This is a very short video about our work and the questions that we ask. Courtesy of the Academy‘s Visualization Studio.
Remember the Deepwater Horizon drilling disaster? It’s been a few years now. This video is a brief update on some of our research on how the spill might have affected, and may still be affecting oysters off the coasts of Louisiana, Alabama and Florida. This has been very slow and difficult work, mostly because we have been monitoring the oysters for several years, and have had to develop protocols for the tissue analyses. But it is now moving toward publication.
One of the central questions of our paper was, “How stable are ecological communities during a mass extinction?” This might seem a bit of a silly question at first glance, with the obvious answer “Not stable at all!” But that is not necessarily the case. Consider yourself standing on the deck of a leaky shop which is filling gradually with water. You know that the ship is going down, but your situation is stable as long as the deck remains level, or at least until the water begins to lap around your knees. We often tend to think of mass extinctions as chaotic dramas, perhaps being influenced by the end Cretaceous event, 66 million years ago (mya), when a 10 kilometer asteroid collided with the Earth and much hell really did break loose. There is also a lot of talk these days about collapsing ecosystems, because we continue to warm up the planet, eat all the fish we can eat, and so on. But what would a Sixth Mass Extinction really look like? Would ecosystems collapse, or wind down slowly to shadows of their former selves? Did the citizens of a Roman city in Gaul turn out the lights one night in the 5th century CE, bid the ancient world farewell and lay out their clothes for the next morning’s Middle Ages? Or did they rather one day, in corner market conversation, question how the heck all those Germans wound up in government anyway? A little bit of both I suspect.
So getting back to our question of mass extinctions at the end of the Permian, some 252 mya, were ecosystems stable before the extinction, collapsing as species extinctions spiralled out of control, or were they whittled down to a hardy core? Did they become more sensitive to smaller insults, such as storms or droughts, or were they hardy cores? Answering these questions depends surprisingly on what you mean by “stability”. The term is used in various ways in ecology, and I’ve even been accused of using it in a rather narrow sense, in contrast to others who believe that there are many kinds of stability. I am not convinced that the latter is really the case, and even if it is, I would argue that there is only one important type of stability, and that is the likelihood that the community will persist, that is, continue to exist in pretty much the same form, under non-extreme environmental conditions. The conditions that have prevailed during the history of a stable community, including seven year droughts, megastorms, the occasional disease epidemic, etc., did not cause the community to collapse or its species to become extinct. This definition encompasses many aspects of stability. Consider again our boat, this time with no leaks. Whether it is at anchor in a calm bay, sailing steadily on smooth seas, heaving rhythmically on rolling waves, or pitching about chaotically in a storm, the most important question is, are you and the boat still afloat the next day? I therefore do not believe that there are many different kinds of community stability, but instead different aspects to the likelihood of persistence, and different ways to measure it.
In our paper we looked at one particular aspect of stability, commonly termed “local”. Let me explain why. Imagine our community is represented by a small ball, and its state is represented by its position on a landscape (Fig. 1; scientists love to imagine states as positions on an imaginary landscape). The landscape is rugged and hilly, and is shaped by the environment. If our ball is on a slope, it won’t stay there for very long, and its state will change. It is unstable. If it is located at the bottom of a basin though, then it will remain there, as long as nothing disturbs it. It is stable. If it is displaced by a small amount, remaining in the basin’s depression, then it will roll downhill and return to the bottom of the basin as soon as the displacing force is removed. Interestingly, with a little care one could also balance the ball on one of the peaks, and it will remain there, but that position is precarious and fragile. Any relatively minor force would serve to start a downhill roll. The basin is an “attractor“.
Now, there are a number of limitations to using local stability to describe the behaviours (dynamics) of which your community is capable. A perhaps obvious one is what happens as you increase the distance by which the ball is displaced. One possibility is that the community does not return to the basin of origin, but specifically what does happen to it depends on the topography of the landscape. A slightly more subtle set of questions, and the ones which we pursued, is what happens to the community between the time at which it is displaced (a little), and its return to the bottom of the basin? Is it a simple, Sisyphusean roll back down to the bottom of the basin? Does it happen quickly? What if the ball is kicked again before it’s finished rolling? These are important questions to ask when the planet is undergoing a slow, persistent environmental meltdown as it did 252 mya.
There are probably many interesting and important transient dynamics between departure and return. These can be very difficult to predict. To appreciate this, let us agree that our community really isn’t a ball at all, but is better described as a large collection of balls (species populations), many of which are connected to each other with ropes, pulleys and springs. The contraption now could even amplify a displacement, weaving about the slope, perhaps shifting to a new basin, or losing species along the way. These transient dynamics might be fairly common in real communities, and communities might in fact never really spend any time at the bottoms of basins, instead rolling about, tracing out complicated pathways in response to displacing forces, according to their system of species, ropes and springs.
So, what did our South African ecosystem do 252 mya as the planet became less and less hospitable?
At the heart of our paper lies a model framework which we devised for analyzing fossil food webs. I stated in the previous post that our main question was “How would those food webs (important parts of the paleoecosystems) have responded to everyday types of disturbances, on the short-term, as the planet was busily falling apart?” We could approach this question in several different ways if we were working with modern food webs. We could conduct manipulative experiments with simple mock-ups of the food web, using for example some of what we believe to be the key species to represent the community. Or, we could conduct large scale manipulations, such as removing a species entirely; but that is very difficult to do, or to obtain permission! Or, we could measure variables such as species population sizes, how species interact with each other, and so forth, to then conduct numerical analyses and simulations. None of those approaches are available to us when dealing with ancient, extinct ecosystems. Therefore, what we did instead was to use the most accurate information that we have for each paleoecosystem, which consisted of categorizing species into “guilds”. Here, a guild is a group of species who shared the same habitat, and potentially shared the same predators and prey. The “potentially” is based on our best interpretations of the ecologies of those extinct species, because without actually being there to witness their interactions (back to that Tardis again), we cannot be sure. The result is usually something like the box figure above. Even with species lumped, you can see how complex and busy the system would have been! And from this guild-level model, we can then construct many many different food webs, tweaking the specific links between species. An example is shown in the second figure.
Now, the number of different food webs that you could generate based on even a modest number of species, say 20, and a few guilds, is astronomically large (in fact beyond astronomical). The important thing, however, is that all of them would be consistent with the guild scheme. Let me give an example. Say we did a guild scheme for the modern African savannah. We would be justified to some extent to place lions and hyaenas in the same guild. We might not know exactly which antelope species (for example) each predator species was preying on, but we would never draw a food web where antelope were preying on lions, hyaenas or each other! So, what we have done for our food webs is constructed a mathematical space that contains all the food webs which could possibly have existed in our paleoecosystem. In other words, we have taken the full set of food webs that could be constructed for a certain number of species, and constrained ourselves to consider only those that are consistent with our accurate knowledge of the guild structure.
This still doesn’t solve the problem of how those food webs would have responded to various types of disturbances in the distant past. And in fact, we really cannot solve that problem, so we did what we think is the next best thing. We asked if there was anything special about those food webs, compared to any others that were not consistent with our guild structure. In other words, what if the ecosystem had evolved a bit differently, and comprised species a bit different from what we actually observe in the fossil record? We considered a number of such alternative models, differing from the real ecosystem in ways such as moving species around in the guilds, or moving guilds and the interactions between them, or removing guilds altogether. And each time we did that, and generated a food web from the new guild scheme, we examined the stability of the food web. Exactly what we mean by stability, and how we measured it, will be the subject of the next post.
One of the main motivations for our most recent paper (available here) was to gain insight into how modern ecosystems might behave in the future as they are subjected to increasing human-driven stresses. “Global change” biology is an emerging field that seeks to understand how the biosphere will change in response to factors such as ongoing climate change, habitat loss, landscape transformation, and so forth. Much of the work in this area rightfully focuses on measuring change, working to understand how modern ecosystems work, and projecting how they might respond in the future. The effort is ongoing, and includes theoretical work, controlled experiments, and uncontrolled impacts on natural systems. A limitation of these efforts, however, is the magnitude of the changes that are available for study. For example, we can observe how species are moving in space right now in response to rising environmental temperatures, or how they are adapting (or not) to drought conditions, but we cannot observe how they will respond in the future as those stressors continue to increase in magnitude. No one realistically expects the responses to increase linearly; we fully expect nonlinear, hard-to-predict, surprises. That was the message of an earlier paper, and a focus of a lot of current work on critical ecosystem transitions. One way to address this concern, and the one that we’ve taken, is to look back into Earth’s past, to times when the planet was similarly undergoing major changes. Those were natural experiments; times when ecosystems were subjected to extreme environmental stresses. The problem there of course is that we don’t have a Tardis, and all our information has to come from evidence that has been preserved in the geological record, and our ability to interpret it. Yes, the natural experiments were performed, but as I like to say, either no one kept notes, or the notebook was chewed up by the family dog before anyone had a chance to read it.
So where does that leave us? It leaves us with an incomplete record yes, but it’s also the only record of how the biosphere has responded to truly dire circumstances. Our challenge is to take this incomplete record, and to extract from it data and ideas that are useful for forecasting how the biosphere might respond to future dire circumstances. In the case of our present study, we were able to take advantage of first-rate field paleontology, first-rate organismal paleontology, recent developments in theoretical ecology, and to combine those with our own methods for reconstructing paleo-food webs. And the main question which we were interested in was, “How would those food webs (important parts of the paleoecosystems) have responded to everyday types of disturbances, on the short-term, as the planet was busily falling apart?”
And the planet really was in trouble at the end of the Permian 252 million years ago. Siberia had opened up in one of the most magnificent episodes of volcanism in the last half billion years. Recent dating suggests the volcanism started about 300,000 years before the marine extinctions, and may have continued intermittently for another 500,000 years after. The knock-on effects probably included greenhouse warming, sulphurous atmosphere, ocean acidification and reductions of oceanic oxygen concentrations. In southern Africa, the location of the terrestrial ecosystem which we studied, the stage was set for a catastrophe of global proportions.
Yesterday, Ken Angielczyk and I published our most recent paper on the Permian-Triassic mass extinction (PTME) in the journal Science. In a nutshell, we examined a series of paleocommunities spanning the extinction, from the Late Permian to the Middle Triassic, and modelled the stability of their food webs. We compared the models to hypothetical alternatives, where we varied parameters such as how species are divided among guilds, or ecological “jobs”, and the numbers of interactions that species have. One of our very interesting discoveries is that the real food webs were always the most stable, or amongst the most stable of the models, even during the height of the extinction! That’s remarkable, given the devastating loss of species at the end of the Permian. Our other discovery is that the ability to remain highly stable during the extinction stemmed from the more rapid extinction of small, terrestrial vertebrate species. That’s not something we would predict given our experience with modern and ongoing extinctions, where larger vertebrate species are considered to be at greater risk. And finally, our last interesting observation is that the early recovery, the immediate aftermath during the Early Triassic, was an exception to the above. That community was not particularly stable, which seems to have been the result of the rapid evolutionary diversification of the extinction survivors, and the arrival of immigrants from neighbouring regions.
Some aspects of the paper are quite technical, and take advantage of fantastic new paleontological data and recent developments in theoretical ecology. Therefore, over the next few posts I’ll go through what we did, and how we did it, using a more “plain language” approach. In the meanwhile, the paper was covered by a number of news outlets, and here’s my favourite!
“5 things we learned from the mass extinction study that’s “the first of its kind”“, The Irish Examiner.
“Habitat Earth“, the new film by the Visualization Studio at the California Academy of Sciences opened this weekend in the Morrison Planetarium. The film documents the ecological interactions that take place continually in natural systems, featuring San Francisco Bay, a northern California kelp forest, and redwood forest watersheds in the northwest of North America. I was one of the science advisers and content persons for the film and am simply in awe of the visualization team. The science is authentic and researched in detail, but most impressive is the sheer amount of data incorporated and visualized. These data range from well-known ecological stories such as the sea otter role in maintaining diversity in kelp forests, to the thousands of food web interactions from my San Francisco Bay food web dataset, to documented tracks of thousands of migrating species and human ship traffic. It’s a masterpiece of science visualization, and I was very happy to be a small part of it. Here is a short trailer to the film, narrated by Frances McDormand. In the next few posts I will link to interviews with a number of the scientists involved. In the meanwhile, enjoy the trailer and, if you are in San Francisco, please stop by and see the film in the world’s largest all-digital planetarium dome!
RESILIENCE AND STABILITY OF PERMO-TRIASSIC KAROO BASIN COMMUNITIES: THE IMPORTANCE OF SPECIES RICHNESS AND FUNCTIONAL DIVERSITY TO ECOLOGICAL STABILITY AND ECOSYSTEM RECOVERY
ROOPNARINE, Peter, Invertebrate Zoology and Geology, California Academy of Sciences, 55 Music Concourse Dr, Golden Gate Park, San Francisco, CA 94118, email@example.com and ANGIELCZYK, Kenneth D., Department of Geology, The Field Museum, 1400 South Lake Shore Drive, Chicago, IL 60605
A central question of the P/Tr extinction is the manner in which Permian ecological communities collapsed and E. Triassic ones were built. The end Permian Dicynodon Assemblage Zone (DAZ) has recently been resolved into 3 phases of the extinction spanning ~120ky, followed by the E. Triassic (Induan) Lystrosaurus Assemblage Zone (LAZ), offering an opportunity to examine the ecological dynamics of extinction and recovery in enhanced detail. We do this with 2 modelling approaches.
The first model assumes that populations exist in an energetic balance between consumption and predation. Communities are modelled as stochastic variants sampled from a space defined by species richness and functional diversity. Paleoenvironmental data from the DAZ indicate an increasingly seasonal, arid and drought-prone environment. The models were perturbed by simulated reductions of primary productivity. Results show that DAZ Phase 0 (Ph0) was a robust community resistant to low-moderate levels of perturbation with a well-defined collapse threshold. DAZ Ph1 and Ph2, however, exhibit highly variable responses and are significantly less resistant. LAZ similarly exhibits highly variable responses across minor variation of model configurations.
The second model assumes that communities are locally stable, i.e. minor perturbations are followed by asymptotic returns to equilibrium. During this return, however, communities can exhibit transient behavior during which perturbations can be greatly amplified. Amplification is likely to be important in unstable environments when the frequency of perturbations is shorter than the return time to equilibrium. Applying this model to DAZ and LAZ communities shows that the Karoo ecosystem became more limited in its responses to perturbation as the P/Tr boundary was approached, with Ph1 and Ph2 communities exhibiting very little transient behavior. LAZ in contrast exhibits increased transience.
The energetics and stability models are reconcilable in a history where the Karoo ecosystem became more ecologically stable as the extinction unfolded, yet more sensitive to cascading effects of species extinction and reductions of productivity. The Induan ecosystem was an unrecovered one, sensitive to both extinction and minor ecological disturbances.