Using movement modelling to improve the design and analysis of vantage point surveys in bird and wind energy studies

Wind energy, although mostly a clean and increasingly efficient energy source, is known to affect communities of flying vertebrates. Mortality by collision with turbines is one of the main impacts on birds and bats associated with wind energy. Soaring birds are particularly vulnerable due to their collision prone behaviours, low manoeuvrability, and their slow population recovery rates. The focus of this thesis is on the identification of areas that are intensively used by soaring birds in order to inform wind turbine placement and minimize collision risk. This thesis is particularly concerned with predictions of bird-use intensity that are based on flight trajectories mapped by observers from vantage points. This survey technique is standard practice during the environmental impact assessment of wind energy facilities, although its virtues and limitations are largely untested. Flight trajectories are counted, timed and mapped during these surveys. However, most assessments ignore the spatial information contained in the trajectories, and mappings are often reduced to metrics such as closest distance to a turbine or whether a particular habitat is visited. In this thesis, I use visual mappings of flight trajectories to estimate the long-term distribution of bird activity using: i) a kernel density estimator adapted to calculate the density of flight trajectories, and ii) modelling flights as being driven by a stochastic process under the influence of a potential field. Acknowledging the subjectivity introduced in the mapping of trajectories by field observers, I also study the discrepancy between mapped and true trajectories. Finally, I showcase the application of the various analytical techniques with a case study, in which I compare collision risk predictions with actual observed fatalities at a wind farm in South Africa. Kernel density estimation proved to be a good exploratory technique, and the estimator designed to estimate trajectory density outperformed other methods that ignore the temporal structure in trajectory data. Nevertheless, kernel methods are limited by its inability to predict bird activity outside areas observed from vantage points. Potential-based models allowed predictions in unobserved areas based on landscape characteristics, and showed promising results identifying areas of high collision risk. I found that the difference between true and mapped trajectories can be substantial, and it should be accounted for in any spatial analysis of vantage point observations. Although based on a single study case, the results are promising and show that the spatial distribution of collision risk predicted with the suite of methods presented in this thesis correlates well with the distribution of observed fatalities. The framework proposed to predict collision risk improves existing procedures in that it uses movement and spatial information contained in the observed trajectories. In addition, it accounts for all known sources of uncertainty throughout the modelling process.

Grantor dc:publisher

University of Cape Town

Year dc:date.issued

2020

Author dc:creator

• Cervantes Peralta, Francisco

Advisors dc:contributor.advisor

• Erni, Birgit
• Theoni Photopoulou, Theoni
• Simmons, Rob

Language dc:language.iso

eng

Handle dc:identifier.uri

http://hdl.handle.net/11427/32547

OAI identifier oai:identifier

oai:open.uct.ac.za:11427/32547