**5. Summary**

236 Autonomous Underwater Vehicles

interpretation of the results. The central method to approach this goal is to analyze videorecordings made during UV transects based on detailed description, categorization and registration of the entire behavior observed with special emphasis to separate human-induced responses from natural behavior, followed by statistical comparisons. Additional data on the biology and ecology of the target species, the physical and biological environment, and the

To reduce complexity, the number of influential variables should be minimized whenever possible. Optimally, the same design models of UV's should be used during all dives that need to be compared. Dive transects, video recordings, and data analysis should follow standardized protocols. During each transect representative size measurements combined with estimates of absolute swimming speed should be obtained from each studied fish species. Visually well identifiable species should be preferably selected for study so to minimize possible informational noise introduced by species differences within composed groups. Groups of closely related species should be used only exceptionally, when *in situ* species identification is impossible and the species have a very similar body structure, hence similar behavior can be expected. Short video or photographic close-ups of each individual fish from problematic species groups should be taken (preferably by a second camera) to visualize diagnostic details helpful for species identification. Advice and assistance from

Use of different UV's in comparative studies cannot be recommended, because it may turn out to be difficult, if not impossible, to adjust for disturbance effects. Most certainly more than a single signal source of disturbance needs to be considered. Experimental manipulation of light, sound, and vehicle velocity, but possibly also the magnetic field, singly or in combination, might certainly assist to better understand the relative importance of these potential sources of disturbance (see also Stoner et al. 2007). However, for full control of disturbance effects from UV's, one would also need to investigate the receiver bias and in particular the sensory equipment (Popper & Hastings 2009) and reaction norms (Tuomainen & Candolin 2010) which may differ considerably among fish species,

The longer the encounter with an UV the more increases the chance of interactions and evocation of disturbance responses. During longer UV stops, odor plumes deriving from collected organisms or bait brought along may be formed and scavengers may be attracted (Trenkel & Lorance 2011). If point observation are made during longer stops of an UV, these data should be treated separately from transect data. Also, when stationary, the vehicle itself may be perceived in quite different ways than when transitionally encountered during transects and disturbance responses may change and in some cases shift from avoidance to attraction or even to aggression (see for instance, Moore et al. 2008). Observations of deep-sea fishes deriving from longer-term interactions with UV's are certainly interesting *per se*, but may not always contribute to properly understand natural behavior. To reduce interactions it may be of advantage to position the vehicle firmly on the ground and switch off the motors for behavioral observations close to the bottom. During point observations in the open water as well as close to the bottom switching off the illumination and use of infrared light combined

As stated initially, investigations of the effects of UV's on deep-sea fish behavior have important implications for many other studies of deep-sea fishes, as for instance, *in situ* assessments of abundances, populations dynamics, habitat associations, community structure, and patterns of biological diversity (Stoner et al. 2008). Hence the suggestions and

effects and possible impacts of anthropogenic disturbance need to be integrated, too.

taxonomists specialized in problematic fish groups should be gathered.

with infrared-sensitive cameras should be considered (Widder et al. 2005).

populations, size classes, and ontogenetic stages.

An important prerequisite for *in situ* ecological investigations of deep-sea fishes using underwater vehicles (UV's) is to distinguish between disturbance responses elicited by the vehicles and undisturbed natural behavior. Nine case studies deriving from ten video transects along deep bottoms of the North Atlantic (Bay of Biscay, Mid-Atlantic Ridge) with a manned submersible and three remotely operated vehicles (ROV's) are presented to demonstrate differences in behavioral disturbance between vehicles, habitats, and species. Three species, roundnose grenadier (*Coryphaenoides rupestris*), orange roughy (*Hoplostethus atlanticus*) and false boarfish (*Neocyttus helgae*), and codling, a group of closely related species (North Atlantic codling, *Lepidion eques*, being the most common), were studied. During each UV transect recordings of disturbance responses and two activity patterns shown by undisturbed fishes, vertical positioning in the water column and locomotion mode, were made. Each behavior was subdivided into several categories and analyzed quantitatively using sample sizes larger than 18 individuals per species/species group and transect. Codling showed no disturbance responses to a manned submersible, while reacting intensely to a ROV during two transects performed in the same area. When the same UV was used, clear differences in disturbance responses were found between both adjacent dive transects and species/species groups indicating habitat- and species-specific responsiveness to signals emitted by the vehicle, in particular sound and light, but possibly also other sources. In three additional case studies, disturbance responses remained rather constant between transects or species, but natural behavior differed. The final study provides the fullest picture with all three behaviors differing, the interpretations being however complicated by the fact that different vehicles were used in different habitats. The findings are discussed emphasizing the significance of *in situ* quantitative behavioral studies of UV-based video transects in deep-sea fish ecology and related research fields. Detailed suggestions and recommendations towards optimization of vehicle-disturbance control and observation techniques are provided.
