**1. Introduction**

26 Underwater Vehicles

224 Autonomous Underwater Vehicles

van der Krogt, R. (2005). *Plan repair in single-agent and multi-agent systems*, PhD thesis,

van Heijst, G., Schreiber, A. & Wielinga, B. (1996). Using explicit ontologies in kbs development, *International Journal of Human-Computer Studies* 46(2-3). von Alt, C. (2010). Remus technology 2020 and beyond, *NATO Naval Armaments Group –*

Yuh, J. (2000). Design and control of autonomous underwater robots: A survey, *Journal of*

*Maritime Capability Group 3 on Mines, Mine Countermeasures and Harbour Protection –*

Netherlands TRAIL Research School.

*Industry Day*, Porton Down, UK.

*Autonomous Robots* 8(1): 7–24.

Fishes have a wide range of perceptual capabilities allowing them to behaviorally respond to various environmental stimuli such as visual, acoustic, mechanical, chemical, and electromagnetic signals. In our "noisy" world of today many artificially evoked signals pass through aquatic habitats, where fishes perceive them and respond to in often unpredictable manner. Proper distinction between natural and artificially evoked (="disturbed") behavior is of utmost importance in ecological studies that try to identify the prevailing factors and mechanisms influencing fish abundance, distribution and diversity.

As we know today, the need to consider human-induced behavioral disturbance as an important factor in ecological studies (Beale 2007) applies even to inhabitants of remote aquatic habitats such as the deep sea. *In situ* studies using various types of underwater vehicles (UV's) have significantly changed the conception that the inhabitants of the deep, dark and mostly cold ocean are less behaviorally active and hence less susceptible to anthropogenic disturbance. While direct observation of deep-sea animals goes back to the time of William Beebe in the 1930s, *in situ* studies of deep ocean organisms and their habitats have become increasingly more common during the last 50 years.

After initial use for exploration and discovery of yet unknown habitats and organisms, UV's were adopted to systematically investigate the ecology of deep-sea organisms, especially the larger and easier observable fauna in the open water and close to the bottoms. In analogy to census studies conducted by divers in shallow waters, vertical or horizontal transects with underwater vehicles were used to obtain density or distributional data of fishes (e.g., Yoklavich et al. 2007, Uiblein et al. 2010). Distinct fish species or closely related taxonomic groups were found to occur at relatively high densities during such transects allowing quantitative behavioral investigations.

Early *in situ* exploration encountered first evidence of pelagic and bottom-associated (demersal) fishes living at depths well below 200 m being behaviorally active similar to shallow-water species (Beebe 1930, Heezen & Hollister 1971). These preliminary behavioral observations were followed by detailed studies of locomotion behavior and habitat utilization based mainly on video equipment employed during bottom transects with manned submersibles (e.g., Lorance et al. 2002, Uiblein et al. 2002, 2003) and later with

Deep-Sea Fish Behavioral Responses to Underwater Vehicles:

Differences Among Vehicles, Habitats and Species 227

Table 1. Overview of dives, vehicles and video transects, with numbers of encountered fish

Close distance

Locomotion behaviour Inactive Drifting Station holding Forward moving

The four species/species groups selected for detailed analysis were the roundnose grenadier *Coryphaenoides rupestris* (family Macrouridae; Fig.1), the orange roughy *Hoplostethus atlanticus* (family Trachichthyidae; Fig.1) the false boarfish *Neocyttus helgae* (family Oreosomatidae; Fig.1) and codling (family Moridae). The term "codling" includes the most common *Lepdion eques* (North Atlantic codling; Fig. 1), its congeners *L. guentheri* and *L. schmidti*, and the slender codling *Halagyreus johnssonii*. Identification of species/species groups was based on the size and form of the body, head and fins, and color patterns and

The recording of all behaviors started immediately after a fish appeared on the video screen. Four main behaviors, overall activity level, disturbance response, locomotion, and vertical positioning above the bottom, each consisting of two or more categories, were recorded for subsequent statistical analysis (Table 2). Fishes visualized on video with high or increasing swimming speed indicating burst swimming in response to prior disturbance by the submersible ("arriving disturbed") were excluded from further-going behavioral analyses. During the subsequent behavioral recordings, the UV frequently got closer to the fishes, with increasing illumination intensity caused by the front lights. If a disturbance response was observed during this process (i.e. a marked change in activity level and/or locomotion behavior), the recordings of locomotion or vertical body positioning were stopped immediately before the occurrence of this behavioral change. The disturbance response during UV approach was split into two separate categories, depending if it happened still at far distance or at close distance to the UV and mostly within the highest illumination radius.

Far distance

bottom Far above bottom

Arriving disturbed

per transect. Samples analyzed are highlighted. For further explanations see text.

Behaviour Category

response

column Close to bottom Well above

distributional data from the respective area deriving from collected material.

Disturbance response No

Table 2. Overview of the behavioral categories studied

Vertical position in water

ROV's (e.g., Trenkel et al. 2004a, Lorance et al. 2006). Quantitative behavioral comparisons conducted with the submersible Nautile clearly showed that fish species differ among each other in the way they swim and in their vertical positioning above the bottom (Uiblein et al. 2003). Moreover, distinct responses to the approaching vehicle were identified which needed to be analyzed in detail so to be able to distinguish natural behavior from responses to anthropogenic disturbance. That underwater vehicles have a disturbance effect on fish behavior has also important consequences for fish density calculations from *in situ* transects, as the data may not reflect natural conditions when disturbance responses are intense and/or occur frequently (Trenkel et al. 2004b, Stone et al. 2008).

Disturbance responses in deep-sea fishes may be caused by a number of factors like noise produced by motors and thrusters, light used for illumination purposes, motion, electromagnetic fields, or odor plumes deriving from the vehicle. Detailed investigations regarding the actual source(s) of disturbance are generally lacking. Here, a description and categorization of disturbance responses is provided and differences between vehicles, habitats, and species are elaborated. These data suggest that disturbance responses are manifold and can – by themselves – reveal interesting insights into the life modes of deepsea fishes. In addition, when disturbance responses are identified, natural behavior (e.g., locomotion and vertical positioning above bottom) can be filtered out and studied independently of artificial evocation.

Here, nine case studies based on manned submersible and ROV video transects in the deep North Atlantic are presented dealing subsequently with differences in disturbance responses between underwater vehicles, dive transects (habitats), and co-occurring species/species groups. In addition, a separate section is devoted to combined analyses of natural behavior and disturbance responses, to show the full picture. These results are discussed referring to (1) novel insights about deep-sea fish artificially and naturally aroused behavior, (2) the need for consideration and integration of all influential factors in the behavioral analysis and interpretation, and (3) future technological possibilities and challenges towards optimizing *in-situ* investigations on the behavior and ecology of deepsea fishes.
