Seabird flight height data collection at an offshore wind farm: final report

Understanding seabird flight heights and behaviour in and around operational offshore wind farms is a priority knowledge gap. Using aircraft mounted LiDAR technology, this study collected data on seabird flight height and shows the potential for using it in offshore windfarm impact assessments.

4. Discussion

In the surveys in June and July (survey 1 and 2) a wide variety of species were detected in the images. In both surveys the most common species detected were guillemots. The detected species assemblage is in line with previous findings in the east of Scotland during the summer months (APEM, 2018).

The imagery-LiDAR system performed well, with a match rate of greater than 81% for all species groups with more than one count (Appendix III). The imagery-LiDAR system appeared to be capable of measuring flight heights of small seabird species such as puffin. Although, not included in the results of this study the imagery-LiDAR system was also able to measure flight heights for tern species when on transit to the study site. This suggests the combined system is suitable for collecting species specific seabird flight heights both over the open sea and within an operational wind farm.

Flight heights for the four species of interest were plotted with violin plots for each survey area. Generally, flight heights were recorded below 25 m above sea surface level for most species in both survey areas (Survey Area 1 = 89% and Survey Area 2 = 80% below 25m). Additionally, large gulls, kittiwakes and fulmars were recorded at flight heights of up to 50-150 m. The results of flight heights in the current study are in line with previous research of those seabird species, suggesting that they are commonly found in flight just above the sea surface (Johnston, et al. 2014).

The methodology for this survey programme used the imagery collected to inform the LiDAR analysis and subsequent data matching. Thus, birds below two metres were not under-recorded, an issue that had been highlighted in earlier studies (Cook et al. 2018).

There was a significant difference for flight heights of all four species in a combined analysis between the two areas. In Survey Area 1, the site with WTGs, birds flew lower (11.32 m (SD= 26.61)) than those recorded in Survey Area 2 (20.03 m (SD= 33.73)), the area without WTGs. This could be driven by species with a higher sample size. However, this also suggests different flight height patterns in the different areas, which could be related to the presence of WTGs. Studies on seabird flight behaviour have also indicated that seabirds may change flight altitudes when approaching wind turbines, this is known as macro avoidance (Cook, et al. 2012). By detecting this change in flight height, it can be assumed that this combined imagery-LiDAR system could potentially be used as a tool to detect the macro avoidance rates within and around wind farms, allowing the potential impact of the wind farm to be detected (Furness, et al. 2013).

Species-specific analysis showed highly significant differences for flight heights between survey areas for kittiwake and gannet and close to significant differences for fulmar. Kittiwake and fulmar had the highest sample size of individuals and a subjectively similar sample size over both survey areas. However, gannet and large gulls had a comparatively large difference in sample size between survey areas which impacts the statistical power of the species-specific analysis.

Histograms plotting the distribution of distance to nearest turbine for each species (Appendix V) demonstrated some discreet differences between species. Guillemots demonstrated some evidence of avoidance, with most individuals recorded 20-30 km away from turbines in both survey months; however, in June a high number were also recorded closer to turbines at 200-5,000 m. The distribution of kittiwakes, gannets, guillemot / razorbills, and great skuas was not consistent between the two survey months, and no behavioural patterns could be concluded from the histograms. A few species appeared to show consistent distributions in both survey months; fulmars were recorded closer to turbines (min 189 m) with numbers decreasing as distance from turbine increased. Herring gulls were mostly recorded >20,000 m away from turbines, suggesting avoidance of turbines. Razorbills, only recorded in June, were mostly distributed 625-15,000 m away from turbines, and fewer recorded outside this distance. Overall, the histograms did not provide evidence of any conclusive trend in respect to the number of birds recorded at different distances to the nearest WTG of the different species. This can partially be explained by the low number of individuals recorded in Survey Area 1. However, no bird was recorded closer than 189 m which suggests avoidance behaviour. Previous studies in this location were specifically designed to investigate seabird displacement from OWF area, meaning they would be more suitable to inform investigations (MacArthur Green, 2019) as the results in this study are based only on flying birds. Avoidance of the area does have implications for collision risk modelling however, as the lower number of individuals in the area reduces the number likely to collide with the WTG. Furthermore, it makes assessing the impact of WTGs on flight heights at this scale difficult. Surveys should be planned, assuming a reduction in densities recorded pre-construction, so that a precautionary approach to survey effort can be planned to ensure a suitable sample size can be achieved.

Flight height in relation to the nearest WTG was studied using different linear models. There were significant results for gannets and gulls. The results of the gannets indicate the birds are flying lower closer to the WTGs. However, the results were largely influenced by one bird being detected far away from the WTGs. Only eight large gulls were detected in Survey Area 1, however a significant effect of those birds flying higher closer to the WTGs was found. There was no significant effect of flight height in relation to the distance of WTGs for the other species.

The sample size of the flying birds detected in Survey Area 1 was low (n = 315) particularly in the area containing WTGs which is likely affecting the statistical power of the analysis. Additionally, the detected birds were far away from the WTG, with the closest bird 44 m away and the second closest 189 m. This could suggest avoidance behaviour of the detected sea birds around the WTG. A larger sample size is needed for achieving meaningful results of the study of flight height in relation to WTGs. This could be achieved by repeating the study in an area with a higher abundance of sea birds or conducting more surveys. However, if seabirds are avoiding WTGs the effect of height on their flight pattern might be difficult to study in close proximity of 0-200 m.

The outputs produced from the combined imagery-LiDAR system could be incorporated into collision risk modelling for use within either Band Option 1 by specifying the proportion of birds recorded at collision risk height (CRH) or within Band Option 2 as the proportion of individuals recorded in one metre flight height bands (Band, 2012, Johnston et al. 2014; McGregor, et al. 2018). The implication of using these values would need to be determined and clear guidance on best practice given, as it may be advisable to undertake linear modelling of the dataset first in order to interpolate between flight bands where there were no individuals recorded flying at those heights. If a representative dataset is collected with a large sample size it can be argued that the raw data could be suitable for use within the CRM.

In order to safely collect data over the wind farm, the aircraft must be at least 305 m above the highest object. Therefore, with the combined imagery-LiDAR system the aircraft altitude was approximately 500 m. There is potential that collecting data at this height could lead to a cone effect whereby higher-flying birds have a lower probability of being included within the imagery, causing a bias in the data. This is not restricted to the imagery-LiDAR system. Provided that an accurate IMU is integrated on the camera system, analysis for actual area sampled at each altitude can be calculated. The effect is likely to be minimal however, as during this survey no birds were found above 200 m suggesting that they may not utilise the airspace at the point where the cone effect may be more relevant. Further study could be undertaken collecting flight height data at differing altitudes to see if there are any impacts on results. There would however be a trade-off in resolution and species identification with differing flight altitudes.

During this study low sample sizes were encountered, therefore it can be recommended that when planning combined imagery-LiDAR surveys previous data on species densities should be analysed to infer predicted encounter rate. Key species for collision risk should be selected to ensure accurate CRM can be undertaken. Furthermore, if future studies are undertaken power analysis can be undertaken to identify the minimum sample size required to detect an effect size and the survey planned accordingly (Cohen, 1988; Maclean et al., 2016).

Overall, the combined imagery-LiDAR system successfully allowed a large sample of different species flight heights outside and within an active wind farm to be measured. Potential macro avoidance of wind turbines was found; however, there was only a minimal statistical difference. A combined imagery-LiDAR flight height dataset has the capability to be used in CRM for EIA, with the advantage of greatly reducing associated error and potential risk.



Back to top