This is going to be a fairly short reading group post compared to the previous few, as I'm rushing to get a bunch of things together before I head out of town for the holidays.

Anyway, the paper we read last week is a new one in GEB by Pagel and Schurr, entitled "Forecasting species ranges by statistical estimation of ecological niches and spatial population dynamics".  In the paper, the authors develop a Markov chain Monte Carlo approach to estimating a hierarchical Bayesian model from distributional data.  The modeling process is very interesting, as it involves the combined estimation of a niche model, a population dynamics model (including dispersal), and a model of observer behavior (the "virtual ecologist").  The authors demonstrate the value of their approach using simulated data sets of different levels of detail and quality.

The paper is fairly dense, but in my opinion the method demonstrates great potential.  It appears to deal fairly well with a lot of issues that are generally problematic for ENM/SDM methods, such as dispersal, species whose distributions are out of equilibrium with the environment, and stochastic population processes on a local scale.  The simulations show the dynamic range model beating the pants off of a more traditional SDM method in most cases.

However, it's not all sunshine and roses.  As far as I can ascertain, there's no readily available package or R code that makes the analysis available to the general public - you'll either have to program it yourself or bash something together in BUGS.  It also seems to be fairly computationally intensive, although for those of us with cluster access that's not a big concern.  More important than either of these issues, though, is that the method doesn't seem to work (or at least was not demonstrated to work) with presence-only data.  That's a shame, as that is by far the most plentiful type of data we have available to us.  

In summary, the method is very promising but it may be a while before there's an application that is ready for most end-users.  I'm all for the idea of a Bayesian MCMC approach to estimating ENMs/SDMs, as I think the level of information about uncertainty that these methods can produce would be a great step forward for the field in general.  I'm excited to see where this research program goes.

Luckily, these biologically iced sugar cookies were not cookies of death. I baked these for my preliminary examinations in the Ecology, Evolution, and Behavior PhD program at the University of Texas. Of course, had I not passed...they might have been. I kid, I kid!

Like all cookies, these cookies underwent a baking metamorphosis. These cookies were once a much more delicious product: dough. Eating raw dough is my favorite part of cookie consumption! However, raw dough also poses one of the greatest risks to one's health when it comes to the entire process of cookie production/gourmandization besides exploded stomachs. Raw dough has raw eggs and these raw eggs sometimes have parasites that can kill you.

I'm a bit of a germaphobe but I like to tell myself that probabilistically there isn't that great of a chance of me contracting a horrible disease from dough because a) there is a low prevalence of germy eggs and b) my immune system could eradicate the pathogens. This post aims to investigate the first of these two justifications.

If you’re like me, you’ve no doubt been warned to not eat raw dough but then ignored the advice. There are several different salmonella species and they infect various animals such as cattle, chickens, amphibians and occasionally people. The species, subspecies and serotype we are interested in is Salmonella enterica enterica enteritidis or S. enteritidis (a discussion of the complication and controversies surrounding taxonomic classification of pathogens will undoubtedly ensue at a later date).  Before the 1970s most cases came from feces on the egg shell surface but now egg shells are sanitized. In the 1980s a new strain appeared that actually infects the ovaries of the hens and is passed down vertically within the egg itself. This is the strain we are worried about.

Most of the publications I found on the topic of S. enteritidis prevalence were from the 1990s. In the northeastern region of the United States, 45% of hen houses were contaminated with S. enteritidis while central and southeastern regions were lower, 17% and 3% contamination respectively.  This information made me feel a little better about my raw egg eating ways since the south has a lower prevalence rate but I was still concerned. The sampling was done on the ceca and not the ovaries. Would this detection method be sensitive enough to detect infection inside the eggs? I also found out that salmonella of some type was detected in 80% of egg-laying houses.  So then of course, I wondered if competitive exclusion took place between the different salmonella species and if infection with “good” salmonella prevented infection with “bad” types. (Ebel et al. 1992)

Hogue et al. also found the northern region of the US to have the highest prevalence of S. enteritidis in unpasteurized egg product and slaughtered hens, 40%. Other regions ranged between 10-12% (Hogue et al. 1997).  Another paper, however, only detected S. enteritidis in 1 out of 42 flocks from the southeast and Pennsylvania when sampling directly from ovaries, a 2.4% prevalence rate (Barnhart et al. 1991).  Aha, I thought, here is my vindication finally!   

And because this is a blog post and not a research paper, I’m going to wrap up the post now. Basically, it seems like if you live in the northern region of the United States, don’t eat raw eggs.  Nearly half the flocks/hen houses are contaminated with pathogenic salmonella. Other areas face a risk but depending on the sampling method, the prevalence will vary.  The real question is what prevalence rate are YOU willing to risk eating cookie dough for? Is 17% too high? Is 2.4% too high? And remember, these studies are dated, what is the actual prevalence now? I would also say that if you are immunocompromised or a young child you should definitely not eat raw dough. I don’t have an answer yet as to whether this information will change my behavior, most likely it will not. Maybe if I move to the east coast it will.

So as the holiday season reaches its apex, remember that not only must you beware of cookie dough but also beware of homemade egg nog, sunny side up eggs, bean sprouts, and frogs. Nothing says Christmas like a holiday tree frog. And apparently, even if your eggs are pasteurized and you buy dough from the store, you may still get sick…from E. coli contaminated flour! (http://well.blogs.nytimes.com/2011/12/12/beware-of-raw-cookie-dough/) 

Here is a great link from the CDC on how to reduce your exposure to salmonella: http://www.cdc.gov/Features/VitalSigns/FoodSafety/ 

Barnhart, H.M., Dreesen, D.W., Bastien, R. & Pancorbo, O.C. (1991). Prevalence of Salmonella enteritidis and other serovars in ovaries of layer hens at time of slaughter. Journal of food protection, 54, 488-491.

Ebel, E.D., David, M.J. & Mason, J. (1992). Occurrence of Salmonella enteritidis in the U.S. Commercial Egg Industry: Report on a National Spent Hen Survey. Avian Diseases, 36, 646-654.

Hogue, A.T., Ebel, E.D., Thomas, L.A., Schlosser, W., Bufano, N. & Ferris, K. (1997). Surveys of Salmonella enteritidis in unpasteurized liquid egg and spent hens at slaughter. Journal of food protection, 60, 1194-1200.
   

Author

Stavana Strutz is a doctoral candidate studying disease ecology in the Parmesan lab at UT Austin.

Okay, I'll admit it.  I forgot to do a Fish Friday today until Stavana reminded me at the UT Integrative Biology holiday party.  Here is a last minute "contribution", in the weakest sense of the word: a Scientific American blog post about a series of photographs of people posing with fish who resemble them.  Dumb as that sounds, the pictures are actually pretty awesome.

It begs the question: what kind of fish are you?  I think I'm probably a garden eel - I don't get out much, but when I do I'm adorable.

Welcome to Fish Fridays, our weekly value added segment about all things fishy. Today I wanted to highlight Moray eels and a really cool morphological adaptation that this critter has. It has to do with their jaws!

Vertebrate jaws are of course a product of evolution, and all vertebrates, even humans, have at some point in development serially repeated array of pharyngeal arches. We retain only the most anterior arch to form our jaw.  Fish have this jaw as well, but they retain the other arches in their hardened form as structures that support the gills, sometimes supporting additional teeth as well.  But imagine if the most anterior arch wasn't the only that developed into a fully functioning jaw? Then you would have a series of jaws! This is only the beginning of why eels jaws are unique. 

Sometime around 2008, Rita Mehta from UC Davis found that eels feed like no other fish.  Like many fish species, they have structures inside their throat called pharyngeal teeth that aid in processing food, but Moray's pharyngeal teeth have evolved to such an extent that they function as a second set of jaws, with the ability to extend forward in the mouth and grab food caught by the primary jaws. This of course is reminiscent of the creature created by Giger that was used in the movie Aliens, and thus the Moray eel has gotten even more attention as an ominous predator.  

Apparently the closest thing to this elaborate set of jaws exists in some snakes.  See the video below with the UC Davis researcher who discovered this.

Ben Labay is a "fish-geek" and research associate for the Texas Natural History Collections at UT Austin

See his fish art at: www.inkedanimal.com

Many of our group are self-described fish geeks and use this taxa group in our modeling research. So we've decided to force our love of fishes onto the readers of this blog with a weekly value-added segment known as Fish Fridays!  Fish are a fantastically interesting and large group that include many classes of organisms, and it was pointed out to me by Dan just last week that for fishes to be a monophyletic group, humans would be fishes as well! So we should have plenty to talk about. This segment is intended to provide brief and fun natural history stories about all things fishy, and so it might not always deal with species in space themes.

For this first Fish Friday, I've decided to geek-out on a relationship between freshwater fish and mussels that I've been thinking about a lot lately. It's a bit of a focus on mussels instead of fish, but it's just such a cool relationship and story of fish mimicry that I couldn't resist. 

As we study freshwater fishes, we have to constantly adapt research approaches to incorporate the distributional constraints that a river network places on these organisms, and with this in mind, we might as well be working with freshwater mussels as their ecology and distributions are fundamentally tied to fishes. This stems from the reproductive ecology of freshwater mussels.

In freshwater, female mussels produce eggs that develop into a larval stage called a glochidium, which temporarily parasitize fish, attaching themselves to the fish's fins or gills. In some species, release occurs when a fish attempts to attack the mussel's minnow or other mantle flaps shaped like prey; an cool example of mimicry. 

Mussel larva, glochidia, are generally known to be species-specific and will attach to any fish gill, but only live if they find the correct fish host. Though from what I gather there is still a ton not known about specific mussel-host relationships.  The glochidia, once attached to the correct host's gills, will live there for a number of weeks before breaking free and dropping to begin an independent and sessil life. This reproductive ecology of mussels tie them directly to the distribution patterns of their host fish, and thus research in distribution of fishes lends well to studying mussels.  I leave you with a video that explains this all using largemouth bass and the genus of Lampsilis mussels.  I think the one featured in the video is a pocketbook mussel, thought I'm not sure.

Ben Labay is a "fish-geek" and research associate for the Texas Natural History Collections at UT Austin

See his fish art at: www.inkedanimal.com

We have a weekly ENM/SDM reading group here at UT, and one of our goals with this blog is to start summarizing our interpretation of the papers we read, and hopefully spark discussion on these papers with the broader ENM/SDM community.  To that end, I'm going to discuss my interpretation of a  paper by Meynard and Kaplan, entitled "The effect of a gradual response to the environment on species distribution modeling performance", Ecography 34:001-011.

We found bits of the paper a little confusing, but I'm fairly sure I understand the overall point.  It's an interesting and useful point to make, so I'm going to try to outline it here.  In the event that I've misinterpreted the paper (a distinct possibility), I'd love to be corrected.  Here is the argument being presented as I understand it.

First, they develop a simulation approach to generate data that treats the probability of sampling from a given environment as a logistic function of some environmental variable, Pi = 1/(1+e^-Yi), where Pi is the probability of sampling in grid cell i and Yi is a function of the environmental gradient.  Yi is calculated as (xi - B)/a, where xi is the value of the variable in cell i and B and a are parameters that determine the inflection point and slope of the species' response to the gradient, respectively (these are beta and alpha in the original paper, but if Weebly has those symbols I don't know how to access them).  So the whole shmear looks like this:

When faced with something like this, I often like to do simple little Excel experiments just to visualize how changes in parameters affect outputs.  I generated a fake environmental gradient and added a bit of randomness to it:

And then I simulated a couple of species.  The first had beta = 1 and alpha = 2:

The second had the same beta, but a much more gradual slope in the species' response to the environmental gradient (alpha = 20):

Here, as I understand it, is their point.  The maximum possible performance of a model will be dependent on the slope of the species' response to the environment.  If the slope is very steep, a perfect model will do a more or less perfect job of distinguishing presences from absences.  For instance, here's one with beta = 3 and alpha = 0.01:

In this sort of system, there's little uncertainty in the expectation of presence or absence across the landscape: almost every grid cell has a near-certainty that you will find the species there, or that you won't.  Contrast that with the previous example (beta = 1, alpha = 5), in which there is a decent probability of finding the species in any grid cell (the minimum probability on that landscape is about as likely as getting heads in a single coin flip), but there's nowhere on the landscape where you're more or less guaranteed to find your species (an absence from the highest probability cell is still about as likely has getting heads three times in a row).

So what's the relevance to model evaluation?  Well, it comes down to the expected maximum performance.  In a situation with a threshold-like response to the environment (e.g., the beta=3, alpha = .01 scenario), it is entirely possible for a decent model to do a great job of distinguishing presence from absence (or pseudoabsence) cells.  In a species that responds more gradually to the environmental predictor (e.g., beta=1, alpha=5), even the TRUE model does a mediocre job of telling you where you should and shouldn't expect to find your species.  This sets a limit on the maximum model performance we can expect, and when alpha is high that maximum performance may be significantly below the theoretical maximum performance of a model for that statistic (AUC, Kappa, sensitivity and specificity).  What it boils down to is, in the authors' words, "the same model that produces a poor prediction in terms of presences and absences may be recovering perfectly well the true probability of occurrence throughout the environment".

It has been remarked before that it is generally going to be more difficult to model habitat generalists than specialists, as their relationship to any one environmental variable will usually be more ambiguous than that of a specialist.  This is essentially putting that argument in the context of model evaluation, which I think is useful - rather than saying "we can't build good models of habitat generalist species", it's saying "we may be able to build good models of habitat generalists, but we can't expect them to perform as well on presence/absence data".


There's actually more to the paper than that, as they tie all of this to issues of species and sample prevalence.  I've got to get back to work right now, but maybe we'll revisit that side of it in a later post.