No Harm to Fish?

The rather dismissive attitude that many engineers and developers have towards fish has generated concerns of fishermen and environmental lobby groups like the ecology action centre. Why the difference in attitudes?

To the credit of the engineers, an effort has been made to see whether or not fish get sliced and diced as they go through an OpenHydro turbine. A video camera was mounted to the side of the turbine so that it looked across the entire face of the turbine. The word around engineering circles is that no one has ever seen a fish go through the turbine.

To give the video footage some scientific credibility, it was handed over to some biologists who published a scientific paper that seemed to show fish hanging around the turbine when currents speed was low (turbine not dangerous) but fish not being observed at all as soon as current speed picked up enough to raise concern for fish welfare.
* Number of fish observed in front of the turbine for different current speeds.
The above plot was obtained by examining 5 video frames that were randomly chosen from the first 2 minutes of observations at the start of every hour. They counted the number of identifiable fish in each frame and then calculated the geometric average over the 5 frames. Thus, every hour we get a single snapshot of how many fish are infront of the turbine. We can do this for every hour of the day because this turbine was installed way up north, where the sun never sets during summer...

A harried, underpaid scientist might quickly glance at the above plot and write, as they did in the State of the Science Report that

"The authors did not observe any collision or strike events"
and engineers would pat themselves on the back and never give the welfare of fish another thought.

The thinking man understands that the above plot really has little to say about whether fish do or do not go through the turbine, let alone whether or not sashimi comes out the other side.

Others might wonder, but what happens during the long winter nights? Or what happens at other places where they have different kinds of fish? Even puting such reasonable doubts aside, there is a far more fundamental problem with drawing conclusions from the above study.

In science, we seek understanding by relating observations to a model. Let's make a model. It's a very simple model that is based upon some ideas that are supported by other observations.

First we ask what happens when the current is fast? Every hour we measure the number of fish in front of the turbine. The volume in which we see fish is the area of the turbine multiplied by the upstream distance that the camera sees. Given that the diameter is 6 m we could guess that the camera sees an average of 4 m upstream so the number of fish that we expect to see in one frame is
Nfast = 24 F /1000000
Taking a biggish value like F=5 fish per square kilometer means that we expect to see Nfast =0.00012 fish in each frame. But each experiment only went for 15 days and there were only 24 views made each day. So we look at only 360 snapshots and only some of those were when current was strong. Adding abundance of fish seen in all snapshots when the current is strong, we expect to see something like 0.03 fish. Little wonder that they didn't see any fish passing through the turbine during fast currents...

Next we ask what happens when the current is slower. Take a typical scale for slow current speed of 0.7 m/s. Fish passing the turbine when the current is slow will hang around it. Let's say that it the current is slow for about 1.5 hours surronding each turn of the tide. Let's also say that the fish will see the turbine from about 10 m and swim towards it. Thus the area that moves with the tide past the turbine when currents are low is A=(2*10+6)*0.7*1.5*3600 square metres. The number of fish that we expect to see in a snapshot becomes
Nslow = A*F /1000000
which is approximately Nslow =0.5. That is a heck of a lot more than the paltry value obtained for Nfast. Given that that the fish distribution is usually clumpy, the results obtained in the above figure are more or less what would be expected. Most snapshots have no fish in them, regardless of current speed. Sometimes a school of fish swims/drift along and "finds" the turbine. Having found it, they hange about and get photographed. Thus, a few snapshots might have 10 fish in them but most have zero, giving an average of about 0.5.

OK, the numbers are rough guestimates. But I hope you can see the essential behavioural asymmetry that explains the results. And I hope you can see that the analysis was, by design, unlikely to measure any fish in front of the turbine when the current was fast.

The correct analysis would have been to look at all the data. Indeed, one should collect video data for a much longer period of time. Of course, it would be impractical for a human to look at each frame. A computational machine could be programmed to do the job...