Predicting the Weather-Dependent Collision Risk for Birds at Wind Farms - Wind and Wildlife - page 9

where v is bird speed, D A - density of flying birds, 2R - rotor diameter,
T - number of turbines, π R 2 - area of the rotor, t day - total daylight time
(in seconds), f night - species nocturnal activity factor, t night - total night time,
Q 2R - proportion of birds flying at risk height.
(e) Probability of collision for a single rotor transit is calculated applying the
approach developed by Band ( 2000 ) and Band et al. ( 2007 ). This approach
incorporates dimensions and speed of turbines and bird species' wingspan,
body length and flight speed. Biometric measurements of birds were obtained
from DOF ( 2012 ) or BTO ( 2012 ) online databases and flight speeds from
Alerstam et al. ( 2007 ).
(f) Proportion of time that the wind farm operates.
(g) Finally, wind farm avoidance rates were applied for a bird movement in rotor
swept area. By reviewing available publications to date on offshore wind
farms and seabirds, Cook et al. ( 2012 ) suggested that most seabird species have
overall avoidance rate (including avoidance within collision risk window) of
99-99.5 %. In our estimates we offer two figures: one representing a pessimistic
scenario with overall avoidance rate of 98 % and the other one representing an
optimistic scenario with overall avoidance rate of 99.5 %.

The final figure of possible collision rates is determined by multiplying the bird
flux through the rotor swept area ('d' above), collision probability ('e'), proportion
of wind farm operational time ('f'), and avoidance rates ('g'):

defg ××× .

Results

Seabirds (Resident Bird Model)

As examples of the results for seabirds, northern gannet, Morus bassana, and
common scoter have been selected based from the data collected at Horns Rev,
North Sea (Fig. 1 ). The GAMM flight model for the northern gannet indicated that
the birds fly higher in tail and side winds in comparison to head winds (Figs. 2 and 3 ,
Table 1 ). They also seem to increase flight height with increasing wind speed and air
pressure and also with decreasing relative humidity. The model had a good predictive
ability with a Spearman's correlation coefficient of 0.70. The adjusted R 2 value
indicated that the model explains 35 % of the variability in the data set. We did not find
spatial autocorrelation in the model residuals of the random effects, which indicated
that the northern gannet model was able to account for the spatial autocorrelation
in the data.
We used the model for predicting the average flight altitudes during one autumn
season, with separate predictions made for head winds, tail winds and side winds.
According to the predictions the northern gannets fly, on average, at rotor height