Contrast degradation theory predicts that increased turbidity decreases the visibility of objects that are visible at longer distances more than that of objects that are visible at short distances. Consequently, turbidity should disproportionately decrease feeding rates by piscivorous fish, which feed on larger and more visible prey than particle-feeding planktivorous fish. We tested this prediction in a series of laboratory feeding experiments, the results of which indicated that prey consumption by two species of planktivorous fish (juvenile chum salmon (Oncorhynchus keta) and walleye pollock (Theragra chalcogramma)) is much less sensitive to elevated turbidity than piscivorous feeding by sablefish (Anoplopoma fimbria). Planktivorous feeding in the turbidity range tested (0–40 nephelometric turbidity units (NTU)) was reduced at high light intensity, but not at low light intensity. Comparatively low (5–10 NTU) turbidity decreased both the rate at which sablefish pursued prey and the probability of successful prey capture. These results suggest that turbid environments may be advantageous for planktivorous fish because they will be less vulnerable to predation by piscivores, but will not experience a substantial decrease in their ability to capture zooplankton prey.

It is often assumed that visual survey data provide more accurate fish counts than conventional extractive gear. As a result, use of underwater vehicles to assess the abundance and distribution of fishes has increased rapidly over recent years. However, a review of observations reported for 48 demersal marine fish taxa showed that almost all respond in some way to underwater vehicles. Whether or not movements or changes in behavior affect survey bias is more difficult to assess. A simple conceptual model is presented to evaluate relationships between stimulus intensity, distances from the vehicle where reactions occur, and survey bias. Largest bias is caused by attraction or avoidance that occurs outside the field of cameras or observers. While light level and vehicle speed have been explored experimentally in a few cases, much remains to be learned about how bias varies among species, age groups, different vehicles, and operating conditions. Given poor understanding of survey bias, we recommend that surveys be conducted with minimum possible variation in operations and that vehicle time is devoted to experimental evaluation of methods. There is no good substitute for direct observations on fish behavior, distribution, and abundance; and survey design can be improved through experimentation.