We begin with a video of  juvenile Chinook salmon behavior.  It’s in high-definition, so watch it full-screen for best viewing.  We show a wide variety of behaviors here because they’re all very interesting, but the most common one in the Chena is drift-feeding, in which fish hold a fairly steady position in the river, facing upstream, and dart back and forth to intercept drifting prey as it passes.  You’ll also notice that these fish are often gathered into schools or shoals, because there are plenty of predators in the Chena and there’s safety in numbers.

It’s slightly unusual to see young-of-the-year salmonids grouped up like this, because normally they inhabit smaller streams, where it’s safe to spread out and defend larger individual feeding territories.  In this presentation, we looked at whether they are still territorial in a large-river setting.  This inquiry addresses a broader ecological question:  Can any schooling/shoaling fish also be territorial at the same time?

To answer that, we’re using 3-D video measurement technology introduced in this earlier post and updated in this poster (PDF download) at Luarca.  It allows us to do things like measure the 3-D configuration of the whole group of fish, and the distances between them (the green arrows):

We can also measure the habitat.  The video below shows prominent logs in brown, the boundary of dense cover in green, the surface in blue, the bottom in beige, and velocity vectors in yellow, which were calculated by tracking naturally drifting particles as they pass by the cameras.

We set out to use this system to detect territoriality in juvenile Chinook salmon.  We we found it, we hoped to learn something about the relative quality of each territory by seeing how much prey each fish catches.  However, when we started watching the feeding fish very closely, we noticed this happening a lot: the fish captures an item and spits it right back out.

Also, they often chase items they decide not to capture:

In fact, most times these fish chase a potential prey item, they get nothing for it.  About a third of the time, they decide not to even capture it.  Of the items they do capture, they spit more than half out.  The percentage of prey capture attempts that end with ingesting actual food is much lower for Chinook salmon in the Chena than for brook trout in still water, which were the subjects of the only other studies we found on this topic.  It makes sense that fish in a flowing river should feed less efficiently than those in still water, because there’s more suspended debris in a river, and because the fish have less time to identify moving objects before capturing them.

This principle may apply to juvenile salmonids in rivers in general, in which case ours is the first clear documentation of a pattern that could be very important for mathematical models of drift-feeding fish behavior.  Failing to account for the high debris error rate would lead current models to overestimate the amount of food these fish eat and underestimate the minimum size of suitable prey.  These errors would then influence predictions of growth, habitat suitability, and abundance.

The high debris error rate also means we can’t judge the relative quality of feeding territories by counting up the foraging attempts.  Two fish could be making the same number of attempts, while one gets mostly food and the other, positioned behind it, is stuck with nothing but debris to spit out.  That’s frustrating for our efforts to understand territoriality, but we can still tell a lot about territories even from failed attempts.  First and foremost, are these fish territorial in the first place?

Here’s the raw data to answer that question.  Each colored dot represents a foraging attempt, and each color represents a different fish.  The yellow arrow shows the direction of the water’s flow from the upstream end of the school.  These data cover 20 minutes of foraging by one school of juvenile Chinook salmon in the 70mm length range on August 14th, 2009.

You can tell there’s some exclusivity to the feeding areas of each fish, but it’s hard to judge exactly how much.  We can learn more by wrapping shapes called convex hulls around almost all the foraging attempts by each fish.  We use an objective algorithm to identify a few outlier points to exclude from these shapes, and the result is a very good representation of the core area each fish was using:

Now we can tell that these fish have very well-separated territories, but those territories look very different from the typical salmonid territories, which are spread across a large area of the stream in a 2-D mosaic.  All of these territories would fit within one typical territory for many other fish species of the same size.  Despite their close proximity, our calculations show that no fish performed more than 45% of its foraging attempts within the boundaries of another fish’s territory, and many had little overlap at all:

Summary statistics for a few territories. Click to view full-size.

This analysis highlights the exciting things we can do with the new 3-D video analysis program we developed, available at http://vidsync.sourceforge.net.  The first 20 minutes of video we’ve analyzed in detail have taught us some interesting things about juvenile Chinook salmon, and probably juvenile salmonids in general.  And that’s only the beginning — we have more than 12 hours to analyze, and will film many more this summer.  Stay tuned!

The village of Luarca, on the north coast of Spain. Photo by J. Neuswanger.

We presented both talks and posters.  Some materials from the talks (including high-definition fish video and 3-D animations of their feeding territories) will be available on this website within the next few weeks.  Click on the links below to download the posters as PDF files.

Wipfli et al 2010: Environmental processes affecting juvenile Chinook salmon in an interior Alaskan river – A summary of some key findings of the overall Chena project.

Neuswanger et al 2010: Improved 3-D video analysis methods, with applications to wild juvenile Chinook salmon behavior

Scientists interested in more information about these projects can email Mark Wipfli or Jason Neuswanger at our email addresses, mark.wipfli and jneuswanger, both @alaska.edu.  People interested in downloading the VidSync 3-D video analysis program, or its source code, can visit http://vidsync.sourceforge.net.  Help files and demonstration data for the program are coming soon.

Fish Video Highlights

These are some of the prettiest moments from my two years filming the foraging behavior of juvenile Chinook salmon.  They include a variety of revealing situations that show how the fish react to important elements of their environment such as food and predators.

Here’s a recording of the VidSync software, with a yellow arrow (part of VidSync’s built-in annotation features) following one fish through the course of a feeding maneuver.  It rises up from its “focal point” where it watches for prey, grabs something tasty, then sees a big piece of debris and scoots over to try it out.  It spits out the debris and goes back to its focal point.

Here’s the 3D digitization of that maneuver:

This is the same thing, viewed from 4 different angles:

Here’s a different prey capture maneuver, from the 2010 symposium on salmonids in Luarca, added to this post on 5/23/2010:

Of course, we don’t study maneuvers like that just to generate slick animations for PowerPoint shows.  The real point is to look at how far fish are moving for prey, which can be combined with information about water velocity and what the other fish are doing to help answer some important scientific questions.  For example, how much energy are they expending to capture prey — energy that could otherwise go toward growth?  Who gets more food for less energy, fish in fast water or fish in slow water?

Another important topic for our Chena research is territoriality.  The upper limit to how many fish a habitat can support depends on how much food there is and how the fish use that food.  When there’s a group of fish, there are millions of ways a limited supply of food could be divided between them.  Studying territoriality tells us something about how many fish are likely to do well in a given situation, and how many might be left out and risk starvation.

For a simple sense of territories, you can see the fairly uniform spacing among the fish in this 3D snapshot of a school of fish as the movie circles around it:

However, one moment in time can’t tell us much.  It gets much more interesting when we look at where the fish are relative to one another over a longer period of time.  The video below, from the VidSync software, shows part of 10 minutes of footage I analyzed to see the feeding territories of 4 fish in this video.  Each arrow marks a prey capture, and fish has different-colored arrows.

Here are the 3D territories.  At first glance on the original video it seems like the fish are just in sort of a mixed-up mess, chasing whatever tasty thing they see.  Watching them carefully and mapping their feeding in 3D immediately reveals some very interesting biological information: they’re respecting each other’s personal space remarkably well, with no intrusions during these 10 minutes.  It’s exciting to see a method reveal such interesting information even without the benefit of the large sample size and statistical rigor that will accompany this video method in practice.

FishDLT and VidSync Software for Scientists

This software is a work in progress: lots of work, and lots of progress.  It works well enough that we’re using it to process real data right now, but the current version of FishDLT has some personality (i.e. glitches) related to flaky behavior of Mathematica’s variable scoping in DynamicModules.  I’m planning to shift most if not all of the user interface stuff (buttons and picture-clicking) form FishDLT into VidSync, which will solve that problem and double (or better) the speed with which most things can be done.  That upgrade will probably be done by the time any other users begin to apply it to their data.  However, if you want to get started right away using the demo video, and you run into problems, contact me and I’ll be happy to help.

Hardware and Software Information/Downloads

Contact

Feel free to email me (Jason Neuswanger) about anything related to this method or these programs.  My address is my first initial dot my last name at sfos.uaf.edu, all lower-case.  (I can’t just write out my email address online, or spam robots will flood my inbox.)

We spent the long Fairbanks winter analyzing last year’s data and preparing results for several scientific conferences.  Most of us are beginning field work this week or next, so in recent weeks we’ve been purchasing, preparing, and organizing equipment at an ever-more-frantic pace.  In this picture, finishing touches are applied to a video calibration box that will help measure fish to millimeter-scale accuracy without even touching them.

Just like last year, Chena River users will notice mysterious contraptions up and down the river, collecting data for this project.  We’ll try to use this website to provide you with explanations of our more visible gadgets and the science they’re supporting.

First, some adult Chinook resting on their way up the North Fork of the Chena:

Juvenile Chinook feeding:

Juvenile Chinook versus the shockwave from an airboat:

All the videos are much lower-quality than the originals due to YouTube’s high file compression. To get a sense of the original HD video, here are some full-resolution stills captured from the HD video. Click a picture to see it full-sized.

Adult Chinook in the North Fork of the Chena, with a grayling and a whitefish in the foreground.

Juveniles foraging in the North Fork:

And in the main stem Chena:

Each of these videos and stills comes from a pair of side-by-side Sony HDR-SR12 video cameras in underwater housings with special wide-angle ports. Using two fixed cameras and some calibration equipment allows us to analyze the positions of these fish in 3D on the computer. That enables us to precisely measure their lengths, feeding positions, spacing, and many other scientifically interesting attributes.

First, here’s what one of our pools looked like in May during the peak of the spring snowmelt runoff:

And here’s the same spot during this flood:

Despite the high water, we managed to catch a few juvenile Chinook at the site above and gather information about their condition.  At least some of them have weathered the storm pretty well.  Some of our equipment (in this case, an insect trap tied to a log “above” the river) did not fare so well:

The water was beginning to creep out of its banks in places where the banks are a bit low:

Here’s the view downstream from the first bridge over the Chena on Chena Hot Springs Road, at what’s usually a popular gravel bar for camping and launching boats:

Even the eagles look bedraggled.

The North Fork is falling and clearing much faster than the main stem.  The high-water mark was about a foot higher than the level in the picture below.  The tree is new.

From July 24th to the 27th, I floated 50 miles of the Chena in a cataraft.  I stopped in prime locations to shoot stereoscopic, high-definition digital video of foraging juvenile Chinook salmon, which I’ll use to analyze the ways that habitat and competition affect their foraging success.  It took months of trial-and-error customization of the video equipment, but I finally got many hours of ideal footage for my analysis. This post simply chronicles the trip, and the fish footage will take a while longer to process, but you can expect to see it on the website soon.

The raft fit all my gear… barely.

Much of it gets unloaded at every data stop:

I spent a lot of time at huge logjams, which seem to be the most important Chinook habitat on the upper Chena.  At this one, the fish live deep in the back of the logjam:

Here’s more Chinook habitat.  The orange flag marks the top of the handle pole on my camera mount.

Spending 72 hours on the river allowed me to both better absorb the “big picture” and pick up on some little details I might otherwise have missed.  For example, I learned that many juvenile Chinook are now venturing far from cover to feed late in the evening, either because they’re safer from predators or because there’s enough food to justify the risk (or both).  This isn’t completely unexpected, but such behavior was seemingly absent early in the summer when the fish were smaller, and the timing of its development may be worth further study.  Here’s a photo from one such site, where the fish were surface-feeding on tiny midges:

It’s hard to spend so much time on the Chena without some golden photo opportunities, and I didn’t pass them up.  Enjoy the moose, eagles, baby mink, and scenery:

Baited minnow traps and unbaited snorkelers were used to count fish. Interestingly, the snorkelers were far more successful in the spring than the minnow traps, and were almost useless in the fall. They were cold both times, and whined occasionally throughout.

Because these different counting methods are not equally effective, it’s more useful to compare the patterns of distribution between spring and fall than the actual number of fish found at each site. Remember, these are raw counts, not estimates of the total fish population at the site.

Spring (trap + snorkel counts):

Fall (traps only; snorkel counts were negligible):

In the lab, these data are being cross-referenced with other relevant numbers, like aerial redd counts. Those maps will get posted later. The most evident pattern is a downstream shift in distribution in the fall. A few of the several hypotheses to explain this pattern include:

1. The slightly warmer temperature downstream is more favorable for growth.
2. The lower water clarity downstream offers more protection from predators.
3. A major mid-summer flood carried most surviving fish downstream.

The study will hopefully reveal which of these (if any) explains the downstream shift.

Update August 1, 2008. New observations suggest there may not actually be a downstream shift.  We used the same methods and sampled similar habitat in the upper and lower river in 2007, but lower-river fish may be more inclined to utilize that habitat than upper-river fish.  We don’t have conclusive proof either way, but this possibility is noteworthy.

This year we’ll be using higher-quality cameras in stereoscopic setups to do 3D analysis of fish lengths and positions, utilizing and advancing the methods described in these two papers:

Hughes, N. F., Hayes, J. W., Shearer, K. A., & Young, R. G. (2003). Testing a model of drift-feeding using three-dimensional videography of wild brown trout, Salmo trutta, in a New Zealand river. Canadian Journal of Fisheries and Aquatic Science, 60(12), 1462-1476.

Hughes, N. F. & Kelly, L. H. (1996). New techniques for 3-D video tracking of fish swimming movements in still or flowing water. Canadian Journal of Fisheries and Aquatic Science, 53(11), 2473-2483.

Here are some highlights. Notice the spawning Chinook on their redds in the top/middle picture, and the large woody debris piles in the pictures on the right, which seem to be especially important habitat for juvenile Chinook.

We’re using these photos right now, along with plentiful GIS data about fish abundance, to narrow down our potential study reaches for this summer.