Sensitivity analysis of collision risk at wind turbines based on flight altitude of migratory waterbirds

Global warming has become a major concern around the world. The Paris Agreement of 2015 (UNFCCC, 2015) mandates reduction CO2 emissions in developed countries, and subsequently renewable energy sources that produce significantly fewer greenhouse gases, such as wind, solar geothermal, have increased rapidly since 2000s. relatively low cost construction maintenance wind power generators makes it commonly used method generation. installed capacity generation globally, 2017, was 20 times greater than 2001 (GWEC, 2017). However, there are concerns farms leads to degradation landscapes natural environment. In particular, impacts on birds bats, is, collisions with turbines, detour costs due avoidance habitat abandonment, serious problems worldwide (Amorim et al., 2012; Erickson 2014; Harrison 2018; Marques 2019; Masden 2009; Smallwood, 2013). Smallwood (2013) estimated 573,000 888,000 bats per year would be killed by 2012, United States, which represented 16.5% world's Bird influenced landscape, terrain weather constructed. They often reported ecologically sensitive locations along flight paths between roosting foraging areas slopes over updrafts occur soaring (Drewitt & Langston, 2006; Johnson 2002; Kitano Shiraki, 2013; Murgatroyd 2021; Peron Avoiding selecting for likely strong impact is part efficient site selection. To this end, development introduction sensitivity maps advance planning been promoted mainly Europe States order prevent bird (Bright Garthe Hüppop, 2004; McGuinness 2015; Retief 2010). Basic information contributes creating includes wintering stopover sites, main habitats key species migration routes Previously proposed were based two-dimensional concerning distributions target Recently, large raptors, collision risk visualized predicting area where they fly below height turbine spatial factors slope distance from nest (Murgatroyd 2021). few research cases reflected three-dimensional altitude, may vary depending environmental characteristics, making impossible identify high-collision-risk areas. Hokkaido Tohoku regions northern Japan considerable potential installation favourable conditions (MOEJ, 2011). Currently, these two account 47% total (MOETIJ, 2020). These many wetlands, including rivers, lakes marshes serve important grounds large, migratory waterfowl geese swans (Mikami 2012), populations number approximately 160,000 2021c). There no reports colliding turbines (Ura, 2015), possibly because farm managers not required report collisions. (Rees, built 2000s 2017), about negative abandonment future (Ura Therefore, an urgent issue. paper, we provide focus altitudes swans. use 3D trajectories influence land topography differences altitude. Additionally, extrapolate statistical models predict showing We four most abundant Japan: Bean goose Anser fabalis serrirostris (BG), Greater white-fronted Goose albifrons (GWG), Tundra swan Cygnus columbianus bewickii (TS) Whooper cygnus (WS). data interviews ‘monitoring 1000’ described include BG treated one. Furthermore, some cases, object observer so great field survey difficult distinguish species. grouped BG. selection sites annual point count surveys at 9000 October April Japan's Ministry Environment This conducted geese, ducks throughout Japan, but Honshu Island. complement MOEJ's surveys, decided also region, ground Field carried out Hokkaido, Aomori, Akita, Yamagata, Niigata Miyagi prefectures, home 80% BG, 90% GWG, 66% TS 67% WS during winter Regarding offshore areas, although flights confirmed study, detailed location could obtained shore too recorded; hence, only targeted onshore Our fieldwork public lands did require application licence. resting flocks located recorded vehicle (Figure 1). A cumulative 60,000 km2 farmland (including pastures, rice fields cropland), addition bodies water, ponds covered. 101 days: 9 days Prefecture (from November 2018 February 2019), 4 Yamagata (in 2019); 10 Akita 2017); 11 December Aomori (March March 5 52 2019). daylight 08:00, after left their roosts morning, 16:00, before roost evening. approximate sunrise sunset study period 04:30 06:30 am 15:30 18:00 (NAOJ, During flock sizes mapped. Birds when passed directly above researcher. ornithodolites (VECTOR21, VECTOR21 AERO MOSKITO manufactured SAFRAN Vectronix; 1σ error: ±5 m, elevation ±0.2°, azimuth: ±0.6°) capable obtaining highly accurate data. devices accurately measure object's azimuth, oblique calculate latitude, longitude altitude using built-in compass infrared laser illuminator 2). When birds, tracking acquired 3–6-s intervals. response landscape can understood. 2019, fixed-point ornithodolite 93 within 1) obtain measured various environments, farmland, water bodies, urban forests mountainous elucidate effects maximum measurable range according device's specifications 12 km structures. swans, 2 individual 3 flocks. ranges spring defined outermost boundary Wild Society Nationwide Census Geese, Ducks, Swans 2021c), Monitoring Sites 1000 2021b) Atlas (YIO, 2021) databases. showed connectivity lacking observational referring satellite-tracked studies (Chen 2016; Kurechi, MOEJ, Shimada Ueta YIO, Esri Arc GIS Pro ver. 2.41 create positional extract topographical (JAXA, MLIT, model drawn continuous latitude centre trajectory set analysis point, extracted. landing excluded movements judged local flights. For related use, 1, 2, buffers generated points, allowed acquisition forests. Farmland fields, vegetable pastures habitats, lakes, rivers habitats. case topography, points. Standard deviations each buffer calculated values representing topographic roughness 3). While analysing determine characterization behaviour variable mountain forest considered process improve explanatory predictability models; changes tracked. considering history (the had recently flown) travelled affects determination lines 3, opposite direction mean (hereinafter, history), highest line modelling, incorporated variables (Table autocorrelation performed ANOVA ID factor. Kruskal–Wallis test Steel–Dwass compare landscape. Statistical created R 3.6.0. LASSO regression (a type regularized linear analysis) package glmnet applied (Friedman 2010; Core Team, 2022). Dimensional compression L1 norm regularization adjusting coefficients applied. It possible easy interpret while preventing overfitting (Ranstam Cook, 2018). our scales. Because influential scale differ, reasonable apply modelling incorporating all variables. regression, less prone overfitting, even involved. (masl) points variable. Environmental standardized 0 standard deviation 1 optimize value λ, determines term, cross-validation divided into sections. avoid final model, λ 1SE minimum error. R2 model. 250 m grid route, extrapolation area, history, transmitters. square, nearest referenced. extracted same map, possibility flying evaluation Hence, relative subtracting average square predicted square. classified turbine. maps, assumed one largest rated output 3200 kW (wind 150 rotor diameter 103 hub 98 m) (SEI, 2020), taking recent increase size (JWEA, 2016). three levels, low, medium high. Medium (45–150 rotating blades collision, (<45 lower medium; high (>150 m), higher turbine, zero S1). Migratory 7168, 3229, 6366 10,929 4). Habitat shown or Miyagi, eastern western previous satellite transmitters 5, Chen Takekawa 2000; overestimation extent routes, observed excluded, except indicated connectivity. Since southern crossing Euroasiatic continent without passing through 2010), route Japan. 185, 221, 170 51 respectively. frequency S2). frequently S3). Analysis variance factor change rather SD 5.9, 7.7, 6.7 7.7 respectively S4). 6; Steel-Dwass test, p < 0.01). other places 0.05). TS, farmlands, No significant WS, flown (sample = though) 0.29). median (<150 blades. All variation forests, differed those flatlands mountains 7). any flat terrain. contrast, roughness. Before mountains, elevation. peaks elevation, sea level high, low. maintained descending, thus times. indicate 0.67 0.79 0.57 0.76 species, positive environments history. smaller scales effect, effect. histories more ones. coefficient, indicating continued ranges. extrapolated map 8). Of altitudes, proportions time spent zones follows: 27.7%, 60.7% 11.6%; 23.9%, 64.3% 11.8%; 21.1%, 18.2%; 48.8%, 47.1% 4.2%. GWG spend zone, greatest turbines. zone One advantages updated strut blade length offers numerical prediction accommodate expected larger future. Differences decision-making terrain, higher, lower. environment broader performance low-load increasing advance. GPS Bar-headed indicus, renowned its trans-Himalayan flights, typically flies valleys, albeit (Hawkes 2012). general, oxygen availability energetic (Bishop Butler, Hawkes winter, food scarce, reproduction, follow maintain minimize consumption. ridges, means sufficiently zone. Several issues identified study. First, density route. Geese choose Klaassen 2004). Assuming preferences involved path density, will able evaluate accuracy. Although focused Lake Shinji, uncharted Eurasian via future, beyond scope present need risks evaluated. unable analyse position insufficient acquisition. plans introduce both offshore. made build off coasts prefectures Wind offshore, 200 taller land. Little available Japanese waters. clarification characteristics assessment MOEJ evaluates (10 grid) five levels score. score distribution several rare selected government included Red Data Book 2020) further reduce environment, region indicator necessary multiple devised. opinion 2014) should particularly rigorous inhabiting breeding ground. reflecting behavioural essential, raptors resolution important. provided project finer (we map) (EIA) site. Biological EADAS 2021a) linked maps. producers EIA, contribute prompt On hand, good cannot avoided. Systems respond stop operation encourage being implemented (Georgiev Zehtindjiev, May 2020; Pescador Proper mitigation essential human–wildlife conflicts. Taito Kamata Tsuneo Sekijima conceived ideas designed methodology; Kamata, Hitomi Sato, Haruka Mukai, Takahiro Sato Shintaro Yamada collected data; analysed Sekjima led writing manuscript. authors contributed critically drafts gave approval publication. completed support Research Technology Development Fund (JPMEERF20164003) Restoration Conservation Agency Yamaguchi Educational Scholarship Foundation. grateful Tatsuya Ura, Yasuhiro Kawasaki Satomi Makoto Hasebe Sarobetsu Wetland Center helpful distribution, Kyokuto Boeki Kaisha, Ltd. lending ornithodolite, Mark Brazil carefully proofreading declare competing interests. peer review article https://www.webofscience.com/api/gateway/wos/peer-review/10.1002/2688-8319.12222. Dryad Digital Repository https://doi.org/10.5061/dryad.73n5tb323). Figure S1. Definitions standards. S2. Positioning counts tracked ornithodolite. S3. Flock S4. trajectories. Table Variance tracking. Please note: publisher responsible content functionality supporting supplied authors. Any queries (other missing content) directed corresponding author article.