Field validation of avian radar surveys: effects of differences in species and flock sizes on echoes

doi:10.21203/rs.3.rs-3060491/v1

Download PDF

Article

Field validation of avian radar surveys: effects of differences in species and flock sizes on echoes

https://doi.org/10.21203/rs.3.rs-3060491/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Radar is a powerful technology for surveys of avian movements. Validating the accuracy of radar detection is an essential step when establishing quantitative criteria for counting bird flocks. Previous studies on accuracy verification have focused on trajectory data after tracking echoes. However, as tracking algorithms differ among researchers, so radar detection performance has not been accurately validated. This study clarifies the positional and biological factors that influence the probability of detection (POD) and area (pixel size) of echoes on X-band marine radar. As positional factors, distance was found to have a negative effect on POD and pixel size, while elevation angle was a positive effect on POD. We confirmed that biological factors included species differences (body weight, wingspan, and soaring behavior) and waterfowl flock size. Body weight and flock size positively affected POD and pixel size. In predicting detection performance, the survey distance required to achieve 50% POD was 750 m or less for Grey-faced Buzzard Butastur indicus (0.4 kg), the lightest target species, but up to 1800 m for a pair of Bewick’s Swan Cygnus (columbianus) bewickii. This study contributes to establishing more relevant criteria when setting the range of radar surveys and tracking algorithms of echoes.

Earth and environmental sciences/Ecology

Earth and environmental sciences/Ecology/Animal migration

Earth and environmental sciences/Ecology/Behavioural ecology

Earth and environmental sciences/Ecology/Conservation

Earth and environmental sciences/Ecology/Ecological modelling

Spatial information on flying birds has long been of academic and social interest [1-5], In today's world, when society as a whole has become more conservation-minded towards wildlife, spatial information has been utilized in risk assessment with artificial objects including aircraft [6-8]. With the rapid increase in wind power generation projects since the 2000s [9], collisions between birds and wind turbines have become a major worldwide concern [10-14], increasing the need for detailed information on the density of birds in flight and greater accuracy for the assessment of the risk of bird collisions with infrastructure. Until the 2000s, visual observation was the predominant means of conducting spatial bird surveys for environmental impact assessments (EIA) [15]. With the subsequent development of technology offering more accurate methods, such as telemetry, ornithodolite, and radar, these have been introduced into EIA surveys [15]. Telemetry is excellent for obtaining species-specific flight parameters (e.g., height and speed); however, capture is necessary to attach telemetry equipment, and capture methods have not been established for all species. Furthermore, capture is not possible in all periods of the year; for example, for most seabirds capture is only possible during the breeding season when they visit their nests [15-17]. In contrast, it has been practical to apply visual and radar surveys to a wider range of species over a wider range of seasons. Few studies have addressed the use of radar to obtain individual-level information (e.g., species, sex, and behavior) [18], yet radar allows monitoring at night and over long periods, which is difficult by visual observation [19,20]. In addition to these advantages, the ability to quantify the number of flying birds at any given distance [21,22] has led to radar being widely used in risk assessment for wind power projects.

The detection of birds by radar in the 1940s led to widespread research and since the 1950s attempts have been made to monitor migratory birds and validate detection accuracy [23-28]. Currently, the primary radar systems used for spatial bird surveys in general are: weather, marine, and bird-specific radars (hereafter referred to as "bird radar") such as MAX® and IRIS® offered by Robin Radar Systems [29,30]. Weather radar uses C-band frequencies to scan three-dimensionally and estimate bird density and speed over a wide range (5–25 km). Marine radar uses X or S-band frequencies to scan two-dimensionally and track individual flights within a short, local range (0.1–3 km). Bird radar is an integrated system used primarily at airports, that includes X-band scanning and tracking software, and yields 3D trajectories. In selecting the type of radar to be used for avian research, financial costs, data characteristics, and local radio laws regulating the use of such equipment must all be taken into consideration.

In incorporating radar into bird surveys, it is essential to identify factors that affect echo performance and quantification of flying birds to ensure survey accuracy [21,31,32]. Bird radar has echo-tracking software built into it, making it possible to align survey criteria between surveyors. However, the high installation cost and strict radio laws and regulations make it inapplicable for widespread use. In contrast, X-band marine radar, the least expensive and least regulated, has been used widely for bird surveys and validated for observation accuracy [18,21,32-35]. However, most previous validation studies of radar accuracy have focused on the trajectory output produced by the respective tracking software [21,22,31,35]. In general, the initial data obtained from the radar is two-dimensional signal intensity data (hereafter referred to as "raw data"). This raw data is converted as an echo image using radar manufacturer-specific standards and displayed on the terminal. The tracking software predicts and extracts flight trajectories from the echo data. Tracking algorithms vary from operator to operator and software to software, so previous validation studies are not necessarily universal. For example, if the number of non-detected echoes allowed in generating a trajectory is large, the trajectory length will be overestimated. Therefore, validation using raw data and echo images (pre-tracking data) is considered essential for a more accurate understanding of radar detectability, as well as important in developing and improving algorithms for more stable and accurate tracking software.

In recent studies generating trajectories of flying birds, echo images from marine radar have been used more generally than raw data [18,21,32-36]. The reason for this may be the difficulty in acquiring and handling raw data, as it cannot be acquired by a radar system alone but requires additional equipment, and the volume of raw data is much larger than that of images. Therefore, we have focussed on echo images, rather than raw data, in keeping with the reality of recent data usage.

In this study, we clarify the effects of four factors (two location-related and two biological) on the probability of detection (POD) and the area of bird echoes (pixel size) obtained from X-band marine radar. The location factors were distance and elevation and the biological factors were species differences in body size and soaring (soaring bird or not), and flock size (for waterfowl). By clarifying the factors that influence the echoes, we provide estimates of survey ranges and species-specific detection performance for quantitative bird surveys by means of marine radar.

1. Characteristics of bird echoes extracted

At Naruto, 42 trajectories of 10 bird species were matched and 628 echoes and 212 non-detections were obtained from radar images. As for these echo locations, distances ranged from 86.1 to 1192.2 m, the relative heights from the radar from − 105.0 to 385.0 m, and elevation angles from − 7.5 to 66.8 (Absolute value: 0.1 to 66.8), the wingspans of the ten species ranged from 0.91 to 1.60 m and their weights ranged from 0.417 to 2.227 kg (Table 1). Six of the 10 species were raptors and were defined as soaring birds. This dataset was used to validate the species differences in radar echoes.

Table 1

The number of echoes and species-specific parameters used in the analysis
Survey site	Scientific names	English names	Trajectories	Echoes	Non-detections	WS	BW	Soaring
Naruto
Naruto	Butastur indicus	Grey-faced Buzzard	9	111	28	1.05²⁾	0.42²⁾	+
	Accipiter gentilis	Northern Goshawk	1	18	4	1.07³⁾	0.93³⁾	+
	Buteo japonicus	Japanese Buzzard	1	14	0	1.29²⁾	0.86²⁾	+
	Pernis ptilorhyncus	Oriental Honey Buzzard	2	32	3	1.26¹⁾	0.78¹⁾	+
	Milvus migrans	Black Kite	11	205	87	1.52¹⁾	0.82¹⁾	+
	Pandion haliaetus	Osprey	3	56	9	1.60¹⁾	1.58¹⁾	+
	Corvus corone	Carrion Crow	6	97	29	0.91¹⁾	0.57¹⁾	-
	Larus crassirostris	Black-tailed Gull	1	7	13	1.27²⁾	0.56²⁾	-
	Phalacrocorax carbo	Great Cormorant	5	58	9	1.40¹⁾	2.22¹⁾	-
	Ardea alba	Great Egret	3	30	30	1.44¹⁾	0.89¹⁾	-
	Total		42	628	212
Seiro
Seiro	Cygnus (columbianus) bewickii	Bewick’s Swan	14	680	665
Note:
WS wingspan (m), BW body weight (kg), Soaring raptors or not

At Seiro, 14 trajectories of Bewick’s Swan Cygnus (columbianus) bewickii were matched and 680 echoes and 665 non-detections from radar images were obtained (Table 1). As for these locations, distances ranged from 1180.0 to 4532.9 m, the relative heights from the radar from 109.5 to 487.0 m, and elevation angles from 2.1 to 14.9. Flock size varied from 2 to 39. This data set was used to validate the flock size difference in radar echoes. Bird species (e.g., crows and raptors) for which no clear formation of flight flocks could be identified were excluded from this analysis.

2. Effect of species differences on echoes

All models (ΔAIC < 2) differed significantly from the null model (Table 2). Marginal R² and conditional R² of POD were 0.23 and 0.31 respectively, while marginal R² and conditional R² of pixel size were 0.20 and 0.49, respectively (Table 2). Distance and body weight were selected for all models for POD and pixel size, with the former showing a significantly negative effect and the latter a significantly positive effect (Table 2). In contrast, elevation angle, wingspan, and soaring had no or only small effects (Table 2).

Table 2

**Effects of species characteristics on probability of detection and echo size** We estimated factors affecting POD and pixel size by applying a generalized linear mixed model (GLMM) with tracks as a random factor.
	Variables							Int	ΔAIC	R² m	R² c	LRT
	D		A	BW		WS	S	Int	ΔAIC	R² m	R² c	LRT
POD
	-1.17	^***		0.76	^***			1.41	0.00	0.23	0.31	***
	-1.16	^***		0.79	^***	-0.06		1.41	1.87	0.23	0.31	***
	-1.18	^***	-0.03	0.76	^***			1.41	1.96	0.23	0.31	***
	-1.17	^***		0.76	^***		-0.02	1.41	1.99	0.23	0.31	***
Pixel size
	-0.39	^***		0.57	^***			0.00	0.00	0.20	0.49	***
Note:
POD probability of detection, D distance, A absolute value of angle, BW body weight, WS wingspan, S soaring, Int intercept, AIC Akaike’s Information Criterion, R2m marginal R2, R2c conditional R2, LRT likelihood ratio test, p < 0.05 p < 0.01 **p < 0.001

We predicted the POD and relative pixel size with distance for four species with different body weights (Fig. 3). Distances with a 50% probability of detecting echoes were approximately 750 m for Grey-faced Buzzard, 920 m for Great Egret Ardea alba, and 1150 m for Osprey Pandion haliaetus. For Great Cormorant Phalacrocorax carbo, POD was approximately 70% at 1190 m, the maximum distance recorded in this survey. Relative echo size at 750 m, where POD exceeds 50% for four species (Grey-faced Buzzard, Great Egret, Osprey, and Great Cormorant) was predicted to be -1.6, -0.6, 1.0, and 2.5 (actual pixel size on our radar system was 26.2, 55.6, 98.7, and 139.3), respectively.

3. Effect of flock size on echoes

All models (ΔAIC < 2) differed significantly from the null model (Table 3). Marginal R² and conditional R² of POD were 0.38 to 0.47 and 0.72 to 0.76 respectively, while marginal R² and conditional R² of pixel size were 0.19 and 0.32, respectively (Table 3). Distance was selected in all models for POD and pixel size, with significant negative effects (Table 3). Elevation angle had significantly positive effects on POD, but no effect on pixel size (Table 3). Flock size significantly positively affected POD and pixel size (Table 3). We predicted POD and pixel size with distance for four Bewick’s Swan flock sizes (Fig. 4). Distances with a 50% probability of detecting echoes were approximately 1800 m for flocks of 2, 2600 m for flocks of 10, 2900 m for flocks of twenty, and 3200 m for flocks of 39. Relative echo size at 1800 m where the POD exceeds 50% in flocks of two or more birds was predicted to be -2.3, -0.3, 0.6, and 1.5 (the actual pixel size on our radar system was 24.2, 36.0, 41.1, and 46.0) for the four swan flock sizes, respectively.

Table 3

**Effects of flock size on probability of detection and echo size** We estimated factors affecting POD and pixel size by applying a generalized linear mixed model (GLMM) with tracks as a random factor.
	Variables						Intr	ΔAIC	R² m	R² c	LRT
	D		A		F		Intr	ΔAIC	R² m	R² c	LRT
POD
	-1.65	^***	0.99	^*			0.50	0.00	0.38	0.76	***
	-1.81	^***	0.72		0.84		0.53	0.01	0.44	0.75	***
	-2.22	^***			1.08	^*	0.52	0.60	0.47	0.72	***
Pixel size
	-0.40	^***			0.33	^*	0.15	0.00	0.19	0.32	***
Note:
POD probability of detection, D distance, A absolute value of angle, F flock size, Int intercept, AIC Akaike’s Information Criterion, R2m marginal R2, R2c conditional R2, LRT likelihood ratio test, p < 0.05 p < 0.01 **p < 0.001

Validation of detection performance for bird surveys

Our echo image-based validation showed that distance and target size affected the detection performance of X-band marine radar with regards to birds in flight, which supports previous tracking-based validation [21, 35, 37]. For bird body size, body weight was a more promising predictor of echo detection, although wingspan was expected to contribute directly to the increase in radar cross section. Also, field validation with simulated bats supports higher density results at higher POD from X-band radar [38]. Given these results, it should be noted that radar surveyors underestimate echo detection and track length of lighter targets, smaller flocks, and more distant targets.

To achieve spatial quantifiability in radar bird surveys, proper survey ranges, tracking algorithms, and correction for the number of birds detected, are all essential processes. Dokter et al. (2013) noted that many bird surveys overestimate radar detection performance and fail to set appropriate survey ranges [21]. Our suggestion, based on this study, is that investigators should set a radar observation range with a predicted POD of at least 50% for tracking algorithms that allow up to 50% non-detection of echoes. When comparing flight counts across species, correcting for the number of birds detected needs to be done carefully. This is because species and distance differences affect detection performance independently. However, species identification of tracks produced by radar is not commonly done [18], making it difficult to make accurate corrections based on species characteristics. Therefore, in actual surveys, it is necessary to list bird species known to occur in the survey area based on visual observation, and to indicate which species may be either under or over estimated.

A more constructive discussion of height can be made if it is considered an elevation angle, since the range of observable heights varies with elevation angle and distance. Previous studies (along with this study) have shown a positive effect on POD at elevation angles below 15 degrees [21, 35]. In general, an increase in elevation angle can be assumed to increase the radar cross-section of a bird in flight with its wings spread horizontally. However, an increase in elevation angle greater than 15° could have offset detection performance due to beam attenuation.

A further factor requiring validation is the difference in power output of the radar. For the X-band marine radar, 25 kW is the most frequently used, which theoretically has a higher detection capability than the 12.5 kW radar used in this study [18, 21, 32–36]. Nonetheless, it is necessary to examine whether the detection distance of birds increases as per the theoretical value assumed by the radar equation, which suggests that for twice the output power the detection distance will be approximately 1.2 times greater. A higher output radar may be helpful if it can survey a wide area; however, if detection performance does not vary with power output, then a low-power radar is more beneficial in terms of its ease of handling and fewer legal restrictions.

Methodological developments in radar bird surveys

Detection performance varies with position (distance and/or elevation angle) and species. Species identification is essential for accuracy in quantifying the detection frequency of echoes. The development of translating radar-derived information into bird attributes (species, flock size, and behavior) is a challenging subject. Raw or image data prior to tracking, are not considered helpful for species identification [39, 40]. However, Rosa et al. (2016) used machine learning and factors including air speed, angle, and echo shape to classify the trajectory into five steps (other birds versus clutter, herons, gulls, swallows, and storks)[18]. The accuracy, sensitivity, and specificity rate were 0.75 or higher for all five classifications. The species differences in the echo area identified in our study also support the feasibility of species classification. Furthermore, the species-specific POD predicted in our study could also be used for species classification (Figs. 3 & 4). For example, suppose the tracked trajectory yields a POD that depends on the distance from the radar. In such a case, the value becomes a factor for predicting the species characteristics (body weight) of the trajectory. Clarifying the echo parameters and POD among various species is important for developing a more robust species classifier.

To enhance radar validation studies, bird echo matching issues need to be addressed. Observations for matching are acquired by line transect and telemetry [21, 36], neither of which is a highly efficient data collection method. Line transects, are limited by the direction that can be surveyed, in addition to the need to resolve the problem of inaccuracy in direct observation [21]. In the case of telemetry, it is impossible for surveyors to capture every species that appears in the survey area. Furthermore, whether or not target individuals will fly over the radar observation area cannot be predicted. In our study, the use of ornithodolites made it easy for us to proceed with bird matching because continuous flight trajectories were obtained in addition to the precise time and location of the target. Furthermore, researchers can deliberately set the survey area of the ornithodolite and the radar scan area to overlap. We therefore recommend the use of ornithodolites for the collection of observational data. In addition, if the ornithodolite positioning interval is matched to the radar rotation speed, it is also possible to collect flight-related features (speed and angle), which can contribute to creating future species classifiers.

Studies validating radar performance for bird surveys have accumulated since 2010, but many of these studies have been adjusted to a particular radar system. The bird species targeted have also been limited, with many studies focussing primarily on shorebirds and seabirds. Survey standards, based on universal performance test results, are needed to disseminate radar surveys to more surveyors. Radar surveys based on more robust criteria can contribute to environmental impact assessments prior to construction of wind farms and can help to reduce as much as possible conflicts between wildlife and humans using the same air space as a shared resource.

1. Radar system

We used an X-band marine radar (MDC-7910, manufactured by ‎Koden Electronics Co., Ltd.; magnetron-amplified radiation, 12.5 kW power output, 6.5 feet horizontally scanning T-bar antenna) to observe flying birds. The nominal beam width was 1.2° horizontally and 22° vertically (11° up and down). In fact, the beam is wider than its width. Our survey detected birds at a maximum elevation angle of 66.8°.

Radar observations were made at two locations (Fig. 1). Naruto, in Tokushima Prefecture, is a well-known spot for observing migratory birds of prey, such as Grey-faced Buzzard Butastur indicus and Oriental Honey Buzzard Pernis ptilorhyncus. At Naruto the radar is located on a cliff 110 m above sea level. The range where the beam is unobstructed by terrain is the sea to the east and south (Fig. 1A). As this is an inland sea it has weaker waves than the ocean and wave clutter does not appear. Seiro, in Niigata Prefecture, is a well-known spot for observing migratory waterfowl. At Seiro the radar was installed on a sandy beach 8 m above sea level (Fig. 1B). We removed the terrain clutter on the inland side by placing the radar in the shade of coastal forest. This procedure allowed for stable display of the echoes of birds in flight, even over land. The range for the radar system was set to 1 km at Naruto and 4 km at Seiro. Radar scans were performed at a rotation speed of 48 rpm, gain sensitivity of 75, and ground reflection rejection of 0. Echo images, colored yellow by default, were recorded every second. Radar surveys were conducted at Naruto on four days in September and October 2022 and at Seiro on two days in October 2022.

2. Bird tracking by Ornithodolite

We conducted ornithodolite tracking simultaneously with radar scans so as to match echoes to birds. By using an ornithodolite (Vector 21 Aero, manufactured by Safran Vectronix, 1σ distance error ± 5 m, 1σ elevation error ± 0.2°, 1σ azimuth error ± 0.6°), the azimuth, elevation, and oblique distance of each flying object could be measured, and its 3-D position calculated. Ornithodolite observations were made at the same locations as the radar observations. In addition, at Seiro, an observation site was added approximately 1.5 km inland from the radar site (Fig. 1). We tracked birds at 2.3-second intervals using an ornithodolite support system (LMS 1.0, manufactured by Kyokuto Boeki Kaisha), and recorded species and flock size.

3. Extraction of bird echoes and image processing

The extraction procedure of bird echoes and the procedure for image processing are complex (see Fig. 2).

We used ESRI ArcGIS pro software (version 3.0) to select bird flight trajectories for matching bird echoes. We identified tracks on terrain clutter by overlapping the ornithodolite-derived tracking maps with radar images. We omitted images including these trajectories from the analysis target because bird echoes on terrain clutter are masked and cannot be extracted.

We examined the radar images frame-by-frame to check for the movement of echoes, and identified trajectories where the time, place, and flight direction matched the movements of the echoes. We then targeted the matched, trajectories and echoes for analysis. Images that echo that crossed or merged with other echoes or separated into multiple echoes were excluded from the analysis.

We used Image J software for processing radar images. The radar image was monochromatically transformed, and we extracted echoes with luminance greater than 100 (range 0–255) and an area greater than five pixels. We calculated the center-of-gravity coordinates and pixel sizes of extracted echoes. For non-detected coordinates, we could not directly calculate the centre of gravity because there were no echoes. Therefore, we estimated the non-detection coordinates to be equally spaced according to the number of radar scans between the extracted echoes. Estimation of non-detected coordinates was kept within the range of detected coordinates because outside the detection echo coordinates, unpredictable factors such as small terrain shadows or sudden changes in flight altitude may be involved in non-detection. It was also assumed that within the range where echoes were detected, the further away the echoes were, the smaller they became and the more sporadic their appearance. Therefore, we considered validation within the detectable range to be sufficient to estimate the impact of beam distance attenuation on probability of detection.

4. Modeling

We constructed generalized linear mixed models (GLMM) using the “lme4” and “lmerTest” package for R ver4.2.1 to explain the echo's POD and pixel size[41–43]. The POD was modeled using a binomial error structure, and the pixel size was modeled using normal errors. The random effect structure was the trajectory ID in both models.

As explanatory variables to validate species differences in POD and echo size, we used wingspan, body weight, and soaring (whether the species was a soaring bird or not) (Table 1). For wingspan and body weight, values found in the literature were used as references [44–46]. The number of birds in each flock was used to validate the effect of flock size on POD and echo size. We used distance and elevation angle for both models as the factors related to the echo position. Distances were calculated from the centroid and radar coordinates of echoes. Elevation angles were calculated from the distance described above and the median altitude of the trajectory obtained by ornithodolite tracking. The absolute value of the elevation angle was used because the output power attenuates as the beam direction moves away from the horizontal.

Pixel size and all explanatory variables were standardized to a mean of 0 and a variance of 1. All combinations of explanatory variables (not considering interactions) were tested using the “MuMIn” package [47], and the AIC was calculated. Models with ΔAIC < 2 were deemed worthy of consideration, and marginal R² and conditional R² were calculated after likelihood ratio tests with the null model. POD and pixel size were predicted using the model with the highest Marginal R² in the models with ΔAIC < 2.

Acknowledgments

This research was performed with the aid of the Environment Research and Technology Development Fund (JPMEERF20214G02) of the Environmental Restoration and Conservation Agency, provided by the Ministry of Environment of Japan, and the Sasaki Environment Technology Foundation. We are grateful to Kyokuto Boeki Kaisha, Ltd. for lending the ornithodolite support system for the ornithodolite survey and Mark Brazil for carefully proofreading the manuscript.

Authors contributions

T.K., T.Sa., and T.Se. conceived the ideas and designed the methodology; T.K., T.Sa., K.T., and K.K. collected the data; T.K. analysed the data; T.K., Y.K. and T.Se. led the writing of the manuscript. All authors contributed critically to the drafts and gave final approval for publication.

Data Availability

The datasets generated and analysed during the current study are available in the Supplementary files. The radar images are available from the corresponding author on request.

Additional Information

Competing interests

The authors declare that they have no competing interests.

Barrington, R. M. The migration of birds as observed at Irish lighthouses and lightships. Edward Posonby (1900).
Clarke, M. E. Studies in bird migration. Gurney and Jackson (1912).
Munro, J. A. A preliminary report on the destruction of birds at lighthouses on the coast of British Columbia. The Canadian Field-Naturalist. 38, 141–145 (1924).
Lewis, H. F. Destruction of birds by lighthouses in the provinces of Ontario and Quebec. The Canadian Field-Naturalist. 41, 55–58 (1927).
Hüppop, O. et al. Perspectives and challenges for the use of radar in biological conservation. Ecography, 42, 912–930. https://doi.org/10.1111/ecog.04063 (2019).
Walter, W. D. et al. Using three-dimensional flight patterns at airfields to identify hotspots for avian–aircraft collisions. Applied Geography, 35, 53–59. https://doi.org/10.1016/j.apgeog.2012.05.002 (2012).
Hu, Y. et al. A birdstrike risk assessment model and its application at Ordos Airport, China. Scientific Reports, 10, 19627. https://doi.org/10.1038/s41598-020-76275-z (2020).
Kranstauber, B. et al. High-resolution spatial distribution of bird movements estimated from a weather radar network. Remote Sensing, 12, 635. https://doi.org/10.3390/rs12040635 (2020).
GWEC. Global Wind Energy Council. [Online] Available at: https://gwec.net/wp-content/uploads/2020/11/GWEC_Global_Wind_2017_Report.pdf (2017).
Masden, E. A. et al. Barriers to movement: Impacts of wind farms on migrating birds. ICES Journal of Marine Science, 66, 746–753. https://doi.org/10.1093/icesjms/fsp031 (2009).
Amorim, F. Factors influencing bat activity and mortality at a wind farm in the mediterranean region. Acta Chiropterologica, 14, 439. https://doi.org/10.3161/150811012X661756 (2012).
Smallwood, K. S. Comparing bird and bat fatality-rate estimates among North American wind-energy projects. Wildlife Society Bulletin, 37, 19–33. https://doi.org/10.1002/wsb.260 (2013).
Erickson, W. P., Wolfe, M. M., Bay, K. J., Johnson, D. H., & Gehring, J. L. A comprehensive analysis of small-passerine fatalities from collision with turbines at wind energy facilities. PLoS ONE, 9, e107491. https://doi.org/10.1371/journal.pone.0107491 (2014).
Marques, A. T. et al. Wind turbines cause functional habitat loss for migratory soaring birds. Journal of Animal Ecology, 89, 93–103. https://doi.org/10.1111/1365-2656.12961 (2020).
Largey, N. et al. Methods to quantify avian airspace use in relation to wind energy development. Ibis, 163, 747–764. https://doi.org/10.1111/ibi.12913 (2021).
Bodey, T. W. et al. A phylogenetically controlled meta-analysis of biologging device effects on birds: Deleterious effects and a call for more standardized reporting of study data. Methods in Ecology and Evolution, 9, 946–955. https://doi.org/10.1111/2041-210X.12934 (2018)
Thaxter, C. B. et al. Sample size required to characterize area use of tracked seabirds: Sample Sizes of Tracked Animals. The Journal of Wildlife Management, 81, 1098–1109. https://doi.org/10.1002/jwmg.21283 (2017).
Rosa, D. et al. Classification success of six machine learning algorithms in radar ornithology. Ibis, 158, 28–42. https://doi.org/10.1111/ibi.12333 (2016).
Mabee, T. J., Cooper, B. A., Plissner, J. H., & Young, D. P. Nocturnal bird migration over an appalachian ridge at a proposed wind power project. Wildlife Society Bulletin, 34, 682–690. https://doi.org/10.2193/0091-7648(2006)34[682:NBMOAA]2.0.CO;2 (2006).
Cohen, E. B. et al. Using weather radar to help minimize wind energy impacts on nocturnally migrating birds. Conservation Letters, 15. https://doi.org/10.1111/conl.12887 (2022).
Dokter, A. M., Baptist, M. J., Ens, B. J., Krijgsveld, K. L., & van Loon, E. E. Bird radar validation in the field by time-referencing line-transect surveys. PLoS ONE, 8, e74129. https://doi.org/10.1371/journal.pone.0074129 (2013).
Schmidt, M., Aschwanden, J., Liechti, F., Wichmann, G., & Nemeth, E. (2017). Comparison of visual bird migration counts with radar estimates. Ibis, 159, 491–497 (2017).
Lack, D., & Varley, G. C. Detection of birds by radar. Nature, 156, 446–446. https://doi.org/10.1038/156446a0 (1945).
Buss, I. O. Bird detection by radar. The Auk, 63, 315–318. https://doi.org/10.2307/4080116 (1946).
Harper, W. G. Detection of bird migration by centimetric radar—A cause of radar ‘angels’. Proceedings of the Royal Society of London. Series B - Biological Sciences, 149, 484–502 (1958).
Tedd, J. G., & Lack, D. L. The detection of bird migration by high-power radar. Proceedings of the Royal Society of London. Series B - Biological Sciences, 149, 503–510. https://doi.org/10.1098/rspb.1958.0089 (1958).
Drury, W. H., & Keith, J. A. Radar studies of songbird migration in coastal new england. Ibis, 104, 449–489. https://doi.org/10.1111/j.1474-919X.1962.tb08681.x (1962).
Gauthreaux, S. A. Weather radar quantification of bird migration. BioScience, 20, 17–20. https://doi.org/10.2307/1294752 (1970).
Nilsson, C. et al. Field validation of radar systems for monitoring bird migration. Journal of Applied Ecology, 55, 2552–2564. https://doi.org/10.1111/1365-2664.13174 (2018).
Robin Radar Systems. (2023). https://www.robinradar.com/iris-counter-drone-radar
Gerringer, M. B., Lima, S. L., & DeVault, T. L. Evaluation of an avian radar system in a midwestern landscape: Avian Radar Evaluation. Wildlife Society Bulletin, 40, 150–159. https://doi.org/10.1002/wsb.614 (2016).
Urmy, S. S., & Warren, J. D. Quantitative ornithology with a commercial marine radar: Standard-target calibration, target detection and tracking, and measurement of echoes from individuals and flocks. Methods in Ecology and Evolution, 8, 860–869. https://doi.org/10.1111/2041-210X.12699 (2017).
Capotosti, S., Scacco, M., Nelli, L., Dell’Omo, G., & Panuccio, M. Hypatia-trackRadar: A software for animal tracking using marine surveillance radars. Ecological Informatics, 53, 100972. https://doi.org/10.1016/j.ecoinf.2019.100972 (2019).
Urmy, S. S., & Warren, J. D. Evaluating the target-tracking performance of scanning avian radars by augmenting data with simulated echoes. Methods in Ecology and Evolution, 11, 559–569. https://doi.org/10.1111/2041-210X.13365 (2020).
Washburn, B. E. et al. Monitoring raptor movements with satellite telemetry and avian radar systems: An evaluation for synchronicity. Remote Sensing, 14, 2658. https://doi.org/10.3390/rs14112658 (2022).
Jenkins, A. R. et al. Combining radar and direct observation to estimate pelican collision risk at a proposed wind farm on the Cape west coast, South Africa. PLoS ONE, 13, e0192515. https://doi.org/10.1371/journal.pone.0192515 (2018.)
Beason, R. C., Nohara, T. J., & Weber, P. Beware the Boojum: Caveats and Strengths of Avian Radar. Human–Wildlife Interactions, 7, 16–46. https://doi.org/10.26077/0FVY-6K61 (2013).
Kreutzfeldt, J. et al. Analytical volume model for optimized spatial radar bat detection in onshore wind parks. PLOS ONE, 15, e0239911. https://doi.org/10.1371/journal.pone.0239911 (2020).
Schmaljohann, H., Liechti, F., Bächler, E., Steuri, T., & Bruderer, B. Quantification of bird migration by radar - a detection probability problem: Quantification of bird migration. Ibis, 150, 342–355. https://doi.org/10.1111/j.1474-919X.2007.00797.x (2008).
Zaugg, S., Saporta, G., van Loon, E., Schmaljohann, H., & Liechti, F. Automatic identification of bird targets with radar via patterns produced by wing flapping. Journal of The Royal Society Interface, 5, 1041–1053. https://doi.org/10.1098/rsif.2007.1349 (2008).
Bates, D., Mächler, M., Bolker, B., & Walker, S. Fitting linear mixed-effects models using lme4. Journal of Statistical Software, 67. https://doi.org/10.18637/jss.v067.i01 (2015).
Kuznetsova, A., Brockhoff, P. B., & Christensen, R. H. B. Lmertest package: Tests in linear mixed effects models. Journal of Statistical Software, 82, 1–26. https://doi.org/10.18637/jss.v082.i13 (2017).
R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/ (2022).
Alerstam, T., Rosén, M., Bäckman, J., Ericson, P. G. P., & Hellgren, O. Flight speeds among bird species: Allometric and phylogenetic effects. PLoS Biology, 5, e197. https://doi.org/10.1371/journal.pbio.0050197 (2007).
Myhrvold, N. P. et al. An amniote life-history database to perform comparative analyses with birds, mammals, and reptiles: Ecological Archives E096-269. Ecology, 96, 3109–000. https://doi.org/10.1890/15-0846R.1 (2015).
Storchová, L., & Hořák, D. Life-history characteristics of European birds. Global Ecology and Biogeography, 27, 400–406. https://doi.org/10.1111/geb.12709 (2018).
Bartoń, K. MuMIn: Multi-Model Inference, https://CRAN.R-project.org/package=MuMIn (2022).

No competing interests reported.

Download PDF

Editorial decision: Major revision
22 Oct, 2023
Reviews received at journal
02 Sep, 2023
Reviewers agreed at journal
23 Aug, 2023
Reviewers invited by journal
23 Aug, 2023
Editor assigned by journal
23 Aug, 2023
Editor invited by journal
25 Jun, 2023
Submission checks completed at journal
25 Jun, 2023
First submitted to journal
13 Jun, 2023

You are reading this latest preprint version

Field validation of avian radar surveys: effects of differences in species and flock sizes on echoes

Status:

Version 1

Abstract

Figures

Introduction

Results

1. Characteristics of bird echoes extracted

2. Effect of species differences on echoes

3. Effect of flock size on echoes

Discussion

Validation of detection performance for bird surveys

Methodological developments in radar bird surveys

Methods

1. Radar system

2. Bird tracking by Ornithodolite

3. Extraction of bird echoes and image processing

4. Modeling

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1