Do you remember the scientific method from grade school? That discrete five(ish) step process that was engrained into your head as the way all great science is conducted. Make observations- develop hypotheses- form an experiment- collect data – generate conclusions.
Yea, it doesn’t always work that way.
Our recent manuscript on riverscape genetics (which you can download here) is a great example of just how messy science can be. I’d like to say that we started this project by announcing that we wanted to do a riverscape genetics study (I’ll explain that term in more detail below), or that we collected data with a very clear hypothesis in mind. But, those would both be lies. I started that project knowing nothing about basic genetic concepts, in a study system I had no intent to work in, and my only goal was to describe what I saw. Nonetheless, the result is a manuscript that I think (or, at least hope) makes a great contribution to science.
So, how did we get there? My PhD position came with no project funding or ideas, only a mutual interest between me and my advisor to study trout. Luckily, we quickly stumbled on a small grant opportunity, and we were pretty sure they would give us money if we 1) would work in the Loyalsock Creek watershed, and 2) study genetic diversity of brook trout. I was very happy to work on brook trout and being new to Pennsylvania I didn’t have any strong feelings about the location of research. But, genetics? I specifically avoided genetics courses in college and actively rolled my eyes at the research idea.
But, in the case of research, money does buy happiness and I needed research projects for my dissertation. So, the first project was set. We were going to describe how genetically related 20 populations of brook trout were throughout the watershed. No hypotheses, no experiments…just go out, collect data, and describe. And, honestly, I’m not sure how much difference the results would have made at the time. Brook trout populations often have low genetic diversity (which happens when fish don’t move between streams and populations become isolated from one another). But, how “bad” low genetic diversity is is debatable. Plus, solving it can be very difficult, not to mention expensive. Understanding population genetic diversity is good information to have, but rarely does it actually help change fish conservation and management.
It took about two weeks to sample fish from those 20 populations we initially agreed to study, but even less time to fall in love with Loyalsock Creek. So, after wrapping up the initial field work, I made plans to do a side project on fish behavior, and while doing it I decided to also collect genetic data at about 15 more sites. Those 15 extra sites were chosen with really little consideration for science, and the decision to collect genetic data at them was a leftover thought.
So, the summer flew past and I had samples from fish at 35 sites. “Samples” in this case refers to little clips from the caudal fin, which we can use to determine the genetic composition of each fish in that lab. That lab work takes a while, and I was about to leave to start research in West Virginia. But, before I left, I needed to provide a short update of our research and provide some preliminary results to the person funding the project. So, I determined the genetic composition of fish from a few sites and ran some basic analyses and found some surprising results. The trout had higher than expected genetic diversity, and didn’t seem to have clear patterns of isolation like we would expect if fish were not moving between streams. But, despite being interesting and unexpected, I was getting ready to start some new projects and didn’t have time to really think more about it.
At this point, I basically stopped working on the genetics project. I did about two more years of data collection in various other states and countries, eventually getting back to Loyalsock Creek to do a fish movement study. If you haven’t read about our telemetry study before, we tracked fish movement for eight months and found that basically no fish moved all summer. But, we found that about 20% of tagged fish have this interesting behavior where they seem to spawn in the small tributaries and then move into mainstem Loyalsock Creek. If you have never been to this area, don’t let the name “creek” fool you. Loyalsock Creek is actually a moderately sized river, and very atypical brook trout habitat.
Now things are getting interesting. My preliminary genetics data suggested that populations might be connected, and now I have telemetry data showing that some fish move to mainstem Loyalsock Creek. But, again, Loyalsock Creek is far from normal trout habitat, and it’s really not that many fish that seem to move. So, there’s two possible fates for fish that move to Loyalsock Creek. They could 1) just die, or 2) use Loyalsock Creek as a movement corridor to make effective migrations. Effective migrations refer to a very specific type of movement where an individual leaves one spot, moves to another, AND spawns there. We need those effective migrations to actually have populations connectivity and increased genetic diversity, otherwise movement does very little to population genetics.
So, how do we determine which of the two fates might be true? We conduct a riverscape genetics study. You can think of a riverscape as the aquatic version of a landscape- it’s comprised of all the habitat features that a fish might encounter as it moves throughout a watershed. In a riverscape genetics study, we determine which of those habitat features seems to be particularly important for increasing or decreasing genetic diversity. And, because genetic diversity is related to fish movement, it basically just tells us which habitats increase or decrease fish movement.
But, it turns out, there were no existing statistical methods that are specific to riverscape genetics, meaning we couldn’t really determine the effect of those habitat features on movement. Riverscape genetics studies have been conducted before, but they always borrowed methods from landscape genetics studies. Now, it’s quite obvious that water and land are very different thing. But, statistical methods in landscape genetics can’t fully capture the unique properties of rivers that influence fish movement.
To visualize this, think about a deer moving across land. That deer is basically free to move in any direction it wants. It probably moves most in forests, and might avoid roads (though, the amount of roadkill argues otherwise), but it technically can move in any direction. And, there’s nothing forcing it to move one more or less than the way or the other.
Now, think about a fish. Unlike that deer, it can only move where there is water. So, it’s forced to use the river network which, compared to land, is a very linear environment. The fish also has to fight against the flow of water, and so movement might be biased in either the upstream or downstream directions. So, to conduct a riverscape genetics study, we needed to use a statistical method that can account for the linear network and can handle the possible bias in movement in one direction (which we call bidirectional gene flow).
And, this is where we got very lucky. A member of my PhD committee was a statistician, who had been thinking of these exact ideas for a few years. So, we had data. He knew how to create the statistical model that would take care of the problems above. And, together, we created the bidirectional geneflow in riverscapes (BGR) model to account for the problems above. The BGR model is designed to be used in any riverine environment, and we tested it first on the data that we collected in Loyalsock help understand how fish were moving throughout the watershed. We used the model to see if a whole suite of habitat variables were important – anything from elevation to stream crossings to stream size. Ultimately, we found that four habitat variables seemed to influence fish movement Loyalsock most:
A useful part of this model is that, once we know which habitat features influence movement, we can create maps of relatively migration rates (relative here doesn’t mean number of fish, just that higher numbers mean higher migration). These maps can help highlight movement corridors throughout the watershed (like the mainstem), but also easily identify areas in the watershed where management and conservation might improve connectivity.
So, what does all of this mean? That “atypical” brook trout habitat might actually be some of the most important habitat there is for population connectivity. Mainstem Loyalsock is only inhabitable by brook trout for about half of the year, but during that time it is being used as a significant movement corridor to increase population connectivity. Anything that threatens habitat in the mainstem- be it climate change, reduced stream flow, water withdrawals, sedimentation, deforestation, barriers, etc- even at very small, local scales, could have very significant effects of brook trout populations across the entire watershed. It also means that conservation of brook trout habitat means more than just habitat enhancements to small streams.
More globally, if you’re conducting riverscape genetics studies of any species, check out our BGR model. It’s easy to run (even if the manuscript doesn’t make it sound like it is) and provides more informative results than other methods that you may be thinking about. The results we provide here are likely to change depending on where you are working and the behavior of the fish you are studying.
And, what I find funny about all of this is that we probably wouldn’t have discovered these patterns had I not randomly gone out to study 15 extra sites that first summer of data collection. It turned out that some of those sites had the highest genetic connectivity, which is what initially tipped us off that fish might be using the mainstem as a movement corridor. So, if I had to revise the scientific method for this project, it would look more like collect data- forget about results for three years- get lucky- communicate findings. And, honestly, I think that’s probably the more likely sequence of events for most scientific professionals.
After determining which habitat features influence trout movement, we can create these maps of relative migration rates in the downstream (top) and upstream (bottom) directions. Zooming in, you can see the effects of barriers (circles) and temporarily dry stream segments (squares) on migration rates. These maps can be used to easily visualize where conservation efforts could be focused to increase movement.