USING A COLLISION RISK MODEL TO ASSESS BIRD COLLISION RISKS FOR OFFSHORE WINDFARMS
( SOSS Guidance: March 2012)
SUPPLEMENT - AVOIDANCE RATES USING THE BASIC AND EXTENDED MODELS
March 2014 - Bill Band
This is a supplement to guidance prepared for the Crown Estate as part of the Strategic Ornithological Support Services programme, project SOSS-02  . That provides guidance for offshore wind developers, and their ecological consultants, on using a collision risk model to assess the bird collision risks presented by offshore windfarms. The March 2012 version of the guidance enabled use to be made of flight height distribution data.
This supplement is an addition to Stage E - Avoidance and Attraction. That section describes how, having used the collision model to calculate the potential collision rate if birds take no avoiding action, one should then apply an avoidance rate A to allow for the fact that many species of birds do in fact take avoiding action, either at long range (macro) or at close range (micro).
Paragraph 80 notes that
'if the extended model taking account of flight height distribution is used, it is important that the calculations on which avoidance rates are based also start with a no-avoidance collision rate derived using the extended model'.
Most of the published literature on avoidance rates is currently based on using the basic model. This supplement shows how such avoidance rates may be modified to enable their application to the extended model.
Avoidance in the basic and extended models
The two models - basic and extended - yield different predictions of the rate of collisions before avoidance is taken into account. The extended model is a more refined model which takes into account the effect of flight height distribution. It takes into account the fact that, for a given number of flights at risk height, a flight height distribution skewed towards low altitude leads to a smaller proportion of birds passing through the rotor, and bird passages through parts of the rotor with less risk, than if the distribution were uniform.
The outputs of the two models may be formally compared if the data input to the basic model on the proportion of flights at risk height (Q' 2R) is derived from the same flight height distribution used in the extended model, as in Option 2 of the spreadsheet accompanying the SOSS guidance. That is, the comparison should be made between the collision rate using the basic model (Option 2) in the spreadsheet, and the extended model (Option 3).
The collision rates (before avoidance) projected by the two models are:
Basic model (Option 2):
C basic = v(D A/2R)(TπR 2)t x Q 2R ' x p average x Q op (guidance eq.5  )
i.e. flux factor x Q 2R ' x p average x Q op
Extended model (Option 3):
C extended = v(D A/2R)(TπR 2)t x (2/π) ∫∫ d(y) p(x,y) dxdy x Q op (guidance eq. 9)
i.e. flux factor x collision integral x Q op
Where the bird flight height distribution is skewed towards low altitude, the extended model prediction C extended is usually less than C basic, because this equation takes full account of the reduction in risk at lower parts of the rotor. Let g be the ratio C extended / C basic , g is thus usually less than 1. The value of g may be obtained by dividing the second of the above equations by the first:
g = C extended / C basic = collision integral / ( Q' 2R x p average ) …. eq. S1
and this is readily calculated from the 'Overall collision risk' spreadsheet
g = cell D35 / ( cell D33 x cell D27 )
The expected collision rate must then take into account the proportion A of birds avoiding the turbines ( e.g. by displacement, or by evasive action), by multiplying the above no-avoidance collision rates by the proportion (1-A) which do not avoid. Values of A are typically in the range 90-100%. It is more helpful to think in terms of the non-avoidance rate A' = 1 - A , such that A' is the small proportion of birds which do not avoid the turbines. The expected collision rate is then
A' basic C basic in the basic model, or .. eq. S2a
A' extended C extended in the extended model. .. eq. S2b
The two models require the use of different non-avoidance rates. The calculation of C extended takes account of the effect of a skewed flight distribution, such that the factor A extended ( = 1 - A' extended ) refers only to genuine behavioural avoidance. The calculation of C basic in the basic model does not, such that any such effect, in the basic model, must be covered by the avoidance factor A basic.
Establishing avoidance rates from reference windfarms
Values of A' basic and A' extended for use in the two models are obtained by monitoring collisions at one or more reference windfarms, and working back from the two models. For either model we have
Non-avoidance rate A' = Actual no of collisions / Predicted number of collisions C.
using basic model using extended model
Actual no of collisions = A' basic x C basic(ref) = A' extended x C extended(ref)
thus A' extended = A' basic x C basic(ref) / C extended(ref)
but g(ref) = C extended (ref) / C basic (ref)
so A' extended = A' basic / g (ref) … eq. S3
A' extended is the non-avoidance rate from the reference windfarm, for use with the extended model. Equation (S3) describes how it is related to the value of A' basic derived using the basic model, using the g factor for this reference windfarm.
Where data from several reference windfarms are used to yield an average A' basic , then the value for A' extended should be the average of A' basic / g(ref) as calculated for each of the reference windfarms.
Applying reference avoidance rates to new or projected windfarms
Avoidance rates, derived from collision studies at one or more reference windfarms, may be used to inform the calculation of collision rate at a new or projected windfarm. The assumption in applying such avoidance rates is that the birds' behavioural response to the new windfarm will be similar to their response to the reference windfarm, and hence the proportion of birds avoiding the turbines of the new windfarm, further to the calculation of a no-avoidance collision rate, is likely to be the same as for the turbines of the reference windfarm.
Thus, having established values A' basic and A' extended for non-avoidance, as derived from the reference windfarm, these same values may be assumed to apply to new or projected windfarms for the same bird species. If C basic(new) and C extended (new) are the no-avoidance collision rates calculated for the new windfarm, the predicted collisions after avoidance for the new windfarm are:
basic model: A' basic C basic (new) .. eq. S4a
extended model: A' extended C extended (new) .. eq. S4b
A' basic is the avoidance rate established from the reference windfarm(s) using the basic model, and A' extended that using the extended model; they are related as in equation (S3).
Dealing with lack of information on g(ref)
Published information on avoidance rates for reference windfarms has not so far included information on avoidance using the extended model, or on g(ref), the ratio between the outputs (before avoidance) of the extended and basic models. Calculation of g(ref) requires information on bird size and speed, turbine parameters, and the flight height distribution at the reference site; however it does not need information on bird density, levels of flight activity, or number of transits. If this limited subset of data is available, then it should be possible to calculate g(ref) for the reference windfarm, for the bird species under assessment.
It is recommended that any future publication of reference avoidance rates, derived from collision monitoring studies, should state both that for use in the basic model and that for use in the extended model. This will require application of both models to the reference windfarm.