Abstract: Animal flight behaviour in response to aircraft could have a profound influence
on the accuracy and precision of aerial estimates of population size but is rarely
investigated. Using independent observers on the ground and in the air we
recorded the presence and behaviour of 17 groups, including 136 individually
marked horses, during a helicopter count in New Zealand’s Kaimanawa
Mountains. We also compared the helicopter count with ground-based
estimates using mark-resight and line-transect methods in areas ranging from
20.5 to 176 km2. Helicopter counts were from 16% smaller to 54% larger than
ground-based estimates. The helicopter induced a flight response in all horse
groups monitored. During flight, horse groups traveled from 0.1 up to 2.75 km
before leaving the ground observer’s view and temporarily changed in size and
composition. A tenth of the horses were not counted and a quarter counted
twice. A further 23 (17%) may have been counted twice but only two of the
three observers’ records concurred. Thus, the helicopter count over-estimated
the marked sub-population by at least 15% and possibly by up to 32%. The net
over-estimate of the marked sub-population corresponded to the 17% and 13%
difference between helicopter counts and ground-based estimates in the central
study area and for the largest area sampled, respectively. Feral horse flight
behaviour should be considered when designing methods for population
monitoring using aircraft. We identify the characteristics of the helicopter
count that motivated horse flight behaviour. We compared our own recent
estimate of population growth from measures of fecundity and mortality (λ =
1.096 with an earlier-published one (λ = 1.182, where r = 0.167) that had been
derived by interpolating between the available history of single counts. Our
model of population growth, standardised aerial counts, and historical estimates
of annual reproduction suggest that the historical sequence of counts since
1979 probably over-estimated growth because count techniques improved and
greater effort was expended in successive counts. We used line-transect, markresight
and dung density sampling methods for population monitoring and
discuss their advantages and limitations over helicopter counts.