Comparisons of measured and estimated distances and angles from sightings surveys

Résumé

Photogrammetric systems using video cameras were used to measure radial distances to sightings during the SCANS-II, CODA and SOWER surveys. These surveys included sightings of a variety of species from harbour porpoise, at distances of a few hundred metres, to large baleen whales at distances greater than 10km. A total of 910 initial sightings with estimated distances from reticles and measured distances from video, using 7 × 50 (636) or 25× ‘Big Eye’ (274) binoculars, were compared. Bearings to sightings were also measured from still images. The CVRMSE in distances varied between 0.19 and 0.33 for reticle binoculars. Comparisons of measured distances to simultaneous sightings by other observers using naked eye gave a CVRMSE of 0.39 for naked eye estimates. There was a consistent, non-linear pattern in all data sets, of over-estimating close distances to sightings of surfacing cetaceans and under-estimating those further away. However, this pattern was not evident from the distance experiments on SOWER to fixed targets which also had a much lower variance (CVRMSE = 0.13). Bearing data from SCANS-II and CODA showed around 5% of estimates had gross errors greater than 20º that were attributed to mistakes. For the remaining values, RMS errors were in the range 5.7º–7.2º for SCANS-II and CODA and 4.9º for SOWER. Both distance and angle errors will make a substantial contribution to the variance of abundance estimates and simulated data showed that the observed non-linear nature of distance errors may cause considerable bias even when linear regressions might suggest little bias. There still remain technological challenges in operating complex electronic systems at sea to measure distances and bearings, but investment in these methods should be a cost effective way of reducing bias and improving precision of cetacean abundance estimates.