Boat-based surveys have been commonly used to monitor sea otter popula
tions, but there has been little quantitative work to evaluate detecti
on biases that may affect these surveys. We used ground-based observer
s to investigate sea otter detection probabilities in a boat-based sur
vey of Prince William Sound, Alaska. We estimated that 30% of the otte
rs present on surveyed transects were not detected by boat crews. Appr
oximately half (53%) of the undetected otters were missed because the
otters left the transects, apparently in response to the approaching b
oat. Unbiased estimates of detection probabilities will be required fo
r obtaining unbiased population estimates from boat-based surveys of s
ea otters. Therefore, boat-based surveys should include methods to est
imate sea otter detection probabilities under the conditions specific
to each survey. Unbiased estimation of detection probabilities with gr
ound-based observers requires either that the ground crews detect all
of the otters in observed subunits, or that there are no errors in det
ermining which crews saw each detected otter. Ground-based observer me
thods may be appropriate in areas where nearly all of the sea otter ha
bitat is potentially visible from ground-based vantage points.