From dusk till dawn: Ecoacoustic monitoring reveals wind energy impacts on roding activity in the European Woodcock (Scolopax rusticola)

From dusk till dawn: Ecoacoustic monitoring reveals wind energy impacts on roding activity in the European Woodcock (Scolopax rusticola) | Authorea try { document.documentElement.classList.add('js'); } catch (e) { } var _gaq = _gaq || []; _gaq.push(['_setAccount', 'G-8VDV14Y67G']); _gaq.push(['_trackPageview']); (function() { var ga = document.createElement('script'); ga.type = 'text/javascript'; ga.async = true; ga.src = ('https:' == document.location.protocol ? 'https://ssl' : 'http://www') + '.google-analytics.com/ga.js'; var s = document.getElementsByTagName('script')[0]; s.parentNode.insertBefore(ga, s); })(); Skip to main content Preprints Collections Wiley Open Research IET Open Research Ecological Society of Japan All Collections About About Authorea FAQs Contact Us Quick Search anywhere Search for preprint articles, keywords, etc. Search Search ADVANCED SEARCH SCROLL Wildlife Biology This is a preprint and has not been peer reviewed. Data may be preliminary. 5 March 2025 V1 Latest version Share on From dusk till dawn: Ecoacoustic monitoring reveals wind energy impacts on roding activity in the European Woodcock (Scolopax rusticola) Authors : Jan Engler 0000-0001-7092-1380 , Michael Bokämper , Stefan Hannabach , Manuela Merling de Chapa , Lena Daum , and Kostadin Georgiev Authors Info & Affiliations https://doi.org/10.22541/au.174117130.00877062/v1 511 views 257 downloads Contents Abstract Introduction Material & Methods Results Discussion Supplement Supplementary Material References Information & Authors Metrics & Citations View Options References Figures Tables Media Share Abstract Renewable energy is vital for reducing carbon emissions, yet its infrastructure poses challenges for biodiversity. While wind power impacts on bats and raptors are well-studied, effects on elusive species remain largely unknown. The European Woodcock (Scolopax rusticola), a nocturnal forest bird, performs roding flights at twilight to mark territories and attract mates. Despite evidence suggesting potential impacts on the species, details regarding the effects of wind energy use in forests on the habitat use of Woodcocks remain unclear. We compared the vocal activity of Woodcocks at 15 windfarms scattered throughout central Germany. We used passive acoustic monitoring in a paired design, with control sites, situated ≥3 km from wind turbines. We took recordings for three hours at dawn and dusk over ten days in June 2023—during peak roding season. We validated presumed Woodcock matches by BirdNET to assess precision for the species accompanied by a stratified screening to estimate recall rates. We compared true vs. false positive BirdNET hits in relation to the twilight periods during dawn and dusk respectively. Further, we compared the presence of Woodcocks as well as their continuity throughout the recording period, separately for dusk and dawn. We found that the Woodcocks’ roding activity was confined by darkness and by the vocal activity of other species (mainly Thrushes), which during the morning/evening chorus partially overlapped with the acoustic niche of the Woodcocks. Detection probabilities of BirdNET were not affected by differences in soundscapes with and without windfarm instalments. Likewise, species’ presence did not change between windfarm and control sites. However, we found that roding activity of Woodcocks was significantly reduced at wind power sites by around 59% (95%CI = 40%-75%). We discuss our findings with regards to environmental management of the species for windpower planning, including the use of passive acoustic monitoring for elusive species. Introduction The transition to renewable energy sources is a pivotal strategy in mitigating climate change impacts and reducing carbon dioxide emissions. However, while renewable energy presents a sustainable alternative to fossil fuels, the development of renewable energy infrastructure can have unintended consequences for wildlife (e.g. Lafitte et al. 2023, Marques et al. 2021). These impacts range from habitat disruption to direct harm to animal populations. Addressing these challenges is vital to ensure that the pursuit of clean energy does not come at the cost of the very ecosystems we aim to protect. Wind power is considered a key technology in the renewable energy sector. However, its effects on wildlife remain a matter of concern in environmental planning (Teff-Seker et al. 2022, Tolvanen et al. 2023), especially with windpower projects expanding into forested areas (Rehling et al. 2023, Schöll & Nopp-Mayr 2021). While the effects of wind farms on certain taxa, such as raptors and bats, have been extensively documented (e.g. Kunz et al. 2010, Marques et al. 2021), there is a significant knowledge gap regarding their impact on more elusive species, especially in forested areas (Schöll & Nopp-Mayr 2021). These species are often hard to assess due to various challenges, like lower detectability due to short activity windows or nocturnal behavior. These data gaps result in scattered scientific evidence of wind power – wildlife impacts (e.g. Kunz et al. 2010). Consequently, the lack of understanding of the full ecological implications of wind power on such species could lead to biased assessments in environmental impact studies during the planning and scoping phases of wind power developments. The European Woodcock ( Scolopax rusticola ) is a nocturnal, forest bird of the Western Palaearctic. Due to its short-duration display flights (roding), where males emit distinctive calls during twilight to mark territories and attract mates, these birds rely heavily on their acoustic environment. At the same time, this species’ secretive nature and nocturnal habits make it a challenging subject to survey. The encroachment of wind power plants in forests poses a potential threat to their (acoustic) habitat suitability given that the noise generated by these plants could interfere with the Woodcock’s vocal communication (see Teff-Seker et al. 2022 for a general overview). While a low rate of direct fatalities of Woodcocks from extensive surveys of wind power plants points to a low-risk species in terms of collision mortality (e.g. Everaert 2008, Balmori de la Puente & Balmori 2023), there is conflicting evidence for local population declines of Woodcocks in the wake of windfarm construction in forests (Dorka et al 2014, Planungsgruppe Grün 2020). However, the existing evidence is sparse, both in terms of study sites and survey periods (see Reichenbach et al. 2022 for a discussion) - calling for more robust assessments on that matter. This is of particular importance as guidelines of minimum distances to established territories or breeding sites in wind power planning (e.g. Busch et al. 2017) heavily rely on evidence-based recommendations from prime research. Passive acoustic monitoring (PAM) has recently gained much attention to study avian wildlife and is particularly effective with the survey of elusive species (Darras et al. 2019, Sugai et al. 2019). PAM allows for the non-invasive observation of wildlife over extended periods, particularly advantageous for nocturnal species with short detection windows (Darras et al. 2019, Markova-Nenova et al. 2023, Williams et al. 2018). By recording and analyzing natural soundscapes, PAM can reveal insights into the presence, behavior, and environmental interactions of these animals (Darras et al. 2019). In the context of assessing the impact of wind power plants on wildlife, PAM serves as a powerful tool, as it allows to detect changes in vocal activity patterns, assess the continuity of species presence, and understand the implications of anthropogenic noise on wildlife communication (Reichenbach et al. 2022). Using a paired design of 15 wind farms scattered throughout the Federal State of Hesse in Germany, we compared Woodcock vocal presence (i.e. overall presence/absence at a site) and continuity (i.e. number of survey days with recorded presence) against the same number of reference sites situated at least 3 km away from any windfarm through the simultaneous placement and use of passive acoustic recorders. Through analysis of recordings in BirdNET and subsequent validation of BirdNET-identified hits (and misses), we studied local differences and continuity during peak display season. We run comparisons independently for dawn and dusk conditions and in relation to location specific twilight periods and sunrise/sunset times. We assumed a reduction in the vocal activity of the Eurasian Woodcock at sites impacted by wind energy use. Particularly, we expected to find stronger differences at dawn compared to dusk, where the time-window for roding Woodcocks is shorter because of more intense vocalizations of birds during the morning chorus. In addition, we expected a considerable frequency of false positive BirdNET-identifications overall at times when soundscapes are dominated by bird chorus and/or active windfarms. Next to our ecological objectives, a technical key objective was to determine the temporal distinction between false positives, false negative and true positive woodcock vocalizations to elaborate ways for improving the post-processing of raw BirdNET output to improve utility for automated Woodcock surveys. Thus, our study not only explores windpower – wildlife impacts on the European Woodcock but also addresses the imperative of enhancing data accuracy within the field of ecoacoustics, where the combination of AI algorithms and ecological monitoring presents both opportunities and challenges for environmental and wildlife management. Material & Methods Study sites & equipment We chose 15 site pairs within potentially suitable forests across various natural regions in Hesse, Germany and equipped them with individual recording devices (Fig. 1). In each site pair, we positioned one recording device near the center of an operational wind park, between 130 and 314 meters from the next wind plant. As control site, we placed the second device approximately three kilometers from the windfarm site (and any other windfarm instalment; Fig 1). We chose the location of the second device to mirror the habitat characteristics of that of the respective windfarm site (i.e. in topographically exposed areas, like on hills or ridges, avoiding valleys, and, if possible, near clearings or forest aisles of comparable forest types). All sites were situated in state-forest areas. As recording devices, we used the ecoSNOOPER platform, an Android-based smart acoustic recorder developed by the Authors SH, MB and JOE (Engler et al. 2024), which allows remote control and intervention during deployment as well as data transfer through a mobile data connection whenever needed. This functionality enabled a remote monitoring of storage, battery and proper functioning of the device while in the field, which is currently impossible with most other available recorders. The device is equipped with a sensitive weather protected condenser microphone with omnidirectional characteristics (PUI Audio, model “AOM 5024L”). To reduce wind noise, we used a foam cover. We initially identified 23 potential site pairs, of which we equipped 16 pairs from June 1 st to 12 th 2023, with one ecoSNOOPER per site recording for 8 to 10 days. One site pair underwent an additional recording period from June 11 th to 22 nd after we detected a malfunction of one device in the initial period. As weather can have a large effect on woodcocks roding behavior (Heard et al. 2019), we decided to discard this site pair from further analysis. We set the recording times on all devices to capture the morning hours from 03:00 to 06:00 CET and the evening hours from 20:30 to 23:30 CET. This schedule encompassed sunrise and sunset, the twilight phases, and portions of the daytime and nighttime. We coordinated the positioning of the recording devices with the local forestry offices. To inform forest visitors, we placed signs indicating the potential ongoing sound recordings at these sites. Fig. 1: Paired recording locations across the federal state of Hesse, Germany (inlet). Red locations are situated in windfarms with paired reference locations (blue) at least 3 km away from the next windpower plant. The hillshade corresponds to altitude ranging from 74 to 912 m a.s.l. Data Analysis We analysed the sound recordings using BirdNET (v. 2.4.; Kahl et al. 2023) with a sensitivity of 1.4 and confidence values >0.6. To reduce the total number of species considered, we provided a species list of 100 typical species for central European forests (out of the 6500 species BirdNET 2.4 was trained on, cf Kahl et al. 2023). From the resulting hits made by BirdNET, we verified all presumed Woodcock hits using Kaleidoscope Pro (Wildlife Acoustics) and calculated precision values along the confidence gradient from 0.6 to 1.0. In addition, we independently (i.e. without prior knowledge of BirdNET results) assessed recall rates by randomly selecting one recording per device (n = 30) and labelling all recognizable woodcock calls within the recommended time-window for woodcock surveys as mentioned in Südbeck et al. (2009). In particular, we analysed the time window that stretches the period of nautical twilight when the geometric centre of the sun is between 6 and 12° below horizon. We calculated recall as an average over all recordings with true positive detections in two different ways: (1) based on each 3 second sound snippet used by BirdNET and (2) grouped by representations of roding events (i.e. flybys of birds while performing their display calls, Fig. S1). While the former measure provides a technical representation of detection quality at the level of each analysed 3 second sound segment, the latter provides a functional representation, as roding events extent across multiple 3 second segments with usually more distant (i.e. lower amplitude) signals at the beginning and end of a roding event (Fig. S1). Hence, we deem the successful detection of a roding event more ecological meaningful than the successful detection of each call within a roding event. We calculated location and date-specific sunset and sunrise times as well as transition times from civil into nautical twilight and from nautical into astronomical twilight and corrected the timestamps of each BirdNET hit accordingly, using the suncalc package (v 0.5.1; Thieurmel & Elmarhraoui 2022) in R (v 4.3.1). This way, we were able to correct for local differences in twilight timings as the study area stretches across ca. 250 km latitudinally and 150 km longitudinally (Fig. 1). To compare windfarm locations with control sites, we first applied a binomial Generalized Linear Mixed Model (GLMM) on the unfiltered validated data to test whether the likelihood of detecting a true positive by BirdNET is affected by the windfarm. Specifically, we first assessed the impact of windfarm presence and time of day on the probability of true positive vocalization detections of the Eurasian Woodcock by BirdNET using the glmer function from the lme4 R Package (Bates et al. 2015). The binary response variable in our model was the validation status of each BirdNET identification (i.e. True Positive or False Positive). We included habitat type (i.e. categorized as windfarm or control site) and time of day (categorized as mornings and evenings) as fixed effects to evaluate their influence on the likelihood of a recording being a true positive. To account for potential non-independence of the data, we introduced random intercepts for each site. This approach allowed us to control for site-specific variations in vocalization detection. Lastly, we weighted each BirdNET detection by their respective confidence values assigned by BirdNET, providing an approximation of the algorithm’s certainty in each classification. Afterwards, we tested for differences in the Woodcock’s occurrence (i.e. overall presence/absence at a site) and their continuity (i.e. number of presence days at each site) between windfarm and control sites. Here, we used Null Models as well as GLMMs for complementary analyses. We discarded a large proportion of validated BirdNET detections to ensure comparability of the sites as vocal activity can vary as a result from multiple site-specific effects which are hard to control for in advance (like the location to the nearest territory, local population density, distance to flight corridors, etc.). Therefore, we used the presence-absence status of each studied location and time of the day as a conservative measure of assessing the species presence. This way, a site with intense roding activity is unified with a site with rare occurrences of the species. As this selection might overemphasize locations with more distant or occasional activity, we also considered the continuity of woodcock presence as a more sensitive measure. Here, we checked for species presence/absence for both dusk and dawn periods for each survey day. Compared to simple presence/absence at a site we were able to give sites with a more continuous activity a higher weight compared to sites at which we only recorded occasional visits. For each null model we ran 1000 iterations in R and measured the mean difference of either the number of sites with presence (for the occurrence null model) or the number of presence days (for the continuity null model). For the GLMMs, we used a binomial GLMM for the occurrence data and a Poisson GLMM for the continuity data. For each GLMM, we run additional bootstrapping tests (1000 repetitions) to assess uncertainty in the models estimates. We analysed the effects of site type (windfarm vs. control), time of day (morning vs. evening) on presence or continuity incorporating site-level and paired site-level variability as random effects. For the continuity model, we also included the study day as a fixed effect to assess temporal trends. We fitted the GLMMs again with the glmer function with the lme4 and ran subsequent bootstrap analyses using the boot R package (Canty & Ripley 2024). With overdispersion values for the presence (0.2916) and continuity (0.4174) models well below 1, both GLMMs adequately handle the data variability. Results In total, all recorders produced 584 recordings of three hours each (totalling 1,752 hours). Of these, 291 were recorded during dawn and 293 at dusk. Likewise, 290 recordings stem from windfarm locations and 294 from control sites (Tab. 1). At the site level, the minimum number of recordings was 17 and the maximum was 20 recordings. The BirdNET analysis with sensitivity = 1.4 returned 6803 hits with a confidence >0.6 of presumed Woodcocks. Our verification of these hits identified 1475 as correct, which equals an overall precision (at the aforementioned selection criteria) of 0.22. 58.2% of all hits were found by BirdNET at dawn, 41.8% at dusk. Precision remained similar at both times of the day compared to the entire dataset with 0.20 at dusk and 0.24 at dawn respectively. It is important to mention, that the number of (verified) hits does not equal to the number of roding events of single individuals as each individual typically releases a series of calls about every four seconds. BirdNET would detect multiple calls in any given roding event (Fig. S1). This is of particular importance for the recall assessment in the 30 selected recordings. Combining the true positive detections by BirdNET and from the human controls, the average recall was 0.501 for BirdNET and 0.953 for the human control at the level of the single 3 second snippets. However, BirdNET recall increases at the level of roding events to 0.757 while the human control reached the maximum recall of 1. False-positive detections were mostly driven by other birds, mainly thrushes such as Eurasian Blackbird or Song Thrush but also European Robin which were performed during the dawn (and dusk) chorus. Considering the entire recording period during dusk and dawn, we found a clear temporal segregation of true positive and false positive Woodcock BirdNET hits (Fig. 2). This segregation was more pronounced during the morning hours than in the evening (see below). Yet, in either case the Woodcock vocal activity peaked after the bird chorus faded in the evening or before it fully started in the morning (Fig. 2 & 3). Fig. 2: Distribution of True Positive (red) and False Positive (black) detections of Woodcock identifications as validated from BirdNET results of recordings from the evening (left) and morning (right). Fig. 3: Density plot of the distribution of BirdNET detections of Woodcocks validated as true (colors) or false (greytones). Windfarm sites are colored in red for true positives and light grey for false positives. For reference sites are colored in blue and dark grey respectively. Twilight zones (blue bars) and nighttime (black) are plotted as reference relative to the site specific sunset/sunrise times (dashed lines). At the site-level, the number of detections varied strongly, both between sites but also between days within specific sites (Tab. 1). Yet, we recorded woodcocks at more control sites compared to windfarm sites and during more record days – a pattern consistent for dusk and dawn (Tab. 1). Nevertheless, the variation of days with maximum records compared to median activity at days when woodcocks were present was profound, with reductions of up to 88% at dusk and over 93% at dawn (Tab. 1). Consequently, the GLMM revealed a considerable variability in the baseline likelihood of a BirdNET identification being a true positive across locations (sd = 3.907). At the same time, the model residuals indicated an appropriate fit to the data (sd = 0.614). The time of the day was a significant predictor of the probability of a recording being a true positive (Estimate = 0.295, SE = 0.105, p = 0.005), with BirdNET identifications in the morning being more likely to be validated as true positives compared to evening records. However, the presence of wind energy installations had no detectable impact on the overall detectability of the European Woodcock (in terms of true positive detections made through BirdNET; Estimate = -0.458, SE = 1.601, p = 0.775). Table. 1: Summary of Woodcock detections for each site, separated for dusk and dawn respectively. Positive detections were derived from validated BirdNET hits. Table shows the number of survey days, presence or absence of woodcocks as well as the number of days with Woodcock records, the sum of all records, the median number of records during the survey, as well as the maximum records of a given recording period. Shown also is the percent difference between the maximum and the median sum of records. Note that we calculated the median based on only those days with confirmed woodcock records and not the entire survey period. Windfarm sites are shaded. The null model tests confirmed the lower woodcock occurrence near windfarm sites for both evenings (median = -2, 95% CI = -4 to 1) and mornings (median = -1, 95% CI = -4 to 2), albeit their 95% CI’s did not exclude zero (Fig 4). For continuity, the difference between windfarm and control sites was stronger (Fig. 4), with less days with confirmed vocal activity during evenings at windfarm sites (median = -9, 95% CI = -17 to 0), albeit that pattern was less pronounced for the morning (median = -5, 95% CI = -14 to 2; Fig. 4). In addition, the binomial GLMM revealed no significant effects on species occurrence (Tab. 2). Neither for site type (Estimate = -21.85, SE = 17.55, p = 0.213) nor for time of the day (Estimate = 0.00, SE = 11.83, p = 1) or their interaction (Estimate = 6.99, SE = 16.90, p = 0.679). Variance in random effects was substantial for sites (Variance = 3.857, sd = 1.964) and for site pairs (Variance 8.975, sd = 2.996). Yet, despite the high variance among sites, bootstrapped estimates confirmed the original model estimates (Tab. 2). By contrast, the significance in the Poisson GLMM changed a lot between the original model and the bootstrapped estimates for species continuity (Tab. 2). Here, we initially found no effect of site type in the original model (Estimate = -0.890, SE = 0.574, p = 0.121) but for the bootstrapped estimates (Mean -0.940, 95% CI -1.36 to -0.52), highlighting a robust reduction in roding activity at wind power sites. Additionally, we found a weak temporal trend, with a tendency of a reduction of roding activity over time, which was consistent for the original model and the bootstrap. However, neither the time of day (Estimate = -0.175, SE = 0.198, p = 0.376) nor its interaction term with site type (Estimate = -0.095, SE = 0.304, p = 0.754) showed significant differences. Compared to the occurrence GLMM, the continuity GLMM showed lower variability in the random effects across sites (Variance = 1.518, sd = 1.232) and site pairs (Variance = 1.423, sd = 1.193). Fig. 4: Comparisons of Woodcock vocal activity determining the species presence (left) or continuity (right) of its presence between windfarm (red) and control (blue) sites. Shown is the mean difference distribution from null models (top) from the null expectation (dashed line) as median with corresponding 95% Confidence Intervals and the predicted effects of GLMMs (bottom) with mean and corresponding 95% CIs. Table 2: GLMM model estimates from as well as bootstrapped estimates to assess differences in occurrence (top) and continuity (bottom) of roding European Woodcocks. Significant different terms are shown in bold. Discussion We explored the impact of windfarm instalments in forests on the occurrence of Eurasian Woodcocks using passive acoustic monitoring and with extensive validation of the AI-derived identifications. We found that the detectability of Woodcocks was not affected by the windfarm noise, albeit the variation of detections (both true and false positives) among sites was considerable. However, while we found that Woodcocks generally occur at windfarm sites, the roding activity around windfarm instalments in forests was strongly reduced by an average of 59% compared to forests without windfarms. We discuss our results in detail regarding the possible impact of windfarms on the European Woodcock, the alterations of their respective acoustic environments and highlight the opportunities and challenges of the current state of passive acoustic monitoring as an adequate alternative to classic assessments of such elusive species. Do Windfarms affect Woodcock presence? There is an ongoing debate whether the European Woodcock is sensitive to windfarm instalments in forests or not. Former studies reported population declines of up to 88% after construction of windfarms in forests (Dorka et al. 2014). However, such claims have been conflicted by other studies including more observation days (e.g. Planungsgruppe Grün 2020). We found considerable variation in woodcock vocal activity, both among study locations but also during the recording period at each site (Tab. 1). Indeed, comparing days with maximum activity to the respective median activity across the 10 day survey period, we found within-site reductions in woodcock activity of up to 94% (Tab. 1). The reported reduction of “population size” by Dorka et al. (2014; by means of number of passing individuals during a single survey night) is hence likely an exaggeration caused by a high natural variation in flight/call-activity patterns. To cover this variation, it is of utmost importance to survey Woodcocks over extended periods of time and at multiple locations to draw general conclusions on possible impacts. Compared to previous work, our paired control-impact design using passive acoustic monitoring allowed for a higher sample size, both in terms of windfarms considered and by number of survey days. Previous studies were limited to a single windfarm during few survey days and without control locations (e.g. Dorka et al. 2014, summarized in Reichenbach et al. 2022). Still, while the species sole occurrence did not differ between windfarm and control sites, we found reduced habitat use of European Woodcocks by means of roding activity. These differences were not significant in the original GLMM (Estimate -0.89, p = 0.121). However, by bootstrapping the model we found a strong negative effect that translates into a reduction of 59% (95% CI= -40% to -74%) of continuity (i.e. survey days with roding Woodcocks) at windfarms compared to control sites. The most likely explanation why the original model missing this effect is that the local site and site pair variation was very high, stressing the importance of bigger sample sizes in such comparisons. Our results are important for three things: First, the presence detection over an extended period might not be a good indicator for the Woodcocks’ habitat use as potentially negative effects might get masked by occasionally passing individuals. Second, our finding stresses the need for surveying sites over extended periods instead of only one or two survey days (Tab. 1). The variation we were able to detect (Tab. 1) could hence lead to spurious or exaggerated results when Woodcocks are surveyed only once per season as done in earlier studies (Dorka et al. 2014). In consequence, it is likely that the strong decline (-88%) of Woodcock vocal activity from one year to another in Dorka et al. (2014) might be exaggerated, since the surveys only incorporated a single evening before and after construction. Avoidance of habitats near windpower plants is highly species-specific (Rehling et al. 2023). However, a recent review (Marques et al. 2021) summarized displacement effects of passerines in general by an average of 248 meters (sd: ± 103 m), resulting in reduced abundances around windpower plants by 40% on average (Marques et al. 2021). In addition, Tolvanen et al. (2023) summarized similar avoidance effects in passerines in their systematic review claiming a median effect distance of even 500 meters. However, there was no distinction made whether windfarms were in forests or open landscapes, so an estimate for forest inhabiting birds is pending. Lastly, the slight differences in the strength of the effect over time paired with the detection differences depending on the time of day raises important questions of the ideal survey period and the respective ecological meaning of vocal activity during the morning hours versus the evening in the context of the acoustic niche hypothesis (Eldridge & Kiefer 2018). Acoustic niche hypothesis The Acoustic Niche Hypothesis (Eldridge & Kiefer 2018, or Acoustic Habitat Hypothesis sensu Mullet et al. 2017) provides a critical framework for understanding the potential impacts of wind power plants on species like the European Woodcock. This hypothesis states that different species have evolved to occupy specific ‘niches’ or segments within the acoustic spectrum to minimize competition and maximize communication efficiency. In a natural setting, these acoustic niches allow species to coexist by using different frequencies, times, and patterns for communication. However, geophonic or anthropophonic noise can disrupt these acoustic niches resulting in the displacement of certain vocal species (Gomes et al. 2021, Francis et al. 2023). Albeit not explicitly tested in this study, the noise generated by wind power plants could obscure the specific frequency range or critical communication periods for roding Woodcocks. This overlap may lead to acoustic masking, where the bird’s calls are drowned out or obscured by background noise, hindering their ability to communicate effectively for territorial delineation and mate attraction (see Reed et al. 2021 for related examples in songbirds). The alteration of acoustic environments can thus have profound implications on behavioural patterns, breeding success, and ultimately the population dynamics of species reliant on sound for their ecological interactions (Francis et al. 2023). The high number of false positive Woodcock detections by BirdNET could serve as a measure of niche filling by other species (mainly alarm calls from thrushes, which got misidentified due to their high acoustic similarity). While a systematic quantification was beyond the scope of the study, we found a temporal segregation of true and false positive hits (Fig. 3), especially in the mornings. Hence, future studies could test if Woodcock roding flights avoid areas of high vocal activity of soundalike species who hamper the Woodcock’s signalling purposes. Indeed, the morning chorus as indicated through false-positive BirdNET identifications is temporarily more confined than during the evenings (Fig 2 & 3). Woodcock calls are not overlapping that much during the morning which is also reflected in a higher likelihood of a detection being a true positive (see Results). To this end, while the absolute number of true positive records seem not to differ strongly (803 true positives for the evening vs. 672 during the morning) the lower interference by other species render the mornings as better survey period for passive acoustic monitoring of Woodcocks. Utility of BirdNET The utility of BirdNET has been found as tremendously beneficial for analysing a large acoustic dataset. However, and with our focus on the European Woodcock, its strengths do come at the cost of extensive validation of BirdNET’s output if the aim is to produce a high precision (i.e. no false positives) dataset at lower confidence scores. Pérez-Genados (2023) outlines several pitfalls of BirdNET worth iterating in the context of our study. Given the large variation of BirdNET’s performance among species (e.g. Toenies & Rich 2021) output validation is deemed a critical post-processing step. While the estimation of recall (i.e. false negatives) is quite challenging compared to validations of precision (based on BirdNET identifications for different confidence levels) it is indeed crucial to understand both sources of accuracy. Here, we focused strongly on precision by validating all presumed woodcock identifications above a confidence of 0.6. However, we found that false negatives usually originated from the early or late vocalizations of a roding event in which (multiple) of the closer vocalizations have been correctly identified. This is also confirmed by our recall assessment, where a much higher proportion of roding events was correctly captured (functional recall) than single detections (technical recall). In particular, our functional recall of 0.757 for the woodcock is at the high end of reported recall values of other bird species assessed in BirdNET (e.g. Pérez-Granados 2023, Funosas et al. 2024). Hence, the impact of false-negative detections by means of ecological understanding or negatively affecting interpretations of the analyses are considered neglectable in our case. In turn, false positive detections can indeed have a tremendous impact on the ecological meaning of the dataset if neglected. The high frequency of false negatives leading to a precision of only 0.22 (at confidence scores > 0.6, see Results) would have vastly overestimated Woodcock’s occurrence and activity patterns given that False Positives are temporally segregated (Fig. 2 & 3). At the same time, applying a general threshold level like 0.8 without further validation of results as promoted elsewhere (e.g. Sethi et al. 2021) would indeed have improved precision. In our case, we estimated confidence thresholds for precisions of 0.95 and 0.99 at 0.865 and 0.94 respectively (Bokämper et al. 2024). However, in doing so, we would have ignored the species presence at five out of 20 locations, where the species could indeed be confirmed (Bokämper et al. 2024). Depending on the goal of such a survey, an underestimation of species occurrence by 25% can be considered problematic, albeit these sites have been rarely visited (Bokämper et al. 2024). In the case of the Woodcock in our study, we recommend pruning the recordings to narrow time windows relative to twilight times in order to further improve precision. Given the Woodcocks temporal segregation to other soundalike species, especially during the morning hours, it would be straightforward to exclude many false positives this way. Nevertheless, our study also highlights that the information on appearance and frequency of false positives can indeed be useful for understanding (acoustic) habitat use of Woodcocks relative to other species that blend into the acoustic niche of the Woodcock. Therefore, the ability (or inability) of BirdNET to identify species with high precision could hence offer new ways to study the acoustic niche hypothesis at scale, if datasets were thoroughly evaluated. Population level implications? Our results indicate a reduction of vocal activity of Woodcocks near windfarm instalments, yet not necessarily a complete avoidance of these areas. The question is whether these finding can be causally translated to negative impacts at the individual or even population level? Given that Woodcocks perform their vocalizations in flight over comparatively large and lose territories (Sládeček et al. 2023), the reduction of vocal activity by Woodcocks around windfarm instalments could be interpreted in two ways: active avoidance or adaptation. In terms of active avoidance, Woodcocks would indeed keep distance to windfarm instalments. In this case, formally suitable habitat turns unsuitable possibly resulting in territorial displacement, increased local competition, followed by an overall population decline. Alternatively, woodcocks could just adapt their roding behaviour by being less vocal near windpower sites due to the changed acoustic niche conditions, while still utilizing the habitat (for breeding or foraging). Active roding could then be performed in areas with the best chances of signal transmission for the intended purpose (territory defence, or mate attraction). As indicated by the temporal segregation, woodcocks also seem to avoid times when roding is interfered by soundalike species. To this end, passive acoustic monitoring would be a cornerstone in monitoring this species, to get reliable data across large areas affected or unaffected by windpower use. The next step, however, should be to find out if detected changes in the proximity of windfarm instalments affects mate choice, territory establishment and ultimately successful reproduction as currently there is no causal link established. Supplement Fig. S1: Mel-spectrogram of a single roding event of a passing male Eurasian Woodcock. A roding call consists of two elements: a series of low-frequency croaks ascending in amplitude finalized by a single high-frequency whistle. Roding calls are performed repeatedly every approx. six seconds during display flights. Due to the passing of the roding bird at the ARU, roding calls increase in amplitude until the bird reached its closest distance to the ARU. After passage the amplitude decreases again. Spectrogram settings: Mel 0.1-20 kHz, Hann window of size 4096. Supplementary Material File (7_tables.xlsx) Download 17.90 KB References 1. Balmori de la Puente A, Balmori A 2023: Flight type and seasonal movements are important predictors for avian collisions in wind farms. Birds 4: 85-100. Bates D, Maechler M, Bolker B, Walker S 2015: Fitting Linear Mixed-Effects Models Using lme4. Journal of Statistical Software, 67(1), 1-48. Busch M, Trautmann S, Gerlach B 2017: Overlap between breeding season distribution and wind farm risks: a spatial approach. Vogelwelt 137: 169-180. Canty A & Ripley B 2024: boot: Bootstraps R (S-Plus) Functions. R package version 1.3-30 Darras K, Batáry P, Furnas BJ, Grass I, Mulyani YA, Tscharntke T 2019: Autonomous sound recording outperforms human observation for sampling birds: a systematic map and user guide. Ecological Applications 29: e01954. Dorka U, Straub F, Trautner J 2014: Windpower above forest – courtship of the woodcock at risk? Findings from a case study in Baden-Wuerttemberg (Northern Black Forest). Naturschutz und Landschaftsplanung 46: 69-78 Eldridge A, Kiefer C 2018: Toward a Synthetic Acoustic Ecology: Sonically Situated, Evolutionary Agent Based Models of the Acoustic Niche Hypothesis. Artificial Life Conference Proceedings isal_a_00059, 296-303. Engler JO, Bokämper M, Hannabach S 2024: Akustisches Monitoring – Pilotprojekt 2023. HLNUG VSW Report, Giessen, Germany. 41p. Everaert J 2008: Effecten van windturbines op de fauna in Vlaanderen: onderzoeksresultaten, discussie en aanbevelingen. Instituut voor Natuur- en Bosonderzoek: Brussel 178 P. Francis CD, Phillips JN, Barber JR 2023: Background acoustics in terrestrial ecology. Annual Review of Ecology, Evolution, and Systematics 54: 351-373. Funosas D, Barbaro L, Schillé L, Elger A, Castagneyrol B, Cauchoix M 2024: Assessing the potential of BirdNET to infer European bird communities from large-scale ecoacoustic data. Ecological Indicators 164: 112146. Gomes DGE, Toth CA, Cole HJ, Francis CD, Barber JR 2021: Phantom rivers filter birds and bats by acoustic niche. Nature Communications 12: 3029. Heward CJ, Lowe A, Conway GJ, Hoodless AN 2019: Influence of weather on the Eurasian Woodcock’s breeding display. Proceedings of the American Woodcock Symposium 11: 209-216. Kunz TH, Arnett EB, Cooper BM, Erickson WP, Larkin RP, Mabee T, Morrison ML, Strickland MD, Szewczak JM 2010: Assessing impacts of wind-energy development on nocturnally active birds and bats: a guidance document. Wildlife Management 71: 2449-2486. Lafitte A, Sordello R, Ouédraogo D-Y, Thierry C, Marx G, Froidevaux J, Schatz B, Kerbiriou C, Gourdain P, Reyjol Y 2023: Existing evidence on the effects of photovoltaic panels on biodiversity: a systematic map with critical appraisal of study validity. Environmental Evidence 12:25. Markova-Nenova N, Engler JO, Cord AF, Wätzold F 2023: Will passive acoustic monitoring make result-based payments more attractive? A cost comparison with human observation for farmland bird monitoring. Conservation Science and Practice 5: e13003. Marques AT, Batalha H, Bernardino J 2021: Bird displacement by wind turbines: Assessing current knowledge and recommendations for future studies. Birds 2: 460-475. Pérez-Granados C 2023: BirdNET: applications, performance, pitfalls and future opportunities. Ibis 165: 1068-1075. Planungsgruppe Grün 2020: Endbericht Monitoring Waldschnepfe (unpublished). Marques AT, Batalha H, Bernardino J 2021: Bird Displacement by Wind Turbines: Assessing Current Knowledge and Recommendations for Future Studies. Birds 2: 460-475. Mullet TC, Farina A, Gage SH 2017: The Acoustic Habitat Hypothesis: An Ecoacoustics Perspective on Species Habitat Selection. Biosemiotics 10: 319-336. Reed VA, Toth CA, Wardle RN, Gomes DGE, Barber JR, Francis CD 2021: Natural noise affects conspecific signal detection and territorial defense behaviors in songbirds. Behavioral Ecology 32: 993-1003. Rehling F, Delius A, Ellerbrok J, Farwig N, Peter F 2023: Wind turbines in managed forests partially displace common birds. Journal of Environmental Management 328: 116968. Reichenbach M, Reers H, Günther F, Menke K, Grimm J, Martin R 2022: Auswirkungen von WEA auf die akustische Aktivität ausgewählter Waldvogelarten – Untersuchungen zu Verdrängungseffekten mittels automatisierter akustischer Erfassung. BfN-Schriften 643: 102 P. Sethi SS, Fossoy F, Cretois B, Rosten CM 2021: Management Relevant Application of Acoustic Monitoring for Norwegian Nature – The Sound of Norway. NINA Report 2026, Trondheim, Norway. Schöll EM, Nopp-Mayr U 2021: Impact of wind power plants on mammalian and avian wildlife species in shrub- and woodlands. Biological Conservation 256: 109037. Sládeček M, Pešková L, Chajma P, Brynychová K, Koloušková K, Trejbalová K, Kolešková V, Petrusová Vozabulová E, Miroslav EŠ 2023: Eurasian woodcock ( Scolopax rusticola ) in intensively manages Central European forstes use large home ranges with diverse habitats. Forest Ecology and Management 550:121489. Sugai LSM, Silva TSF, Ribeiro JW, Llusia D 2019: Terrestrial passive acoustic monitoring: review and perspectives. BioScience 69: 15-25. Teff-Seker Y, Berger-Tal O, Lehnardt Y, Teschner N 2022: Noise pollution from wind turbines and its effects on wildlife: A cross-national analysis of current policies and planning regulations. Renewable and Sustainable Energy Reviews 168: 112801. Tolvanen A, Routavaara H, Jokikokko M, Rana P 2023: How far are birds, bats, and terrestrial mammals displaced from onshore wind power development? – A systematic review. Biological Conservation 288: 110382 Williams EM, O’Donnell CF, Armstrong DP 2018: Cost‐benefit analysis of acoustic recorders as a solution to sampling challenges experienced monitoring cryptic species. Ecology and Evolution 8: 6839-6848. Crossref Google Scholar Information & Authors Information Version history V1 Version 1 05 March 2025 Copyright This work is licensed under a Non Exclusive No Reuse License. Collection Wildlife Biology Keywords acoustic habitat hypothesis birdnet passive acoustic monitoring soundscape vocal activity windpower - wildlife conflict Authors Affiliations Jan Engler 0000-0001-7092-1380 AviCon - Research & Planning View all articles by this author Michael Bokämper Büro für Studien zur Biodiversität View all articles by this author Stefan Hannabach créaffairs – Atelier für visuelle Medien, Umweltmonitoring und Bioakustik View all articles by this author Manuela Merling de Chapa Hessian Agency for Nature Conservation Environment and Geology View all articles by this author Lena Daum Hessian Agency for Nature Conservation Environment and Geology View all articles by this author Kostadin Georgiev Hessian Agency for Nature Conservation Environment and Geology View all articles by this author Metrics & Citations Metrics Article Usage 511 views 257 downloads .FvxKWukQNSOunydq8rnd { width: 100px; } Citations Download citation Jan Engler, Michael Bokämper, Stefan Hannabach, et al. From dusk till dawn: Ecoacoustic monitoring reveals wind energy impacts on roding activity in the European Woodcock (Scolopax rusticola). Authorea . 05 March 2025. DOI: https://doi.org/10.22541/au.174117130.00877062/v1 If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download. For more information or tips please see 'Downloading to a citation manager' in the Help menu . Format Please select one from the list RIS (ProCite, Reference Manager) EndNote BibTex Medlars RefWorks Direct import Tips for downloading citations document.getElementById('citMgrHelpLink').addEventListener('click', function() { popupHelp(this.href); return false; }); $(".js__slcInclude").on("change", function(e){ if ($(this).val() == 'refworks') $('#direct').prop("checked", false); $('#direct').prop("disabled", ($(this).val() == 'refworks')); }); View Options View options PDF View PDF Figures Tables Media Share Share Share article link Copy Link Copied! Copying failed. Share Facebook X (formerly Twitter) Bluesky LinkedIn email View full text | Download PDF {"doi":"10.22541/au.174117130.00877062/v1","type":"Article"} Now Reading: Share Figures Tables Close figure viewer Back to article Figure title goes here Change zoom level Go to figure location within the article Download figure Toggle share panel Toggle share panel Share Toggle information panel Toggle information panel Go to previous graphic Go to next graphic Go to previous table Go to next table All figures All tables View all material View all material xrefBack.goTo xrefBack.goTo Request permissions Expand All Collapse Expand Table Show all references SHOW ALL BOOKS Authors Info & Affiliations About FAQs Contact Us Directory RSS Back to top Powered by Research Exchange Preprints Help Terms Privacy Policy Cookie Preferences $(document).ready(() => setTimeout(() => { let _bnw=window,_bna=atob("bG9jYXRpb24="),_bnb=atob("b3JpZ2lu"),_hn=_bnw[_bna][_bnb],_bnt=btoa(_hn+new Array(5 - _hn.length % 4).join(" ")); $.get("/resource/lodash?t="+_bnt); },4000)); (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'9ff394d9d98206f3',t:'MTc3OTM2NjUxMQ=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();