The devil isn't always in the details: how system properties can inform ecology

Selection on stability across ecological scales. Jonathan J. Borrelli, Stefano Allesina, Priyanga Amarasekare, Roger Arditi, Ivan Chase, John Damuth, Robert D. Holt, Dmitrii O. Logofet, Mark Novak, Rudolf P. Rohr, Axel G. Rossberg, Matthew Spencer, J. Khai Tran, Lev R. Ginzburg. 2015. Trends in Ecology & Evolution, http://dx.doi.org/10.1016/j.tree.2015.05.001.

This paper in TREE  on selection at higher level systems has been on my must-read list since it came out a few weeks ago, and it was worth the wait. It does what the best TREE papers do - makes you think a bit more deeply about a common topic. In this case, it develops an approach to understanding complex ecological systems (communities, ecosystems) that is blind to the details that ecologists often focus on.

The search for generalities and commonalities drives modern ecology. In short (though this paper deserves an in-depth read), this paper argues that we can learn much by considering stability and feasibility in complex ecological systems. That is, we can also study community structure or trophic webs by considered whether specific configurations of the system are stable. This is in contrast to a context-centric study of a system, where the usual list of proximate causes (productivity, niche availability, connectivity, etc, etc) may be used to understand why the system looks as it does.

The authors' premise is that nonadaptive (e.g. unstable) ecological systems will be unfavourable and selected against, and the resulting selective process “can produce many of those recurrent ecological patterns that have been observed in nature over large scales of space and time.” This requires that you accept a few underlying concepts: first, that large scale systems also experience selection (whether one prefers selection be in parentheses is up to the reader), in that unstable systems will be lost at faster rates leading to greater frequency of stable systems; and second, that this process of selection is determined by the properties of the system alone, not the specific conditions ecologists often focus on.

As an illustration, consider four possible food webs depicting intraguild predation that vary in their interaction strengths. All configurations are possible, but A-C are likely to lead to exclusion of the intraguild predator. D is most likely to be stable since the strong interaction between the resource and prey results in negative feedbacks between the densities of all species (i.e. when the resource is low, the prey should also be low, reducing the predator density as well) and thus more likely to be observed in natural systems. 

From Borrelli et al 2015.

A more specific example looks at attack rates and handling times in predator-prey interactions. When stability is considered, it seems that although predator-prey cycles may occur, it should be uncommon to have such extreme oscillations that populations reach dangerously low levels where stochastic extinctions may occur. Data suggests that oscillatory dynamics are less common in predator-prey relationships, but do occur particularly for specialist predator/prey pairings. Theory (Rosenzweig-MacArthur predator-prey models) predict that such pairings should be most stable if prey are weakly self-limited and predators have high attach rates/long handling times. Empirical evidence for this prediction supports it surprisingly well.

From Borrelli et al 2015.

 
Other related approaches consider feasibility across food webs, communities, and ecosystems. A community perspective might consider interactions across all species, perhaps using a network approach. Networks should tend towards formations that are the most stable – e.g. short chains rather than long ones. The commonness of nested network structures may reflect these constraints. 

Such an approach to ecology is not entirely new (Robert May's weak interactions comes to mind). But it provides perhaps the best potential explanation I’ve seen for ‘generality’ focused approaches in ecology, including ecological allometric relationships, macroevolutionary patterns, and network approaches. Macroecological patterns have often captured, rather than tidy linear relationships, occupied versus unoccupied parameter space. Thinking about feasibility as a macroecological ‘mechanism’ for ecological patterns at the system scale might lead to new research directions. 

#ibs2015 – Confronting uncertainty, biases and the unknown

The 2015 meeting of the International Biogeography Society just came to an end, and even for someone who wouldn’t traditionally consider themselves a ‘biogeographer’ there were many interesting topics and talks to see.

The focus of most talks was on biological patterns over space and/or time (or ‘deep time’, which is a fun phrase to throw around), and the talks emphasized how sophisticated statistical methodologies for such questions have become. The extent and complexity of approaches for making inferences from limited existing information, be it phylogenetic, distributional, or fossil and pollen records, is pretty amazing.

Such complicated inference needn't and shouldn't come at the cost of careful scientific work, and must include recognition of uncertainty and biases. The final sessions of the conference acted as an excellent (and at times provocative) reminder of this. For example, Joaquin Hortal advocated the development of ‘maps of ignorance’, which instead of showing the typical distributions of known information, highlight where information is missing and new sampling should be emphasized. Not only is information sometimes missing, but its value degrades over space and time. The value of a sample declines the further away you get from that site or the more different the spatial scale; samples over 50 years old may not represent current conditions any more. Predictions should consider or even incorporate this uncertainty.

Catherine Graham, David Nipperess, and Jon Chase all gave talks similarly emphasizing how fundamental consideration of scale and extent is. This is as true for phylogenetic community analysis (Graham, what extent or size of tree should be considered for analyses of community phylogenetics?); for rarefaction of phylogenetic diversity (Nipperess); or for measures of beta-diversity (Chase). Without this context, we are likely to be misunderstanding our results.

Finally, David Currie gave a damning critique of macroecology. Unfortunately, he said, macroecology seems to be a field where hypothesis testing is rare and conclusions are drawn based on spurious correlations with little explanatory and even less predictive ability. For example, why has the study of latitudinal gradients in richness progressed little beyond a list of possible correlates after more than 30 years of attention? Though Currie was focused on his own field, his comments were relevant to many ecological approaches. Currie expressed concerns about areas where scientific methods were being given short shift. In particular, he noted a lack of appropriate hypothesis testing and strong inference. Instead there is a tendency for studies to look for evidence in support of a hypothesis of interest, rather than attempting to falsify a hypothesis. Supporting evidence, sadly, does not actually increase the probability that a hypothesis is true, since the evidence could equally support some other, currently unconsidered, hypothesis. Further, correlations between variables of interest are at best a weak test of a hypothesis. The most important suggestions were that macroecologists and others should be testing the predictive ability of their hypotheses on new data sets: model fitting, in his opinion, is too often confused with model testing.

The contrasting effects of habitat area and heterogeneity on diversity

“How extremely stupid not to have thought of that!” (Thomas H. Huxley, commenting on the obviousness of Darwin’s theory of natural selection)

Sometimes I read a paper and Huxley’s famous quote seems exceedingly appropriate. Why I say this is that a new idea or concept is presented which seems both so simple and at the same time a potentially powerful explanation of patterns in nature. This was my reaction to a recent paper from Omri Allouche and colleagues published in the Proceedings ofthe National Academy of Science. The paper presents a simple conceptual model, in the same vein as Connell’s classic intermediate disturbance hypothesis, which accounts for large-scale diversity patterns based on aspects of species niche requirements as well as classic stochastic theory. Merging these two aspects is a critical step forward, as in ecology, there has been a tension in explaining diversity patterns between niche-based processes requiring that species exhibit differences in their needs, and stochastic (or neutral) explanations that ignore these differences, but seem to do well at large scales.

The classic stochastic model in ecology, the theory of island biogeography, simply predicted that the number of species increases with the size of an island or habitat, and ultimately is the balance between species colonizing and going extinct. Allouche et al. also assume this stochastic colonization and extinction, such that in a uniform environment, the number of species increases with area. However, they then add the fact that species do not do equally well in different habitats, that is they have specific environmental niches associated with a particular environment. Thus as you increase the amount of heterogeneity in a landscape, you increase the total number of species, because you’ve captured more niches. However, there is a trade-off here. Namely, as you increase the heterogeneity in a landscape, the amount of area for the dominant habitat type decreases, thus reducing the number of species. So if you increase the heterogeneity too much, the individual habitat types will be too small to support large numbers of species and the numbers of species will be less than regions with less heterogeneity –paradoxically.

Their heuristic prediction is that diversity is maximized at intermediate levels of heterogeneity, as long as species have intermediate niche breadths (i.e., they could perhaps use a couple of different habitats). However, if their niche breadth is too narrow (i.e., they can only exist in a single habitat type), then diversity may only decline with increasing heterogeneity. Conversely, if species have very broad niche breadths (i.e., can survive in many different habitats) then the tradeoff vanishes and heterogeneity has little effect on diversity.

They tested this exceedingly simple prediction using European bird data and found that species richness was maximized at intermediate heterogeneity (measured by the variation in elevation). Further, when they classified species into different niche width classes, they found that the relationship between richness and heterogeneity changed was predicted (i.e., strongest for intermediate breadth).

This is a great paper and should have a large impact. It will be exciting to see what other systems fit this pattern and how specific studies later the interpretation or mechanisms in this model.

Allouche, O., Kalyuzhny, M., Moreno-Rueda, G., Pizarro, M., & Kadmon, R. (2012). Area-heterogeneity tradeoff and the diversity of ecological communities Proceedings of the National Academy of Sciences, 109 (43), 17495-17500 DOI: 10.1073/pnas.1208652109

Putting ecological niche models to good use

Banta et al. 2012. Climate envelope modelling reveals intraspecific relationships among flowering phenology, niche breadth and potential range size in Arabidopsis thaliana. Ecology Letters. doi:10.1111/j.1461-0248.2012.01796.x.

I won’t be the first or the last person to state that I find ecological niche models (ENMs) a bit problematic. In their simplest form, ENMs are statistical models correlating species presences or presences and absences with climatic factors. These models can then be used to predict the location of suitable habitat either elsewhere in space or later in time. They can be used to examine how species’ ranges may shift with climate change, to predict where invasive species’ ranges will expand, or to suggest appropriate locations for new reserves. Over the last while, they’ve faced a fair amount of criticism. For example, most fail to incorporate biotic interactions and so they capture a species’ realized niche: this means that it might not be accurate to extrapolate the model to areas where the biotic environment is different. There are also questions of what is the appropriate spatial scale for environmental data; the problem that many populations’ dynamics (especially invasive species) are not at equilibrium with the environment, so their observed relationship with climatic factors may not represent their niche; statistical and data-quality issues; and the difficulties of validating predictions that may be made for changes in habitat 50+ years in the future. Like many new techniques, ENMs became popular quickly, before they developed an appropriate foundation, and so they were subject to misuse and inappropriate conclusions. But this is a typical pattern – the development of ecophylogenetic tools has followed a similar path.

While this period of early growth has tarnished some people’s view of ENMs, it would be a shame to disregard them altogether when there are people still using them in interesting and inventive ways. A great example is Banta et al. (2012), which combines a model organism, intraspecific phenotypic variation, and spatial structure of genetic variation with ecological niche modelling. Banta et al. focus on the problematic assumption of such models that intraspecific variation in climatic tolerances is minimal or unimportant. One approach to exploring this issue more is to develop intraspecific ENMs using genotypes, rather than species, as the unit of analysis.

Banta et al. take advantage of the fact that the model organism Arabidopsis thaliana is genetically well understood, allowing them to identify ecologically different genotypes, and is widely distributed across highly varied habitats. The authors looked at genotypes of Arabidopsis that varied in flowering time and asked whether these ecologically differentiated genotypes had different niche breadths and potential range sizes. They also looked at the classic macroecological question of whether niche breadth and range size are correlated (in this case, intraspecifically). To answer these questions, they identified 15 single locus genotypes for flowering time (henceforth “genotypes”), and developed ENMs for each, looking at the climatic conditions associated with each genotype. Using the output from these models, Banta et al. calculated the niche breadth (measured based on how much suitability varies among habitat types) and the size of potential habitat (the sum of units of suitable habitat) for each genotype.

The authors could then look at how intraspecific variation in flowering time related to differences in niche breadth and range size among the different Arabidopsis genotypes. They found that genotypes tended to differ from each other in both niche breadth and range size. This is important because ENMs assume that small amounts of genetic variation among populations shouldn’t affect the accuracy of their results. In fact, even differences in a single gene between genotypes could be associated with differences in niche breadth and potential range. In general, late flowering genotypes tended to have smaller potential ranges. The authors suggest a few explanations for this, including that late flowering genotypes may be adapted to harsher conditions, where flowering late is beneficial, but unable to compete in less stressful habitat. Regardless of the particular explanation, it shows that single locus differences can drive phenotypic differences among individuals, which in turn have notable macroecological effects.

From Banta et al. 2012. Relationship between potential range size and flowering time/niche breadth

Similar to the pattern found in a number of interspecific studies, the authors found a strong correlation between potential range size and niche breadth. This matches the oft-quoted statement by Brown (1984) that generalist species should have large potential ranges compared to specialist species, which should have small potential ranges since they only tolerate a narrow range of environments. It should be noted that this explanation is based on the assumption that habitat types are equally common: should a specialist species be adapted (only) to a widespread habitat type, the correlation between niche breadth and potential habitat size would be weakened. Because this study didn’t incorporate competition or other biotic interactions, it is not possible to conclude that there are differences in climatic tolerances among genotypes rather than differences in competitive abilities, for example. Inferior competitors may be exclude from ideal habitats and so appear to be specialized to harsh conditions (and the authors note this). This is always the difficulty with interpreting observational patterns, and further, the ongoing difficulty with defining a species’ niche based on observational data. In any case, this study does a nice job of exploring the underpinnings of macroecological variation and uses EMNs in an informative way, and suggests many interesting extensions.

The European Ecology Federation Congress, day 2

Day two of the conference, and still many great talks. I mainly stayed in the session on synthesizing community ecology, phylogenetics and macroecology. This has turned out to be a great conference and Avila is a great venue.

Carsten Rahbek. His talk is on merging the fields of macroecology to better understand patterns of diversity. Different models explain variation differentially at different scales. For example, climate models do well for wide-ranging species but not for scarce species. A model of evolution may do much better for scarce species, but not for wide-ranged species. Statistical tests confirm a correlation, but not necessarily a mechanism. One could get different conclusions if one were to compare to a null model. He advocates a spatially explicit species assembly model that integrates macroecological models with community assembly. It is scale invariant and can explain spatial and temporal variation in assemblages. In an example, he shows that, based on small scale sampling, species distribution models will over-predict richness. Need to combine macroecological models with distribution models, because acroecological models do well to predict richness but not composition while distribution models predict composition but not diversity.

Jens-Christian Svenning. He talked about paleoclimatic influences on ecological patterns and function across scales. Past climates have shown masive changes and different groups of species have evolved during these events, while other species have gone extinct. The velocity of climate change was highest in northern Europe and North temperate North America, and higher velocity results in lower endemism since it is quicker for species to migrate than diversify. Higher velocity results in lower specialization in hummingbirds. He finishes with a note about current regions undergoing fast climate change; these are not necessarily those same regions that had the most change in the past.

Adreas Prinzing. His talk was about how niche conservatism can inform our potential solutions for changing environments. Specialists are declining in changing environments and how does this apply to specialist clades of closely related species? Specialist species tend to occur in specialist genera. However, niche conservatism does not tell us everytng about species differences/similarities because closely related species clearly coexist and exhibit substantial trait differences. Species coexist within niches by key divergences.

Kenneth Kozak. He presented a way that phylogeny illuminates the origin of climate-richness relationships. Only speciation, extinction or dispersal can change richness, and many models do not ask how these processes change. He examined salamader diversity and evolutionary history using 16000 occurrence records in North America, and examined climate variables for occurrences. Diversity was highly associated with cool, moist places. Richness is strongly correlated with evolutionary time of colonization of climatic conditions. For example, evolution of warm species is recent, hence fewer species. Diversity does seem to be saturating, and so time is limiting factor, and more species can probably still emerge.

David Vietes. He gave an interesting talk on the amphibians of Madagascar, which is a diversity hotspot for amphibians. There were 132 described species n 1999 and now 263 with about 200 still needing to be described. Many are endemic to small regions of Madagascar (the whole family is endemic to Madagascar). He discussed many aspects of the distribution of these species, and looked at phylogenetic patterns. Some interesting observations include: older species pairs are further separated in space and smaller species have smaller ranges. Also, there appears to be a predictable pattern of richness hotspots, but endemism hotspots are more idiosyncratic.

Joaquin Hortal. He discussed the effect of glaciation on richness, functional diversity and phylogenetic diversity for European mammals. The hypothesis he explored was that current distributional patterns driven more by past changes since glaciation than current climate. He compare several different types of measures and it turns out that current climate is more important for explaining patterns of co-occurrence and relatedness, with more closely related species occurring together at northern locales.

Catherine Graham. She also explored patterns of richness, functional and phylogenetic diversity, but was looking at hummingbirds diversity patterns across elevation gradients in South America. She compiled an impressive dataset with several morphological traits and co-occurrence patterns. Broadly, close relatives co-occur at high elevations and more distantly at lower where competition is stronger. In local communities a mix of environmental filtering and competitive dispersion seem to be operating. At high elevations, both functional and phylogenetic diversity are high.

Rob Dunn. He gave a fantastic talk on the species on the human body and in our lives and homes. He told us about projects that involve citizen scientists from across the USA and had them sample their homes and bellybuttons. Amazingly, Dunn’s group has so far identified 1400 species in belly buttons, and many of them are unknown species –which could not be classified into known species groups. He looked at many factors like ethnicity, geography, cleanliness, but none of these explained this diversity well. A subset of these species are bellybutton specialists and dominate bellybutton floras within and among people, and are phylogenetically clustered, evidence that the bellybutton habitat is a conserved trait.

Cecil Albert. I ran to another session to see this talk on intraspecific variation in species traits. She eloquently showed that for plant assemblages, there was substantial intraspecific variation in traits. Some species showed high variation and some showed almost no variation. Importantly, she showed that this variation could substantially change our ability to explain how functional traits link to abundance and coexistence. She simulated different levels of variation and looked at the strength of the correlation between expectations from mean trait versus the actual trait that varies. The strength quickly declines for some traits as variation increases, meaning that with variable traits, the explanatory ability of using a mean trait is weak.

Sally Keith (Flash talk*). She examined the Mid-domain effect (where, because of range sizes, maximal diversity is found in the centre of a geographic landmass by random chance), and process based models to test mechanism for middomain prediction. She showed that these models seem to have limited success. Perhaps environmental gradients and species interactions could be important. But when she added interactions to the model, it then predicts humped shaped pattern predicted by the mid-domain effect.

Tamara Munkemuller (Flash talk). She examined phylogenetic relationships as a way to examine niche patterns and coexistence. She hypothesized that there should be strong filtering under stressful conditions. She examined thousands of plots across elevational gradients, and plots that were in stressful locations tended to be phylogenetically clustered, meaning certain groups of species exist there.

Susanne Fritz (Flash talk). She was looking at diversification patterns in birds. Using lineage through time plots one should expect that the rate of diversification should decline trough time, which perhaps equates to niche filling. For species in tropical Asia, she found that there is not much leveling off of diversification rate. Though interestingly, groups that have not dispersed (for example, birds of paradise) do show a plateau in diversification. Globally, diversification slows down more in more speciose regions.

Jake Alexander (Flash talk). He had a very interesting talk on elevation gradients in richness in non-native plants invading mountain habitats. Most species have narrow elevational ranges in lower and mid elevations and high ranges for those species found at high elevations. The explanation is that these non-natives are generalist and that they originate form lower elevations –where human activity dominates, and must spread up the mountain to get to the high elevations.

*Flash talks are 3 minutes long, and a great way for people to communicate new and exciting results.