A Monte-Carlo Model for Bird/Wind Turbine Collisions

Citation

McAdam BJ (2005) A Monte-Carlo Model for Bird/Wind Turbine Collisions. Master of Science. University of Aberdeen.

Abstract
Wind power generation projects can pose a risk to bird populations through deaths caused by collision, as well as other factors such as displacement. Models of collision likelihood are necessary to evaluate this risk before construction. Previous models have assumed that birds are uniformly distributed in height and travel at fixed speeds, which is unrealistic for most bird species. I created a Monte Carlo simulation to compute the collision rate. This is based on the Band collision model. The model uses species specific height distributions and variation in flight speed according to real wind data. The model is explored by investigating various scenarios relating to a proposed off-shore wind farm in the Moray Firth. The collision rates computed in this way are different from the rates using the basic Band model. The basic Band model underestimated the collision probability for birds flying in near the hub of the turbine and overestimated for other birds. A variation of 20m height could double the collision probability for the bird modelled (on a 63m radius turbine). The Monte Carlo model showed that flight direction relative to prevailing winds also affects collision probability. Birds flying into the prevailing wind were twice as likely to be struck as the safest direction. My model also has a more realistic model of the turbine's blades (with variable pitch) which affected collision probability. To realistically model variation in flight height, it is necessary to take field measurements. A practical technique based on stereophotography is described and tested, complementing the theoretic modelling work. This method allows one person to take all field measurements, and does not require independent measurement of camera characteristics such as focal length. The method quantifies the uncertainty in height estimates, which is especially important for distant birds. The use of a prior distance distribution based on the size of the bird image improves the accuracy of bird distance and height estimates.