Observer precision and bird conspicuousness during counts of birds at sea

Simultaneous transect counts of seabirds by pairs of experienced, ship-based observers were compared to estimate count precision, defined as the proportion of birds counted by both observers. Precision was low; at most 51 per cent for all birds combined, with values for different taxa ranging between 0 and 57 per cent. High abundance of birds and large body size (greater conspicuousness) tended to increase count precision, whereas ship-following behaviour decreased precision. Two main sources of imprecision were identified: failure to correctly categorize birds (in or out of the transect area, passing or following the ship) and failure to detect birds. Incorrect categorization could be avoided by adequate training. Biases caused by specific differences in detectability could be reduced by increasing vigilance and by making use of the best possible vantage. Calculation of values of observer precision permits an estimate of the reliability with which different species are detected.