The ecoSCOPE and dynIMAGE: Microscale Tools for in situ Studies of Predator Prey Interactions

Uwe Kils

(Christian Albrechts Universität Kiel)

(accepted April 23 1992 -
Arch Hydrobiol Beih 36: 83-96)

(this publication is a sketch of the technical setup and some recordings of transparent juvenile fish, performed  from the ATOLL iceSTATION Laboratories in the Baltic, Kiel, Germany - manuscript for the international meeting @ Lake Lacavac on quantitative imaging and acoustics - invited speaker - mirrored here partially for discussion and teaching purpose only - for citation go to the original printmedia)


The behavior and microdistribution of juvenile herring Clupea harengus and the copepod Acartia discaudata have been investigated in Kiel Fjord. For in situ measurements at high magnifications in real-time, two tools have been developed: the "ecoSCOPE", an imaging system allowing for an approach close enough to view the evasive predators as well as their prey via an endoscope, and the "dynIMAGE" program, an image processing software allowing for compensation of system-vibrations and microturbulences. The dynamic processes inside of a zooplankton-microlayer are analyzed at the mm-scale to investigate how small-scale physics affect feeding success. At microscales the maximum organism-concentrations were several orders of magnitude higher than derived from conventional net samples. Juvenile herring (mean length 38 mm) grazed down the copepod patches (up to 850 individuals per liter) within minutes, attacking each copepod separately. The mean frequency of this particulate feeding was 2.4 captures per second at the individual level and up to 84 per liter per second at school level, indicating the extraordinary grazing success of clupeid predators in micropatches caused by interactions between ocean physics and biology.


The importance of microdynamics and microdistributions of predator and prey for the ecology of the early life history of fish has been emphasized by Lasker (1975, 1988), Vlymen (1977), Steele (1978, 1980), Wroblewski (1984), Mackas et el. (1985), Rothschild (1986, 1988), Rothschild & Osborn (1988) and others. Patchiness of prey as well as patchiness of predators - mostly referred to as schooling - are regarded to be key factors in determining the successful recruitment of many commercially important pelagic fish species. The importance and implications of behavioral studies have been demonstrated by Blaxter & Hempel (1963), Brooks & Dodson (1965), Rosenthal & Hempel (1970), Blaxter & Staines (1971), Rosenthal & Fonds (1973), Blaxter (1974), Hunter & Thomas (1974), Hunter (1980), Janssen (1980), Coughlin & Strickler (1990) and others from laboratory investigations. Several working groups now apply optical systems in situ (see discussion) and their results clearly indicate the need for more detailed studies at microscales with very high resolution in time and space. This paper describes the function of two tools permitting a closer view of the very complex processes when highly motile predators encounter extremely alert and evasive prey. These tools have been developed within a project on the orientation of pelagic fish in environmental gradients. One objective of the project is to learn how herring-schools orientate under the changed environmental conditions of the Baltic: how do they search and encounter their prey under the drastically reduced visibilities of the enhanced phytoplankton blooms observed in the Baltic during the last decade? Some preliminary quantifications of feeding dynamics of juvenile herring and the escape capacity of their prey are presented and discussed.

Material and Technology

In an area of the Baltic Sea known to be a major feeding ground for juvenile spring spawning herring, optical and acoustical in situ systems were deployed for behavioral studies. During days with moderate wind speeds, copepods concentrated directly at the halo- and thermocline below the flume of the Kiel Canal in front of the locks of Kiel-Holtenau (54.22.05 N, 11.09.05 E). In late summer, schools of juvenile herring (Clupea harengus, body lengths between 3.5 to 6 cm) migrate through this area into the open Baltic. The visibility is generally very poor due to massive plankton blooms. In this location, the swimming laboratory "ATOLL" (a 27 m doughnut-shaped structure) is anchored (Kils 1986), allowing for long time monitoring of this process: In front of underwater observation windows, optical sensors measure the microdistribution and the behavior of predators and their prey in relation to ocean physics and biological gradients. In calm weather conditions, the copepods concentrated in thin microlayers at a depth of 1.2 meter. For example copepod concentrations (Acartia discaudata, > 95 % adult females, some A. bifilosa) increased from less than 1 ind. / l to more than 850 ind. / l over a vertical distance of 30 cm. The vertical extension of these microlayers was a few decimeters. Although the visibility was very poor (Secchi depth 1.8 m) juvenile herring schools found these micropatches within half an hour from the time they first formed, and within minutes they grazed them down.

The ecoSCOPE

The optical and acoustical sensors are deployed from a small remotely operated vehicle (ROV). Although this ROV is quite small and quiet it was impossible to approach feeding herring closer than 40 cm (Fig. 1). From this distance, however, the copepods in front of the herring are invisible due to the deflection of light by phytoplankton and microparticles in the highly productive waters of the western Baltic. By imitating the long and thin snout of the garfish protruding into the security sphere of the alert herrings (Fig. 1), we designed an endoscope with a tip diameter of 11 mm. The endoscope is camouflaged to reduce the brightness-contrast against the background: the top is black and the sides are silvery. Additionally, the front of the ROV is covered by a mirror, reflecting a light gradient resembling the natural scene and making the instrument body virtually invisible to the animals. The "ecoSCOPE" allowed observation of the feeding herring from a distance of only 4 cm. The predators are illuminated by natural light, the prey by a light sheet, projected via a second endoscope from strobed LEDs (2 ms, 100% relative intensity at 700 nm, 53% at 690 nm, 22% at 680 nm, 4% at 660 nm, 0% at 642 nm). A second sensor is imaging other copepods, phytoplankton and particles at very high magnification (Fig. 1 CCD-array b, Fig. 2 shows the front part of probably Acartia discaudata). A scanning SONAR-sensor is used to find the microlayer and the herrings from 40 m distance. Details of the optics are described in Kils (1989). Another advantage of these small "optical probes" is the minimal disruption of the current-field in the measuring volume, allowing for less disturbed surveys of microturbulence and shear.

The dynIMAGE Software

The signals from the scanned images are transferred to the memory of the board-unit (three Motorola 68000, Euro-bus/VME, 8 Mhz, 12 Mbyte RAM) for storage and evaluation. During the measurements the images can be observed in false colors in real time, strobed or frozen. The last six seconds can be played back on demand from memory. This direct observation can serve only for a first rough detection and estimation of high organism concentrations because the images pass too fast for the human eye when three coordinate-systems are constantly moving: the optical system vibrates, the objects move, and the water is turbulent in mm- and dm-scales. This prohibits a direct evaluation of the camera signals by the human eye, especially at high magnifications which result in considerable angular speeds of light from the focused objects. For further analysis, the software package "dynIMAGE" has been developed: After a measurement, the processing system displays the image-sequence of an attacking herring frame by frame to the operator. In the raw image-data, a quietly hovering prey or a stationary floating particle shows up on every frame at quite different locations, because the optical system has moved in the time elapsed between frames due to microturbulences in the dm-range accelerating the ROV. Such a hovering particle is marked by the operator with a cross-hair-cursor driven by a track-ball at one of the frames as a reference as well as on the frames before and after. The derived x,y coordinates are used by the program to shift all images to the first reference point, correspondingly. In this new film, the system vibrations are filtered out: the marked hovering particle is displayed stationary now and the eye can concentrate better upon the prey, the approaching predator and the remaining microturbulence-field.

Optionally each frame can be contrast-enhanced, changed in brightness and cleared from background-distortions and noise before the system recalculates the animation. This "shifted" scene can be played back on the board unit already in real-time or slow-motion, forward and reverse.

Alternatively dynIMAGE can synchronize the animation to the movements of the operator`s hand: moving the mouse to the right switches frames forward, moving to the left backwards. This tool serves as an interface to the image processing capabilities of the human eye and brain for a better interpretation of the fast dynamics by focusing the interest on any detail of the scene at optimal replay-speed.

If other particles adjacent to the marked particle now show up more or less stationary too, the selection was all right; if not, the reference-particle had its own propulsion - like a swimming zooplankton - so a new reference point should be selected.

Subroutines interactively calculate distances between particles, sizes of the organisms, areas, vectors of movements, angles between organisms, contrasts against the background, velocities of predator and prey and water velocity-shear by tracking algal cells or detritus. In the image-processing, markers can be left behind on the monitor for tracking and counting of organism frequencies. The results, the coordinates and the tracks can be stored as LOTUS 123-, ASCII- and TIFF-image files for export to MS-DOS systems. The processed animations are stored on magneto-optical disks. The images can be transferred to other computer systems for further evaluation (Meta-, TIFF- or EPS-format on IBM, Macintosh, NeXT or VAX). The animations can also be transformed to a standard video signal for tape recording (RGB or composite video in PAL or NTSC, 50 or 60 hz).

With respect to memory requirements and high processing speed a resolution of the images of 320 times 200 pixels at 16 intensity levels was selected for the behavioral analysis (DELTA and JPEC compression routines additionally reduce each 32 kilobyte images to approximately 2 - 25 kilobytes). This relatively low resolution is a fair compromise for evaluation of dynamic processes, considering that an animated sequence appears to the human eye much sharper than the single frames do. Due to the deflection of the light by phytoplankton, in the plankton-rich waters of the Baltic the sharpness of in situ images is not much higher anyway. For applications which do not need the full 50 frames per sec images can be scanned at 640 times 400 pixels in true colors (the CCD-chips have 440 000 pixels).

It will be only a mater of time until portable computer systems will be available to handle the high resolution in real-time.

Biological Results

Fig. 2 shows an image of the prey-sensor (see Fig. 1 CCD-array b), probably Acartia discaudata.

Fig. 3a shows an image of the predator-sensor (see Fig. 1 CCD-array a), it is image no. 21 of an animation as delivered by the system printer. The image shows the encounter of a copepod (marked as A, probably Acartia discaudata) and an approaching juvenile herring. The copepod is escaping at a speed of more than 72 mm per second over a distance of more than 8 mm. At the end of this escape reaction it tumbles for about 440 ms. However, the predator does not attack this staggering copepod directly in front of his mouth but is aiming at another one (Fig. 3b copepod B) more than 10 mm above his eyes. Another 160 ms later the herring is catching copepod B from below by opening the mouth (Fig. 3c) and 40 ms later spreading the opercula (Fig. 3d). The fish then returns to its horizontal position (Fig. 3e). In Fig. 4, the outlines of the silhouettes of the head of the approaching predator are drawn in a time-series (40 ms intervals) derived from images no. 17 to no. 22 indicating the high escape speed and the tumbling. The following milliseconds are drawn in Fig. 5 (images no. 23 to no. 30). Fig. 6 shows the velocities of the escaping copepod A and the results of the evaluation of another animation (C). In the first few seconds after the school detected the microlayer the average frequency of this particulate feeding was 2.4 per second (for a single fish). Four to six attacks could be observed in succession, as demonstrated in the time-series of Fig. 7. During this process, the predators come higher and higher in the water column due to the oblique attack direction. After swimming down without feeding at an angle between 40 to 70 degrees and a speed of 4 to 6 body lengths per second the herring immediately start a new feeding sequence.


With the help of an endoscope and high speed macro optics, it is possible to approach a feeding school of juvenile herring close enough to roughly image their prey as well. The dynamics inside such feeding assemblages are fast and complex. By "shifting" single images the system vibrations are compensated for a "filtered" playback of the processes to the human eye. This facilitates a better analysis of predator/prey interactions. The use of an LED light sheet permits imaging of transparent copepods and a rough quantification of prey-abundances as well as swimming distances, speeds and directions. Although the light of the LEDs is visible to the human eye, no noticeable responses of herring or copepods could be observed when they entered the light sheet alone. For herring the optimal sensitivity is around 510 nm with a minimal response at 660 nm (Blaxter 1964). At 660 nm the LEDs produce only 4% of the radiation emitted at 700 nm.

Herring predation act can be broken up as described for other species by Holling (1959), Gerritsen & Strickler (1977) and O'Brien (1979) into a series of discrete components: location, pursuit, attack and capture. The attack is directed upwards corresponding with the laboratory findings of Blaxter & Holliday (1958), Janssen (1978), Batty et al. (1990) and Gibson & Ezzi (1990). The observed mean frequency of 2.4 ingestions per second for a single fish is comparable to frequencies (1.9 - 2.3) found by Janssen (1976) for particulate feeding by Alosa pseudoharengus at high prey concentrations in the laboratory. The ram-jaw feeding mechanism of herring on a copepod is very similar to the blue-green puller fish Chromis viridis feeding on Eucalanus crassus as described by Caughlin & Strickler (1990), although in herring the process is considerably slower. The protrusion/retraction time for C. harengus is 40 to 80 msec which is close to the 30 to 60 msec Motta (1984) found for Chaetodon miliaris and in the range of 40 - 100 msec given for ten species by Motta (1988).

The wide opening and protrusion of the mouthparts as well as the steady position of the attacked copepods indicate that the herring uses little suction to avoid shear which would alert the prey (Janssen 1976, 1978; Drenner et el. 1978; O'Brien 1979; Launder 1980; Liem 1980; Launder & Liem 1981; Vinyard 1982; Wright & O'Brien 1984; Kettle & O'Brien 1987; Yen 1988). As indicated in Fig. 3 d, C. harengus did not seal its opercular cavities but spread out the opercula, possibly to allow water to flow out as water moves in the protruding mouth surrounding the prey. The observed escape speeds of A. discaudata (60 - 72 mm per second, flight distance 8 mm, body length = 0.8 mm) are similar to those reported for A. fossae (65 - 90 mm per second, flight distance 8 mm, body length = 0.7 mm) by Yen (1988), Cyclops strenuus (34 - 49 mm per second, body length = 0.6 mm) by Rosenthal (1972), and Cyclops scutifer (69 - 85 mm per second, body length = 1.5 mm) by Strickler (1975), but slower than those reported for Diaptomus sp. (132 mm per second, body length = 1 mm) and Cyclops scutifer (272 mm per second, body length = 0.7 mm) by Strickler (1977) or for the bigger Calanus finmarchicus (161 - 369 mm per second, flight distance 12 - 18 mm, body length = 2.2 mm) by Haury et al. (1980). The described successful escape of copepod A in Fig.4 indicates that a darting of 72 mm per second can be sufficient for a 0.8 mm copepod to escape from an attack of a 38 mm juvenile herring and portrays them as highly evasive prey as discussed by Singarajah (1969, 1975), Confer & Blades (1975), Swift & Fedorenko (1975), Drenner et al. (1978), O'Brien (1979), Drenner & McComas (1980), Haury et al. (1980), Kerfoot et al. (1980) and Vinyard (1982).

These preliminary and rough results indicate an extraordinary grazing success of clupeid predators when the random distribution of copepods is altered by ocean physics. Such extreme prey concentrations occur with a somewhat higher chance in calm weather conditions, at strong stratifications, and at horizontal edges like convergences or shorelines. Although these patches are exceptional in time and space, their encounter yields much energy to cruising schools of juvenile clupeids and should be of considerable importance for the early life history of pelagic species. From the measurements with these new tools a theoretical maximum grazing rate of up to 84 particulate feedings per liter per second at the school level can be estimated.

Denman & Gargett (1983), Sundby (1983), Yamazaki & Osborn (1988), Costello et al. (1990), Haury et al. (1990), Marasse` et al. (1990), Osborn et al. (1990), Sundby & Fossum (1990), Yamazaki & Kamykowski (1991) and Donaghay & Sieburth (1992) clearly show the biological importance of structures and dynamics in the mm- and msec-scales. More effort should be invested in further developments of imaging systems as tools for in situ ecological studies: Besides multi-nets (Wiebe et al. 1976), Longhurst-Hardy-Plankton Recorders (review by Haury et al. 1976), high frequency acoustics (Holliday et al. 1989; Smith 1992), optoelectronic counters (Herman & Dauphinee 1980), advanced in situ imaging instruments with high resolution in time and space like GULF III type plankton samplers with photo- or video- cod ends (Ortner et al. 1981) and the directly imaging Video Plankton Recorder (Davis et al. 1992) are needed for detailed descriptions of organism abundance and distribution. However, as a next step we should also enhance the instruments capabilities to quantify behavior like the in situ Crittercam, the ROVs of Bergstrðm et al. (1990) or Marschall (1988), the high speed in situ cameras of Motta (1988) or Kils (1989), the free-falling in situ camera with multi-strobes by Kils (1981) or the three-dimensional video system of Hamner et al. (1988). For a better understanding of microstructures ethological knowledge can be of great help, because the slightest alteration in behavior can cause patchiness (Fraenkel & Gunn 1961; Vlymen 1977; Hamner 1988; Price et al. 1988; Ramcharan & Sprules 1989). This research is envisioned of a developing effort on microscale methods and investigations which are planned under the umbrella of Global Ocean Ecosystems Dynamics (GLOBEC). One of the objectives of GLOBEC is the "Concentration of First Principles", the understanding of how important processes at the level of individual organisms control population abundance: "Special emphasis needs to be given to assessing the little-studied roles of ocean physics in feeding success, growth rates, reproductive output, and mortality rates including losses to predators. In both the physical and the biological disciplines, extensions of improvements in sampling from the mesoscale into the fine- and microscale environments are needed" (Muench 1989).


I thank Thomas Seidel and Helmut Thetmeyer for the software work, Alex Herman, Tom Osborn, Gus Paffenhoefer, Dietrich Schnack, Rudi Strickler, Hideka Yamazaki and Jeannette Yen for comments on the manuscript,
ATARI, NeXT and SONY for their friendly sponsoring and cooperation, the VOLKSWAGEN-FOUNDATION for funding, the DEUTSCHE FORSCHUNGSGEMEINSCHAFT (DFG) for support, Baudirektor G. Brandenburg and the administration of the Kiel-Canal.

Authors` address:

Institut für Meereskunde & Institut für Toxikologie

Christian Albrechts Universität

Kiel - Deutschland

current address:

Institute of Marine and Coastal Sciences
RUTGERS, the State University of New Jersey
New Brunswick, New Jersey 08903-0231, POB 231


update 020130

unter wasser mikroskop under water microscope cyber technology cyber scope remote sensing remote imaging high resolution underwater imaging 3D imaging quantitative imaging vrco virtuelle realitaet geeichte ueberblendung unterwasser optik robotics rov rovs fisch fish plankton usa germany danmark daynat markus uwe kils ecopost stativ halterung post arm arms titan kohlenstoff glass mikro maschinenbau micro engineering anguilla rostrata anguilla anguilla eel aal glassaal spawning endoscope flow laminar flow system rotation flow breeding aquaculture aqua culture cryo deep distribution micro micro manipulation stereo force feedback force feedback natureplus scienceplus obis fred grassle gotthilf hempel mildred scheel loki schmidt