Biological Posteriors

Classics in geographic ecology

2012-01-30T07:24:00.000-08:00

The conceptual foundations of ecology are unique because ecology is a relatively young field. Many of the early ideas in ecology were put on paper less than 100 years ago, which means that some important, foundational works are easily accessible.

An important idea in ecology is that the distribution of species are constrained by external factors. Why would anyone be interested in this topic? In 1894, C. Hart Merriam explained:

We will need to go back to von Humboldt's work to trace these ideas further (I don't have a copy of his Essai sur la géographie des plantes yet... anyone have a copy?). But already it is apparent that researchers recognized the importance of climate in shaping the geography of species.

In 1917, Joseph Grinnell began critically examining this question with specific reference to the California Thrasher (Toxostoma redivivum). He stated:

With this work, we saw the first use of the term "niche" in an ecological context. Additionally, Grinnell's approach to understand "these critical conditions ... through an examination of the bird's habitat" set the stage for most of the current work on species distribution modeling.

Grinnell expanded his ideas in 1924 by thinking about the role of environmental constraints in speciation and extinction:

Of course, Grinnell's explanations for the data that he collected were not entirely environmental; he recognized the potential importance of "faunal conditions" (i.e., influences from other organisms) in his 1917 work. As he explained in 1924:

These works formed a foundation for some classical theory that emerged in the 1970s and the ideas outlined in them continue to be used. We still have very few cases where we can definitively state "how [species] are checked in their efforts to overrun the earth." Fortunately, these classic works are a great resource for thinking about what we still need to know.

The pulse of a tree

2012-01-23T12:23:00.000-08:00

An important component of the scientific method is measuring observations. These measurements form the basis of how we learn about the world. For example, when I go to the doctor I usually have a series of basic measurements taken that give the doctor an indication of my general health.

My pulse provides an estimate of the health of my heart. If the doctors were concerned about a myocardial infarction, they might do a more detailed measurement than a simple pulse reading, like an ECG to monitor the status of my vascular system. If some doctors were interested in how my heart responds to the stress of strenuous exercise, they might pop me on a treadmill and measure what happens to my heart rate. This type of physiological monitoring is standard procedure for learning about the health of humans.

What if we wanted to monitor the health of trees to understand how they respond to stresses? Of course, trees are very different than people (they don't have blood, among other things), but in some ways they are very similar. Just like humans, plants have an intricate vascular system connecting all of their organs. This system supplies water to the plant organs to maintain cellular processes, so maintaining water flow is essential.

Fig 1. Monitoring water flow with sap flux probes, note similarity to an ECG setup.

We can monitor the rate of water flow through the vascular system of a tree with sap flow probes, which produce data that look like ECG results (see below). There are other methods (like taking point 'pulse' measurements) but sap flow probes allow us to monitor water flow over long periods of time to observe how plants respond to different conditions.

Fig 2. Water flow through trees over one season - differences among trees of the same species from two sites.

These measurements allow us to monitor the health of trees. Because of the importance of maintaining water flow for tree health and survival, we are concerned about situations that cause flow to decrease. In this example, the upland site is significantly drier than the lowland site (by about 55%, t₃₅₃₂₆ = 189, P < 0.0001) and we know that dry soil can cause the xylem to cavitate, which is essentially like blocking the vascular system (McDowell et al. 2008). If water flow decreases enough, the plant may experience the equivalent of a heart attack.

Around the end of September (approx. day 270), the conditions at these sites got really dry. But, it looks like at least one of the trees from the upland site (which is drier overall) may have maintained sap flow better than the lowland site trees. Fortunately, the drought was not enough to cause mortality and it rained, which allowed the trees to resume normal function.

Fig 3. Average water flow throughout the day.

The ability to tolerate the drought comes at a cost for the upland trees, which had much lower rates of sap flow on average. It also took the upland trees longer to 'get going' in the morning and sometimes they needed to keep moving water after sunset to keep their internal water levels sufficiently high...

As you can guess from what I've said above, these measurements offer a nice overview of the health status of trees. The questions that we could ask with this information are quite broad. For example, this data could help us understand what types of conditions different trees can tolerate (with stress tests) or what risk factors may lead to mortality, which are questions that remain unanswered for many species and conditions. So now we just need to come up with some questions, hook up some trees to the probes, and test them!

Some thoughts on mechanisms

2012-01-17T12:17:00.000-08:00

"An appraisal of plant distribution requires information on ecophysiology."
- Lambers, Chapin, and Pons (2008)

"[O]ne man's mechanism is another man's black box."
- Suppes (1970, cf Jon Elster 1983)

How far down the mechanistic rabbit hole should one go to get an answer to any question?

I am interested in what limits the distribution of plants, which could be examined from multiple different scales. Ultimately, the presence of a plant species in a location depends on the maintenance of a population, which means that births must be greater than deaths. But, mechanistically, we need to know a little bit about why plants are born or die.

As Lambers, Chapin, and Pons point out, plants are born and die for physiological reasons. For example, a plant will not reproduce if it does not have enough carbon resources to allocate to reproductive tissues. As conditions become more extreme, a plant will die if the whole plant cannot maintain a positive carbon balance or water flow. But, why do plants fail to maintain positive carbon balance? Certainly there are genetic mechanisms that control when plants begin to shut their stomata. In that sense, maybe I should measure DNA sequences instead of carbon stores to understand what limits population growth. At lower (or higher, depending on where you stand) levels, physiology is controlled by physical laws (I'm looking at the scaling people here...).

So the reductionist program proceeds. And so Peter Turchin (1999) stated that "[i]n practice, however, it is not always possible or desirable to follow such a reductionist program to the logical extreme, and we find that some level of abstraction can be quite productive." Science requires mechanistic explanations that come from lower levels of organization, but how far to go seems pretty subjective.

Pragmatically, I am inclined to agree with Turchin and say that I'll only go deeper if it is completely necessary (for example, if competing hypotheses with different outcomes demand a more detailed explanation). But, like Lambers, Chapin, and Pons, I suspect that plant population regulation is tightly related to the physiology of individuals. So why not measure those responses so that they can inform my view of the world?

I suppose at some point it just depends on what is feasible. I'd never finish my PhD if I wanted to do everything all the way down, no matter how interesting and informative it may be. But, I have a hunch that at least some physiological measurements could really improve my story.

How do you draw the line of abstraction between mechanism and black box?

Ecology directions

2012-01-11T05:59:00.000-08:00

After a LOOOOONG holiday break, I'm still catching up on emails and journal articles. In the coming weeks I will start to post more frequently as I start to get back on track. But, in the mean time, I wanted to share this great work from the EEB & Flow blog. They've created a word cloud of ecological journal articles from 2011. I think that it highlights some important directions for the field. I am not surprised that climate change is less prevalent than it was in their 2010 assessment, nor am I surprised that community ecology and range limit topics have surged in interest.

It is wonderful what we can learn about ourselves with some bibilometric research!

Keeping it real

2011-11-19T09:25:00.001-08:00

Against my intentions for this semester, I have neglected the blog for the past few weeks. But, I have two unrelated thoughts that I wanted to post today, which may kick-start me into posting more often.

Part A. Keeping it real... in the bigger scheme of things

When I start digging into research I usually find myself deep in the hole and realize that I haven't been looking around at where I am going. Of course, in research, we don't know exactly where we are supposed to be going. But I have difficulty with just going along and letting the science take me where it will. I think I'm afraid that I'll end up in some dark ivory tower that isn't relevant for anyone.

To help myself get over this concern, I've been trying to continually reidentify my position on the map and see if I am near anything good. Radiolab and Science Friday, the blogosphere, and major news outlets (like the NY Times) have been important sources for reality checks, but that is still limited to a science-y crowd. I still need to find out how to tie my rope to ideas that someone that I meet on an airplane would find interesting.

How do you keep it real? Do you worry about this like me?

Part B. Keeping it real... in the blogosphere

I came across a great new article in PLoS Biology that discusses the opportunity of merging teaching and research. I have had an interest in this for a long time, since I took a class that was designed like this in my undergraduate. I highly recommend the article, but I'm not going to talk about it here, because I want to focus on what their thoughts mean for the blog.

This summer I came to the realization that the blog can be useful for me to reflect on research. I think that it can be an efficient vehicle for formulating my ideas and sharing them with others.

This made me think about how I can use the blog to communicate information and to do research (just like Kloser and others want to do for education). So, I've got a series of posts planned that will be framed around research topics. Maybe they'll end up like bits and pieces of literature review that I can use for my dissertation. And then maybe some day you'll get to see snippets of data that relate to these blog posts! What an exciting idea.

Onward with the blog as a research and communication tool....

Visions for ecological research

2011-10-01T12:51:00.000-07:00

Society, culture, and technology influence the research questions and subjects that scientists study. The values that underlay research fit within the discipline but certainly vary among individuals. For ecologists, this divide is often over the role of ecologists in policy. The frequently-invoked division between "applied" and "theoretical" research is grounded in differing views of what problems are important and how research creates change. The need to identify these differing "visions" is the basis for Mark Neff's (2011) cool analysis of how ecologists frame their science.

Neff identified four dominant visions for ecological research:

Document problems and communicate findings to compel change
Ecological theory that includes humans will guide policy, when communicated
Stick to the science; build theory, but don't interact directly with policy
Ecology should inform restoration and management

These visions demonstrate "different logical or value structures about what constitutes worthwhile research." They represent a spectrum of views on how research informs policy and what are important ecological problems.

The survey results are general categories and many people probably move back and forth between groupings depending on specific topics or issues. Thus, Neff notes that "we must therefore be cautious when trying to classify a person as being a 'type'." But I am going to throw caution to the wind and try to classify my own values. (Like a "what type of ecologist are you?" online quiz...)

I took Neff's survey (based on the questions that Neff supplied in the text) and drew correlations between my rankings and the rankings of the four main types. See my worksheet here and the text for survey details.

Those of you who read the blog often may not be surprised to find that my replies correlated most strongly with the 3rd vision for ecological research (r = 0.576, prob. of being greater than zero P < 0.01, t = 3.86, df = 30). At least I was not very surprised. My views frequently fall into a theory-building/basic-science framework. The specific problems that I perceived as being important depart from others who took the survey, too. An important point of Neff's analysis is that these interests vary among individuals based on personal perspectives, even though we usually think that these views come directly from our science.

As Neff notes, these data are not extensive enough to do a demographic analysis of the ecological community, which is what I would like to see next. Do any of these types dominate the current ecological scene? I think that Neff kind of alludes to the need to think about this in the discussion, by stating that we need to talk about what the consensus of research priorities are in the field. Additionally, and perhaps most importantly, we need outsiders to help us define what are these priorities. Without this input, we risk spinning our wheels on topics that are not socially relevant, which could marginalize the public perception of our field.

Neff, M. (2011). What research should be done and why? Four competing visions among ecologists Frontiers in Ecology and the Environment, 9 (8), 462-469 DOI: 10.1890/100035

Ecology in Science

2011-09-22T16:45:00.000-07:00

30% of the reports in this week's Science magazine are ecological (five out of 17 reports... okay, maybe six if you include the one talking about antibacterial resistance, bacterial niches, and microenvironments)!

One that I think will be important is a report from a slew of great ecologists (Adler et al. 2011) focused on the hypothesized relationship(s) between biodiversity and primary productivity. The key quote: "We found no clear relationship between productivity and fine-scale (meters^-2) richness within sites, within regions, or across the globe." Within sites, the relationship between biodiversity and productivity is just a giant cloud. Across sites, there are some more patterns (I was really excited to see a quantile regression drawn on the data!), especially when contrasting "natural" and "modified" environments. Congrats to the NutNet project for contributing some cool analyses!

Populations in the wild - stochasticity and determinism

2011-09-21T12:48:00.000-07:00

Population persistence and extinction are controlled by the population growth rate (r). If births are greater than deaths, a population will grow (ignoring meta-populations for now...).

r = births - deaths

But, populations in the wild are more complex than this deterministic generalization, which requires us to think about how to add stochasticity into the equation. This is easily done with some additive error

r = (births - deaths) + N(0,ε)

but masks the components (and all of the cool ecology!) that contribute to this variation.

Melbourne and Hastings (2008) outline these components simply and eloquently in the introduction to their article on extinction risk. They identify three sources of stochasticity:

Demographic stochasticity - the probabilistic nature of individual birth and death,
Demographic heterogeneity - (a) variation in the age/stage structure of a population and (b) variation among individuals within a population to common environmental variables, and
Environmental stochasticity - variation in time or space in population-level birth and/or death rates due to variable environmental conditions.

Melbourne and Hastings argue that incorrectly characterizing this uncertainty structure can lead to erroneous predictions of extinction risk. In line with prediction concerns, breaking apart this uncertainty is essential for understanding the dynamics of the system (Clark 2009).

So now we begin to envision ways to make a hierarchical structure (NOTE: this is very rough and could have errors in the way I am thinking about it). For demographic stochasticity, each birth and death has a probability

births ~ Pois(BirthRate)*N, and

deaths ~ Bin(N, 1 - SurvivalRate).

For demographic heterogeneity, birth and survival rates vary among individuals, and so the population parameters follow a distribution,

BirthRate_i ~ Gma(shape,rate), where the parameters are influenced by the hyper-parameter birth rate (β), and

SurvivalRate ~ Beta(shape1, shape2), again where the parameters are influenced by hyper-parameters for the population (maybe σ for survival).

Environmental stochasticity can be incorporated by adding variability in the hyper-parameters, where the population level parameters depend on environmental variables.

β = ƒ(Env, θ_b), and

σ = ƒ(Env, θ_s).

It seems common to exclude this last part, because it is hard to figure out how the Environment (capital E because it is the combination of environmental variables) impacts demography (its shape...). This is equivalent to identifying the niche. It is being done with the Soay sheep in Scotland (Coulson et al. 2001), but it still seems like a hard row to hoe (how the heck do you identify "Environment"? for relevant discussion: Hallett et al. 2004).

Learning how environmental and demographic stochasticity influence population growth rates depends on correctly structuring the hierarchical effects and correctly identifying the functional form of how each variable responds to the stochastic parameters. Demographic heterogeneity seems to increase extinction risk more than equal amounts of environmental stochasticity (Melbourne and Hastings 2008) and can promote coexistence among species by reducing interspecific competition (Clark 2010). If we think about temperature effects on survival, whether that response is concave or convex will influence whether environmental variation will increase or decrease survival in the population (Gravel et al. 2011). Drake (2005) showed that, when the temperature response of population growth is concave, variance can increase overall population growth rates. But, what if the response is not concave? If the response is hump-shaped, the effect of variability will depend on where on the curve we are looking.

All of this is emphasizing that we just don't know how these curves are supposed to look (a sentiment echoed by Charles Canham at his presentation at ESA this year), which leaves us unsure about the consequences of stochasticity in wild populations.

Lessons from battles between climate scientists and society

2011-09-20T18:48:00.000-07:00

There are great comments at the New York Times about the difficulties of convincing the public about the reality of human-caused climate change. Some of the points raised by the readers have implications for scientists of any field.

Mark Bessoudo from Toronto points out that climate change skeptics will not change their mind despite the continual stream of evidence produced by scientists. I agree with this assessment. There is plenty of evidence suggesting that humans are undeniably causing our climate to change. The disconnect is partially in the complexity of the system - the Earth's climate is really complex and it is hard to explain to people who do not have advanced experience in physics, chemistry, and biology. This is an issue raised by Vaughn Gilbert from Pennsylvania. It is compounded by the fact that few social leaders (and scientists, for that matter) are making efforts to sway public opinion (and some are pushing the other way).

But additional problems arise because we only have one Earth. Our sample size is always going to be one. This concern is more subtle than most laypeople would articulate, but I think that it is a bigger issue than the last two concerns that I mentioned (those issues could be resolved with effective communicators, social outreach, and public policy work). I think that most people have an intuitive skepticism about n = 1, and for good reason.

Theories about complex, large-scale systems are rarely amenable to experimentation. The "solution" to this approach is to conduct work at many scales, doing reductionist experiments to study the components, coming up with explanations that we can weave together into systems of equations that produce numbers that we can compare to our monitoring data, and comparing the intermediate parameters that the computer spits out to other observations and experiments. This observation-model-validation cycle is not only hard work but it can never provide conclusive evidence (as Fredric Morck from California pointed out). No matter how many validations are performed, how often the forecasts turn out to be correct, or how many other models were rejected to select our best working explanation for the world, there can always be another explanation. I think this could explain why, despite mounds of evidence and decades of research, large segments of the public are skeptical of human-caused climate change.

This idea came up today in a conversation I had with a colleague where he pointed out that learning from time-series and observational data ONLY is really hard. Unfortunately for climate scientists, this is a reality that they must live with (even the evolutionary biologists have it easier!). So how does one move forward in the face of complexity? I think that it depends on the goals. If it is too hard or unethical to perform critical tests and experiments, observational analyses are the only way to learn. Many of the issues where this is the case (climate, evolution, some diseases) require our best understanding of the system to manage the consequences as best as we can. Our ability to discriminate among hypotheses is reduced, though, and the output of these analyses is always subject to critique on aspects that were not considered. And there are always aspects that were not considered (unless we attempt something like this).

Fortunately for people like me (ecologists), there are loopholes. When the sample size is greater than 1, one can do somewhat-controlled comparative analyses. Even large, complex systems like ecosystems or populations can theoretically be subject to experiments (a beautiful example). Or, one can devise clever ways to test the assumptions and secondary output of our models with experimentation as a way to test the foundations of our theories. This follows with a recommendation for myself from myself; don't try to test the theory (unless I have good reason to suspect that it is wrong), test its assumptions to advance my understanding of the theory. Those questions are always more amenable to experimentation....

On "making it"

2011-08-31T11:31:00.000-07:00

How does someone (for example, a biologist) become successful? Does working long hours lead to success, or are other factors at play?

This week's issue of Nature hits this question head on with an article on work ethics in science. The article profiles a research laboratory where long hours are the expectation, channeling the "protestant work ethic" model to success. The article's subject took a "path from 19-year-old illegal immigrant from Mexico, labouring in the fields of California, to neurosurgeon at one of the United States' leading research hospitals." The author continues, "He did not get there by working 9 to 5."

The approach of working hard and long is something that I associate strongly with the ideal of America, that if a person works hard they will be rewarded for their efforts. It is very democratic. Of course, by working more a person limits their ability to achieve other goals in life, a point raised in the original article and in a reply article. But, if one wants to be successful, they should be able to work hard and wait for the rewards to flow.

Unfortunately, I think that this view is limited by concerns about efficiency and the likelihood of failure.

The first point, on efficiency, is also brought up in the original articles and in this recent article at The Atlantic. Even if one can train their brain "like an athlete" (as the subject of the Nature article suggests), each person has their limits, and efficiency may be the first thing to go at those limits, which ends up requiring even more time input to complete tasks. And time is finite, so at some point it is more effective to drop a task and pick it up later. At some point, it just comes down to whether the job is done.

Additionally, it is easy to fall prey to survivorship bias when thinking about scientific success (a point eloquently laid out by Nicholas Taleb in his book Fooled by Randomness). Just because the subject of the Nature article is successful because he works hard does not mean that working hard will lead to success. This should be especially clear when thinking about the path from illegal immigrant to researcher at Johns Hopkins University, but is apparently not clear to this subject.

While working long hours is a possible route to success, it is certainly not the only way and is undoubtedly not any more definite than other paths. But, it is great that issues like this are highlighted in major journals like Nature - it has at least given me a chance to reflect on my place in the spectrum!

Full disclosure: I will be leaving work at 5 today to go to the gym and then go home to cook a homemade meal.

Designing a research project

2011-08-25T15:29:00.000-07:00

Choosing a good research project is important but also very hard. Early career scientists and graduate students are tasked with developing projects that will produce useful outcomes (socially and scientifically). But, as Scott Lanyon (1995) notes, "research is not an activity with guarantees."

Personally, I am trying to hone in on an answer to the question what am I going to study? By Googling the subject of "choosing a research topic", one can find pages of resources from all walks of science and industry. While I will not attempt to cover even a broad swath of this literature, here I bring together a couple of points made by other biologists to produce a synthetic understanding of a successful research project.

Outline
Lanyon (1995) identifies some points on how to choose a dissertation project and provides a good starting point for my discussion. He describes how graduate students must consider a number of factors when designing a project, including

logistic feasibility, size, and scope,
originality (ability to generate new knowledge), and
ability to launch a career (social and scientific interest).

I am going to develop an outline for choosing a scientific problem based on these three concepts and their relationship to difficulty. The reason I am interested in difficulty is because I can envision a perception that a good research project is a difficult one, and this is a perception that I want to challenge.

Figure 1. Proposed relationships between key variables of a research project and the project's difficulty.

General interest
Lanyon explains how it is good to start by identifying broad issues and concepts. This matches Peter Medawar's assertion in "Advice to a young scientist" that
any scientist of any age who wants to make important discoveries must study important problems.

Doing so fulfills the requirement of attracting attention from other people, whether they are other scientists who just study different organisms or people who do not think about scientific research on any average day of their life.

The relationship between interest and difficulty could be all over the board. However, I think that there is probably some trend where more difficult problems are slightly more interesting and vice versa. So, while a goal may be to choose more interesting topics (and one can push the upper limits of "interestingness" with certain topics), topics that are not as difficult may be slightly less interesting (Figure 1, dotted line).

While it is simple enough to say that one has to do "important research," figuring out what constitutes an important issue is not that easy! Lanyon recommends beginning by reading widely in one's field. I would extend that to include topics and publications outside of science. I think many research projects could probably be born by reading various national and international newspapers or periodicals...

This does not mean that a project should necessarily be broad. Richard Reis quotes Cliff Davidson and Susan Ambrose of Carnegie Mellon University who state that

The most successful research topics are narrowly focused and carefully defined, but are important parts of a broad-ranging, complex problem.

Knowledge gain
Once one has begun to digest a body of work, Lanyon recommends identifying questions and issues. This requires a great amount of creativity (Lohele 1990) and can be challenging. But Leslie Reperant points out what I think is a really great way to approach this step. She states that

Most of the time, projects with the highest impact are not the most complicated high-tech and extraordinarily obscure projects. Many high impact papers in Nature or Science deal with questions with rather obvious answers. Yet those questions had never been scientifically answered before. So, one way to find a good topic is to find those “conventional wisdoms” that appear in all the introductions of papers on your system of predilection, check out the references they give (if they give any) to see whether it is a real fact, or a widely accepted assumption. If it is an assumption, then ... think of a simple way to test it and prove it right or maybe even ... prove it wrong!

I really like this concept. As with general interest, the relationship between knowledge gain and difficulty can be all over the board; even though simple problems tend to have less impact, occasionally they can substantially increase our knowledge. Additionally, it is possible for really difficult problems to push our knowledge frontier only slightly. Uri Alon (2009) suggests that researchers can choose projects that maximize knowledge gain along the "pareto efficiency front" - only engage in projects that produce small gains in knowledge if they are easy to do and reserve more difficult projects for those that will produce substantial gains in knowledge (Figure 1, dashed line). For most graduate students, this means choosing a project somewhere along the axis that will maximize knowledge gain while minimizing difficulty (since the dissertation project should be on a TIMELINE!). This leads me to the last component of choosing a research project.

Feasibility
In the quest for a project topic, feasibility is very important. Unfortunately, feasibility is inversely related to difficulty (Figure 1, solid line). Some questions are REALLY IMPORTANT but cannot be answered yet with current methods or data. Considering feasibility should be done throughout the project and can include choosing appropriate study organisms for the question and re-evaluating and modifying methods as data is collected and analyzed (Lanyon 1995).

Synthesis: payoff
We can synthesize these three components into a "payoff" value, where

Payoff ∝ Intererst × Feasibility × Knowledge gain

The payoff is basically a likelihood of success. The most important understanding from this relationship is that there is a hump-shaped response in payoff to difficulty (Figure 2). Some projects are easy to solve, but provide little advances and are not interesting. Others are extremely interesting and would revolutionize the field, but are not feasible. There is a "sweet spot" somewhere at intermediate levels of difficulty where payoff is greatest.

Figure 2. The "payoff curve" derived from equation 1.

This area is what Craig Lohele (1990) calls the "Medawar Zone" (after Sir Peter Medawar, who I quoted earlier), and it provides a guide for where success can come in science. Lohele uses Robert MacArthur as an example of someone who did this effectively (an example I love, since I idolize MacArthur):

Robert H. MacArthur was a prominent ecologist, active in the 1960s and 1970s, who died young. MacArthur was known not for being right all the time, but for having an unerring creative instinct for discovering interesting problems that were solvable and for extracting the essence of complex problems so that they became solvable. What is notable about such people is that even when they were wrong, they were wrong in an interesting way and on an interesting topic.

While not formally described by Lohele, the payoff curve can be derived by considering the general interest in the problem, the feasibility of the question, and the knowledge gained by the answer. These three components are essential for choosing a good research question.

Literature cited

Alon, U. (2009). How To Choose a Good Scientific Problem Molecular Cell, 35 (6), 726-728 DOI: 10.1016/j.molcel.2009.09.013
Lanyon, S. (1995). How to Design a Dissertation Project BioScience, 45 (1) DOI: 10.2307/1312534
Loehle, C. (1990). A Guide to Increased Creativity in Research: Inspiration or Perspiration? BioScience, 40 (2) DOI: 10.2307/1311345

Variation in observed changes in species range limits

2011-08-24T18:01:00.000-07:00

There is a good report in last week's Science Magazine on changing species range limits by Chen et al. (2011). It has been almost 10 years since the last comprehensive meta-analysis of species range shifts in response to climate change (Parmesan and Yohe 2003), so these authors performed a new analysis with a large diversity of species and regions.

Compared to previous analyses which suggested rates of change around 6.1 km poleward or m uphill per decade (Parmesan and Yohe 2003), Chen et al. (2011) found average values that are substantially higher: 16.9 km poleward/decade and 11.0 m uphill/decade.

Latitudinal shifts matched the expected response to temperature changes (the distance needed to "keep up" with temperature), but elevational shifts did not. This is very interesting, but I'm not going to talk about it. Instead, I am more interested in the actual distribution of responses that they observed.

The spread
Despite a statistically-greater-than-zero response on average, the spread of responses was pretty large and 22-25% of species shifted in the opposite direction to the expectation (data redrawn from Chen et al. 2011):

This is because there are very diverse responses among species in their pattern of change. Not all species change their range to track temperature (a finding that I posted about earlier on the blog). Chen et al. (2011) identify a few reasons for why this happens:

delayed ability to respond to the changed environment
different physiological sensitivities to temperature changes among species, or
other variables beside temperature (including moisture, biotic interactions, land use) constrain responses.

These hypotheses are not new and are extremely broad. But, they are consistent with other recent research (e.g. Ibáñez et al. 2009, Crimmins et al. 2011), and they should guide our efforts to figure out what is causing these limits so that we can understand why they change or not.

It is important to remember that there is no mechanism in this report, just a compelling presentation of observed changes. Not surprisingly, the findings led the authors to conclude that we need "more detailed physiological, ecological and environmental data ... to provide specific prognoses for individual species." This is a great call to the broader scientific community emphasizing that work on the ecology behind these observations is ripe for development.

I think that the most important result from this paper is the acknowledgement of the diversity of responses among species. There is not a single "biological response" to changes in the environment. The authors' call for more data echoes that of Gaston (2009), who states that empirical data lag far behind theories. The challenging step is to connect observed responses to the proposed hypotheses.

There is also a nice accompanying audio interview with coauthor Chris Thomas, which is worth a listen.

To be continued in the future with a discussion of the "state of the art" for range limits, an exploration of where we are at with understanding what maintains range limits.

"What do you study?"

2011-08-22T12:12:00.000-07:00

Today I officially begin the next step of my education (the PhD) at Duke University. After a long three-day trek across the country, I've landed in Durham and I am ready to get things started, which means that the blog will likely be shifting into a vehicle for developing and testing my dissertation research ideas.

Unfortunately, now I need to deal with the inevitable, hard-to-answer question of "what do you [want to] study?" How do I answer this? I know that whatever I say will likely change over the course of the next few years. Do I just dive into my dream ideas or do I guard what are certainly half-baked concepts? How careful should I be about what I say as far as defining "who I am" for people?

Ultimately, I think that people ask this question because they don't know what else to ask and just want to learn about what interests me. This is a corollary of how one describes their work to someone they meet on an airplane (an issue that keeps me awake at night...). But scientists are notoriously bad at making small talk, a point that was further reinforced at the breakfast get-together for the new PhD students that I attended. Discussion often quickly turns to talking about science as a way to learn about the other person.

No matter. I think I will just resort to a canned, jargon-laden description of what I find interesting right now (basically, the sum of this blog) and hope that people do not mis-interpret my illegible ramblings about species, populations, resource limitation, etc.

How many of you are dealing with this question? How have those of you who have gone through this answered this type of question? Does anyone else struggle with this?

Visualizing ESA 2011 in Austin

2011-08-06T07:00:00.000-07:00

I am planning my 2011 ESA visit and have read discussion around the internet about blogging on which presentations to see or tweeting at the meeting (#ESA11 is the handle...). As a complement to those endeavors (I'm not going to blog at the meeting), I decided to have some fun with the ESA "personal scheduler" search to get an overall feel for the meeting (and maybe the discipline as a whole!).

Think of it like a meta-analysis of the ESA Austin meeting.

UPDATE: added a rough selection of the number of presenters from a few American universities. Check it out at the bottom. ABB 08/10/11

First, I had to get an idea of how many presentations will be given (since I can't find that metric anywhere on the website!). I tried a few searches before I settled on searching for "2011" since I realized that all of the presentation descriptions had the presentation date written out. This gave me 3732 presentations, which I think could be a pretty good ballpark estimate. Then, I let my boolean searches go wild!

What organisms are people studying?
I searched for different study organism categories (see search criteria at bottom) and found that most people are studying plants, followed by animals, microbes, insects, and biogeochemicals. These categories accounted for 94% of my total number (pretty good coverage).

At what scales do people study these topics?
This one is really interesting! There is a nice distribution from small to large scales, with the majority of presentations at the community and ecosystem levels.

Also, these groups exceed the total number of presentations, meaning that there are a good chunk of people who are talking about multiple scales of organization. Pretty cool...

My presentation (shameless plug: Tuesday, Aug. 9, 8:40am, Rm 6A) is in the small minority of presentations (1.5%) that study biogeochemistry at a global scale.

Research approach
There are slightly more people outing themselves as empiricists (n = 1730) than as modelers (n = 1148). And a good group of people who mentioned both (n = 689, 24% of all mentions).

Scope of study
How are people framing their research? Time is hotter than space this year at ESA (52% greater), and there are almost 3 times more self-identified "applied" presentations than "theoretical" presentations.

Just for fun (since I just blogged about niches), what about niche versus neutral? There are over twice as many presentations that discuss niche (n = 181) as neutral (n = 76) topics.

Over 25% of the presentations are about biodiversity or diversity (n = 1006) and 20% are about global warming or climate change (n = 754) (201 presentations are about both!), while only 13% of the presentations are about some aspect of the meeting's theme (n = 469) (Earth Stewardship: Preserving and enhancing the earth's life-support systems) and 6% of presentations are about education (n = 224).

Where are these people from?
The majority of presentations have authors from a "university" (77%) while smaller numbers of presentations have authors from a governmental agency (17%) or "college" (11%).

I've also selected a biased sample of a few American universities to get a feel for where people are from.

Synthesis: Many of us are studying communities and/or ecosystems (especially through time), focusing largely on plants with a substantial group of animal ecologists also.

Where do you fit into the mix? What other types of questions would you have asked in this analysis?

The messy details
All of the boolean search terms that I used.

Study organism
Plant - (plant)
Animal - (animal OR fish OR amphibian OR reptile OR bird OR mammal OR marsupial OR primate)
Microbe - (microorganism OR microbe OR microbial OR bacteria OR pathogen OR virus OR protist)
Insect - (insect OR invertebrate OR arthropod)
Biogeochemical - (biogeochemical OR biogeochemistry)

Study scale
Cellular - (molecular OR gene OR genetic OR cellular)
Organism - (physiological OR physiology OR organism OR organismal)
Population - (population)
Community - (community)
Ecosystem - (ecosystem)
Landscape - (landscape)
Global - (globe OR biosphere OR global) NOT (warming OR (climate AND change))

Approach
Modelers - (model OR simulation OR modeled OR simulated)
Empiricists - (empirical OR field OR data OR observation OR observational)

Subject
Spatial - space OR spatial
Temporal - time OR temporal

Applied - applied OR management
Theoretical - theory OR theoretical

Diversity or biodiversity - (diversity OR biodiversity)
Global warming or climate change - (global AND warming) OR (climate AND change)
The ESA theme - (stewardship OR (life AND support) OR preservation OR sustainability) NOT (workshop)

Where from?
University - (university)
College - (college)
Government - (USDA OR EPA OR NPS OR (Forest AND Service))

Finding the niche

2011-08-03T13:10:00.000-07:00

The niche is an important idea in ecology (as I discussed previously). It is interesting to me (and many other ecologists) because it is a theoretical abstraction that allows us to think about how organisms relate to their environment. Despite its importance, our understanding of the niche has evolved throughout history, which can lead to confusion about what a niche really is. The Merriam Webster dictionary frames the confusion quite nicely with their definitions.

niche (noun):
a: a place, employment, status, or activity for which a person or thing is best fitted.
b: a habitat supplying the factors necessary for the existence of an organism or species.
c: the ecological role of an organism in a community especially in regard to food consumption.
d: a specialized market.

These definitions conflict somewhat. To understand the niche better, it is helpful to consider some historical context.

Beginnings: Joseph Grinnell

The history of the idea that a species has a specific role or place extends well before Joseph Grinnell (ranging from the bible to Darwin and Wallace). But, Grinnell (pictured at right) was the first scientist to formalize its definition in 1917, so we will begin with him.

Grinnell defined the niche as all of the habitat conditions that are required for a species to exist (Grinnell 1917). He thought that each species occupied their own niche that could be related to specific environments. He developed this idea around the relationship between the California Thrasher and its chaparral habitat, and refined it in future years with his studies of the environmental associations of mammals in Yosemite National Park. For each species, one could identify habitat conditions that allowed it to survive (Fig. 2). This view of the niche focused on the species and its requirements, like definition b above.

Figure 1. Grinnellian niche. Pluses indicate presence and oh's represent absence of the species in "environmental space." The dotted line indicates the species' niche boundary. Modified from Pulliam (2000).

Redefinition: Charles Elton

Ten years later, Charles Elton (pictured at right) promoted an alternative view of the niche in his textbook on Animal Ecology. He focused on the role that a species plays in a community (Elton 1927), similar to definition c above. In doing so, he proposed that a species might occupy more than one niche and that a niche might be occupied by multiple species (maybe even in the same community). Additionally, a community may have "empty niches" that are not occupied by any species.

This can be explained with some examples (c.f. Colwell and Rangel 2009). Elton observed that both the arctic fox and the spotted hyena eat bird eggs and dead animals, so these organisms occupy "the same two niches" (the egg-eater niche and the scavenger niche). Additionally, when considering how multiple species fit into the same niche, he stated that "we might take as a niche all of the carnivores that prey on small mammals." If there were no egg eaters in a community, that could be an "empty niche."

Modernization: G. Evelyn Hutchinson

For both Grinnell and Elton, the niche was associated with a particular place. Their perspectives fit with the prevailing view of community ecology at the time, where the community was viewed as a super-organism that was associated with a specific habitat (e.g., the chaparral community). They contrast with one another by considering whether a species' role is related to its requirements or its impacts.

A modernization of the idea came from G. Evelyn Hutchinson (pictured at right), who envisioned the niche as a "multidimensional hypervolume" where many variables restrict the persistence of a species (Hutchinson 1957). Hutchinson's niche is defined as the conditions that enable a population to have births greater than deaths. With this definition, Hutchinson removed the niche from the place and pulled the focus to the species. Thus there are no "empty niches" that could be filled; the niche is an attribute of the species (fitting with the individualistic view of the community that was defended the year before by Robert Whittaker). "Species, not environments, have niches" (Pulliam 2000).

Hutchinson went on to address the issue of requirements versus impacts with the fundamental and realized niches. Each species has some set of habitat conditions that allow it to exist (similar to the Grinnellian niche), which are the fundamental requirements for that species. But, when other species are present, they can affect the species directly (through prediation or parasitism) or indirectly (by consuming limiting resources), thereby reducing the portion of the niche that is realized. Even though Grinnell's Thrashers can theoretically persist in chaparral environments, they may not exist there in reality if, as suggested by Elton, there are predators that consume their eggs (Fig. 2).

Figure 2. The realized niche. Symbols are the same as Fig. 1, except that the second dotted line bounds a competing species' niche. Modified from Pulliam (2000).

Hutchinson's niche removed the species from a physical place and put it in a niche space and defined persistence with both biotic and abiotic conditions.

Refinement: metapopulations and dispersal
In the last 30+ years, ecologists have refined the niche idea by recognizing that populations are constrained by history, do not exist in isolation, and are not genetically homogenous (Pulliam 2000, Holt 2009), with different consequences for how niches are realized.

Many species do not occupy all of the locations that match their niche simply because they have not had the opportunity to get there either from physical (mountains, oceans, etc.) or biological barriers (short/slow dispersal) (Fig. 3).

Figure 3. Limited niche occupation (symbols as in Fig. 1). Modified from Pulliam (2000).

Populations also often exist in networks (metapopulations), which allow for the flow of individuals and genes among populations. A consequence of this is that individuals may be able to survive outside of the "niche" but (by definition of the niche boundary) not maintain population viability. These sink populations are supplied from source populations inside the niche, either through movement or recruitment. In this case, the realized niche can be greater than the fundamental niche (Fig. 4).

Figure 4. Sink populations can exist outside of the niche boundaries (symbols as in Fig. 1). Modified from Pulliam (2000).

Populations can also be different from one another, and there is evidence that niches vary in space and time (Pearman et al. 2008). So, for a single species, there may be sub-niches among populations which have more restricted requirements than others (Fig. 5). Additionally, the niche requirements of a species or population may change through time, leading to some potentially complex dynamics.

Figure 5. Populations may have different niche constraints than others within the same species (indicated with smaller dotted lines) (symbols as in Fig. 1). Modified from Holt (2009).

The future niche

I vew the niche as an overarching concept in ecology (despite works like Hubbell 2001) that provides a framework for understanding how species relate to their environment.

Despite its power, I think that many aspects of the niche have been relegated to theoretical discussions. Because of the complexities of the idea, it is difficult to identify and measure the niche empirically. When this exercise has been done, it is often for individual species because of the labor and a priori, species-specific knowledge that is required (Holt 2009), which can limit the generality of insights.

Rather than building a catalog of niche boundaries for specific species (which can be valuable for those species that are of particular interest - disease vectors, endangered species, etc.), I wonder if advances can be made by trying to test hypotheses proposed by niche theory? For example, do species shift their place or shift their niche? If an environment changes (à la Loarie et al. 2009), the fundamental niche of a species should shift on the landscape but it might be difficult for the species to shift with the environment. There are likely some cool connections between changing environments and changing niches (Colwell and Rangel 2009) which could give us insight into how often species change their physical place versus their niche space. There might be certain types or rates of changes that favor a niche shift versus a place shift. Related to this question (although on a more theoretical note), how many dimensions are needed to define a species niche? Do species differ in their dimensionality? Are lower-dimensional species more likely to experience a physical shift and higher-dimensional species more likely to experience a change in their niche?

The niche has not been a static idea through time and it will continue to change depending on how we wish to use it. By reflecting on the evolution of the idea, I think that we can gain a better appreciation for its importance and identify new pathways to explore in the future.

Literature cited

Colwell RK, & Rangel TF (2009). Hutchinson's duality: the once and future niche. Proceedings of the National Academy of Sciences of the United States of America, 106 Suppl 2, 19651-8 PMID: 19805163
Elton, CS (1927). Animal Ecology University of Chicago Press
Grinnell, J (1917). The Niche-Relationships of the California Thrasher The Auk, 34, 427-433
Holt, R. (2009). Bringing the Hutchinsonian niche into the 21st century: Ecological and evolutionary perspectives Proceedings of the National Academy of Sciences, 106 (Supplement_2), 19659-19665 DOI: 10.1073/pnas.0905137106
Hutchinson, GE (1957). Concluding remarks Cold Spring Harbor Symposia on Quantitative Biology, 22, 415-427
Pearman, P., Guisan, A., Broennimann, O., & Randin, C. (2008). Niche dynamics in space and time Trends in Ecology & Evolution, 23 (3), 149-158 DOI: 10.1016/j.tree.2007.11.005
Pulliam, H. (2000). On the relationship between niche and distribution Ecology Letters, 3 (4), 349-361 DOI: 10.1046/j.1461-0248.2000.00143.x

Ecology in the era of big data

2011-08-02T07:33:00.000-07:00

In this month's Frontiers in Ecology and the Environment, Dave Schimel editorialized about the "era of continental-scale ecology," its relationship to NEON, and what it means for future ecologists. As the main advocate of NEON, Schimel is understandingly focused on the need for ecology at the continental-scale.

His argument is that the ecological issues of this generation and the next require an understanding of patterns and processes at large scales. The argument is reminiscent of earlier calls for large-scale ecology.

But this time, the rallying point is different. The focus is not on meta-analysis by individual scientists (or research groups) or on a loose federation of individual research sites. Instead, ecologists will finally have access to a substantial infrastructure for examining and answering their questions. When it is completed, NEON will fundamentally change the way ecological research is conducted in the United States. Ecology is joining the ranks of other sciences that rely on big data, and I am excited for it. I think that it will be a good shift for the discipline.

Schimel emphasizes that "young investigators in ecology and environmental science should be encouraged to explore opportunities for increased training in the mathematical and computational tools needed for analyzing large datasets" and that "mentors and promotion committees should encourage and reward young faculty for working with such 'big' data, community models, and collaborative science, and be sure they don’t overly emphasize hands-on collection of data and site-based science." I agree almost wholeheartedly. I think that ecologists' experiences in site-based science (or at least the experiences of past generations) will help us to use the NEON infrastructure constructively while retaining a vision of what types of questions can (and cannot) be asked and answered observationally at the continental scale.

There will be more about NEON at the Ecological Society of America meeting in Austin next week, including the poster session on Monday, August 8 at 4:30pm "Development of the National Ecological Observatory Network (NEON): Long-Term, Continental Scale Data and Information to Enable Ecological Understanding and Forecasting" and probably many others.

Environmental change and infectious diseases

2011-07-20T12:20:00.000-07:00

The idea that infectious disease distributions and dynamics may change with changes in climate (IPCC 2007) is particularly potent. It suggests that we must be vigilant about the emergence and expansion of vector- and water-borne diseases in areas where they may not have been a problem historically. But, without taking account of some critical ecological theory, these forecasts have the potential to be simply wrong.

Current perspectives
Researchers have taken advantage of "environmental niche models" to predict the potential distributions of infectious diseases. For example, climate change is predicted to cause an increase in malaria vulnerability in the southeast United States (Rogers and Randolph 2000, van Lieshout et al. 2003). Based on the bio-climate envelope of Anopheles gambiae, an important malaria vector, the area of vulnerability could move northward in the United States at rates approaching 5km per year. There are many problems with this analysis, especially for infectious diseases, since so many factors besides climate (or the environment in general) are important (Lafferty 2009). Barriers to dispersal and biotic interactions can further restrict an organisms distribution beyond the "fundamental" environmental niche, and infectious diseases face many barriers to dispersal (like disease control efforts).

This is not to say that infectious disease ecology is an "empty room" without accomplishments. There has been amazing work on disease dynamics and the ecology behind them in the last 30 years (some examples here and here). But, I think that most of this has been quite theoretical and hasn't seeped into the epidemiology and public health spheres very much, especially related to understanding how normal dynamics map onto climate change forecasts (although I would enjoy being proven wrong on that statement! like this example).

If a gap exists between our theories for the ecology of infectious diseases and our forecasts for the future, there is a great opportunity for ecologists to develop and expand their ideas with infectious diseases as a model. In fact, examining our common assupmtions about organismal responses to environmental change (like a shift in species distribution with a change in the environment) with disease models could be really useful for theory in general. So, tying our existing theory to disease data could help us resolve conflicts in our theories.

A conclusion
In a critical analysis of disease-environment relationships in East Africa, Hay et al. (2002) state, "The more certain climatologists become that humans are affecting global climates, the more critical epidemiologists should be of the evidence indicating that these changes affect malaria." This statement rings true for all studies of the consequences of environmental change for organisms and biodiversity. As we continue to develop theories about the processes that cause changes in organisms with changes in the environment, the family of hypotheses grow. However, the possibility that researchers will cling to their pet hypothesis also grows, which restricts our progress on the issue. I think that studies that equally consider multiple hypotheses and assess the relative effects of different processes will help.

A few remaining questions in my mind:

How susceptible are the distributions of infectious diseases to changes in the environment?
Are there situations where a level of environmental change could override extrinsic control efforts (for example, by increases in local population sizes) and increase the likelihood of a "disease invasion"?
How can physiology and population dynamics be incorporated more explicitly in disease models?
How do we incorporate both environmental conditions and extrinsic factors in predictions of vulnerability?

Gregor Mendel's birthday

2011-07-20T06:15:00.000-07:00

Happy 189th birthday to the man who studied variation in plants and was fortunate enough to uncover the basic principles of inheritance!

I just learned that during the seven years of his pea experiments, he cultivated and tested over 29,000 plants. Talk about impressive sample size!

Model organisms

2011-07-07T12:31:00.000-07:00

Model organisms have been used by biologists to make some of the most important scientific breakthroughs in recent history, ranging from our understanding of genetics and evolution to development and disease. They are routinely employed in a laboratory setting to conduct replicated experiments more quickly and more ethically than could be done in the wild.

Generally, ecologists have been slow to utilize model organisms when compared to their other biologist kin. Of course, MANY ecologists have successfully used model organisms to test ecological theories. But, comparatively, I think that ecologists do not use them nearly as often as they could. This could be due to the reluctance of some ecologists to move into the laboratory or a skepticism about the relevance of results from model organisms for "natural" ecological systems. Or maybe it is just because ecologists are not creative enough in thinking of how they can use model organisms to answer the questions that they have. Jeremy Fox at the Oikos Blog has outlined some related ideas about the use of microcosms in ecology and has done a much better job than I could discussing the value of using model organisms.

I am increasingly interested in how I could use model organisms in my own work. Some of this interest was rekindled a couple of months ago when Todd Levine from Murray State University posted on the ECOLOG listserv soliciting information about model organisms for population ecology research, and the feedback was pretty eye opening about the possibilities! Here is a run-down of a few of the ones that I think are pretty neat, have been used successfully by ecologists in the past, and should be used more often in ecology in the future.

Daphnia magna - aquatic flea, raised in water, traditionally used to study toxicity, could easily be adapted to study organismal and population responses to the environment. Cool ecological example: (Drake and Griffen 2010)

Tribolium castaneum - flour beetle, raised in vials with flour medium, traditionally used in genomic studies, easily adapted to study species distribution dynamics. Cool ecological examples: (Melbourne and Hastings 2008, 2009)

Drosophila melanogaster - fruit fly, raised in vials with instant potato medium, traditionally used in genetics and inheritance studies, easily adapted to study organismal or population responses to the environment. Can be used with closely related species and parasites to examine community dynamics and species interactions. Cool ecological example: (Davis et al. 1998)

Other notable examples:
Lemna minor - asexually-reproducing aquatic plant, used frequently to demonstrate population growth and density dependence. Cool ecological example: (Jefferies 1993)

"Wisconsin Fast Plant" Brassica rapa - fast life cycle mustard plant, produces flowers in 2 weeks, can be used to demonstrate environmental impacts on plant traits.

Protists - Jeremy Fox (Oikos blog and University of Calgary) uses them to study community dynamics. Plasmodium falciparum is a protist that causes malaria. They sound pretty useful!

Model organisms have much to contribute to ecology, most notably as a sensible approach to testing ecological theories. Many of them can be purchased relatively inexpensively from biological supply companies (e.g., Carolina Biological Supply) or harvested in the wild. Their usefulness is only limited by our creativity, with applications in extinction, distributions, demographics, physiology, temporal change, etc. How can model organisms enhance your research?

Tenure trials

2011-07-06T15:53:00.000-07:00

At first glance, academic tenure seems pretty scary. A committee of peers reviews your on-the-job progress to assess your fate over a number of years. It is viewed as a challenging hurdle on the way to academic independence and a stable future. I just read a first-hand account of dealing with tenure denial on the Chronicle of Higher Education website. I've previously discussed the probability of getting a tenure-track position on the blog, but this article got me thinking about what are the odds of tenure once you have a job?

A report by Dooris and Guidos (2006) catalogues the number of faculty entering on a tenure track from 10 American universities (Florida, Illinois, Iowa, Maryland, Michigan, Northwestern, Penn State, Pittsburgh, Rutgers, and Wisconsin) who were ultimately awarded tenure. It presents a particularly grim portrait of tenure, with only 52% of the entrants becoming tenured. Additionally, the distribution (I fitted it to a beta distribution) is pretty narrow!

Is tenure really that unlikely? To look into it further, I grabbed some information from the American National Center for Education Statistics. Their 2004 National Survey of Postsecondary Faculty has data from 63 American universities on tenure outcomes. They report that 91.6% of individuals who are considered for tenure are granted tenure. This looks much more favorable... once an individual makes it to the final stage, there is a pretty good chance that they will get tenure.

Can we reconcile the difference? First, it is important to remember that tenure is a process, not a one-time decision, so there are periodic reviews throughout. Dooris and Guidos (2006) also report recommendations for continuation in the year 2004 for 2nd (96% recommended for continuation), 4th (86%), and 6th (89%) year reviews. Based on the limited data from Penn State, not all individuals who get a tenure-track job make it through each review, which can get us closer to 50% (actually around 75%). Then, we need to account for individuals who leave the university even though they have a tenure-track position and may be recommended for continuation. For the individuals who entered into tenure-track positions in 1998, only 58% were reviewed in 2004, which means that many of those individuals had to leave in another way.

In sum, tenure outcomes seem to be pretty favorable if the individual can continue to make progress on reviews and stay in their position. But, across institutions, the probability of obtaining tenure in the final review has a pretty fat tail (as shown in the second figure), which means that there are many chance events that can produce an unfavorable result.

Forecasting species' range changes

2011-07-06T09:28:00.000-07:00

Predictive knowledge of which species are likely to exhibit range changes in response to changes in their environments would be extremely useful for conservation biologists and resource managers. We are not quite there yet, though, in part because of the general inadequacy of "habitat-niche models" to create accurate scenarios of range changes (Araujo 2005, Beale et al. 2008).

In the newest issue of Ecology Letters, Amy Angert and others examine whether species' traits (e.g., habitat preference, life history characteristics) could fill the gap and increase our ability to predict range shifts (Angert et al. 2011). Their hypotheses were that dispersal potential (e.g., dispersal mode), growth/reproduction rate (e.g., generation time, offspring number), and ecological "generality" (e.g., diet width, mating system) would impact the size of range shifts. They tested these hypotheses with data on North American birds, British Odonata, Swiss alpine plants, and western North American mammals by constructing models that link the species' traits to the magnitude of observed change.

They were able to identify variables that impacted range change for each taxa (obviously... that is like saying that a person found a pair of pants to wear by trying on all of the pants in their closet). However, they could not find any specific variables that explained the patterns across all datasets. This is a profound conclusion because it suggests that range changes are not strongly influenced by specific species' traits across taxa.

They closed the article by examining the prospect of predicting range shifts, acknowledging that changes in ranges are clearly individualistic (i.e., there is no "globally coherent" magnitude of range shifts). Does the lack of support for traits in general mean that we are limited to making prediction on a taxa-specific basis? I hope not, and I don't think that Angert and others think so either. Their conclusion is that we need "a better understanding of the process of range shifts," which I take to mean a better understanding of the environmental sensitivity of population dynamics that lead up to changes in the range boundary.

Our ability to predict becomes more complex when we consider the likelihood that multiple climatic variables can have an impact (Berdanier and Klein 2011), especially when considering inherently uncertain processes (e.g., dispersal; Clark 1998). It is unlikely that we can increase forecast information even if we can make detailed parameterizations of these types of variables (Clark et al. 2001). But, we may be able to improve forecasts by identifying the "slow" variables that forewarn of changes (coming back to my idea about comprehensive population dynamics, à la Doak and Morris 2010). I also like the authors' suggestion to "use niche modelling to quantify predicted range shifts for each species, and then relate species traits to a relative range shift metric (e.g. the difference between observed and predicted shifts)." I think that this type of study should yield some interesting results.

In sum, Angert and others present an excellent analysis that provides more ecologically meaningful information on range changes than has been given to date with simple environmental correlations. This article is a substantial step towards improving our knowledge of range changes. The specific factors that cause change and thresholds of influence may vary from species to species or taxon to taxon, but generalities may still be lurking. A useful next step may be to move away from variables that would be uninformative even if we understood them (e.g., long-distance dispersal) and toward examining the ecological processes that these variables impact (e.g., net population dynamics). Then it might be possible to identify patterns that forewarn of change (Drake and Griffen 2010) across taxa.

Reference
Angert AL, Crozier LG, Rissler LJ, Gilman SE, Tewksbury JJ, & Chunco AJ (2011). Do species' traits predict recent shifts at expanding range edges? Ecology letters, 14 (7), 677-689 PMID: 21535340

The niche is important

2011-05-31T10:03:00.000-07:00

The ecological niche is a foundational concept in ecology. Based on the french word nicher (for "nest"), it is a theoretical construct used to describe how an organism or species fits into its environment. The niche is generally attributed to a publication by Joseph Grinnell in 1917, and was further defined by Charles Elton in 1927 and G. Evelyn Hutchinson in 1957. I will save a discussion of what is a niche and how its definition has changed through time for a future blog post. Today, I want to focus on the importance of the niche in the ecological literature.

To examine this, I searched Web of Science for publications with "niche" as a topic. I looked at how the number of publications with the topic "niche" has changed through time and how this change is related to the total number of ecological publications. I searched over the years 1957 to 2010, to identify changes in the importance of the niche since Hutchinson's publication.

More "niche" publications

As seen in the figure above, the number of publications indexed by Web of Science with the topic "niche" have grown dramatically (pretty much exponentially) since 1957. This could be because the niche is becoming more important, or it could just be an artifact of the number of journal articles that are indexed on Web of Science.

To account for this issue, I looked at the fraction of "niche" publications in ecological journals, to provide an index of the number of ecological publications that examine niche ideas.

Is "niche" becoming more important in ecology?

Yes, it is. For the above data, I searched for the number of publications with the "niche" topic in 20 top journals for the field of ecology (based on impact factor: Publication Name=(Ecology Letters OR Global Ecology and Biogeography OR Global Change Biology OR Ecological Monographs OR American Naturalist OR Proceedings of the Royal Society B OR Journal of Ecology OR Conservation Biology OR Functional Ecology OR Ecology OR Ecography OR Diversity and Distributions OR Journal of Applied Ecology OR Journal of Biogeography OR Ecological Applications OR Ecosystems OR Oikos OR Oecologia OR Journal of Animal Ecology OR Journal of Plant Ecology)). I divided the resulting number by the total number of publications for those 20 journals.

The fraction tells us about the relative importance of the "niche" in ecology. Interestingly, this fraction increases as well. It is not quite exponential, but it is definitely a dramatic increase in recent years, with now over 5% of publications in ecology journals examining niche topics.

The niche is a very important idea in ecology. It seems to be growing in importance. This is likely because of the relevance of the niche for addressing questions of environmental change, which are particularly "hip" in ecology right now. Additionally, niche topics are growing in use in other subjects, such as geography and evolution, which could be encouraging an increase of its examination in ecology as well.

How have the ideas of the niche evolved through time? Are current uses of "niche" ecologically accurate? Are "niche" ideas useful for other subjects besides biology? These are questions that I will examine in the future.

Reforming undergraduate science education

2011-05-20T10:08:00.000-07:00

Two awesome articles on inquiry-based science education came out this month in the journal Science. The authors propose ways to increase learning by reconsidering the classic approaches to post-secondary science education: the lecture (Deslauriers et al. 2011) and laboratory (Moskovitz and Kellogg 2011) settings.

Interactive lecture
The first article focused on modifying a traditional lecture to encourage students to "think scientifically". In their experiment, the authors taught one of two sections of an undergraduate physics class with a "deliberate practice" approach, which requires students to practice scientific reasoning and problem solving on challenging questions and tasks. They completely did away with the lecture. Instead, they had students work in small groups to answer questions throughout the class (both multiple-choice "clicker"-style questions and short answer responses that were used for participation grades). To prepare for these tasks, the students were assigned short reading assignments before class and were given pre-class online T/F quizzes (to encourage reading before class!).

This approach worked. Deslauriers et al. (2011) report that the average scores on an exam after the experiment was 41% for the control group versus 74% for the experimental group. Given that random guessing could give a student 23%, students in the experimental setting performed twice as well as students in the control setting!

Laboratory writing
The second article focused on building "inquiry-based writing" into college laboratory courses. They challenge the traditional cookie-cutter "lab report" - where students lack a research agenda and are given methods that they must parrot back to the instructor in their reports on exercises that lack the "discovery" aspect of authentic research environments - and propose many alternatives.

First, they suggest that instructors can design laboratory experiments that allow students to have something meaningful to say. Traditionally, instructors know the expected outcome of an experiment. Students know that instructors know the outcome, so they have no motivation to write persuasively. If instructors develop exercises that could have multiple potential outcomes (and ideally would be "double blind" so the instructor doesn't know what the outcome will be either), students can write more like scientists who must convince a skeptical audience (either the instructor or their peers). This suggestion requires students to write persuasively, but allows them to do meaningful and engaging tasks, and requires instructors to evaluate how clearly and convincingly each case is made, but allows them to respond to tasks as scientists instead of evaluators.

Second, instructors can create assignments that reflect what students can actually gain given their educational level. Traditionally, students are instructed to write reports in the IMRaD style, but students are not equally prepared to write all sections of a scientific article. Moskovitz and Kellogg (2011) suggest having advanced students that are conducting original research focus on introductions (because they have a research agenda), intermediate students that are designing (or at least modifying) their own experiments focus on methods, and introductory students who are generating data focus on presenting and discussing results. For example, introductory students could produce "a single, well-designed page" including "a main claim supported by key results, appropriate visuals, analysis of error, and so forth." This suggestion requires thinking beyond a single exercise or course to the entire progression of classes in the degree, but would allow students to produce authentic scientific rhetoric. Additionally, since these assignments would be shorter than the traditional lab report, students would be forced to write selectively and compellingly, much like researchers must do for a "letter" or "short communication" publication.

Inquiry-based science education
I like both of these works as a demonstration of the value of inquiry-based science learning. In the first article, the authors reject the traditional "lecture" and suggest using 1.) pre-class reading assignments and quizzes, 2.) in-class "clicker" and short-answer questions with student-student discussion, and 3.) targeted in-class instructor feedback. In the second article, the authors reject the traditional "lab writeup" and propose tasks that allow students to 1.) present more "authentic" scientific findings and 2.) focus on distinct aspects of scientific research at different stages of their education.

I particularly like the second article because Moskovitz and Kellogg's (2011) suggestions could be implemented by graduate teaching assistants relatively easily.

Inquiry-based learning approaches are hallmarks of the education that students receive at small liberal arts colleges (like my alma mater, Beloit College) and it is great to see their implementation at larger universities. Wider use of these methods will undoubtedly advance science education and general scientific understanding in the United States.

References
Deslauriers, L., Schelew, E., & Wieman, C. (2011). Improved Learning in a Large-Enrollment Physics Class Science, 332 (6031), 862-864 DOI: 10.1126/science.1201783

Moskovitz, C., & Kellogg, D. (2011). Inquiry-Based Writing in the Laboratory Course Science, 332 (6032), 919-920 DOI: 10.1126/science.1200353

Causation and inference

2011-05-18T11:48:00.000-07:00

I was reading the wikipedia entry on epidemiology and came across some interesting text on inference. I think that it is an eloquent description of the approach of most science.

It is nearly impossible to say with perfect accuracy how even the most simple physical systems behave beyond the immediate future [...] ; "Correlation does not imply causation" is a common theme for much of the epidemiological literature. For epidemiologists, the key is in the term inference. Epidemiologists use gathered data and a broad range of [...] theories in an iterative way to generate or expand theory, to test hypotheses, and to make educated, informed assertions about which relationships are causal, and about exactly how they are causal. [...] If a necessary condition can be identified and controlled [...], the harmful outcome can be avoided.

I was thinking about this topic this morning as I heard a report on NPR about a study suggesting that coffee consumption can reduce the risk of prostate cancer in men (while drinking my coffee!) and saw a rebuttal on a blog criticizing the study based on two ideas: 1. the difference between a population and individuals (a distinction understood clearly by almost all scientists) and 2. the question of correlation versus causation (another distinction understood clearly by almost all scientists).

In science we are interested in causation, but absolute causation is a difficult condition to establish. Rather than reject all information that is "not causal," we can use that information to increase our general understanding of how the world functions. I like the idea of a working hypothesis (hopefully chosen from a family of hypotheses) as the best explanation that exists for a pattern. Although clearly they are often inadequate and simplified, these working hypotheses are useful. The cautious but pragmatic application of working hypotheses can enable us to avoid "harmful outcomes" - which is really a major purpose for scientific research.

Climate sensitivity of species ranges

2011-05-16T09:40:00.000-07:00

There is great concern that changes in climate will have substantial impacts on biological systems, including extinctions due to limitations on species distributions (Thomas et al. 2004). Many reports suggest that climate changes are already impacting organisms and ecosystems (Walther et al. 2002; Parmesan and Yohe 2003). A common prediction in this literature is that, as the climate changes, organisms will need to migrate to follow their climate "niche" or face extinction (Laurie et al. 2009). Indeed, the climate envelope models used by Thomas et al. (2004) are based on this assumption.

Are species really that sensitive to changes in the environment? This is a complex question that will require much more work than a blog post. But, we might be able to learn a little bit about this question by reviewing how species ranges have changed in the past.

There are a few examples of studies that have examined changes in species' entire ranges (as opposed to the plethora of studies that only examine changes at a single site, or along a single range margin), which I've summarized below in a table indicating how many species' ranges shifted, expanded, contracted, or saw no discernible change:

Table 1. Analysis of species range responses to environmental change.

An interesting finding from this analysis is that, even though nearly half of the species have followed this "shift or contract" pattern, almost a quarter of species have not changed their distribution at all. In fact, a large chunk of species have expanded their ranges. These findings are consistent with other results suggesting that similar fractions of species are either non-responsive to climate change or have responded in a direction opposite to that predicted (Parmesan et al. 2005).

There are multiple hypotheses for why this pattern might happen, ranging from a simple lag between when the environment changes to when the species responds (e.g., maybe the species who are not responding or are expanding will shift or contract in the future) to complex responses including physiological plasticity and stronger sensitivity to local habitat conditions (Crimmins et al. 2011) or species interactions (Clark et al. 2011) than to broad climatic patterns. As I've discussed previously on the blog, climate envelope models are less than ideal for predicting species responses to climate change. A major reason for this shortcoming is that these models assume that species will track large-scale changes in climate conditions. The fact that many species do not track their climate "niche" emphasizes the important limitation of this approach to understanding the responses of ecosystems to environmental change.

A basic conclusion from this analysis is that species are not equally sensitive to changes in climate. This conclusion certainly complicates predictions of species extinctions that are based on assumptions about how species respond to large-scale environmental change (à la Thomas et al. 2004) and justifies efforts to 1.) assess which species are more or less sensitive to changes in climate, and 2.) identify why these differential responses exist.

Brief literature cited

Loarie, S., Duffy, P., Hamilton, H., Asner, G., Field, C., & Ackerly, D. (2009). The velocity of climate change Nature, 462 (7276), 1052-1055 DOI: 10.1038/nature08649
Parmesan, C., & Yohe, G. (2003). A globally coherent fingerprint of climate change impacts across natural systems Nature, 421 (6918), 37-42 DOI: 10.1038/nature01286
Parmesan, C., Gaines, S., Gonzalez, L., Kaufman, D., Kingsolver, J., Townsend Peterson, A., & Sagarin, R. (2005). Empirical perspectives on species borders: from traditional biogeography to global change Oikos, 108 (1), 58-75 DOI: 10.1111/j.0030-1299.2005.13150.x
Thomas, C., Cameron, A., Green, R., Bakkenes, M., Beaumont, L., Collingham, Y., Erasmus, B., de Siqueira, M., Grainger, A., Hannah, L., Hughes, L., Huntley, B., van Jaarsveld, A., Midgley, G., Miles, L., Ortega-Huerta, M., Townsend Peterson, A., Phillips, O., & Williams, S. (2004). Extinction risk from climate change Nature, 427 (6970), 145-148 DOI: 10.1038/nature02121
Walther, G., Post, E., Convey, P., Menzel, A., Parmesan, C., Beebee, T., Fromentin, J., Hoegh-Guldberg, O., & Bairlein, F. (2002). Ecological responses to recent climate change Nature, 416 (6879), 389-395 DOI: 10.1038/416389a

Abundance in ecology

2011-05-13T09:52:00.000-07:00

Abundance - the relative amount of a species in a particular ecosystem - is an important concept in ecology. Measures of abundance, which are approximated by counting the number of individuals in a sample area, are used to indicate population health and well-being. Thus, abundance has a central role in many ecological theories, including the limitation of species ranges and geographical patterns of speciation.

However, many of these ideas are based on a basic hypothesis about how abundance varies across a species' range. Specifically, it is generally assumed that abundance is greatest at the center of a species range and declines towards the edge of the range (the 'abundant center' can look much like a normal distribution). This simple assumption is often justified by the idea that abundance is tied to environmental gradients; at some point, the environmental conditions become less favorable and population numbers decline and, eventually, populations are no longer sustained because conditions become too extreme. In other words, local abundance is a reflection of how well a particular site meets the needs of a species. It approximates the niche. However, the 'abundant center' hypothesis assumes that niche conditions are spatially correlated and gradually decline toward range edges. This assumption is used extensively in theoretical and applied ecological research. It might also be inappropriate most of the time (Sagarin et al. 2006).

Why wouldn't abundance exhibit a uniform pattern across a range?
Essentially, abundance is controlled by many factors other than gradual environmental gradients. Canham and Thomas (2010) showed that trees in North America showed very little variation in local abundance across the species' ranges. Similarly, Sagarin and Gaines (2002) found that only 39% of the (145 diverse) cases that they examined even closely resembled a pattern of peak abundance in the center of the range. This is probably because, once an organism has colonized a site, factors like competition with other organisms and heterogeneity in local environmental factors like soil type, soil moisture, or topography are what modulate local abundance (Clark et al. 2011).

Is abundance still a useful indicator for ecology?
Given this issue, it is likely that abundance is not as useful as we thought for predicting the climate sensitivity of a species. This is an issue given that many models that predict species responses to climate ('climate envelope models') assume that abundance is a direct indicator of climate response.

Abundance is still valuable. In its basic definition it is still relevant as an indicator of population health. However, it is unlikely that most species will meet the assumption of an abundant center with gradual declines towards the edges. Consequently, we need to re-test many of our hypotheses about how species range dynamics function (Sagarin and Gaines 2002). To make abundance useful for predicting species responses to climate, it will be important to include other variables like physiological condition or population growth rates to get the whole picture (Clark et al. 2011). However, our realization that real abundance distributions are much more complex and varied provide an opportunity to re-evaluate theoretical concepts and develop better hypotheses to explain species dynamics (Sagarin et al. 2006). The possibilities are nearly endless!

Literature cited

Canham CD, Thomas RQ. 2010. Frequency, not relative abundance, of temperate tree species varies along climate gradients in eastern North America. Ecology 91:3343-3440.
Clark JS, et al. 2011. Climate change vulnerability of forest biodiversity: climate and competition tracking of demographic rates. Global Change Biology 17:1834-1849.
Sagarin RD, Gaines SD. 2002. The 'abundant centre' distribution: to what extent is it a biogeographical rule? Ecology Letters 5:137-147.
Sagarin RD, Gaines SD, Gaylord B. 2006. Moving beyond assumptions to understand abundance distributions across the ranges of species. Trends in Ecology and Evolution 21:524-530.

Happy Mother's Day

2011-05-08T07:26:00.000-07:00

For all of the moms in the world.

Epidemiological approaches

2011-05-06T09:57:00.000-07:00

There are many approaches to study population risk factors and sensitivities to different variables with observations (rather than strict experimentation). Many of these approaches have been extensively developed in the field of epidemiology. Unfortunately, few people have carefully considered the implications of these approaches for ecological research, where they are very useful and are frequently utilized unknowingly. In ecology, epidemiological approaches can inform ecosystem management by identifying risk factors and determining optimal approaches to prevent undesirable consequences. Additionally, they can provide tests of theoretical concepts about the impacts of variables on the population. Here I compare a few of these approaches with the idea that thinking about these approaches can help me design better ecological studies. For an example, I ask the question "Does a warming climate affect plant survival and reproductive output?" and think about how each type of study would differ in approaching this question. Thanks to Wikipedia for help with some of the definitions.

The "ecological study"
In this case, we are not talking specifically about the field of ecology. Instead, we are talking about using a population as a unit of analysis, as opposed to an individual (how unfortunate that it has the "ecology" label!). Examples of this approach include comparisons between variables in different countries or species. A valuable aspect of ecological studies are the ease of performing them; they can be quick and inexpensive since they often use data that is already available. They can form the foundation for future, more-carefully-designed studies and generate useful hypotheses. However, they are subject to the issue of "ecological fallacy" (i.e., inferring individual responses with aggregate population statistics, assuming that individuals have average characteristics) so they are often viewed as inferior to cohort and case-control studies. For our example question, we could compare the average survival and reproduction of different plant species in different counties of the United States and relate it to the average temperature in those places.

The "case-control study"
In this setup, we identify individuals that are at risk and match them with individuals that are not at risk (the controls). We then examine what characteristics of the past or future are different between the groups to identify how those factors impact the response variable. In this case, we could identify a sample of plants that have reduced survival and reproductive output and compare their temperature histories to those that do not have reduced survival or reproductive output.

The "cross-sectional study"
This design observes the entire population (or a representative subset) at a defined time period. It primarily describes spatial variation in the population, which can then be related to a variable of interest (a correlation in space). This type of study can provide data across the entire population about absolute risks and describe population characteristics. It is a step up from the "ecological study" since the unit of inference is the individual. However, it is limited in that it can confound space for time, and correlations in space may not translate to changes in time. In our example question, we could examine survival and reproductive output for one year across a species range and relate it to the temperature that those individuals experience in that year to assess risk to warming.

The "cohort study"
This design involves a longitudinal study of a small group of individuals, tracking their responses through time. Each cohort is only a small sample of the population, but they can provide specific information about individual risks through time. For our example question, we could take a group of plants that are exposed to a warming climate and a group of plants that are not exposed and follow them through time to note dif ferences in plant survival and reproductive output. A more rigorous approach to this study is the "randomized control trials" approach, where a control (placebo) is used as well.

Better theory of species range limits

2011-04-28T11:55:00.000-07:00

I'm reading "Geographic range limits: achieving synthesis" by Kevin Gaston (2009). Gaston paints a rather grim picture of our understanding of species range limits, and identifies a few areas that need exploration.

First, Gaston says that empirical studies lag substantially behind theoretical developments. We need empirical analyses of species ranges!

Second, he emphasizes the connections between genetics, physiology, and population dynamics, and the fact that many different factors can limit a species range at the same time. To address this, Gaston suggests that we need "studies that document variation in all of the parameters [immigration, birth, death, and emigration]" (and to identify organisms that would be suited to this work - challenging!).

He ends with this sentence: "[I]t has often proven frustratingly difficult to explain what determines the limits of a particular species at a given place and time." What a dim ending to an important article on such an important topic (with implications for biological invasions, habitat loss, climate change, emerging diseases, and food security)!

I think that combining some of the theoretical concepts around range boundaries with empirical data from a range of species could provide some great information about what is going on and hopefully shine some light on the determinants.

Gaston, K.J. 2009. Geographic range limits: achieving synthesis. Proceedings of the Royal Society B 276:1395-1406.

A great divide and scientific progress

2011-04-27T09:33:00.000-07:00

Why do economists tend to be optimistic and ecologists tend to be pessimistic? That is certainly an unfair blanket statement, but some recent articles have made me really interested in the question.

Yesterday the New York Times published a review of the work of Robert W. Fogel, an economist at the University of Chicago who has written a book on the connection between "Health, Nutrition, and Human Development in the Western World Since 1700." In it, Fogel explains how human bodies have grown more over the last few centuries than they ever have in previous history (basically, people are taller and weigh more). Fogel argues that this change is driven by advances in technology which have allowed us to improve food production and public health, and have resulted in modern humans being distinct from our ancestors.

This is not only a success story of human history, though. When faced with the question of what "overnutrition" will mean for our futures, Fogel said that "he remained an optimist at heart [emphasis added]. The human body is enormously flexible and responsive, he said, a fact that fills him with confidence that 'the trend of larger bodies and longer lives will continue into the future.'" Optimism is a quality that I view as being disciplinarily aligned with the study of economics. It is similarly reflected in the work of Hans Rosling and the information available from the Food and Agricuture Organization.

In a very different article at SEED magazine, scientists were asked what information they would pass on if only one statement of scientific knowledge could survive. Here are some selected quotes:

The "economists"

"The dazzling diversity of species and biological adaptations over 3.5 billion years of life on Earth owes its existence to 'adaptation by natural selection,' which requires just three simple conditions to operate: variation, differential selection (the best performing traits survive and reproduce more effectively than others), and replication of successful traits by subsequent generations, via a double helix of molecules that code for proteins as biological building blocks, or among more complex animals, via imitation or cultural transmission of methods and knowledge."
—Dominic Johnson is a reader in politics and international relations at Edinburgh University.

"Knowledge is a public good and increases in value as the number of people possessing it increases."
—John Wilbanks is vice president of science at Creative Commons.

The "ecologists"

"The scale of the human socio-economic-political complex system is so large that it seriously interferes with the biospheric complex system upon which it is wholly dependant, and cultural evolution has been too slow to deal effectively with the resulting crisis."
—Paul R. Ehrlich is president of the Center for Conservation Biology at Stanford University.

"Humans have a tendency to fall prey to the illusion that their economy is at the very center of the universe, forgetting that the biosphere is what ultimately sustains all systems, both man-made and natural. In this sense, 'environmental issues' are not about saving the planet—it will always survive and evolve with new combinations of atom—but about the prosperous development of our own species."
—Carl Folke is the science director of the Stockholm Resilience Centre at Stockholm University.

The distinction could not be more clear between the pessimism of the ecologists and the optimism of the economists. This distinction is not necessarily due to the focus of study, either. For example, both Fogel and the "ecologists" mentioned above are concerned with human well-being and the future of our species. And I could just as easily argue that (like Fogel's statement about the responsiveness of the human body) the Earth and its ecosystems are extremely flexible and resilient, which gives me confidence that they will persist into the future.

Ecology seems to have adapted the "crisis discpline" perspective of conservation biology and the environmental movement while economics has adapted the "continual growth" perspective of the financial system and development sector. This divide is definitely rooted in history (for an excellent example, see the Simon-Ehrlich wager).

Our disciplinary norms constrain our ability to clearly and objectively answer the science (with a lowercase "s") questions that we ask. But Science (with a capital "S") progresses as individual perspectives contrast with one another, evidence accumulates, and time progresses. I wonder if efficient progress is made by bouncing back and forth between disciplinary extremes or whether answering questions from an interdisciplinary perspective could facilitate faster insight?

Eric Berlow: How complexity leads to simplicity

2011-04-18T15:14:00.000-07:00

complexity != complicated;

Eric Berlow: How complexity leads to simplicity | Video on TED.com

Range and change

2011-04-18T10:49:00.000-07:00

I have been reading a lot about biogeography recently, which has made me realize the importance of the distribution of plants and animals for some pretty influential topics (biodiversity, climate change, infectious disease, invasive species) as well as the large gaps in our abilities to predict these distributions.

Most of biogoegraphy comes back to a simple question: Why are certain plants and animals found in some places while others are not? This question is set in the backdrop of some volatile conditions, as described nicely in Dennis McCarthy's book, Here be Dragons:
"Continents rift along volcanic ridges; oceans flow into gaps; islands spew into existence from volcanoes of the deep; and mountains thrust upward as continental regions converge and plates begin to fold. These earthly upheavals lead to changes in climates: sea levels rise or fall, glaciers advance or retreat; deserts become fertile; or rainforests turn barren. The evolving landscapes and new barriers divide various species of plants and animals, insects and fish, isolating certain groups in new environments, pushing them into different climes, mixing them with new predators and food sources, creating new hazards and eliminating old ones. Organisms must continuously adapt or perish on this always dynamic and often violent Earth."

The environment is always changing, creating risky situations for organisms (Loarie et al. 2009). Individual strategies and responses to cope with change (e.g., moving, adapting, or dying) determine the distribution of the species. The dynamics of these responses in space and time - in other words, the range and change - influence the spread of infectious diseases and invasive species as well as the configuration of local ecosystems.

Range
Organisms have many physiological mechanisms to deal with multiple limiting resources, which define the species niche. The limits of a species distribution are defined by these conditions. However, there is often not a definitive boundary between a suitable environment and an unsuitable environment. In reality, these boundaries are very fuzzy due to individual variation in dispersal, establishment, and survival. Individual variations influence a species niche and how they relate to other species (Clark 2010). Examining the subtle complexities that define the fuzziness at species' limits could be useful for predicting whether an organism will live in one place versus another.

Change
While the distribution of a species is a very spatial issue, the impacts and risks of changes in species distributions for humans are primarily temporal. How quickly and effectively species respond to changes in their environment will determine whether the species will spread or decline. As we increase our understanding of which variables are important for the geographic distribution of a species, we need to continue to ask how predictable are these patterns (Pearson and Dawson 2003)? Examining the distribution of responses within a population could be useful for predicting the species-level response to changes in time. For example, individual responses to short-term weather variation could illuminate population responses to long-term changes in climate.

Do we think we know more than we do?

2011-03-15T08:34:00.000-07:00

I saw this quote today from Proofiness by Charles Seife (Viking, 2010, pp. 54-55):

Our minds revolt at the idea of randomness. ... Our minds, trying to make order out of chaos, play tricks on us. Casinos make so much money because they exploit this failure of our brains. ... Our minds seize on any brief run of good or bad luck and give it significance by thinking that it heralds a pattern to be exploited. Unfortunately, the randomness of the dice and of the slot machine ensure [sic] that there’s no reality to these patterns at all. Each roll of the die, each pull of the lever gives a result that is totally unrelated to the events that came before it. That’s the definition of random: there’s no relationship, no pattern there to be discovered. Yet our brains simply refuse to accept this fact. This is randumbness: insisting that there is order where there is only chaos – creating a pattern where there is none to see.

and it made me think about determinism versus stochasticity. This issue is something that ecologists have to deal with often since ecology occupies an interesting place between a deterministic physical system and a random economic system (Bjørnstad and Grenfell 2001).

Are ecologists fooled by randomness? I think more often than not we fall into the trap of determinism, but determinism certainly exists (much more so in ecological systems than in economic systems). Really, what this comes down to is a question of how certain our predictions can be. Great stochasticity leads to great forecast uncertainty, which means that the information that we have gives us less information than we think it does (Clark et al. 2001).

Derman-Wilmott Oath

2011-02-11T12:36:00.000-08:00

(aka the model maker's Hippocratic oath)

Emanuel Derman and Paul Wilmott

Aleks Jakulin and Andrew Gelman over at the Statistical Modeling, Causal Inference, and Social Science blog report on the "model maker's Hippocratic Oath", a set of pragmatic rules-to-live-by for anyone who builds models. With modification by Gelman, they originate from an article by Emanuel Derman and Paul Wilmott at Bloomberg Businessweek (which is why I have called them the "Derman-Wilmott Oath"). You can check out the links above for associated discussion and content, but I really like them and think that they'd be useful for any scientist, so without further ado...

The Derman-Wilmott Oath

I will remember that I didn't make the world and that it doesn't satisfy my equations.
Though I will use models boldly to estimate values, I will not be overly impressed by mathematics.
I will never sacrifice reality for elegance without explaining why I have done so. Nor will I give the people who use my model false comfort about its accuracy. Instead, I will make explicit its assumptions and oversights.
I understand that my work may have enormous effects on society and the economy, many of them beyond my comprehension.
[added by Mark Palko:] Our model only describes the data we used to build it; if you go outside of that range, you do so at your own risk.

Very nice.

Chinese drought, agriculture, and nitrogen

2011-02-04T09:59:00.000-08:00

The NYTimes reports that Northern China is experiencing a drought, which is bad news for their winter wheat crop. Nearly all of China's wheat supply is domestic (avg. 97% since 2000), so if something goes wrong (like the occurence a drought), that could spell t-r-o-u-b-l-e (or at least i-m-p-o-r-t). But, how did China get here?

In 1961, China produced only 6% of the world's wheat, whereas today they produce nearly 20%. Yet, the amount of area that China dedicates to wheat production hasn't changed since 1961 (F[1,47] = 0.0013, P = 0.97). China (shown in Blue) has gone from underperforming in wheat yields (production per area) to overperforming when compared to the world average (shown in Orange) and the United States (shown in Red).

Wheat yields (tonnes / ha)

How did they do that? Simply, with nitrogen. China has dramatically and significantly increased their average fertilization rates per area since the 1960s (F[1,46] = 1213, P < 0.001). In fact, this rate of increase is significantly greater than the rate of increase for the world average and the United States!

Avg. fertilization rate (kg N / ha)

And we have evidence that the 2008 fertilization rate estimate of 200 kgN/ha is an underestimate for wheat in China (it is just the national average for all crops). The actual value could be closer to 300-350 kgN/ha (Zhao et al. [2006] Agron. J.; Ju et al. [2004] Ambio), or almost 4 times the global average!

What is the point here? The NYTimes suggests that China's drought problems are largely their own, since they depend heavily on their own crop production and an increase in imports wouldn't substantially affect the United States' market (although, admittedly, Chinese imports could affect prices for poorer countries). Beyond this, though, China's agricultural choices certainly impact the entire world. The consequences of China's dependence on their own crop supplies (for "national security reasons") extend beyond the impact on global trade. China's continual reliance on increasing N inputs seems unsustainable and will likely contribute to increased emissions of agricultural greenhouse gases. Perhaps, in a different political and social situation, China could import wheat from regions of the world that are actually successful at producing high-yielding wheat without intense N inputs?

Salaries in the liberal arts

2011-02-03T09:26:00.000-08:00

I have been blogging a lot about academia recently because I would like to build a career at a college or university and because I have just been very interested in academic topics recently. To continue with this trend, I analyzed some information about salaries for professors at small liberal arts colleges (the type of place where I did my undergraduate) and learned some interesting things.

First, professor salaries increase but also become more variable with academic rank.

The first half of that sentence is not surprising. But, the second part is very interesting. The coefficient of variation for professor salaries increases from 15.5% for Assistant Professors to 23.1% for Full Professors, meaning that most people enter in at a similar salary, but professors at some institutions have more upward mobility.

Another interesting finding based on the density distributions of the data is that the distributions are bimodal. There seems to be an upper tier of liberal arts colleges where professors are paid more across the board.

Second, professor salary is strongly correlated with the US News college rankings.

I think that this relationship relates to perceived prestige. The ranking correlated better than either tuition or acceptance rate alone. There are many variables that go into the US News rankings. From their website:
Academic reputation: 22.5%
Student selectivity: 15%
Faculty resources: 20%
Graduation rates: 20%
Financial resources: 10%
Alumni giving: 5%
Graduate performance: 7.5%
Of these, faculty salary only makes up 7% of the total weight, and total academic finances (e.g., endowment) only makes up 10%. There is more going on here than just money. The overall prestige of the institution seems to weigh heavily on salary for liberal arts college professors.

Third, gender disparities increase with professor rank also.

Across the board, men are paid more than women. But, the greatest disparity is seen for Full Professors, where men make 5% more annually than women.

Click here for a Google Spreadsheet of the full dataset.

Futile doctorate?

2011-01-27T09:20:00.000-08:00

I just got around to reading the Economist's "The Disposable Academic" article from last month. I find it to be a rant from a former PhD student who was unlucky in their quest for a job in their field. Regardless, the points that they raise are interesting.

Their issue comes back to a question of whether working hard affords a person success or not. Clearly, completing a PhD demands dedication, hard work, and often years of one's life. The author considers doing a PhD "a waste of time" because

the number of PhDs is much greater than the number of academic jobs, and
obtaining a job outside of or within academia is not guaranteed by completing a PhD.

For the author, there is just too much chance in it. I do not see how this is much different from any other line of work. There are few jobs in academia and, consequently, those jobs either go to the most academically-productive or the luckiest PhDs (i.e., those who are in the right place at the right time), or maybe just to the PhDs who are both.

Just as going to law school does not guarantee that one will pass the bar or become a lawyer, or getting into medical school does not guarantee that one will become a doctor, going to graduate school does not guarantee one a position as a faculty member at a university. I would guess that the ratios of success/failure are at least comparable in those other fields. If so, then getting a PhD is no different than any other professional field. Or even better yet, how about a field like acting, where only a small fraction of those who attempt it (and maybe even fewer who actually study it for four years in college) become successful? That does not mean that trying to become an actor is a waste of time.

Only the PhD is a waste of time?
"Many of those who embark on a PhD are the smartest in their class and will have been the best at everything they have done. ... few will be willing to accept that the system they are entering could be designed for the benefit of others, that even hard work and brilliance may well not be enough to succeed, and that they would be better off doing something else." The assumption that one would be better off doing something else just because they did not obtain an academic job right out of their doctorate is only justified by the assumption that one SHOULD obtain a job right out of their doctorate given the amount of work they have completed. Accepting the reality that there are always those who are better than you at a given task/profession (as well as those who are not as good) is a part of the simple realization that there is more to success than working hard and there is more to happiness than extreme success. This is not something that is exclusive to the PhD.

By the numbers
In the United States in 2009 across all fields, 51.7% of graduating PhDs had "definite employment commitments" in academia, and 69.2% had definite commitments in something (source: National Science Foundation). That looks pretty good to me. The author cites statistics that say only 16% of new PhDs get tenure-track jobs (based on the estimate of 100,000 new PhDs and "only 16,000 new professorships"). That would be 31% of PhDs who go into academia! It is not so surprising to me that only the top tier would get a professorship straight out of their PhD. The others could fall into many categories: research academic positions (of which there are MANY; "professor" is not the only academic job), adjunct professorships, or postdocs. And out of the 48% who don't go into academia there are certainly those who didn't want an academic job in the first place! According to the National Research Council, on average only 35% (±16% SD) of graduate students have "definite plans" for an academic career. Those numbers make it look like nearly all of those who had definite plans got some sort of job in academia.

Advising in graduate school

2011-01-26T10:58:00.000-08:00

Today's edition of Nature has a wonderful editorial about having a successful student-advisor relationship in graduate school. They emphasize that the burden is not on the advisor to make a student succeed in graduate school... it is on the student herself. Or, at least it is a two-way road.

The fact that "it is my thesis and I need to be the one doing things with it" is a fact that I think I learned pretty quickly during my Master's degree, but it is not something that I think most people understand when they are fresh out of their undergraduate. I do not know what factors helped me to realize it, but it is an important idea for new graduate students to get. I guess the point is that one's successes and/or failures are their own (maybe with some influence from Lady Luck).

From the article:
"One of the secrets of looking after your adviser is working out what they want — and what most advisers want is a student who comes to them with suggestions and solutions as well as problems, gets things done and makes the job of advising easier."

Am I lactose intolerant?

2010-12-22T09:31:00.000-08:00

Lactose intolerance is a common condition; 70% of humans experience lactose intolerance worldwide, with abdominal discomfort, bloating, gas, and diarrhea coming from the consumption of dairy products (like milk, cream, cheese, or yogurt). But, its prevalence varies widely. I have a hunch that I might be lactose intolerant, but I do not know, so I'm going to learn a little bit about lactose intolerance and do a study to assess the correlation between my eating habits and abdominal issues.

Lactase
Lactase is an enzyme that is located in the microvilli of the small intestine. When lactose is ingested, lactase splits and hydrolyzes the molecules into glucose and galactose, which can be carried across cell membranes and removed from the intestine.

Intolerance/deficiency
Deficiency of lactase enzymes (hypolactasia) causes these molecules to remain unabsorbed. This concentration gradient causes fluid to osmotically enter the bowel lumen, causing nearly 3 times the "normal" amount of fluid to be present in the intestine.

Secondarily, unabsorbed lactose enters the colon. The colon contains high densities of bacteria (~10 trillion bacterial cells per mL luminal contents) and can contain a high diversity of species (>500 species identified). The intestinal microflora ferment the unabsorbed lactose, producing carbon dioxide gas.

These two processes cause the common symptoms of lactose intolerance: bloating, gas, and loose stool. Symptoms often present 2 hours after ingestion and the most common levels required for a reaction are around 8-12 g (or oz) of dairy.

"Normal" human lactase activity
Lactase activity decreases naturally after weaning in all mammals. For humans, there is a continuous decline in lactase activity throughout life. However, different human populations have highly variable incidence of intolerance. Europeans (my ancestors) have very high tolerance (around 5% hypolactasia). It is possible that nearly 15-20% of people in the United States have hypolactasia, while African and Asian populations tend have a very high prevalence (nearing 100% hypolactasia).

High tolerance in European populations is thought to be the result of their cultural/evolutionary history. These populations continued the use of dairy products after weaning, leading to an evolutionary advantage for individuals who could tolerate lactose. There is evidence against the alternative hypothesis that these populations used high amounts of dairy products because they had a pre-existing tolerance (Burger et al. 2007).

Why I think I might be intolerant
I think that I get stomach aches more often after eating dairy. (That is reason enough to do a study!) If the culture-historical hypothesis is true, a correlated hypothesis could be that a reduction in dairy use could induce lactose intolerance (okay, I know I'm confounding an entire population and an individual... but bear with me). I was vegan (strictly no dairy) for 5 years before I started eating dairy again a year or two ago. If I had a predisposition to hypolactasia, there is a chance that maybe my veganism triggered my symptoms more quickly (caused my lactase activity to decline more abruptly).

Second, symptoms are closely related to the intestinal microflora. If I've had any change in my microflora, symptoms could change. I don't know much about intestinal microflora but it is VERY interesting to me. Our bodies are like small ecosystems that sustain large populations of microorganisms. Just like natural communities (e.g., a grassland) can have dramatic species changes, especially with disturbance, human bodies can have species turnover also, with consequences for the "ecosystem properties" (in this case, the prevention of bloating in my body).

Third, I am primarily of European descent. Since these populations have high levels of tolerance, it also can take longer for lactase levels to drop to the point where hypolactasia presents, sometimes through their 20s, which is my age group.

The study
So, we're going to do a study. I'm going to record what I eat every day, the amount of dairy that I consume (on a scale from 0-3), and the degree of stomach issues (on a scale from 0-10). Then we're going to statistically analyze the data and see where we stand. Am I lactose intolerant? We'll see....

References:
Burger, J., Kirchner, M., Bramanti, B., Haak, W., & Thomas, M. (2007). Absence of the lactase-persistence-associated allele in early Neolithic Europeans Proceedings of the National Academy of Sciences, 104 (10), 3736-3741 DOI: 10.1073/pnas.0607187104

Lomer MC, Parkes GC, & Sanderson JD (2008). Review article: lactose intolerance in clinical practice--myths and realities. Alimentary pharmacology & therapeutics, 27 (2), 93-103 PMID: 17956597

Swagerty DL Jr, Walling AD, & Klein RM (2002). Lactose intolerance. American family physician, 65 (9), 1845-50 PMID: 12018807

Real-time results

Ecological games?

2010-12-13T10:57:00.000-08:00

In San Francisco this week for the American Geophysical Union meeting, so I have some down time to blog!

The New York Times reports about the possibility of using video games to solve scientific problems. Gamers spend hours in virtual worlds solving puzzles: "creating Wikipedia took eight years and 100 million hours of work, but that’s only half the number of hours spent in a single week by people playing World of Warcraft." So, why not harness that effort and focus to solve real-world issues in a virtual-world setting?

This question is easily related to social and psychological topics. But, can video games solve ecological problems too?

Games could provide data on complex networks and connections and interactions between organisms. We could examine diversity in different communities that have "evolved" much like natural communities, with disturbances and multiple influences from external conditions, and learn about community stability and stochasticity (are more stable communities likely to persist? or is there some randomness in persistence). Games could even offer insight on the spread of organisms. And what a great way to explore organismal and community responses to climate/environmental change! I think that we could learn a lot about classic "natural" ecological questions in a virtual game setting.... maybe it is time for me to learn how to make video games.

Brewing biology II: yeast recap

2010-12-02T07:34:00.000-08:00

I would say that my monitoring of yeast activity was a success, despite some issues in data collection (I missed peak activity measurements). However, I was able to learn some about yeast growth in beer... basically, there are a lot of yeast cells. But they aren't packed to capacity. Rather, it looks like they just eventually run out of food (at which point the sweet wort has been converted to dry, alcoholic beer).

Growth begins exponentially; since yeast are living the "good life" in the sugar-rich wort they are reproducing rapidly. This appears continuous because individual yeast cells divide at different times, causing population growth to be smooth in reality (Briggs et al. 2004). But, as time continues, sugar concentrations decrease and by-products of metabolism increase to "an inhibitory concentration." At this point, the growth rate is dependent on the amount of sugars and enzyme kinetics, which is represented mathematically as
μ = μmax - [S]/(K+[S]),
where μ is the growth rate, μmax is the maximum growth rate, [S] is the sugar concentration, and K is the half saturation constant (from Michaelis-Menten).

Thus begins the "stationary phase" where the population remains at a constant level (my beer is almost there as of 12/02/10). But not all is lost, as Briggs et al. (2004) note, "this apparent inactivity is illusory since metabolic activities of various kinds continue." When no further growth is possible (because of substrate limitation) cells undergo changes that maximize their chances of survival: shutting down unnecessary metabolic pathways, increased stress tolerance, thickened cell walls to prevent enzymatic degradation, etc. All of these changes begin based on the yeast sensing changes in their environment.

The stationary phase is what makes bottle carbonation possible. The yeast that remin in suspension at bottling time can resume growth with the addition of more sugars.

I will repeat these measures in future brews which will allow me to ask some other interesting questions related to what controls these rates, like those that Noam raised in his modeling work on the former post. How do different yeast strains behave? How does the composition of the wort affect yeast activity? How does temperature play into this? Is it important to control all of these conditions in a brewing environment to produce good and consistent beer (i.e., how sensitive are yeast to changes in these variables)? These are all awesome questions that are probably being explored by brew scientists and biochemists currently. Maybe we can contribute a little bit to that...

Reference: Briggs, DE, et al. 2004. Brewing: Science and Practice. Woodhead Publishing Limited, Cambridge.

Why has American beer production stagnated?

2010-12-01T10:21:00.000-08:00

The commercial side of beer production is related to the biological aspects. Brewing styles change with market conditions, which alters the biological/scientific processes that are utilized in production. Additionally, production and consumption of beer might be an indicator for social conditions.

To continue with the brewing theme, I investigated US production and consumption of beer, which has produced some interesting results.

1. US beer production has slowed and decreased since around 1980.

The blue line shows US beer production (tonnes/yr), and the red line shows US beer consumption (tonnes/yr, calculated with FAO import/export data).The first question that I had from this is, why is US production decreasing relative to consumption? That leads to the second insight:

2. Imports of international beers have increased dramatically in the last 20 years.

This is not surprising, as the economy is globalized and people are introduced to new, international styles. Could all of these findings be due to changes in population?

3. Beer consumption (kg/yr) per person over 20 years old has decreased since 1980.

That is a 21% decrease since 1981... This is surprising. Why would this happen? Are people happier and thus drinking less? Maybe more people are making their own? Or, perhaps, these values do not account for craft breweries (which made up 4% of beer sales in 2009, and have been steadily increasing with around 9% growth per year). Maybe Americans are tired of the major breweries.

Other interesting findings:
4. Beer hoppiness has not changed with trends in production/consumption.

There is no trend in the national consumption of hops relative to production of beer (hops / beer = hoppiness). Hoppiness (the bittering part of beer) nationally seems to be sensitive to changes in hop production (which is tied to weather, etc.) more than changes in beer production. There is no discernible impact of craft brewing (which tends to be more hoppy) on the national scale, since it is currently such a small market share.

I am still left with the question, Why has American beer production stagnated? As a follow up question to this, Will craft brewing turn this trend around?

Brewing biology I: yeast population biology

2010-11-29T20:21:00.000-08:00

Yeast fermenting.

Brewing beer is an extremely scientific process, from the extraction of sugars to the production of alcohol. To produce beer, Saccharomyces cerevisiae yeast convert the sugars extracted from grains (like barley: Hordeum vulgare) into ethanol through fermentation. I brew beer and I think that this is a really neat process. Aside from the biochemistry involved, there is also some very interesting population biology. Yeast reproduce inside the fermentation vessel in response to the large amount of available food (sugars). But, do fermentation vessels have a carrying capacity? That is, is there a point where there are still enough sugars but just not enough space.

In an attempt to answer this question, I am monitoring yeast activity inside a beer fermentation vessel. This beer fermentation vessel, to be exact:

As a surrogate for measuring yeast population (since I do not have the resources at home to directly estimate yeast populations), I am counting airlock bubbles (the airlock is the tube device at the top of the vessel), which will give me a good indicator of yeast activity and, subsequently, yeast population size. With the instantaneous measure of yeast activity (bubbles per minute), I am estimating cumulative yeast activity (total bubbles produced). Ideally this will tell me something about the biology of what is going on inside there!

Below are the continuously updated charts plotting yeast activity (as I measure it). Once fermentation is complete, I will repost with an analysis of the results.

Some notes on the data collection: There are already some collection anomalies. First, there is a dip in yeast activity at 17 hours because I chilled the vessel to reduce the fermentation temperature (it was getting a little too hot in there). After 30 hours, there is a 36 hour break in data collection because of a bubble "blow over" which required me to remove the airlock and install a removal tube. But, now data collection is back on!

Credibility v. confidence: how much does an elephant weigh?

2010-11-03T11:04:00.000-07:00

Today I was asked "What is a credible interval? Is it directly comparable to a confidence interval?" These are questions that plague statisticians and make interpreting results somewhat tricky.

First, here are some definitions (they might help us understand what is going on):

95% Confidence interval: If we take more samples, 95% of the time the ("real") average value will be within the interval that we calculate. This interval assumes that the parameter (in this case, the average) is fixed and that the observed data are uncertain.
95% Credible interval: Given the data, there is a 95% probability that the value is within the interval. This interval assumes that the data are fixed (i.e., real) and that the parameter is uncertain.

Are the different intervals directly comparable? Sometimes. In one sense, they both represent uncertainty around a measure. It is possible to calculate credible intervals that equal confidence intervals. However, they do not "say" the same things. In fact, most of the time what we are most interested in is a credible interval and people often confuse the two terms by interpreting confidence intervals as credible intervals.

An illustration: How much does an Asian elephant weigh?

We could estimate how much an Asian elephant weighs by taking a sample of the population. Luckily for us, Hile et al. (1997) already weighed 63 adult Asian elephants. I calculated the 95% confidence interval and credible interval for this data set. For both calculations, the average weight was 3658 kg.

I used the equation x_bar ± 1.96*(σ/√(n)) to calculate the 95% confidence interval. I estimated the confidence interval as (3501, 3816), shown in black lines below. This means that, if I went out and weighed more elephants, 95% of the time the average weight would be in the interval.

Then, I calculated the 95% credible interval with a Gibbs sampling routine (with JAGS in R; details available upon request). I estimated the credible interval as (2406, 4911), shown in red lines below. This means that 95% of the elephants that I measured fall within the interval.

The reason the credible interval is so much larger than the confidence interval in this case is because the confidence interval is concerned with the average elephant weight - we assume that some "real" weight exists and that any deviation outside of the interval is due to other factors. Conversely, I estimated the credible interval as the actual observed weights - we assume that each weight is "real" and that variation within the interval is due to other factors.

They are both measures of uncertainty, but their interpretation depends on the question. Which interval to use mainly depends on what you hope to get from the data and your realm of inference.

For our question "how much does an Asian elephant weigh?" the credible interval is more useful: Based on our sampling, there is a 95% probability that Asian elephants weigh between 2406 and 4911 kg. For this question, the credible interval is closer to the 95% quantiles of the distribution than it is to the confidence interval. However, if we were asking "how much does the average Asian elephant weigh?" we could use the confidence interval: If we collected more samples from the greater population (or another population), 95% of the time the average weight would be between 3501 and 3816 kg. Formulating the credible interval to address this question results in a similar interval, but allows us to say: Based on our sampling, there is a 95% probability that the average elephant weight is between 3498 and 3818 kg.

How to get published

2010-10-22T15:25:00.000-07:00

Writing is hard. Since I'm in the process of writing and attempting to get some work published, I think it is timely to reflect on how to get published.

There is a nice article in last week's Nature about how to "publish like a pro" (Powell 2010). The article contains some great suggestions for new authors like me, including: 1) be resilient, 2) find the right journal, an appropriate editor and reviewers, and 3) identify an appropriate audience.

Clearly, a lot of this starts in the beginning by designing a good research project, but some of it just needs to come out in the writing process - good writing can make an average data set publishable, and bad writing can make a stellar data set unpublishable.

My favorite point from the article:

[L]ess is more for junior scientists crafting manuscripts. The introduction need not cite every background article gathered, the results section should not archive every piece of data ever collected, and the discussion is not a treatise on the paper’s subject. The writer must be selective, choosing only the references, data points and arguments that bolster the particular question at hand.

Writing is definitely an art. Unfortunately many scientists are not too good at it (as my English major fiancée likes to point out). Luckily, articles like these give new scientists hope (bravo, Nature). Learning some of the specific "tricks" of scientific publishing (maybe from a good mentor) plus taking a good writing course should be on every scientist's to-do list.

What do your priors say?

2010-10-19T14:29:00.000-07:00

I am trying to estimate a parameter in a model with MCMC. Ok, that might be scary. Before you run away, take a look at this picture:

Now for the messy details (if anyone is still here). I'm thinking that a Beta distribution is the best way to make a prior for a value with an average of 0.01 and a range of 0.003 to 0.03. So I've created a Beta distribution where the shape parameters are α = 3 and β = 230 (see the density plot above). I have drawn the 50%ile of the distribution as a solid black line, and the 95%iles as dashed black lines. The red lines are the values that I am trying to approximate.

So, it is not perfect. How much will this introduce bias? Anyone have any ideas about how to get better estimates for the shape parameters?

Changing range boundaries in a changing environment

2010-10-14T13:39:00.000-07:00

There are two neat articles about species distributions in the October issue of Ecology Letters, and they complement one another in an interesting way.

The first, by Murphy et al. (2010), examines the abundance of eastern North American tree species at their northern and southern limits. The idea here is that abundance at the limits may give us some information about the potential of a species to respond to changes in the environment. High abundance at the northern limits could increase capacity to migrate at the "advancing edge" while high abundance at the southern limits could allow populations to survive changes at the "trailing edge" of environmental change.

Based on their analysis, most species had greater abundances per area at the northern portions of their range (relative numbers within cells) suggesting higher "climate suitability" in the north. But, > 50% of the species also occupied fewer areas (lower "occupancy" - the number of cells) at both limits, which the authors say could negatively affect the species' ability to expand. So, they concluded that

"the most common response is likely to involve range erosion in the south and limited range expansion in the north, possibly leading to an overall reduction in range size."

Why might that be? The explanation is partially explained in the study by Burton et al. (2010), who discuss the importance of evolution at the advancing edge. These authors modeled dispersal, reproduction, and competitive ability and found that dispersal and reproduction are selected for at the advancing front (making the populations more suited for survival in a novel environment/area), at the cost of decreased competitive ability. This means that, when a competing species is present, a species ability to expand could be restricted because the edge populations are trying to maximize their ability to "get by" in their new surroundings.

The articles mesh nicely because the empirical observations of abundance by Murphy et al. (2010) can be directly related to the theoretical predictions by Burton et al. (2010) to explain how a species distribution could respond to changes in the environment. For these authors, I recommend collaboration to do some mechanistic modeling of North American tree species!

Murphy, H., VanDerWal, J., & Lovett-Doust, J. (2010). Signatures of range expansion and erosion in eastern North American trees Ecology Letters, 13 (10), 1233-1244 DOI: 10.1111/j.1461-0248.2010.01526.x

Burton, O., Phillips, B., & Travis, J. (2010). Trade-offs and the evolution of life-histories during range expansion Ecology Letters, 13 (10), 1210-1220 DOI: 10.1111/j.1461-0248.2010.01505.x

Bee decline by virus and parasite co-infection

2010-10-08T12:47:00.000-07:00

Bromenshenk, J., Henderson, C., Wick, C., Stanford, M., Zulich, A., Jabbour, R., Deshpande, S., McCubbin, P., Seccomb, R., Welch, P., Williams, T., Firth, D., Skowronski, E., Lehmann, M., Bilimoria, S., Gress, J., Wanner, K., & Cramer, R. (2010). Iridovirus and Microsporidian Linked to Honey Bee Colony Decline PLoS ONE, 5 (10) DOI: 10.1371/journal.pone.0013181

A group of United States researchers think they have identified the agents that cause Colony Collapse Disorder (CCD) in US bee colonies: an RNA virus (invertebrate iridescent virus, IIV) and a fungal parasite (Nosema sp.). The researchers came to this conclusion by 1.) analyzing samples from colonies around the country and 2.) conducting laboratory survival tests with infected bees.

Based on their national sampling, the researchers found a high prevalence of both the virus and the parasite in colonies that had either collapsed or were failing. The virus was present in 100% of collapsed and failing colonies, and the parasite was present in 90% of failing colonies. This contrasted with the strong colonies, which had generally lower prevalences of the infections than the others.

Figure 1. Concentration and prevalence of different infections in US bee colonies, showing that IIV and Nosema infections are most prevalent and concentrated in collapsed (n = 8) and failing (n = 11) colonies, as compared to strong (n = 13) colonies, based on data from Bromenshenk et al. (2010).

The infections were still both present in over half of the strong colonies. So why do they think that these infections caused CCD? Some evidence comes from the relative concentration of the organisms in the colonies (based on measured peptide counts; Figure 1). Even though the infections were present in the strong colonies, the highest concentrations of the infections were found in the failing colonies. This suggests that the diseases are having a greater effect on the failing colonies than the strong colonies. Additionally, the relatively low concentrations and prevalences of other viruses, regardless of colony status (Figure 1, panel 3), implicate IIV as having an important role.

The researchers backed this up with a nice laboratory study where they infected bees with one or both of the agents and compared their long-term survival to a control group. The control (non-infected) bees had the highest survival rate after two weeks while the combination of Nosema and IIV infection caused the greatest mortality (Figure 2).

Figure 2. Survival of bees infected with IIV, Nosema, or both, compared to a control group, from Bromenshenk et al. (2010).

What is it about co-infection with these two organisms that leads to colony collapse? According to Bromenshenk et al. (2010):
"The mechanism by which these two pathogens interact to potentially increase bee mortality is unknown. It may be that damage to gut epithelial and other host cells by the N. ceranae polar tube allows more robust entry of the virus. Alternatively, replication of N. ceranae in honey bee cells may cause a decrease in the bees' ability to ward off viral infections that normally could be controlled."

This is an interesting study, though, because the cause of CCD in the United States has been a mystery; researchers had not been able to identify a single cause. This evidence suggests that multiple infections could play a role. Although this explanation adds a level of complexity, it seems to have more weight than any widespread, single cause.

Chili pepper

2010-09-21T08:17:00.000-07:00

There is a cool article in the New York Times about human's fascination with chili peppers despite the burning pain that their chemical component offers.

Habanero peppers © NY Times

The author references many researchers and peer-reviewed articles in an excellent example of effective science writing. Check it out.

Evolution of ecology article

2010-09-07T14:24:00.000-07:00

Simon Levin discusses the past, present, and future of ecology in a recent opinion article for The Chronicle Review. He traces the development of the discipline through the contributions of numerous scientists (from Charles Darwin and Vito Volterra to Levin's own work), which provides a good, narrative history. He also points out ecology's natural (sometimes foundational) connections to other scientific disciplines like demography, epidemiology, systems and computational biology, and resource management.

The meaning of ecology
What I especially liked about the article was Levin's treatment of what ecology is. Levin states that:
"[e]cology is a scientific discipline, like physics or molecular biology, whose practitioners are driven by the search for patterns and process in nature. Their findings can inform political decisions about resource use, pollution, climate change, and other environmental issues; but advocacy regarding decisions about our environmental priorities is outside the discipline of ecology."

He goes on to recognize that:
"[s]till, for many people, 'ecologist' became a term applied to anyone who wanted to save the planet, or selected parts of it, which made no more sense than calling someone who marvels at the night sky an astronomer."

This succinctly summarizes my own view on the topic and raises what I think is an important issue. The boundary between the science of ecology and the use of the science seems to be particularly blurry for many people.

Levin calls for a stronger unification of ecological "science and the social sciences and humanities" (with a strong emphasis on economics). He brings up the mutual benefits of collaborations and interactions between disciplines (a truly ecological concept) with an example of his research on how banking and financial networks are essentially ecological systems that are sensitive and prone to disruption (a favorite of mine, published a few months before the financial crisis).

Some critique
This is a nice overview article. I appreciate the distinction between science and advocacy. But I feel like the distinction is contradicted somewhat by a cliche statement like "Society has become increasingly aware that we are losing crucial parts of our ecosystem, and that the activities of human beings are threatening the sustainability of the biosphere as a life-support system for humanity." The reality (i.e., the ecology) is much more complicated.

Ecology has been recognized by some as a crisis discipline, especially since its association by the general public with environmental issues (e.g., works by Rachel Carson and Paul Ehrlich). This is a restrictive (and I would argue unscientific) perspective. It is also one that Levin reasons against in his conclusion, when he discusses the utility of an ecological approach to understand any type of complex system (disease, finance, social systems, etc.). That is the ecology that I know.

Another quote of note: "... ecology was the first subdiscipline of biology to become deeply quantitative."

Back to the important questions

2010-09-02T12:22:00.000-07:00

I think that identifying important questions is really important for any scientific venture. I think it is also pretty difficult, since most good questions require some sort of inspiration. Luckily, the internet makes collaborative brainstorming much more feasible.

So, I'd like to start an ongoing discussion dedicated to important questions in biology. This will be an interactive blog exercise!

For it to work, just enter a question topic into the form below and send it off... The response will be stored in a spreadsheet which I will check periodically. You can come back and enter topics/questions as often as you'd like. There will be a permanent link on the right side bar for easy access.

There are only two rules:
1. It must be a question that can be answered scientifically.
2. It must be somehow related to biology.

I'll work with the responses, do some visualization, and we'll start seeing what people think are important topics in biology. Thanks (and hooray for interactive research)!

For some inspiration, check out the 2005 anniversary edition of Science Magazine:
So much more to know... and What we don't know

Link to view responses

Loading...

The ecosystems within us

2010-08-18T12:19:00.000-07:00

De Filippo C, Cavalieri D, Di Paola M, Ramazzotti M, Poullet JB, Massart S, Collini S, Pieraccini G, & Lionetti P (2010). Impact of diet in shaping gut microbiota revealed by a comparative study in children from Europe and rural Africa. Proceedings of the National Academy of Sciences of the United States of America PMID: 20679230

In the guts of each of us there are trillions of microbes. They provide us (their host) with "enhanced metabolic capabilities, protection against pathogens, education of the immune system, and modulation of gastrointestinal development" (De Filippo et al. 2010). In fact, the human body may be one of the most species rich ecosystems in the world.

The diversity of these organisms can play a role in the future development of disease; researchers have found connections between imbalances in gut microbiota and inflammatory bowel disease and obesity. So, what makes our guts so diverse? As De Filippo et al. (2010) report, one main factor appears to be diet.

They examined the gut organisms of children (ages 1-6) from a rural village in Burkina Faso (BF) and compared them to the gut organisms of children (ages 1-6) from Florence, Italy. The diet of the rural village consisted "mainly of cereals, legumes, and vegetables" with high levels of fiber which resembles the diet that humans evolved with before the agricultural revolution, while the diet of the European city is very westernized; "high in animal protein, sugar, starch, and fat and low in fiber."

Almost 95% of the samples from both populations belonged to four common bacterial phyla (Actinobacteria, Bacteroidetes, Firmicutes, and Proteobacteria). But, there were substantial differences in their distribution:

Table 1. Average proportions (%) of common bacterial phyla in the guts of two populations (traditional: BF; western: Italy).

Phyla/Population	Burkina Faso (BF)	Italy
Actinobacteria	10.1	6.7
Bacteroidetes	57.7	22.4
Firmicutes	27.3	63.7
Proteobacteria	0.8	6.7

What does that mean? The two populations experienced dramatically different bacterial colonization of the human gut. Primarily, the traditional population had much more Bacteroidetes, while the western population had much more Firmicutes. De Filippo et al. (2010) cite previous studies which have found that a high Firmicutes:Bacteroidetes ratio has been linked to obesity in humans, and the bacterial species associated with a high-fat and high-sugar diet promote obesity in mice.

Interestingly, the gut bacteria of the two populations were similar in the youngest subjects of the study, due primarily to breast feeding at a young age. Once solid foods were introduced, though, the two populations diverged in their resident microbes.

Who is richer?
The Burkina Faso samples had higher gut species richness and biodiversity (Shannon index) levels than their European counterparts, which De Filippo et al. (2010) attributed to a high-fiber diet in the BF samples. They note that in "[b]oth the Western world and in developing countries diets rich in fat, protein, and sugar, together with reduced intake of unabsorbale fibers, are associated with a rapid increase in the incidence of noninfectious intestinal diseases."

Assuming that the microbes and the host have coevolved throughout history, we can think about some of the benefits that come from the bacteria that colonize the gut. The diversity of the microbes in the guts of the BF individuals appears to play a role in preventing the establishment of pathogenic intestinal microbes like Shigella and Escherichia, which were present in significantly lower levels than in the European hosts. And the connection between exposure to microbiota and obesity and autoimmune disorders makes me want to reconsider brats and tater tots for dinner...

A question that remains for me is what does the turnover of gut microbiota look like? How long does it take to establish new communities? I think that the gut is a pretty dynamic ecosystem (probably moving at light speeds compared to the traditional terrestrial ecosystems that we think about). But, is it possible to adjust my gut bacteria for the better?

Modern ecological research: what, where, how?

2010-08-17T11:45:00.000-07:00

May, R. (1999). Unanswered questions in ecology Philosophical Transactions of the Royal Society B: Biological Sciences, 354 (1392), 1951-1959 DOI: 10.1098/rstb.1999.0534

(because I like to give myself some perspective every now and then)

01. Introduction

Just over ten years ago, Robert May published an article considering "the most important unanswered questions in ecology" (May 1999). This perspective piece offered some direction to a young field that was expanding rapidly. It focused on a few key themes,

nonlinearity,
interactions, and
complexity,

which have defined a large (huge!) body of ecological research in the last decade. For example, how does scale affect ecological processes? What about space and time? How do biogeochemical cycles work? Which species should be conserved/protected?

These three themes and the subsequent questions are important. But, I would argue that they have been examined extensively by many highly intelligent people and they are losing their luster. Certainly, other works have been written more recently on important research questions in ecology (Lawton 1999; Hastings et al. 2005; National Research Council 2009; Sherrat and Wilkinson 2009). But, where is ecology going in the next thirty years (or so)?

02. Redefining research for the '10s

Ecology is rooted in the early 1900s when doing natural science meant going out into the field for months at a time and recording analog observations. This defined the scope of questions that ecologists could ask. Now, however, individual cell phones have more processing power than national labs did in the past. Google has exploded mapping, taking away the hold that ESRI had on geography and giving the world a sense of scale. A century on, we no longer need to think at a plot level.

So, beyond specific questions, the biggest change that ecologists can make is to step away from the field. Data collection no longer requires long periods of time in remote locations. In fact, so much data already exists that there are opportunities to do research primarily from a "dry lab" (i.e., not out in the rain!).

Stepping away from the field in the 2010s necessitates a different type of question than ecologists asked in the 1910s. The transition from natural history in the mid-1800s to experimental ecology in the mid-1900s required biologists to ask more focused questions that were tailored to an experimental setup at a single site. More recently, attempts at hammering out the questions of space and scale have been done by replicating this type of study writ large at multiple sites. But, we are no longer asking if space affects biological processes or whether a global scale is important (it does, it is). The next step requires synthesis and compilation of existing data and/or global collection of information. It is simply not tractable for a single researcher at a field site to answer these types of questions.

03. Questions that are relevant

While many "classic" questions in ecology are still relevant in a general sense, many more are becoming outdated. This is because of the close interplay between questions and feasibility. While research is driven by important questions, the questions that are asked are constrained by whether we can answer them or not. Synthesis, compilation, and global data collection offer ecologists an opportunity to ask questions of particular societal relevance that have not been answered effectively in the past (due to research constraints).

The questions that I see as important in ecology revolve primarily around change. Within this category, I think that organismal physiology and species geography are particularly relevant right now. Here are a few examples.

Disease ecology
Infectious diseases change rapidly and are strongly influenced by their environment. Biologists can no longer study disease dynamics without considering their ecological context. This research need provides a wonderful opportunity for collaboration between ecologists and cellular and molecular biologists.

Organismal distribution
Why are organisms found where they are? Will a disease expand into a new region? Can we predict whether a species will invade? How will organisms shift in response to climate changes? Many of these questions depend on the organismal response to the environment.

Physiology
Organisms and species respond physiologically to changes in their environment and an individual's physiology offers impressive predictive power for many important questions at the levels of populations, species, and ecosystems. How will food production in x place respond to climate change? Why does influenza peak in the winter in the N. hemisphere? Will mountain pine beetles spread to other places in the Colorado Rockies?

Species diversity
Many interesting questions in ecology come back to issues of species diversity. Why are ecosystems so resilient to "collapse" in response to environmental change? Does it have anything to do with the species that reside there? How do different assemblages of species arise, what maintains them, and does it matter if anything (individuals, species, environment) changes?

Environmental change offers many grand natural experiments (the biggest one being our CO₂ experiment) that can be used quite effectively to learn about how the world works. As we gain a handle on the biology behind these questions, their social relevance demands that we improve our ability to forecast probable outcomes (Clark et al. 2001). Ecologists can and should use their knowledge and experience to provide information (i.e., data) to the public and to offer expert advice.

04. How to make it happen

Step away from the field
As I mentioned, ecology is in a position where the next advances will require scientists to step away from the field and put more effort towards synthesis and compilation. Collaboration is helpful with this. There are troves of data sitting in tables and figures in old (and not so old) publications with easy access in Web of Science or Google Scholar. New web databases are being created every day which offer open access to more data than could be collected in many lifetimes (e.g., GBIF, AmeriFlux). Additionally, new research initiatives like NEON will make continental and global data collection "away from the field" easier.
Drop ANOVA designs and learn some 'new' statistics
Our ability to learn from data has advanced rapidly in the last 30 years, especially with the advent of inexpensive computing power. Researchers no longer need to rely on simplistic research designs which offer limited inference. We can now learn from the data that we have with models that mean something biologically (Hobbs and Hilborn 2006). It is no longer acceptable to ask yes/no questions; answering important questions depends on estimating effects and responses, and accurately characterizing uncertainty in those estimates.
Go big or go home (See reconsideration of this point below - 8/23)
The emerging scope of ecology is to address issues with large amounts of data and at a global level. With the availability of fast and cheap technology and easy to use tools for analysis, there is no excuse for a sole reliance on plot-level studies. The most important direction for modern ecology is to go big or go home.

P.S. Also, embrace rapid/online/open-access publication. The days of 40 page monographs are over and data are no longer hard-won. The concepts are what matter and what will make a difference. Greater advances can be made by publishing important and concise documents and making information widely available on the web.

Literature cited

Clark, J.S., S.R. Carpenter, M. Barber, S. Collins, A. Dobson, et al. 2001. Ecological forecasts: An emerging imperative. Science 293:657-660.
Hastings, A., P. Arzberger, B. Bolker, S. Collins, A.R. Ives, N.A. Johnson, and M.A. Palmer. 2005. Quantitative bioscience for the 21st century. Bioscience 55:511-517.
Hobbs, N.T., and R. Hilborn. 2006. Alternatives to statistical hypothesis testing in ecology: A guide to self teaching. Ecological Applications 16:5-19.
Lawton, J.H. 1999. Are there general laws in ecology? Oikos 84:177-192.
May, R.M. 1999. Unanswered questions in ecology. Philosophical Transactions of the Royal Society B 354:1951-1959.
National Research Council (U.S.). 2009. A new biology for the 21st century. National Academies Press, Washington, DC.
Sherratt, T.N. and D.M. Wilkinson. 2009. Big questions in ecology and evolution. Oxford University Press, Inc., New York.

Sankey Diagrams in R

2010-07-30T12:48:00.000-07:00

I need to create Sankey diagrams. And I like R (because it is free and can create high-quality, publication worthy graphics). But, there was no way to create them in R. So, I modified an existing function for Matlab (called drawSankey) to work in R. I call it SankeyR!

Sankey diagrams are beautiful visual representations of the balance of in and out flows. The function takes inputs of these values and will produce various output formats (currently either bmp or pdf, or simply a plot in the R graphics device).

To officially launch the function publicly, I've created an example of flows in the global carbon cycle (data taken from Schlesinger's Biogeochemistry text [1997]).

It shows flows into and out of the biosphere and demonstrates the current imbalance due to excess fossil fuel emissions (the cause of current increases in atmospheric [CO₂]).

To decipher the abbreviations: GPP is gross primary production, Ra is autotrophic (plant) respiration, Rh is heterotrophic (animal) respiration, and LULCC is land use and land cover change.

Also, please download/distribute/use the R function, Sankey.R, and give me comments/suggestions if you have them!

UPDATE (8/10/10):
Thanks to a helpful comment from a reader, I've updated the code to accept a single input value. Also, I haven't created an installable package for the script. So, to load the function into R, you can:
1. download the R file to your computer and run source('filename') to load (I prefer source(file.choose()) for that sort of thing).
2. copy the text from the function, paste it into a new script window, and run the entire thing.

UPDATE (12/01/2011):
I've hosted the file on GitHub.

Global tree height

2010-07-26T09:54:00.000-07:00

After a two week blog-vacation...

Colorado State University scientist Michael Lefsky has used NASA satellite data to create a global map of tree canopy height (see low-res version below).

This map provides unique information because it was created with a consistent methodology (as opposed trying to compile and standardize information from elsewhere), and will offer scientists an opportunity to learn a lot about habitat structure, carbon storage, and other neat ecological variables. Check out more about it and see high-resolution images here.

Poisson plants, part II

2010-07-07T10:17:00.000-07:00

NOTE: if you want more posts like this, or want me to dive more deeply into any of these topics, please comment!

So, how many trees were in the forest? Previously, I estimated how many trees were in each meter squared (average 2.815). I also showed how the Poisson distribution can be used to estimate the discrete probability of obtaining a given number of trees, based on our random sample of 27 plots.

The easiest way to estimate how many trees are in the forest is by multiplying the average number of trees per meter squared by the total area (= 2.815 x 2077), which gave me 5846.37 trees. That is not too bad, but it would be nice to have an estimate of uncertainty with that value (would you like fries with that?!).

To estimate uncertainty, I used Gibbs sampling (a special type of Markov Chain Monte Carlo simulation) with the JAGS program in R. First, I used the observed samples shown in the previous post to estimate the Poisson distribution of tree densities (with a non-informative Gamma "prior" distribution). Then, I used the distribution of λ values to estimate the total number of trees. I traced the MCMC chain for 10,000 iterations. The resulting information allowed me to estimate the posterior distribution for the number of trees in the forest (Figure 1).

Figure 1. MCMC chains (left) and posterior distribution (right) for the total number of trees in the forest patch. Solid black line is the mean and dashed lines are the 95% credible interval for the posterior distribution. There were three chains with a burn-in of 10,000 iterations.

Posterior distributions are useful because they offer a measure of uncertainty. Previously I estimated 5846.37 trees. The hierarchical model calculated a value of 5917.33 ± 702.21 trees, with a 95% credible interval of 4616.63 to 7372.75 (which means that the probability of the actual number being in that interval is 95%). These are numbers that could be used to scale up other processes, like biomass storage, carbon sequestration, etc.

The next step would be to test this model. One approach would be to develop a competing model (maybe the Binomial distribution, instead of the Poisson? or a model that doesn't assume random tree spacing e.g. Moore P.G., Ecology 1954) and compare the results. Another would be to test my parameter estimates (particularly the λ with some more samples). I could just count all of the trees for an ultimate test.

The code/details
The hierarchical structure of this model is:
P(λ,ρ,σ|y) ∝ P(y|λ) x P(λ|ρ,σ) x P(ρ) x P(σ)
where ρ is the rate and σ is the shape of the Gamma distribution.

The JAGS model code is as follows:


model{
  for(j in 1:length(y)){
    y[j] ~ dpois(lambda[j])
    lambda[j] ~ dgamma(rate,shape)
  }

  rate ~ dgamma(0.001,0.001)
  shape ~ dgamma(0.001,0.001)

  density <- rate/shape
  trees <- density*area
}

Special thanks to Dr. Tom Hobbs for assistance with the hierarchical structure.

Poisson plants

2010-07-01T13:55:00.000-07:00

A common problem in biology is estimating a value when the actual number is unknown and unobservable. An example of this issue could be estimating how many trees are in a forest patch. The first step in approaching the problem is to obtain some observed samples and estimate how many trees one might expect to find.

With Beloit College Plant Ecophysiology students, I sampled a small (2077 m²) forest patch on Beloit College's campus.

We took 27 random samples of tree density counts (number of trees per meter squared). On average, our samples had 2.815 ± 1.210 trees per meter squared (mean ± SD). While this is nice, it is impossible to have 0.815 trees, because they are discrete units.

The Poisson distribution can offer a better estimate of how many trees are in each plot without violating our whole-tree constraint.

The Poisson
First, a little bit about the distribution (not the French fish). The Poisson is a discrete probability distribution that was developed by Siméon Poisson in 1838.

One can calculate the probability of obtaining any value (n) with the mean of our samples (2.815), which is called λ, and the probability mass function for the distribution, f(n|λ) = e^-λλⁿ / n!, where n! is a factorial.

I did this in R

dpois(seq(0,10,1),2.814814815)

and produced the grey values in the plot below. The black bars are the observed proportions, showing their (relatively) close association to the predicted values.

Figure 1. Observed proportions and predicted probabilities of tree density in the forest patch on Beloit College campus.

The modal value of this distribution is 2, with a 23.7% probability of coming up in any sample. This Poisson distribution will help me estimate the number of trees in any given plot.

Next... I need to perform the calculation for the total number of trees in the 2077 meter squared patch...

Neat data visualizations 1

2010-06-24T14:12:00.000-07:00

I am very interested in seeing how people choose to represent data. Visualizing data is at the heart of communicating information. I thought it would be appropriate to highlight examples of unique data presentation that I come across in a series of posts. The first image that I will be highlighting on the blog is from a research article on plant roots:

Cahill Jr., J.F., et al. 2010. Plants integrate information about nutrients and neighbors. Science 328: 1657.

First, the figure and legend:

Fig. 1. The annual plant A. theophrasti was planted into six combinations of soil heterogeneity (uniform, patch-center, and patch-edge) and competition (alone versus with a competitor) treatments. (A) Alone uniform, (B) alone patch-center, (C) alone patch-edge, (D) competition uniform, (E) competition patch-center, and (F) competition patch-edge. Hatched areas denote nutrient patches (when present). Plant illustrations indicate the location of focal and competitor plants. Red data obtained from the focal plant; blue, data from the competitor (when present). The horizontal bars represent average root breadth (±1 SE), and the lines at the bottom of each frame indicate the proportion of replicates with focal plant roots in each location in the pot.

What is going on there?
Each panel is like a pot. The pots in the left column have only one plant, while the pots on the right have two plants. The first row has an even distribution of nutrients, while the second and third have nutrient patches concentrated in the grey shaded areas.

The idea is that plants adjust where they grow roots based on nutrient availability AND whether other plants are present or not. For this species, when it was planted in isolation, the roots were evenly distributed no matter where the nutrients were located.

BUT, when it was planted with a neighbor, things changed. When nutrients were only in betwee the two plants (panel E), the two plants grew together. But, when there were ample nutrients throughout, they "kept their distance" from the other plant. COOL!

Why I like the visualization
The dots on the graph show the proportion of replicates that had roots at that position. The bars on the graph show the average root distance for each plant. The x-axis is spatial, the lower half of the y-axis is numeric, and the upper half of the y-axis is categorical. Plus, as icing on the cake, they drew cartoons of where the plants were located! Very visual (without that, we'd need to read the figure legend again). Good work, Cahill et al.

Continuing agricultural intensification is unlikely

2010-06-21T13:25:00.000-07:00

UPDATE (6/24/10): I stand with my foot in my mouth, thanks to an interesting policy perspective article in Science:

Glover, J.D., et al. 2010. Increased food and ecosystem security via perennial grains. Science 328: 1638-1639.

In it, the authors discuss the value and possibility of perennial grains. Perennial grain crops would require less water, fertilizers, and herbicide than annual grains, among other benefits. Sounds great! But, as one of the authors put it, perennial grains would be one of the largest innovations in the history of agriculture. So, when will they be ready? The authors estimate that commercially viable perennial grain crops will be available in 20 years! Check it out and never underestimate the potential of innovation...

-~+~-
Burney JA, Davis SJ, & Lobell DB (2010). Greenhouse gas mitigation by agricultural intensification. Proceedings of the National Academy of Sciences of the United States of America PMID: 20551223

Agriculture contributes a substantial amount (10-12%) of of anthropogenic greenhouse gas emissions. Reducing this input is a major priority for mitigating climate change. However, we do not have very good information about what the best management strategies are to prevent emissions. Burney et al. (2010) add a unique piece to the puzzle by estimating the impact of agricultural intensification (i.e. increases in crop yield per area) on greenhouse gas emissions.

First, the trend
Since the 1960s, global production of crops per unit area (yield) has increased significantly.

Figure 1. Global increase in agricultural crop yields through time (F_1,46 = 6623, P < 0.0001), and year to year changes (horizontal line is the 1961-2008 average).

This means that, for every unit of agricultural land, we are getting more crops than we were in the 1960s. Or, agriculture is "intensifying."

How has this happened?
As Burney et al. (2010) put it succinctly, "higher-yielding crop varieties, increased use of pesticides and fertilizers, and improved access to irrigation and mechanization."

The authors suggest that this intensification has avoided up to 161 gigatons of carbon emissions! They calculated this value by comparing the observed emissions to a "what-if" scenario where, instead of increasing crop yields per area, we just increased the amount of land for agriculture (by a land area the size of Russia between 1961-2005, to be exact). Their main conclusion is that the emission-prevention benefit is due largely to investment in agricultural research which has contributed to the advancements mentioned above.

What does this mean?
There are many sources of greenhouse gas emissions in agriculture. Directly:
1. soils emit N₂O due to high fertilizer application,
2. rice cultivation and livestock manure application emit CH₄, and
3. burning crop residues emit both N₂O and CH₄.
We have a pretty good handle on those emissions. We're a little bit more fuzzy on the indirect inputs, which include:
1. production of fertilizers and pesticides,
2. production and operation of farm machinery, and
3. on-farm energy use.

Common solutions to prevent these emissions include improvements in agricultural buildings, energy supply, transportation, and industrial activities (IPCC 2007). Burney et al. (2010) calculate that, per dollar spent, agricultural intensification has the potential to prevent much higher amounts of greenhouse gas emissions. This is novel because we rarely think of agricultural intensification as a major contribution in the effort to mitigate climate change.

Is it possible?
This is my main question with this study. Even though increases in crop yields have been fairly even through time (approx. 0.07 tonnes per hectare per year), we are not seeing substantial advances in crop output (a la the Green Revolution). In fact, there has actually been a steady decrease in crop yield "advances" per dollar invested in agricultural research (as used in Burney et al. [2010]), which means that we're getting worse at realizing these advances, regardless of how much money we throw at them.

Figure 2. Decrease through time in crop yield advances (g/ha/yr) per millions of dollars invested in agricultural research (F_1,44 = 2.881, P = 0.097).

This trend is probably largely due to some of the major benefits that nitrogenous fertilizers provided in the 1960s, and the subsequent decrease in the usefulness of continuing to apply them. Regardless of the exact cause, banking on continued advances in crop yields seems like a bad idea, even if they have historically prevented substantial emissions of greenhouse gases.

Infectious art

2010-06-15T07:58:00.000-07:00

Beautiful glass creations by artist Luke Jerram are featured in the New York Times today (for an exhibition at the Heller Gallery in New York this month).

What fantastic work! These creations will probably have a greater educational impact than many biology textbooks. From the color (real viruses aren't colorful) to the form, the biological realism and careful detail in the designs are awesome. How inspirational...

Realism in species distribution models

2010-06-14T13:54:00.000-07:00

Buckley LB, Urban MC, Angilletta MJ, Crozier LG, Rissler LJ, & Sears MW (2010). Can mechanism inform species' distribution models? Ecology letters PMID: 20482574

Predicting species distributions is not easy. Current approaches can be broken into two broad categories: "correlative" or "mechanistic" models. Buckley et al. (2010) do something very unique by comparing the relative accuracy of these two approaches (a total of 5 models) for two species (a butterfly and a lizard, see image). Their findings are interesting and very informative, but their conclusions lack some potential insight and they miss some important opportunities to advance our ability to predict.

To begin, what do "correlative" and "mechanistic" mean? Correlative models are based on current species distribution and environmental conditions. Mechanistic models are based on organismal responses to the environment, and can use factors like energy gain (model #3 below...), reproduction/survival based on climate observations (#4), or reproduction/survival calculated by energy gain (#5).

THEIR MODELS
Correlative - require species location and env. data:
1. Maximum Entropy (maxent)
2. Generalized Linear Models (GLM)
Mechanistic:
3. Biophysical threshold - requires trait thresholds and env. heat variables.
4. Life history - requires demographic data and env. data.
5. Foraging energetic - requires trait data and env. heat variables.
(Nice work on the models, Buckley et al.!)

What was my prediction for the study? That the mechanistic models would outperform the correlative models because of increased "biological realism" (seemed reasonable). And, what did Buckley et al. (2010) find? The correlative models performed just as well as the mechanistic models for predicting current species distribution... In fact, my rough calculation from their results table has the correlative models correct 77% of the time, but the mechanistic models correct only 70% of the time! (bummer)

Why did that happen?! Buckley et al (2010) point out two reasons for why the mechanistic models might not do as well as I expected:
1. The constraints in the mechanistic model might not be the most important variables, and
2. parameter estimates for the mechanistic models might have errors.
I think that the first part is relatively easy to get handle on with more background knowledge for the species in question. The second part is a little bit more tricky, though. Buckley et al.'s (2010) solution for the second issue is to extend "sensitivity analyses" for parameter values, since we know that there is uncertainty in the parameters. There is a great (as yet unexplored) opportunity here to develop species distribution models that consider some of this uncertainty!

The bottom line:
My rough, correlative prediction for the current and future distributions of Acer saccharum (shown in green and transparent pink above, with observations as filled points) might not be as bad as I thought (data from GBIF and LifeMapper - using the BIOCLIM correlative model). But, it would be nice to incorporate some mechanism into it. This is especially important considering that the correlative models in Buckley et al. (2010) produced VERY DIFFERENT forecasts than the mechanistic models.

As Buckley et al. (2010) point out, we are on our way to making these models really useful. All that is needed is a better accounting of what mechanisms are important, and how uncertain we are about those mechanisms. (Easy, right?!)

Data science

2010-06-07T18:35:00.000-07:00

There is an interesting essay at the O'Reilly Radar (emerging technologies) blog about data science. I do not suspect that it will be read by many biologists, but I think we would be well served to think about the coming data transition more closely.

I think that understanding (learning) and using the wealth of information will be critical for scientific research in the VERY NEAR future -- no matter what discipline.

The scientists of the future will no longer be "collecting" data in the traditional sense. Our primary role will be to compile, interpret, and communicate knowledge from the tonnes of data that already exist and are (metaphorically) collecting dust. The author is spot-on.

My key quote: "Data science requires skills ranging from traditional computer science to mathematics to art."

I also liked: "data jujitsu."

My only complaint: data are, not ~~data is~~!!!

Why is running good for you?

2010-06-02T11:54:00.000-07:00

This weekend I am running in the Kettle Moraine 100, in a 31 mile leg of the 100 mile relay. Given that event, this is my first running science blog post - a slight tangent from ecology...

Lewis GD, Farrell L, Wood MJ, Martinovic M, Arany Z, Rowe GC, Souza A, Cheng S, McCabe EL, Yang E, Shi X, Deo R, Roth FP, Asnani A, Rhee EP, Systrom DM, Semigran MJ, Vasan RS, Carr SA, Wang TJ, Sabatine MS, Clish CB, & Gerszten RE (2010). Metabolic signatures of exercise in human plasma. Science translational medicine, 2 (33) PMID: 20505214

Where do the health benefits of running come from? Physiologically, people's metabolism changes. Or, more specifically, certain metabolites increase in the body, which trigger cellular responses (for example, fat burning).

Scientists from around the Boston area (Lewis et al. 2010) identified over 200 metabolites and measured how they changed when the subjects ran, either on the treadmill or in the Boston Marathon. From their list, they identified 21 compounds that increased with running (10% is not too bad!). Then, they analyzed muscle tissue samples in culture to identify the mechanistic connection to the "health benefits" that we usually consider. A set of the metabolites (glycerol, niacinamide, glucose-6-phosphate, pantothenate, and succinate) were linked to pathways that break down stored sugars (glucose) and fats (lipids) in skeletal muscle through the expression of the nur77 protein. Some of the metabolites were also associated with the regulation of blood sugar.

Aside from the potential for these findings to "eventually lead to dietary supplements that boost athletic performance or invigorate patients suffering from debilitating diseases" (study author Robert Gerszten), I think it is pretty awesome that we can identify specifically how exercises like running lead to better health.

So, next time you're out running, think about all of those metabolites that are ramping-up to trigger the proteins and genes in your legs, burn fat stores, and regulate your blood sugar!

Further reading: The metabolic secrets of good runners: Nature News

The nature of risk assessment

2010-05-29T18:19:00.000-07:00

The New York Times reports on strains of E. coli that are "largely ignored." They mention six rarer strains of the bacteria that are also toxic to humans and can have substantial health impacts, although

Few food companies test their products for the six strains, many doctors do not look for them and only about 5 percent of medical labs are equipped to diagnose them in sick patients.

Is this a problem? When highlighting individual case studies where the admittedly-rarer E. coli strains infect humans with serious consequences, it seems outrageous that these strains are under the radar. But, is the risk big enough? This is a good chance to think about conditional probabilities.

Lets say that the probability of finding contaminated food at any given visit to a restaurant (undercooked, infected beef or unwashed vegetables that have the bacteria on them) is somewhere between 0.2-0.7% (the bounds estimated in the NYTimes article). Just from this step, statistically, I would say that this is not a significant risk. But, then we get to combine this risk with the probability that a person becomes infected after coming into contact with the bacteria. With a rough estimate of the upper probability of infection at around 50% (Cassin et al. 1998), the risk is cut in half to 0.1-0.35% (0.002x0.5 to 0.007x0.5), or -at maximum- 2 in 1000. THEN, there is the probability that an individual is infected and the infection becomes severe (beyond a mild case of diarrhea), which might be 5%. Now, our probability becomes 0.00005 to 0.000175, or around 1 in 10,000. This value probably even overestimates the probaiblity that the chef cooking your food didn't cook the food at a proper temperature or wash the vegetables properly (most of them probably do, right?).

Every time you go to a restaurant, there is a risk that you will be infected with a strain of rare E. coli that will make you VERY sick. But, each time you go out to eat, that risk is EXTREMELY small.

Conditional probabilities are so helpful.

Global food security

2010-05-25T09:20:00.000-07:00

What does the global food supply look like? What does it look like compared to 50 years ago?

Figure 1. Food supply time series (1961-2005) by country (data: faostat).

This figure shows one measure of food supply (kg/person/yr) for each country in the world (data: faostat). I've used the amount of cereal grains as an indicator of global food supply to think about where we are at now and where we have come from.

There are three styles of lines. Red lines identify countries where cereal supply has decreased significantly per capita through time. Blue lines are where cereal supply has increased significantly per capita. Black, dashed lines are where there was no significant change through time.

So, what are the numbers?

77 countries (50.66%) are increasing,
46 countries (30.26%) are decreasing, and
29 countries (19.08%) have not changed significantly.

That is not too bad... in fact, I'd say it is a pretty optimistic picture. The biggest problem is the 1/3 that has decreased (I will expand on which countries those are in the future).

In a snapshot comparison of the whole globe, we have significantly increased each person's food supply (t = -2.218, df = 302, P = 0.027):

Figure 2. Increase in food supply per person from 1961 to 2005 (data: faostat).

But it hasn't been by much. In 2005, each person on Earth had only 8.9% more cereal grains than they had in 1961, which comes out to a change of about 29 grams per day... maybe the equivalent of one slice of bread.

Corn dynamics

2010-05-25T08:35:00.000-07:00

Corn is certainly a remarkable creature, as Sean Carroll highlights for the New York Times today.

With our help, the creature overcame remarkable odds to expand its global range from its probable origins as teosinte in "the tropical Central Balsas River Valley of southern Mexico." Carroll identifies a few of these accomplishments, which he credits to the early plant breeders:

The most crucial step was freeing the teosinte kernels from their stony cases. Another step was developing plants where the kernels remained intact on the cobs, unlike the teosinte ears, which shatter into individual kernels. Early cultivators had to notice among their stands of plants variants in which the nutritious kernels were at least partially exposed, or whose ears held together better, or that had more rows of kernels, and they had to selectively breed them.

However, along the way, the creature (Zea mays) lost its ability to reproduce in the wild (it is now entirely dependent on human re-planting every year). It is also dependent on a high supply of water and nutrients, exhibiting what we call a very low resource use efficiency. Luckily for Z. mays, we're still willing to play with it. In fact, corn has risen to be a key player in nearly all packaged foods in the United States.

Carroll mentioned that corn comprises 21% of human nutrition globally. Over the last 50 years, we have seen a linear increase in corn yields globally:

Figure 1. Global corn yields (g/m²) through time (data: FAOSTAT)

This increase was matched by a significant increase (P < 0.001) in land area for corn production (FAOSTAT).

An important question, however, is whether or not this can continue. We have produced steady increases (t = 1.895, df = 46, P = 0.064) in yields every year (mean 6.73 g/m²/yr).

Figure 2. Rate of increase in global corn yields (g/m²/yr; data: FAOSTAT)

But, this increase has been entirely dependent on an oversupply of nitrogen fertilizer (often synthetic) and an abundance of available water - two factors that could use reconsidering. The future of corn may lie in its humble beginnings as a wild grass and on our ability to breed strains that can thrive in harsher conditions than at present.

Measuring ecosystem metabolism

2010-05-21T08:43:00.000-07:00

The metabolic rate of an ecosystem is the amount of carbon uptake or release from the organisms within the ecosystem collectively. For ecosystems, we call this rate the net ecosystem productivity (NEP) - the net amount of carbon exchanged by an ecosystem at any given time. NEP can be measured with a flux tower (see right) by the eddy covariance method. This method provides an instantaneous measure of carbon exchange between the biosphere and the atmosphere.

There is a lot that we can do with these values. To start, I will demonstrate how to decompose NEP into its primary components from an example site (Duke Forest). Then, I will demonstrate how these values can be related to environmental variables. But first, lets take a look at NEP. We can plot NEP by day of the year to see how carbon exchange changes through time.

Figure 1. Daily daytime net ecosystem productivity at the Duke Forest Hardwood flux tower (2005).

This pattern is characteristic of many seasonal environments; NEP increases in the summer months when plants are most active and decreases in winter.

So, what goes into NEP? Carbon enters an ecosystem through photosynthesis (plant growth). This is called gross primary productivity (GPP). Carbon leaves an ecosystem through respiration. The combined respiration of the autotrophs and heterotrophs is the net ecosystem respiration rate (Re). We have NEP values, and we want to estimate GPP and Re.

Luckily, plants only photosynthesize during the day, while all organisms respire all the time. So during the daytime, NEP = GPP - Re, and during the nighttime, NEP = Re. With a simplifying assumption that daytime respiration equals nighttime respiration (which we can test later...), we can estimate Re with the nighttime NEP. Then, we can estimate GPP by adding daytime NEP and nighttime NEP (for daily values, I've averaged 5 hours of 30-min subsamples, calling "daytime" 11:00-15:30 and "nighttime" 23:00-03:30).

Figure 2. Daily gross primary productivity and net ecosystem respiration rate at the Duke Forest Hardwood flux tower (2005).

How do these fluxes relate to the environment? We can compare the observed values to the expected temperature response of photosynthesis and respiration.

Figure 3. Expected response of photosynthesis and respiration to temperature (copied from Todd Dawson).

Figure 4. Observed response of gross primary productivity and net ecosystem respiration rate to temperature at Duke Forest.

The observations match fairly well with the expectation, although there is still a lot of scatter, probably due to estimation error.

But, temperature is not the only environmental variable that influences carbon exchange. I will leave this post with the relationship between GPP and soil water, which (very roughly) agrees with the curvilinear pattern that I would expect.

Figure 5. Response of gross primary productivity to soil water content at Duke Forest.

This analysis is only scratching the surface of what is possible with publicly available data sets and there are many opportunities to learn about how ecosystems function with similar methods.

About the data: All data from the AmeriFlux website.

About the site: The site that I examined here was the "hardwood" site in Duke Forest. The site is an 80-100 year-old mixed hardwood forest dominated by oak and hickory species. It is 163 m above sea level and is located between the Coastal Plain and the Piedmont Plateau near Durham, North Carolina (36º 2'N, 79º 8'W). Mean annual temperature is 15.5 ºC and mean annual precipitation is 1140 mm. In this analysis I used data from 2005 only.

Malaria in a changing world

2010-05-20T10:28:00.000-07:00

Gething PW, Smith DL, Patil AP, Tatem AJ, Snow RW, & Hay SI (2010). Climate change and the global malaria recession. Nature, 465 (7296), 342-5 PMID: 20485434

Infectious diseases straddle a fuzzy boundary between ecology and epidemiology, making the prediction of future disease dynamics difficult. Scientists need to consider how environmental change will affect the infectious organism (and sometimes the vector) as well as how humans respond to the disease. These predictions carry special weight, given their importance for human health and well being, and getting them right is essential. Unfortunately, they are not very good right now.

Globally, scientists and governments are intensely studying and working to control and reduce the impact of malaria. Key components of this effort are predictions that climate warming may increase the range and intensity of Plasmodium falciparum in Africa and other parts of the world. But, how meaningful are these predictions? This is the question that Gething et al. (2010) addressed head-on by comparing historical (1900) to the current (2007) distributions of the disease.

The authors raise two important assumptions about existing malaria predictions (which are really applicable to all species distribution studies):

Assumption 1: The connection between the current spatial distribution of a disease and surface temperature allow us to extrapolate and predict the future range and intensity of the disease.

This is the classic, correlative "bioclimate envelope" approach that is the essence of many distribution models. Briefly, one can 1.) draw the species boundary, 2.) identify the climate where the organism lives, 3.) predict the climate in the future (e.g. with a suite of GCMs), and 4.) redraw the boundary based on how the environment changed.

However, despite clear global warming over the last century, Gething et al. (2010) documented a "marked, global decrease in the range and intensity of malaria transmission" - in contrast to the bioclimate assumptions of numerous other analyses, which would have predicted an expansion of malaria over the last 100 years.

Assumption 2: Climate/organism relationships remain stable through time.

Central to predicting organismal distribution with environmental correlations is an assumption of uniformitarianism - essentially, nothing changes. Unfortunately, this doesn't always work, especially when humans are involved. Since the 1900s, people have confounded the climate/malaria relationship directly by disease control efforts and indirectly by urbanization and economic development. Given these factors, analyses of current distribution based on climate may not capture the whole picture.

In closing, the authors clearly address these two major concerns about predicting future species distributions with a globally-important organism. Climate can and will influence the distribution and abundance of P. falciparum in the future. But, as the authors conclude, the counteracting effects of economic development and control programs have been dominant factors over the last 100 years and will likely have an increasing presence in the future.

What controls where a species lives?

2010-05-20T08:38:00.000-07:00

Pigot AL, Owens IP, & Orme CD (2010). The environmental limits to geographic range expansion in birds. Ecology letters PMID: 20412281

Or, what prevents that species from expanding further? Generally, we assume that the environment has something to do with controlling a species' distribution. But, our understanding of this topic is surprisingly limited, given centuries of scientists prodding the question. Part of the problem has been:

a limited number of species analyzed with small spatial extent, and
scientists promoting their "pet" hypotheses of which factors are important.

Pigot et al. (2010) approach this question elegantly in a refreshing analysis that stands out from previous work on the topic. First of all, they examined data on the global range of all living bird species, which gives their results substantial weight. Secondly, they consider multiple working hypotheses - an approach to science championed by Thomas C. Chamberlin (Beloit College grad) in his seminal work on scientific methods. These two factors combined allow Pigot et al. (2010) to present a novel view of an old question.

For their analysis, they used a "process-based model of range expansion" analogous to spreading dye to simulate each species range given different environmental constraints:

mean annual temperature,
mean annual precipitation,
elevation,
net primary production, or
entirely random.

They compared these simulations to the observed shapes of each species' range and found that a combination of temperature and precipitation provided the best fit. This reinforces our suspicions that climate is important (and provides a very nice test of the hypothesis). Elevation provided the second best fit, emphasizing the importance of local-scale variation in the environment.

Perhaps most interestingly, though, was that their random distribution models performed worst out of the set. This challenges previous ideas about neutral models for species distribution and suggests that range shape can be explained by factors that we (at least partially) know about and understand!

Their analysis has some shortcomings that I think are important. First, which they acknowledge, there is no way to build in "biotic interactions" to these models. So, the role of interspecific interactions, which we think should be important, especially in the tropics, is still uncertain. Second, this study just begs for a more complex consideration of randomness. While they've done a good job of considering many alternatives, their list is by no means exhaustive. It will be important to consider more combinations to really prod this question, especially sources of stochasticity in observation (range size/shape, environmental measurements) and in process (randomness at the edge, in what factors constrain and in how they constrain distribution).

Overall, this is a nice paper. The author's approach to science is a good model for untangling the complex, unanswered questions in ecology.