Introduction:   The quantitative data on game fish behavior were obtained in the open water season of 2019, essentially between June and December of that year.  The results came from 38 trips on regional lakes.  Of these outings, 19 were on Lake Geneva in southern Wisconsin; I consider Lake Geneva to be my home lake.  The other 19 outings were on lakes in southeastern or northern Wisconsin, or the upper peninsula of Michigan.  Time of the year comparisons were limited to Lake Geneva because I had trips there during all the months of the open water season.



      (a) Fishing Cameras – Small cameras have been developed so fishermen can lower them into the water to watch fish come to their baits.  If attached to a monitor in the boat, the fisherman can observe in real time when the fish either bites, or turns away.  Or the events can be stored on a micro SD card to be viewed later.  Sometimes the camera can be trolled behind the boat, with the lure trailing in back, so the fish can be watched as they come to the lure.  If the camera is attached by cable, then the fisherman can watch in real time.  If attached to a fishing line, without a cable, then the video events can only be stored on micro SD cards because WiFi doesn’t work underwater.  In order to be trolled on a fishing line, the camera has to be small and streamlined so it will orient properly behind the boat.

      One of the first line-attached trolling cameras to come on the market was the Okuma Waterwolf system.  I purchased 2 of these and used them extensively during the 2018 open water season.  They were very streamlined and easy to use.  In fact, with the Waterwolf cameras I developed most of the trolling and data recording methods used for the present project.  However, it turned out that the resolution of the resulting videos wasn’t optimal.  I found I needed better resolution for viewing the videos on larger screens, producing snapshots from the video frames, and identifying individual fish under less than optimal viewing conditions underwater.

      The state of the art for trolling cameras appears to be what are called action cameras, such as the GoPro system.  Whereas there is a commercial trolling housing made for GoPro cameras, it generally needs to be attached to the boat by a downrigger or teaser cable.  I did purchase one of these devices and used it to some extent, but there were problems.  It was very large and heavy, and difficult to troll with a fishing line.  It was also cumbersome to remove the camera for battery changes.  In addition, the angle of the camera could not be adjusted enough to account for different diving depths of the attached lures.  I needed to design my own lightweight housings for GoPro cameras.

      (b) Custom Underwater GoPro housings – I generally used GoPro Hero 5 Black cameras for this project.   This camera is configured as a rectangular box 2.4x1.7x1.0 inches with the lens set off somewhat to the left facing forward.  Because the lens is on the largest face of the rectangle, it is awkward to troll.  The camera itself weighs 4.2 ounces.  Also, the Hero 5 Black is waterproof down to 33 feet without a case.  But sometimes the cameras would end up deeper than that.  Therefore, they need a waterproof case.  These are made by the GoPro company and are called super suits.  They fit the cameras nicely, and have a footing on the bottom for attachment to other surfaces.  I needed a housing that was small and light, but tough enough to protect the camera from logs or rocks or whatever.  And it had to be streamlined enough to move through the water without wobbling or swimming like a fishing lure. 

      After some thought it seemed as if the right size plastic funnel might work for the hydrodynamic part.  Then I used a thin brass plate to attach a footing for the super suit, and connected that to the distal (wide) end of the funnel (at the bottom) with a piece of one-eighth inch threaded brass rod – attached vertically across the wide end of the funnel.  I connected the threaded rod with a piece of 300 pound stainless steel trolling wire that led out through a plug at the funnel’s small end, and twisted into a loop for attaching the line.  I drilled small holes in the brass footing plate to attach leaders for the lures.

      The finished product didn’t look like much – just a black funnel with the super suit and camera sticking out somewhat from the wide end.  The finished funnel housing without super suit and camera weighed just 4 ounces.  The camera and super suit added another 8.2 ounces for 12.2 altogether.  It turned out that the weight of the camera and super suit were concentrated at the bottom of the housing, and this property caused the camera to orient properly all the time.  Because the super suit protruded somewhat from the funnel part of the housing, it could be angled enough to keep the lure centered in the camera’s field of view.  This applied to lures that dove down (like shad raps or husky jerks), or ones that ran straight (like worm harnesses).  Also, I could angle the super suit down enough to remove the camera for battery changes without taking the device apart.  By chance, when trolled these housings ran about 15 feet below the surface, which turned out to be optimal for this project.  I have since made similar housings with a diving light incorporated for use at night.

      (c) Still cameras on the bottom.   I also made camera housings to sink to the bottom in good spots for videoing fish there.  These consisted of a short pole-like section that fit into a flag holder base.  The latter was made of cast iron and heavy enough to sink the whole rig to the bottom.  The pole had a GoPro footing attached at the top; here I placed an underwater light, and then the camera in its super suit above that.  On top of all this was a line that led to a float above.  These cameras were placed into areas where I thought the fish might congregate.  Then I usually fished around there to see if anything was biting.  The results didn’t always correlate with what was going on in the videos.



      (a) Water Clarity in Freshwater Lakes – The water in even the clearest freshwater lakes still has crap suspended in it.  The worst offenders are microscopic algae, particularly Cyanobacteria species, which use pigments such as chlorophyll A to generate energy from sunlight.  They tend to tint the water an unpleasant shade of green.  And worse, many of the fish (especially muskies) are often more or less green, too.  This is probably not a coincidence, and this factor tends to reduce the contrast between the water and the fish.  Therefore, I needed to find the clearest lakes possible for my video outings.  There are lists of the clearest lakes on the internet, but I also knew of some from previous outings.

      (b)  Lake Geneva – I had fished in Lake Geneva, (located in Walworth County, Wisconsin) in the 1990s.  Then it was primarily known as a good lake for smallmouth bass, big northern pike, lake trout, and panfish.  When I first started using the Waterwolf cameras in 2018 my first trip was to Lake Geneva.  At that time the Wisconsin Department of Natural Resources (DNR) listed the following species for the lake:  Smallmouth bass (abundant), largemouth bass (present), panfish (present), northern pike (present), trout (present), and walleye (present).  Muskie were not listed, but I wanted at least to get started with some other fish species. 

LK Geneva1LK Geneva1

      For my first trip that year with the Waterwolf cameras I trolled around at the deep edge of the flats area; this trip revealed on the videos 24 smallmouth bass, 14 largemouth bass, 12 rock bass, 5 northern pike, 4 walleyes, and many bluegills and pumpkinseed sunfish.  I came back a couple of weeks later to fish for the panfish, and got a real surprise.  A big muskie followed a hooked bluegill to the surface, then turned off.  Back at the landing I mentioned the muskie to another guy there, and he said he’d had the same thing happen earlier that day. 

      It didn’t take much of an internet search to learn what was going on.  Turns out the Wisconsin DNR had been stocking Lake Geneva with muskies since 2010.  These were the Upper Chippewa River and Upper Wisconsin River strains.  Plus, some of the local fishing clubs had been doing their own muskie stocking.  This effort has been called the Rosset Stocking Program.  It is named after the late Dean Rosset of the Chicagoland Muskie Hunters chapter of Muskies, Inc.  With financial support from local fishing clubs and private donors they are stocking Lake Geneva with the Leech Lake strain that is known for growing to large size.  To say that I owe them a debt of gratitude would be a serious understatement.

      In 2018 I made a number of trips to Lake Geneva, but did not see any muskies on the videos until September.  Then they began to turn up fairly frequently.  It appeared that late September through at least mid-November were the best months to encounter muskies on Lake Geneva.  At least on my videos.  In any event, this information was very useful in planning my outings for the 2019 season with the GoPro camera housings.  Lake Geneva is also closer to my home in Milwaukee than most of the other clear lakes, so I make many trips there for the video outings.  My favorite trolling locations on the lake are the flats area, Williams Bay, west of Conference Point, east of Cedar Point, Geneva Bay just northeast of Maytag Point, and Fontana Bay.  However, there are some really good fishing guides on Lake Geneva, and I'm sure they have lots of other good places.



      (a) Trolling Set Up - The GoPro cameras in their custom housings are light enough to troll with regular fishing tackle, but they are not exactly small.  They are larger in profile and heavier than the Waterwolf cameras.  However, I found that they could be trolled using fairly heavy rods equipped with line counter reels and 100 pound test Spiderwire braided line.  The heavy line was used mainly to get the cameras off when they became stuck on bottom objects.  The lures were attached to the camera housings with an 18 inch piece of fluorocarbon monofilament – usually 6 pound test, but sometimes heavier. 

      Lures used on these outings were generally limited to crankbaits, such as Rapala husky jerks or smaller Rapalas, and 2-hook worm harness rigs with either nightcrawlers or minnows attached to the hooks.  For the crankbaits the treble hooks were removed so that a fish wouldn’t get hooked and jerk the camera around.  When this happened, it would generally mess up the recording.  On the other hand, to keep bait on the worm harnesses the hooks needed to be intact.  It turned out that muskies were definitely attracted to the worm harnesses, but they never seemed to strike them.

      (b) Trolling Techniques - Usually I would troll two cameras at once, except for lakes in Vilas and Oneida counties in northern Wisconsin – where one is limited to a single trolled line when fishing alone.  The GoPro camera settings were as follows:  2.7K, 4:3 aspect ratio, 30 frames per second.  This aspect ratio was used because the fish often went up and down behind the lures, rather than side to side.  The 30 FPS setting was used to improve the light available underwater.

      I generally let the cameras back 60 feet, and trolled at 1.75 miles per hour.  The lures seemed to work best at that speed.  Since the housings ran about 15 feet deep, I usually stayed in 20-25 feet of water, often near drop-offs.  However, I often strayed into deeper water and sometimes (unfortunately) into the shallows where entanglement with weeds was a real problem.  Weeds are difficult because they flap around in front of the lens and trash the video.  I used a weed-catcher to prevent surface weeds from sliding down the line to the cameras.  This device consisted of a big 3-way swivel about 3 feet up the line (which went through the center of the swivel), held in place by a 2-way swivel below.  It worked pretty well unless I hit into deep weeds near the bottom.



      (a) Recording the Fish Seen - After each trip I would take out the cameras, charge the batteries, and download the micro SD card files.  The GoPro videos would come out as MP4 files, and were divided into segments of 8 minutes 52 seconds each (with newer model cameras the segments would be 11:49).  I would label each one as to the camera it came from and its number in sequence.  For each trip and camera I noted the date, lake, location on the lake, the lure, and any other pertinent information.

      Each evening I would view the video segments using the VLC program.  Screenings were done at 2.45X speed.  If a fish came up the video was viewed at a slower speed, and then the species (and number if more than one), the time on the video, and anything else interesting would be noted on a log sheet for each trip.  Later I could go back and check different parameters for each fish seen on the videos.

      (b) Documenting Behaviors -  A number of characteristics in the ways the fish approached the lures and cameras could be analyzed, as discussed below.

            (1) Follows – As an arbitrary rule, a follow was defined as the fish pursuing the lure for at least 5 seconds.  Muskies and sometimes other fish would appear to be attracted to and follow the cameras as well.  This behavior was also considered a follow.  By going back to the original video I could check to see the time at which the follow began, and when it ended.  Because the boat speed was kept to about 1.75 miles per hour, the follows could be expressed as both duration and distance traveled.  Then the data could be compared for the various fish species.

            (2) Muskie Depth and Angle of Approach – Because the cameras stayed at about 15 feet below the surface when trolled as described above, I could estimate the depth from the appearance of the bottom.  Obviously these are only estimates.  However, at 15-20 feet the camera would either be right at the bottom, or at least the bottom features would be very clear.  At 20-25 feet the bottom would be visible somewhat, but more as a yellowish discoloration than distinct features.  Deeper than 25 feet the bottom would not be visible at all.  Therefore, when a muskie first appeared, its depth could be approximated using these measures.

      When a muskie was first seen, the angle at which it approached the camera and lures would also be noted.  That is, did it come up from below, down from above, or on the same plane?

            (3) Strikes – At times it was difficult to determine what actually constituted a strike.  However, if a fish actually opened its mouth on approaching the lure, or physically bumped into it, I considered that a strike.

            (4) Assessing Curiosity – Some fish such as bluegills appeared only interested in the lures.  Others, particularly muskies, seemed attracted to the cameras as well.  To quantitate this behavior, I went back over every fish seen on the videos and determined if the fish approached only the lure, or if they also came up to the cameras as well. 

                (5) Statistics - Quantitative data were expressed as mean + SD, with significance determined by an analysis of variance and Tukey's tests for individual comparisons.  For numerical data, significance was determined by Chi Square analysis; if more than two comparisons were made, the Bonferroni correction was made to the p value required for significance.

      (c) Snapshots - The VLC viewing program allowed for a frame by frame review and the ability to save a good one as a snapshot.  The GoPro files were fairly big – as noncompressed TIFF files each frame represented 21.9 MB (note that as noncompressed files they were all the same size).  In comparison, the Waterwolf system yielded much smaller files for individual frames.  Each Waterwolf frame as a noncompressed TIFF represented 3.7 MB.

     Since I had recorded where each fish turned up on the videos, I could go back and look for good frames that might yield decent snapshots.  I calculated that from all the GoPro videos generated during the 2019 open water season, I had over 15 million individual frames – hopefully enough to choose from.  However, It actually was difficult to find really good frames because there always seemed to be something wrong.  Most often there would be no fish, or if they were present they would be too far away or not in the right part of the field.  If there were any weeds on the lure or camera, that would screw up the frame.  And motion was a real problem.  If the fish were to be attacking the lure too quickly, it would look blurry at the 30 frames per second speed.  The same issue held true for the lures.  Some just had too much motion and always looked blurry.  For some reason 5.5 inch husky jerks and the worm harnesses with night crawlers or minnows looked the best.  I could pick out frames where the lure looked as if it were just laying still.  But then the fish had to be close and look good in the same frame, too.  Not always easy, but on the other hand there were a lot of frames from which to choose.