Environmental Engineering Reference
In-Depth Information
used as a grouping factor, thus accounting for the autocorrelation within the tracks.
At first a model was fitted including all variables, whereafter variables not contrib-
uting to the model fit were eliminated based on the GCV value (Wood
2006
). The
residuals were assessed using a correlogram with 10 lags (1 lag was the defined
nearest neighbourhood of 250 m) to inspect whether the model was capable of
accounting for the spatial autocorrelation in the model residuals. For calculating
the “Moran's I” (measure of spatial autocorrelation) the R package “spdep” (Bivand
2009
) was used.
The predictive accuracy of the models was evaluated by splitting the data into
two data sets, a calibration set (70 % randomly selected from all data points) and
an evaluation set (30 %). The model was fitted on the calibration data and predicted
on the evaluation set. Thereafter the agreement between observed and predicted
altitudes was checked by plotting the predicted values against the observed, and the
Spearman's correlation coefficient was estimated (Potts and Elith
2006
). The models
were used for predicting the average flight altitude of birds entering the wind farm
areas using mean wind and weather parameters.
Collision Models
Collision models were applied to estimate potential collision rates for the selected
species using the detailed flight trajectories obtained by combined radar and range-
finder techniques. The radar and rangefinder data were used empirically, i.e. the
modelled flight altitudes were not used as input into the collision models. A collision
is here defined as the proportion of birds/flocks exposing themselves to a collision
by crossing a scale-specific collision conflict window. To calculate collision risks,
several parameters need to be considered. Technical parameters are in this case the
measurements/dimensions of rotor structures and wind farm design. Given these,
the number of birds flying within the collision risk area, defined by the design of the
wind farm can be estimated from the measured two- and three-dimensional flight
trajectories and total numbers of tallies of birds passing the wind farm area.
Migrating Bird Model
The collision model used for migrating birds (Band
2012
) is based on the assumption
of a single transit through the turbine array by any one individual.
The model is based on the availability of the following data for the target species or
species groups:
(a) The proportion of the number of birds entering the wind farm, calculated from
horizontal radar and rangefinder data;
(b) Proportion within horizontal reach of rotor-blades in each turbine row. The value of
this parameter corresponded to the proportion that the swept area comprised relative
to the area of the so-called risk window in each turbine row. It is assumed that flight
