrocution. Although no differences in total bird electrocution rates were detected between poles with and without perch deterrents, when bird size is considered, deterrents reduced the mortality rate of small birds, while they were ineffective for medium-sized birds. Despite the widespread use of perch deterrents in the Mongolian power line network, there is still an alarming electrocution rate. This strategy is ineffective and some mechanisms, such as brush perch deflectors, may increase the electrocution rate for some medium-sized birds. Finally, we propose strategies to minimize the avian electrocution rate in the Gobi Desert.


Background
The development of power line networks is both a consequence and a driver of a country's economic progress (Chaurey et al. 2004), but it is also an important source of environmental impacts (Sánchez-Zapata et al. 2016). Electrocutions and collisions with power lines are among the leading non-natural causes of bird mortality (Bevanger 1998;Haas 2005), drastically reducing the populations of some endangered species (García-del-Rey and Rodríguez-Lorenzo 2011), especially raptors (Meretsky et al. 2000;Real et al. 2001;López-López et al. 2011;Angelov et al. 2013). Three factors have been linked to the increased risk of accidents between birds and power lines. The first one is related to the pole and crossarm configuration or the wire arrangement (e.g. grounded steel and concrete poles and crossarms, or the increase of jumper wires; Tintó et al. 2010;Guil et al. 2011), the second one is associated with the ecology and biology of birds (e.g. size and wingspan, low maneuverability, narrow visual field, hunting behavior; Janss 2000; Lehman et al. 2007;Martin and Shaw 2010;Guil et al. 2015) and the last one is related to the environment (e.g. absence of natural perches, frequency of fogs, location on migratory routes; Harness et al. 2008;Dixon et al. 2018).
Electrocution occurs when a bird simultaneously contacts two differently energized phases or one energized phase and one grounded area (APLIC 2006). Thus, mitigation strategies (from now on "retrofitting") used to prevent bird electrocution can be: (i) by "separation" between potential contacts, which is the most recommended strategy since it is permanent and does not require maintenance; (ii) by "insulation", which requires strategically covering energized or grounded contacts, it is the most used strategy and needs maintenance; and (iii) by "redirection" of birds to perch in safer places, which is the cheapest strategy and also requires maintenance (APLIC 2006). By contrast, bird collisions can be mitigated modifying surrounding habitats, removing overhead shield wires, burying lines, and increasing visibility to birds by marking wires (Eccleston and Harness 2018). Although these mechanisms are considered effective in reducing electrocutions (Tintó et al. 2010) and collisions (Alonso et al. 1994;Barrientos et al. 2011), few studies have evaluated their long-term effectiveness in the field and with different avian populations (Janss and Ferrer 1999).
Asia is undergoing the fastest economic growth and many of its countries are exponentially increasing their power line network (Hammons 2011). In addition to this sharp increase, most power lines have little or no retrofitting (Dixon et al. 2013a), as most Asian countries are limited in their economic resources. However, avian electrocution and collision studies in this continent are scarce (Lehman et al. 2007;Bernardino et al. 2018;Slater et al. 2020), so there is an urgent need to focus efforts on the continent, especially in important bird areas, to understand the magnitude of these impacts. In the last two decades, Mongolia has sharply increased its power line network (Amartuvshin and Gombobaatar 2012). Studies carried out so far show a severe environmental impact as a result of the interaction between rich and abundant raptor populations, an electrical network with hazardous designs (grounded steel and concrete poles and metal crossarm with few or obsolete retrofitting, usually redirection) and habitat conditions without natural perches (Harness et al. 2008;Amartuvshin and Gombobaatar 2012;Dixon et al. 2018). One of the largest concerns is the high mortality rate of the Saker Falcon (Falco cherrug; Dixon et al. 2020), which is globally endangered (BirdLife International 2021) and a large part of its breeding population is in Mongolia (Gombobaatar et al. 2004). Despite previous research about birds and power lines carried out in Mongolia (Harness et al. 2008;Amartuvshin and Gombobaatar 2012;Dixon et al. 2013a), most studies have been located in the central and eastern part of the country (Ganbold et al. 2018). By contrast, large areas in southern Mongolia have remained understudied. The Gobi Desert extends in this area, an extremely arid ecosystem in which many species adapted to these extreme ecosystems survive. Although the abundance of species that are more sensitive to power lines impacts, such as birds of prey, is slightly lower than in other areas previously studied in the country, the presence of potential natural perches is even lower, so the interaction between birds and power lines can be high (Gombobaatar et al. 2004).
In this study, our main goal was to study bird mortality at power lines in the Gobi Desert (southern Mongolia). Specifically, we aimed to: (1) evaluate bird mortality by electrocution and collision with power lines, (2) explore the factors that increase the risk of bird electrocution, such as the crossarm configuration and the size or abundance of the bird species in the area, and (3) evaluate the effectiveness of the retrofitting measures used.

Study area
Our study area was located across several provinces (aimags) of Mongolia, between the Khogno Khan National Park in the north (47°19′ N, 103°41′ E), and Dalanzadgad city in the south (43°34′ N, 104°25′ E) of the study area ( Fig. 1). Power lines cross the Gobi Desert, an undulating landscape mainly composed of arid steppes and scattered mountains ranging from 706 and 2825 m a.s.l. (Begzsuren et al. 2004). Short bunch grasses are dominant in steppe zone lacking trees and natural perches. In the desert zone vegetation cover generally reaches values of < 10%, often only 1-2% (Walter 1993). This region has sharp temperature contrasts reaching 40 °C in summer and − 49 °C in winter, concentrating 125 mm rainfall annually between April and September (Pfeiffer et al. 2003).

Fig. 1
Map of the study area across several aimags (provinces) in the Gobi Desert, Mongolia. We show road car transects (354.6 km) performed to estimate raptor and corvid abundance and systematic transects (132.9 km) conducted along power lines to record bird mortality from both electrocution and collision

Data sampling
On July 2019, we systematically covered a total of 132.9 km of 15 kV power lines, checking a total of 1092 poles to record bird mortality from both electrocution and collision. Due to the favorable visibility conditions for finding dead birds in the study area, we used car transects. We drove under the power lines at a speed of 15-20 km/h. Two observers, one on each side of the car, searched for remains, and two additional people noted the field data (i.e. dead species, pole characteristics and causes of death: electrocution or collision) and GPS positions. In those places with worse visibility (e.g. with vegetation) the searching was carried out on foot. We checked for dead birds within a 5-m radius around each pole, and 8 m on each side under the lines. All dead birds found were examined for burn marks or trauma. Additionally, we recorded all casualties found outside the systematic transects, but these data were excluded from analyses and mortality rate estimates.
All power lines were constructed with concrete poles and metal crossarms. We recorded the crossarm configuration (type and number of isolators, number of jumper wires), the presence of retrofitting measures (i.e. redirection), such as perch deterrent, and their type (e.g. brush perch deflectors or rotating-mirrors perch deterrents). No bird collision deterrent was recorded in any of the monitored power lines.
Estimation of diurnal raptor and corvid abundance (i.e. number of individuals of each species) were performed by road car transects (Viñuela 1997). Transects were performed from two hours after sunrise to one to two hours before sunset, at an approximate speed of 30 km/h. Two observers were identifying and counting birds (raptors and corvids). Transects were conducted up to a maximum distance of 50 km from the sampled power lines. This threshold was considered a representative measure of the maximum distance that highly mobile resident birds, potentially sensitive to power lines, could move (e.g. Dixon et al. 2017a,b;Reading et al. 2020). We sampled a total of 354.6 road km during 5 days and lasting a total of 14.9 h.
For all the species recorded in the road and the mortality surveys, we gathered information on the species conservation status (IUCN 2020) and its mean wingspan (del Hoyo et al. 1994), which was used to classify them into two categories (small-wingspan ≤ 70 cm, medium-wingspan 70-160 cm) according to Dixon et al. (2019).

Statistical procedures
We calculated the "electrocution rate" as number of electrocuted birds per 100 poles and per 10 km, to compare it with that of previous studies. We only used the dead birds found during systematic sampling for the statistical analyses. We also calculated the "collision rate" as number of collided birds per 10 km.
To test if large or abundant raptors and corvids are more vulnerable to electrocution by relating species abundance and wingspan with "electrocution rate" (birds/100 poles), we used Generalized Linear Models (GLMs), where the "electrocution rate" for each species was the dependent variable and the species-specific (1) total abundance or (2) mean wingspan were the predictors. As we could not normalize the data, we used a randomization approach, where the predictor was kept constant but the values for "electrocution rate" were randomized. We repeated the analyses 1000 times and we checked if the coefficients of the predictor in our model were within the range of values of the coefficients of the models with the randomized variables.
We assessed the mortality of different crossarm configurations by comparing the observed and expectedby-chance proportion of birds electrocuted below each crossarm configuration (one-pin insulator, double pin insulator and strain insulators with one jumper) using Chi-square tests.
To evaluate the effectiveness of retrofitting in reducing bird mortality, we used the Chi-square test, which compares the observed and expected-by-chance proportion of electrocuted birds below full-corrected crossarms (two or four brush perch deflectors) and uncorrected crossarms (without brush perch deflectors) in general, and per crossarm configuration. Partially retrofitted crossarms (i.e. lost one perch deflector) were excluded from the statistical analyses because we could not determine if birds had been electrocuted in the part of the crossarm where the perch deflector was present or where it was not. We also excluded the rotating-mirror perch deterrents from the analyses because of the small sample size (n = 28 poles).
Finally, to test if there was a correlation between the mitigation measures effectiveness and the bird size, we used the Chi-square test to compare the proportion of electrocuted birds below the full-corrected poles (with two and four brushes) and the uncorrected ones by wingspan (small and medium). All the analyses were run in R 3.6.0 (R Core Team 2019).

Results
We found a total of 76 electrocuted birds of 7 species at 55 poles (5% of the total sampled poles) during 132.9 km of systematic transects. The most affected species was the Common Raven (Corvus corax; n = 33), followed by the Upland Buzzard (Buteo hemilasius; n = 11; Table 1). The electrocution of 5 individuals of the endangered Saker Falcon is also noteworthy. Electrocution rate was 6.96 birds/100 poles or 5.72 birds/10 km (see Table 1 for electrocution rates by species). We also detected the electrocution of 12 individuals of 6 species found in non-systematic sampling, where a new electrocuted species was recorded, the Eurasian Eagle Owl (Bubo bubo; Table 1). Also, we found no relationship between species-specific electrocution rates and their abundances (GLMs, coefficient ± SD = 0.094 ± 0.249, p = 0.223) or their wingspan sizes (GLMs, coefficient ± SD = -0.002 ± 0.003, p = 0.21). The most abundant species in the transects were Black Kites (Milvus migrans) and Himalayan Griffons (Gyps himalayensis) which had low or non-existent electrocution or collision rates (Table 1). In addition, power poles with strain insulators had a much higher electrocution rate than the other crossarm configurations (χ 2 = 16.56, df = 2, p < 0.001; Fig. 2).
Bird species affected by electrocution were different from the species affected by collision. Regarding collisions, only eight individuals of five species were recorded, the most affected being the Pallas's Sandgrouse (Syrrhaptes paradoxus) with three individuals (Table 1). Total collision rate was 0.6 birds/10 km.
About 76.1% of the sampled poles had some retrofitting mechanism, of which 96.6% used brush perch deflectors and only 3.4% used rotating-mirror perch deterrents. Some 62.2% of crossarms were retrofitted with two brush perch deflectors, whilst 13.4% had lost one and 24.4% had lost all brush perch deflectors. In contrast, 83.3% of the crossarm poles with rotatingmirror perch deterrents conserved all mirrors, 10% had one and 6.7% had lost all perch deterrents.

Table 1 Avian mortality by electrocution and collision in 15 kV power lines in the Gobi Desert (Mongolia)
The table shows the number of electrocuted birds in systematic (Syst.) and non-systematic sampling (Non-Syst.); birds colliding with power lines (Collision); recorded birds in transects carried out in the area of the power lines studied (Abundance/10 km); electrocution rate measured in number of birds per 100 poles (Birds/100 poles); electrocution or collision rate measured in number of birds per 10 km (Birds/10 km). For the "Total", we have only used electrocuted birds. For electrocution and collision rates we have only used birds recorded in the systematics samplings; abundance was estimated only for diurnal raptor and corvids; global threat level according to the International Union for Conservation of Nature (IUCN)  Mortality of small raptors in the corrected poles were significantly lower than expected (37.5% observed vs. 71.8% expected, χ 2 = 4.59, df = 1, p = 0.032), while for medium-sized birds (raptors and corvids), there were no differences between corrected and uncorrected poles (χ 2 = 0.60, df = 1, p = 0.438).

Discussion
Bird electrocution in Mongolia remains a severe problem that seriously threatens endangered species. At least 30 electrocuted bird species have been recorded in this country, which has one of the largest electrocution rates worldwide. Although our study does not reach the highest rates recorded in Mongolia such as 21 birds/100 poles (Dixon et al. 2017a) or 65 birds/100 poles (Dixon et al. 2018), our mortality rate (6.96 birds/100 poles) remains similar (Harness et al. 2008) and even higher (Ganbold et al. 2018) than studies conducted further north. Through our study, we reaffirm the high bird electrocution rate in the understudied region of the Gobi Desert.
Power line bird collisions have been much less studied than bird electrocutions in Mongolia. Our results support that bird mortality by collisions with power lines is much lower than by electrocution. Interestingly, the bird community affected by collision and electrocution are different. While electrocution affected mainly corvids and raptors, collision affected mainly steppe birds, including Charadriiformes. In particular, the high collision rates of the Pallas's Sandgrouse is alarming. Amartuvshin and Gombobaatar (2012) found 407 collided individuals in Mongolia, and they cite a record from the Veterinary Institute of Mongolia of 2300 collided individuals of this species, also in this country. This also occurs in Spain where large number of sandgrouses (Pterocles orientalis and Pterocles alchata) have been found to collide in the Peninsula (Barrientos et al. 2011) or the Canary Islands (Gómez-Catasús et al. 2020). These species usually fly at dawn and dusk in search of water in very remote areas and this could make them more susceptible to collision, as well as many species of Charadriiformes that migrate at night. Although not threatened, populations of the Pallas's Sandgrouse may be declining due to high collision mortality. Detailed studies of how the population of this species is being affected would be desirable. Besides, the bird mortality rate shown in this study should be taken as a minimum, as the scavenger consumption rate may be high (Dixon et al. 2017a,b;Orihuela-Torres et al. 2021), especially for small birds (Gómez-Catasús et al. 2020).
In the Gobi Desert, we have shown that the most abundant species are not the most electrocuted, suggesting that susceptibility to electrocution depends on other non-evaluated factors such as behavior. Common Ravens, Upland Buzzards and Saker Falcons showed a high sensibility to electrocution likely due to their preference for roosting and nesting on poles (Dixon et al. 2013a,b). Even though the Saker Falcon is globally classified as "Endangered", its high electrocution rate has been known in the country for years (Dixon et al. 2020). This incidence has also been found in our study, indicating that it does not seem to show a decrease on a national scale despite the conservation actions promoted by the government and private entities. Given this situation, it is likely that this cause of mortality is dramatically affecting its population dynamics as some studies point out (Dixon et al. 2020). In other hand, the presence of jumper wires over the crossarm are directly related with a higher electrocution rate (Guil et al. 2011). Fortunately, the configuration of the power lines reported in the study area is very homogeneous, the pylon material is always metal and the crossarm configuration has no more than five versions, so the potential designs are limited. This fact contrasts with what happens in other areas of Europe such as Spain where configurations can exceed a hundred (Guil et al. 2011). This could simplify decision making when it comes to implementing safer mitigation measures.
Despite the widespread use of perch deterrents in the Mongolian power line network, our study has shown that this is a low-effective mitigation strategy, even more, some mechanisms such as brush perch deflectors could be increasing the electrocution rate of some medium and large species (Harness et al. 2008;Amartuvshin and Gombobaatar 2012). Our results show that the presence of perch deterrents reduced the mortality rate of small birds, while are ineffective to reduce the medium-sized birds. This unexpected result could be explained by the fact that the height of the brushes coincides with the head of the small birds, disturbing them. The same brush may approximate larger birds closer to the wire, increasing their probability of being electrocuted. This mitigation strategy was widely used in the 1980s and 1990s in the US, but its use is declining because of its questioned effectiveness (Slater et al. 2020). However, there are places where it is still applied because of its low cost (Janss and Ferrer 1999;Prather and Messmer 2010). In addition, brush perch deflectors are often not placed properly, or have no maintenance and last little time in good condition, becoming death traps . Almost 40% of sampled retrofitted poles had lost at least one of the perch deterrents. By contrast, despite the low number of covered poles, we found no electrocution under poles with rotating-mirror perch deterrents. This is in line with the study by Dixon et al. (2019) that proved the effectiveness of this mitigation strategy, suggesting that this strategy is much more effective than brush perch deflectors, although it also needs maintenance. Despite this, mitigation strategies such as the modification of crossarm to a save design and the insulation of conductors, which are been proved more effective, an extended life-time and lower maintenance and installation failures (APLIC 2006;Guil et al. 2011), should be applied.
It is important to underline that our study does not account for the imperfect detection of the electrocuted birds, which may be an important problem in areas with dense vegetation and very rocky areas that difficult finding the dead individuals (Stevens et al. 2011;Gómez-Catasús et al. 2020). However, our study area was mostly covered by bare ground with almost no vegetation or rocks and the electrocuted individuals could be easily seen from the distance.

Conclusions
Our study confirms that the Gobi Desert has a worrying avian electrocution rate, especially for the endangered Saker Falcon, and that mitigation strategies are partially effective and show a short lifetime. We suggest avoiding brush perch deflectors as mitigation strategy, and retrofitting the most dangerous crossarm configuration (strain insulators with one jumper) by installing effective insulation systems or changing the design using, for example, suspended insulators (Tintó et al. 2010). In parallel, governments should increase the surveillance and impose penalties on utilities that do not meet certain avian friendly power line requirements. These strategies have been taken in the United States, South Africa or some European countries (Antal 2010;Schürenberg et al. 2010), which has led to saving species that were on the verge of extinction (López-López et al. 2011) if these standards are met.