Bluebreast Migration

Bruce are you seeing these shifts driven by seasons or other freerunning forces?

Basically driven by seasons. During low water, low flow in the summer much of the stream bed is covered by emergent plants. This begins to shift in October with autumnal rains and higher flows as the plants die back and are physically scoured out of much of the stream bed. Emergent plants come back strong in May. The total dissolved solids is always about the same, ~180 ppm, and pH drops from near 8 in warm weather with higher rates of photosynthesis taking up available CO2, down to about 7.4 in the colder months. Different macroinvert taxa have different life cycles, and from the four or five months of gut contents analysis we've done we see that the telescope & scarlet shiners are opportunistic rather than selective with feeding (not a surprise). The one surprising thing to me with shiner diet is that they seem to eat a large amount of benthic insects such as waterpennies and relatively large water beetles, rather than a mostly drifting diet like I originally assumed. I would've expected that of darters, whose guts we haven't examined (another research project!). Once we have a year's worth of data we could copy Todd and do some form of CCA for shifting availabilities. I used to have a Fortran version of NMDS, but I'd have to think (review, really) about if that would be appropriate. We'll see...

So is NMDS non-parametric multidimensional scaling?

So is NMDS non-parametric multidimensional scaling?

Or, non-metric multidimensional scaling. But it is a non-parametric technique using isotonic regression with ordinal data.

So is NMDS non-parametric multidimensional scaling?

The main reason I mentioned more reporting exact values and confidence intervals is I've been on a kick away from null hypothesis testing. As you said, replicates and larger sample sizes are often lacking, but I've seen in some of my own stuff where supposed strong differntial use of a resource was probably ecologically meaningless because the mean+-SE of the two samples were almost too precise. Like is a mean depth of 25 cm really different than a mean of 29? It is when sample sizes are large, but in the natural world probably not so much, especially when it is probably related back to discharge and availability. The big table of tukey comparisons is compelling but probably presents a stronger case in the end if it can be referenced to sample means and a measure of precision or error.

Since you're using near bottom velocity a measure of substrate roughness might provide some valuable information. Just conceptually, bottom velocities at the point of a Nothonotus observation are often influenced by cobble, boulder presence, in comparison to flabellare, which can be on the smooth as glass shallow gravel. Their water column and near bottom velocities are probably inverse because of depth, but especially roughness.

If you have the chance to read Analysis of Ecological Communities definately take advantage of it (rather than listing a bunch of pubs on using NMS/NMDS instead of PCA, CCA, DCA, etc.) As far as stats books go its actually really cheap! I like PCA and CCA as a resource/community screening tool because it can also provide visual aide through a plot (instead of a Pearson/Spearman matricie or pair wise comparisions). Since you mentioned potentially having measurements across multiple scales NMS or MRPP (multireponse permutation procedure) may be more appropriate than the old horses of factor rotations. All sounds good though!

We'd never have any time to post anything here, that's for sure.

Thanks for taking the time to type all of that up. I'm gonna buy the book today, I like where you're headed with your thoughts.

I've been really thinking about about that discharge thing. Depth, I think, is important, since flow is the same at that sample point as it is in a shallower sample point... I'm finding that they just don't care for anything faster than 0.6 m/sec except in cases of really large variatum which will tolerate up to 0.8 m/sec. And then maculatum was found in "slow" flows. But this is just because of compensation within Area = width depth - and I need a way to think about in terms of volume. Now, how to easily pick that all apart? Maybe I compress depth and flow. That seems simple enough.

Todd

Hi all, pulled a couple things apart, some I think will be of interest.

We did another CCA by individuals that were distinctly male (M), female (F) or juvenile (J).

 darter_so.jpg   83.22KB   7 downloads

We also ran a trial of NMDS to classify our sites and see how the species fell among them. It helped us to visually determine that Raccoon Creek was an outlier (it visibly was the least like the others, with tons of Ohio Black Shale in the bedding). I need to understand this procedure better and get a handle on distance, and what it means.

 all.jpg   53.89KB   9 downloads

And thus got a look again without the influence of nigrum.

 allnoraccoon.jpg   52.7KB   8 downloads

We also tried this by males only, but I can only tell you that it gave us a similar pattern with indistinct "shifts". That's a nice technique tho Matt. I did some reading, and that book came in yesterday.

Another book that I've found useful is this one: http://home.centuryt...t/~mjm/book.htm

Looking forward to a morning of Mantel tests and Sorensen tomorrow

Todd

The CCA shows ontogenetic shifts for most of the species. Is Fflab not on the graph, or didn't I find it? And I think you're on to a major usage of NMDS, purely to sort out groups in a very relative way. The best way to think of the array of groups within NMDS is like a large mobile by the artist Alexander Calder:
 calder.jpeg   33.13KB   2 downloads
A point within an area of the matrix generated by NMDS has a rotational relationship to neighboring points like one of the parts in a Calderesque mobile. I don't know if that makes immediate sense, but that's what the "dimensional scaling" of NMDS establishes. In a word, it's not a rigid structure but parts move both independently and relative to each other.

Yup that is the (same) book that I was talking about. It is great, easy read and you can't beat the price. The software is pretty easy to use too. I do what I can; you can mention to Jon that was my idea...

With the exception of variatum, I do see the ontogenetic pattern Bruce mentioned. I think that notion of juvenilles using a lesser extreme of available habitat compared to their adult counterparts is pretty well supported. I found it to a lesser extent multivariately with tanasi, but it was undoubtedly clouded because of my huge sample size and their confound behavior to school. Univariately it was very strong, just when you throw it all together in the real world configuration that is habitat variables in a 3-d interconnected world it tends to muddle. The species locations are means correct? So a 95% confidence interval about those means could be interesting...if they don't overlap, you've essentially done the pairwise comparison within multivariate space (see example in Skyfield and Grossman 2008). Interesting to see female variatums out on an island of the fastest velocity...

0.8 m/sec velocity at near bottom is ripping! I know I pretty much stopped observing any darters around 0.5 m/sec and if it was any faster with moderate depth it probably ment it wasn't very effective snorkeling either. God I wish I would have taken measurements on other species...

Bruce, Fflab wasn't in there because there's no good way to tell juveniles from females with what I've done on this so far, and really the only in that group that will look different are super males. I have their lengths, so I may try and analyze them with length in mind. That could be another interesting analysis. One of these days, I'd like to do this in the NW AL streams with some of the ridiculously "endemized" among Catonotus.

Matt, don't worry, I already told him, even before you brought it up

Yeah, those 0.80 m/sec are super fast, wide, shallow gravel runs in the complex. I guess that's what habitat you get to hang out in when you have sails for pectoral fins. And yes, I'm going to look at the variegate species cloud to see how much overlap there is. It's just as interesting to me that there's no apparent ontogenic shift as that crazy line you see for spotted darter!

The other thing I find interesting is what seems to be a higher affinity for juvenile cosmopolitan species for sand.

Another thing I thought of when watching my stream table last night was compressing depth and flow into a single "volume" factor. That might "linearize" the species a little more (accounting even more for the differences between variegate and spotted habitat, for example).

Again, thanks so much for your time looking at this. I really appreciate your thoughts and comments.

Todd

Todd, your research and data set have me looking at our primary research site in a new way. The riffle/run system we're working at in Estill Fork of the Paint Rock River has the same basic microhabitat partitions you describe, and I realize there are ways we could more precisely work within that framework. We observe surface flow rates of about 1m/3seconds, which is slow compared to your peak benthic rates. But it has me thinking...

Well, that's certainly one of the nicest compliments anyone can give

How are you working out flow? By giving a 1m / sec reading, I'm going to guess you're using the "egg of science" neutrally buoyant object? I think that's what's been most key to this puzzle was having the Marsh McBirney, but they cost a ton. One of the other high school teachers in my fellowship program just bought a much less expensive electromagnetic flow meter, I can look into that and what he thought of it, if you're interested.

Todd

We're measuring flow in a very 19th century kind of way, we drop a ping pong ball onto the current and time it moving alongside a meter stick in the area where we put our driftnet. It's really not precise and is probably a low estimate, but we're primarily interested in calculating about how much water flows through the driftnet in the 80 minutes or so that we let it collect drift. (I've always been inspired by the use of oranges floating downstream as a way to measure current flow...)

I always heard lemons were better than oranges but oranges were more readily available...seriously. A pygmy gurley type or some variation that you can get in Bed Meadows or Forestry Suppliers would measure as effective as the neutrally buoyant object and not be as expensive as an electronic meter like a Marsh McBirney.

Matt, you got me looking at flow meters at the various catalog sites, and I hit the wall of sticker shock. Something cheap is $750... the ping pong ball or maybe a lemon look better than ever.

Yeah the "cheap one" that the teacher used was somewhere around $1100, and yeah, that starts screaming finding someone with deeper pockets. Do you have internal grants for undergrad research? That'd be the first place I'd look. They really are indispensable, and you're just throwing away money if it's something less than $1000. I mean, c'mmon, we do this all the time for "lab grade probe" random number generators / boat anchors. At least you'll get a number you can trust!

Perhaps you can put a water quality spin on it (using these species as a "response variable") and get some funding help from EPA or a local watershed group? (or all of the above)

The object that we liked best when using your method was a plastic easter egg. You could fill it to the perfect level. Thus our "Egg of Science" theme

Todd

Edited by farmertodd, 22 December 2009 - 08:10 PM.

I'm in Alabama, the plastic Easter egg seems like a viable option too. But I admit I'd like to get $2700 for one of the slick electronic flow meters, and from there I can go on to $6000 for a Root backpack electroshocker. Pretty soon we'll be talking real money. On the other hand, if I could prove that all of this was for biotech development that will make everyone rich (as has been the promise for the last 30 years or so, and will be for the next 30 years too) money might flow like water.