A review of existing methods to collect data on seabird flight height distributions and their use in offshore wind farm impact assessments

Feather, A.; Burton, N; Johnston, D.T.; Boersch-Supan, P.H. (2025). A review of existing methods to collect data on seabird flight height distributions and their use in offshore wind farm impact assessments (Report No. BTO Research Report 780). Report for British Trust for Ornithology (BTO).

TY - RPRT
TI - A review of existing methods to collect data on seabird flight height distributions and their use in offshore wind farm impact assessments
AU - Feather, A
AU - Burton, N
AU - Johnston, DT
AU - Boersch-Supan, PH
AB - Significant growth within the offshore renewable energy industry (e.g. wind farms) is expected in the coming years and environmentally friendly development is a priority. Accurately predicting, mitigating, and compensating the ecological impacts of offshore development is therefore crucial to ensuring development can continue at pace without significant negative effects. Seabirds are a key component of marine ecosystems, and many species are potentially vulnerable to marine development. Offshore wind farms might impact seabird populations in several ways (e.g. displacement, barrier, and indirect effects) but direct mortality caused by collisions with turbines is of particular concern for several key species (e.g. Gannet, Kittiwake, and large gulls).For the purposes of environmental impact assessment, collision rates with offshore wind farm turbine rotors are predicted using collision risk models (CRMs) which provide a quantitative estimate of risk. Information on species’ flight height distributions is fundamental to predictions and continuous flight height distributions provide more robust estimates of mortality risk, compared to wider discrete flight height bands. Flight height data can be collected using a variety of sampling methods (e.g. visual surveys, photogrammetric digital aerial surveys, animal-borne tracking devices, radar, LiDAR), with which there are uncertainties (e.g. sampling error, measurement error) and logistical constraints (e.g. environmental conditions, species-specific behaviour).This document presents a review of existing methods for collecting seabird flight height data and their potential to produce flight height distributions that might be used in CRMs. The strengths, weaknesses, and limitations of different methods are identified and sources of measurement and sampling error, uncertainty and bias assessed. Best practice recommendations are provided for prominent methods and how data might be best utilised to inform stakeholders is considered.None of the methods reviewed could provide species-specific flight height distributions that were fully representative of the populations of interest under all relevant environmental conditions (i.e. biotic, abiotic) and across all ecologically important temporal scales (e.g. daily, seasonal, annual) of variation. Aside from animal-borne technologies, all sampling methods are vulnerable to systematic over- or underestimation of flight height in particular height bands (e.g. close to the sea-surface) which consequently impacts estimates of collision risk. Bias (positive and negative) is introduced via observer error (e.g. visual surveys), technical challenges (e.g. rangefinders), equipment limitations (e.g. radar, LiDAR, animal-borne tracking devices) and failure to account for variable detection probability within the surveyed area (all methods).Measurement errors are generally better understood than sampling errors. Accuracy of individual flight height estimates from several methods (e.g. rangefinders, stereophotogrammetry, high frequency (< 20s sampling interval) animal-borne tracking devices, LiDAR, microphone array) is generally within 10 m of true flight heights in favourable conditions. Measurement errors (precision) however can be more than 100 m (e.g. visual surveys, photogrammetric digital aerial surveys, low frequency animal-borne tracking device) due to equipment characteristics (e.g. sensor accuracy, sampling frequency), human behaviour (e.g. height estimates) and from interactions between height measurements and supplementary data (e.g. sea level pressure, natural body size variation, and other reference values). Increasingly complex sampling methods (e.g. animal-borne GPS, aerial imagery) were found to simultaneously incorporate multiple sources of error which can interact to alter (e.g. inflate variance, introduce bias, distort shape) flight height distributions with consequences for CRMs.The lack of robust analytical procedures for determining heterogeneity in each method’s detection probabilities prevents the effectively sampled volume from being calculated for most if not all available methods. The true frequencies with which flight heights are distributed is therefore rarely estimated. Developing procedures to determine each method’s detection probability is therefore a priority, particularly for methods for whichSampling in a temporally non-biased manner (e.g. with respect to weather conditions, daylight hours) was noted as a particularly widespread challenge and was therefore also highlighted as a research priority. Telemetry studies that focus on quantifying species-specific relationships between temporally varying conditions (e.g. weather, time of day, behavioural state) and flight height are best placed to improve understanding, but novel sampling designs and the incorporation of remotely sensed data will likely be required. However, no one method provides information that is representative of all environmental conditions or of spatial variation, for a given species; thus, the integration of information across multiple measurement methods is likely to be required to provide more representative flight height distributions.The continued development and assessment of methods for estimating seabird flight height distributions has significantly improved current understanding (e.g. limitations, uncertainty, collision risk). The potential accuracy of flight height estimates appears to be sufficiently high (< 10 m) to allow inferences at the vertical scales of interest (air gap, RSZ) and advanced statistical techniques (e.g. state-space models, nonlinear models) have allowed for a more rigorous quantification of uncertainty by describing the underlying distributions and providing confidence estimates. However, some methods (e.g. low-frequency GPS, aerial photogrammetry) exhibit large flight height measurement errors (> 50 m) under some, or all observation conditions and it is therefore crucial that measurement uncertainty is considered routinely. There remains a lack of agreement in the flight height estimates produced by different methods and the way many observational studies are designed is a key driver of uncertainty. Most methods were not originally designed to sample flight height distributions (e.g. radar, LiDAR), many datasets were not originally collected to describe flight height distributions (e.g. animal-borne tracking devices) and environmental limitations (e.g. rangefinders) regularly require last minute changes to experimental designs which reduce their effectiveness. There is consequently a pressing need for the development of best-practice guidelines to help ensure studies are designed robustly and data collection/reporting is standardised. Technological advances are generating a wide range of novel opportunities for flight height studies. Continued progress will require clear documentation of all practical steps (e.g. methods, analysis) and data (i.e. raw) involved to be freely available to all stakeholders.Rangefinders, LiDAR, and animal-borne tracking devices (high frequency GPS) provide species-specific flight height distributions that are accurate and precise such that the underlying distributions can be statistically modelled. They are also capable of sampling prior to wind farm construction (i.e. baseline data collection) and may be scaled to regional/national operations. We therefore suggest that field validation of these methods is a useful research priority for the ReSCUE project. Other methods however can add value to current understanding (e.g. by being able
DA - 2025/03//
PY - 2025
SP - 57
SN - BTO Research Report 780
UR - https://www.bto.org/our-work/science/publications/reports/bto-research-reports/00780-review-flight-height-estimation-methods
LA - English
KW - Wind Energy
KW - Fixed Offshore Wind
KW - Collision
KW - Birds
KW - Seabirds
ER -

Export Citation to RIS

Abstract

Significant growth within the offshore renewable energy industry (e.g. wind farms) is expected in the coming years and environmentally friendly development is a priority. Accurately predicting, mitigating, and compensating the ecological impacts of offshore development is therefore crucial to ensuring development can continue at pace without significant negative effects. Seabirds are a key component of marine ecosystems, and many species are potentially vulnerable to marine development. Offshore wind farms might impact seabird populations in several ways (e.g. displacement, barrier, and indirect effects) but direct mortality caused by collisions with turbines is of particular concern for several key species (e.g. Gannet, Kittiwake, and large gulls).
For the purposes of environmental impact assessment, collision rates with offshore wind farm turbine rotors are predicted using collision risk models (CRMs) which provide a quantitative estimate of risk. Information on species’ flight height distributions is fundamental to predictions and continuous flight height distributions provide more robust estimates of mortality risk, compared to wider discrete flight height bands. Flight height data can be collected using a variety of sampling methods (e.g. visual surveys, photogrammetric digital aerial surveys, animal-borne tracking devices, radar, LiDAR), with which there are uncertainties (e.g. sampling error, measurement error) and logistical constraints (e.g. environmental conditions, species-specific behaviour).
This document presents a review of existing methods for collecting seabird flight height data and their potential to produce flight height distributions that might be used in CRMs. The strengths, weaknesses, and limitations of different methods are identified and sources of measurement and sampling error, uncertainty and bias assessed. Best practice recommendations are provided for prominent methods and how data might be best utilised to inform stakeholders is considered.
None of the methods reviewed could provide species-specific flight height distributions that were fully representative of the populations of interest under all relevant environmental conditions (i.e. biotic, abiotic) and across all ecologically important temporal scales (e.g. daily, seasonal, annual) of variation. Aside from animal-borne technologies, all sampling methods are vulnerable to systematic over- or underestimation of flight height in particular height bands (e.g. close to the sea-surface) which consequently impacts estimates of collision risk. Bias (positive and negative) is introduced via observer error (e.g. visual surveys), technical challenges (e.g. rangefinders), equipment limitations (e.g. radar, LiDAR, animal-borne tracking devices) and failure to account for variable detection probability within the surveyed area (all methods).
Measurement errors are generally better understood than sampling errors. Accuracy of individual flight height estimates from several methods (e.g. rangefinders, stereophotogrammetry, high frequency (< 20s sampling interval) animal-borne tracking devices, LiDAR, microphone array) is generally within 10 m of true flight heights in favourable conditions. Measurement errors (precision) however can be more than 100 m (e.g. visual surveys, photogrammetric digital aerial surveys, low frequency animal-borne tracking device) due to equipment characteristics (e.g. sensor accuracy, sampling frequency), human behaviour (e.g. height estimates) and from interactions between height measurements and supplementary data (e.g. sea level pressure, natural body size variation, and other reference values). Increasingly complex sampling methods (e.g. animal-borne GPS, aerial imagery) were found to simultaneously incorporate multiple sources of error which can interact to alter (e.g. inflate variance, introduce bias, distort shape) flight height distributions with consequences for CRMs.
The lack of robust analytical procedures for determining heterogeneity in each method’s detection probabilities prevents the effectively sampled volume from being calculated for most if not all available methods. The true frequencies with which flight heights are distributed is therefore rarely estimated. Developing procedures to determine each method’s detection probability is therefore a priority, particularly for methods for which
Sampling in a temporally non-biased manner (e.g. with respect to weather conditions, daylight hours) was noted as a particularly widespread challenge and was therefore also highlighted as a research priority. Telemetry studies that focus on quantifying species-specific relationships between temporally varying conditions (e.g. weather, time of day, behavioural state) and flight height are best placed to improve understanding, but novel sampling designs and the incorporation of remotely sensed data will likely be required. However, no one method provides information that is representative of all environmental conditions or of spatial variation, for a given species; thus, the integration of information across multiple measurement methods is likely to be required to provide more representative flight height distributions.
The continued development and assessment of methods for estimating seabird flight height distributions has significantly improved current understanding (e.g. limitations, uncertainty, collision risk). The potential accuracy of flight height estimates appears to be sufficiently high (< 10 m) to allow inferences at the vertical scales of interest (air gap, RSZ) and advanced statistical techniques (e.g. state-space models, nonlinear models) have allowed for a more rigorous quantification of uncertainty by describing the underlying distributions and providing confidence estimates. However, some methods (e.g. low-frequency GPS, aerial photogrammetry) exhibit large flight height measurement errors (> 50 m) under some, or all observation conditions and it is therefore crucial that measurement uncertainty is considered routinely. There remains a lack of agreement in the flight height estimates produced by different methods and the way many observational studies are designed is a key driver of uncertainty. Most methods were not originally designed to sample flight height distributions (e.g. radar, LiDAR), many datasets were not originally collected to describe flight height distributions (e.g. animal-borne tracking devices) and environmental limitations (e.g. rangefinders) regularly require last minute changes to experimental designs which reduce their effectiveness. There is consequently a pressing need for the development of best-practice guidelines to help ensure studies are designed robustly and data collection/reporting is standardised. Technological advances are generating a wide range of novel opportunities for flight height studies. Continued progress will require clear documentation of all practical steps (e.g. methods, analysis) and data (i.e. raw) involved to be freely available to all stakeholders.
Rangefinders, LiDAR, and animal-borne tracking devices (high frequency GPS) provide species-specific flight height distributions that are accurate and precise such that the underlying distributions can be statistically modelled. They are also capable of sampling prior to wind farm construction (i.e. baseline data collection) and may be scaled to regional/national operations. We therefore suggest that field validation of these methods is a useful research priority for the ReSCUE project. Other methods however can add value to current understanding (e.g. by being able