MIGUEL BOYAYANBiologists are now managing to discover and to understand with precision the impacts of climatic changes on animals and plants, the expansion of insects harmful to agriculture and the real danger of viruses that cause diseases. The forecasts about situations like these, which endanger the future of humanity, are coming up from a series of computer programs and comprise a relatively new area for research, called predictive modeling, in which stands out the work of a discreet American biologist: Andrew Townsend Peterson, from the University of Kansas, Natural History and Biodiversity Research Center, United States.
At the age of 38, with 13 articles already published this year, Peterson took part in the development of SpeciesAnalyst, a program that integrates electronically the biological collections (of animals and plants) from research institutions from all over the world. In the two months that he was in Brazil, the Ohio-born biologist worked with the team from the Reference Center for Environmental Information (Cria) – the institution responsible for the maintenance of SinBiota, the information system of Biota-FAPESP, the program that is mapping the flora and fauna of the state, – on the development of the São Paulo model for predictive modeling: SpeciesLink, conceived to gather together the research material kept at 12 São Paulo institutions, and, at a later stage, to put Brazil into contact with the incipient world-wide information technology network for diversity.
His stay – the longest since he started to work with Brazilian researchers in 1999 – made it clear how modeling can help to solve some of the country’s specific problems, by revealing, for example, the shift of leishmaniasis, which is leaving the poor and rural areas and approaching the cities. In this interview, granted to Carlos Fioravanti at Cria’s offices in Campinas, Peterson tells how this area of studies is expanding and sometimes making corrections to the path of research all over the world.
What is predictive modeling?
Basically, any abstraction from the natural world with some power for foreseeing events on the basis of general principles. We use specifically the term ecological niche modeling, which concentrates its power for foreseeing on phenomena that refer to the distribution of biodiversity.
In one of your more recent articles, published in Nature, you guarantee that the forecasts about world-wide changes in climate, which are already not very good, are underestimated. Could you explain why?
Actually, the forecasts of the impact of climatic changes on biodiversity were estimated inappropriately. There is a general biogeographical relationship according to which altitude and latitude are more or less equivalent. So, something like 100 meters in altitude is more or less equivalent to a shift of 800 kilometers closer to the pole. So that the rise of one degree in the Earth’s temperature is essentially the same thing as pushing the climatic zones in the direction of the poles.
From then on, the climatic changes on Earth are estimated by groups of species from an ecosystem, indicating where the climatic zones are going to and the size of the populations before and after the rise of one degree, let’s say. The problem with this approach is that the ecosystem of Amazon, for example, is not homogeneous, but complex, made up of millions of species of plants, animals and microorganisms. One of the most important points in the article in Nature was to show that species have very particular reactions. If the temperature goes up, they should migrate to the poles, but this does not always happen. Some go literally in the direction of the equator, others to the east or west, so that we cannot successfully foresee the behavior of them all, when the climate of an ecosystem is altered.
20,000 years ago, at the end of the last ice age, there was a forest of beeches and firs in the United States, in an area equivalent to the state of São Paulo. hen the climate changed, these species moved to different regions, and nowadays the firs are in the north of the United States and south of Canada; and the beeches, in the east of the United States. There is no place where beeches and firs can be seen together.
Is it now possible to tell what is going to happen with each species?
It is possible to estimate whether the impact from global warming would be small or large, analyzing the number of species that would go into an ecosystem or leave it. In Mexico, the extinction rate is 3%, but the number of species that would change community is frightening. In some cases, then, the effects of climatic changes can even be less serious than used to be foreseen, as happened with the estimates for extinction. But, on other occasions, far more serious effects can be expected.
I also explored this effect in flat areas vs. mountainous areas to discover the relationship between the effects of climatic changes and topology, and I discovered that plants from mountainous areas may lose ground with climatic changes. This is the case of the high altitude fields at the top of the Mantiqueira mountain range, which are running the risk of disappearing, for being at the top of the mountains. If the climatic zones go up, where will the fields go to? There is nowhere higher. On flat land, as in São Paulo or in the Amazon, it looks as if the problem is one of movement.
Could you explain it?
Imagine that we are looking at a tree at the Amazon. The most suitable climatic conditions for this tree may not change, but move 400 kilometers away, for example. In this case, the tree, which is not able to move, may find itself in a place outside its favorite climate. So we are beginning to understand the negative effects of climatic change on certain kinds of topography. Mexico, Canada, the south of India, the United States, and now Brazil, are areas in which we are analyzing the kinds of effects of the changes in climate and the impact that can take place on conservation plans. We have done preservation plans based on the present, but the rearrangements made by the climate can be so severe that the current patterns of diversity may change entirely.
By taking climatic changes into consideration, a totally different picture may emerge of conservation priorities. For example, the distribution of the cerrado (kind of wooded savanna) in São Paulo is regarded as peripheral and degraded. However, climatic changes can make the climate in the state more favorable for the development of this kind of vegetation, making the remains of the cerrado in the state far more important for the conservation of the biome.
What is the most important concept for understanding environmental changes?
In my opinion, the key is to understand how one species in particular interacts with the environment. It is what we call an ecological niche, which is the specific habitat in which each species lives. There are researchers who give priority to the processes of the ecosystem and are more interested in studying the behavior of ecosystems like the Atlantic Rain Forest or the Amazon. They want to know how each environment behaves, how it manipulates oxygen and carbon dioxide and what it excretes. They handle each ecosystem as a great organism. One model does not exclude the other. Both are important and complementary.
How can predictive modeling help in another area, the dynamics of diseases? In another recent article, you show how a group of kissing bugs, the triatomas, could transmit Chagas’s disease…
When an organism sets itself up in anther one where it shouldn’t be, disease arises. If they shouldn’t be in our organism, we get sick. It is different from what happens with the Escherichia coli bacteria, which have a longstanding relationship with us and no longer cause disease. The problems arise when we come across a strange parasite, like the rabies virus. The natural reservoir for rabies is chiefly the vampire bat, to which the virus does not do any harm, for having a long relationship with it by now. When the rabies virus infects human beings, who do not have this old relationship with the virus, we get sick and may even die. Reconstructing the models of interaction between the reservoir the vector, we can discover the dynamics of the disease. Afterwards, it becomes easier to define the areas of risk.
What is needed to apply modeling to the study of diseases?
You have to have the environmental data and electronic maps of climate, topography and vegetation, the so-called geographic information system, and the spots where the species occurs. Here in Brazil, or rather, here in São Paulo, we can see an interesting situation with leishmaniasis, a disease that always attacks poor rural areas, and suddenly is becoming a worrying disease in poor urban areas as well. In addition, it is moving southwards. I am working with Jeffrey Shaw, from the University of São Paulo, to understand the dynamics of the disease better.
I am also working with people from the Oswaldo Cruz Foundation on Chagas’s disease, but with another group of triatomas. This group of insects is regarded as a single species, though there are doubts about this, and my colleagues at Fiocruz were interested to know whether the members of this group belong to the same ecosystem. We analyzed populations that were a little different morphologically, and we showed that each one of them was ecologically different. We are beginning to explore the details of the ecology and the geography of diseases, with the objective of determining the correct units of management and defining a program for controlling triatomas in the Northeast.
Is a new approach to ecology arising?
The foundations of this approach were laid in 1917, with Joseph Grennel, the founder of Berkeley Museum. The approach, then, is not a new one, what is new are the data and the tools. For the scientific community, it has been a long learning process. When I decided to begin to explore these methods, in 1995, I sent four proposals of work to the National Association of Sciences of the United States. They were all rejected.
Is the worst over now?
Yes. The battle for public access to data on biodiversity is almost won. Until a few years ago, very few researchers really believed that information about biodiversity is a patrimony that belongs to us all. In those days, the predominant idea was to restrict access to information, in order to stress the importance of the museums. Today, people who think like that are an exception. But there still is the scientific battle, which is to establish this new methodology of ecological niche models. Either we present good scientific models and people are convinced, or we fail. I am not bothered any more to make people share data with me. I am the curator of a collection with some 100,000 birds from all over the world, but I regard them as part of a world patrimony. Actually, this collection does not just belong to my institution, but also to the countries where they specimens came from.
And what about the limits to modeling?
There are computing limits, of time and speed. I would love to make models of the whole world with a resolution of one meter and a thousand environmental covers. There is noting to prevent us, except speed and time. In practical terms, there is a restriction on the area for analysis of one million pixels. I have already done 23 million pixels, but the program gets very slow. We are limited to a million pixels and 20 to 30 environmental layers, but we can improve. Ricardo Scachetti Pereira, from Cria (the Reference Center for Environmental Information) is helping to create DesktopGarp, a more friendly version of Garp (Genetic Algorithm for Rule-set Prediction), created ten years ago by David Stockwell at the Environment Australia agency and perfected at the San Diego Supercomputer Center.
Garp provides for different phenomena with limited distributions and in different places, but it was very difficult to use. We often talk about physiological tolerance, which is the tolerance to temperature of a given species, but we can rarely ask how these tolerances change in accordance with the distribution of the species. In work on the Ebola virus, we used these new tools for the first time and noticed that the range of activity of the virus was determined in some places by precipitation and in others by temperature. Garp permits us much greater complexity for characterizing the area of incidence of a species. I can define the parameters for analysis, leave the computer working on its own for a month, and afterwards have 10,000 ecological niche models. No other program has this capacity.
What is your work with Cria like?
The purpose of my visit is to develop applications of interest to Brazil, in collaboration with local researchers. In these two months that I’ll be staying here, I hope that we will least be capable of having a good look at the state of São Paulo and start to study the Atlantic Rain Forest and the cerrado. Cria has been a real leader, by collecting a fantastic set of data about biodiversity in São Paulo, without the need for physically gathering the data together in a single place. The data stays in the universities and museums, but it is integrated by the Internet.
What stage is SpeciesLink at?
SpeciesLink is a distributed data network, conceived at Cria to connect 12 research institutions of the state of São Paulo and to integrate the data with that of other institutions around the world. When it is ready, in a few years, it will allow data to be shared about the species occurrence , which is a critical resource that we needed, to take these models forward. The group from Cria adapted SpeciesAnalyst technology, to handle the specific needs of São Paulo, but SpeciesLink also works in an independent manner, with the idea that all the networks are going to talk amongst themselves.
How many networks of this kind are there at work in the world?
Five. One in Australia, another in Mexico, a third in Europe, Species Link in Brazil, and SpeciesAnalyst, developed at the University of Kansas. They are very different networks, but we are at a stage of integration in which the solutions converge into a single solution. think it is heartening to see people from five countries, including some from the so-called developing countries, talking to each other on an even footing. The representatives from each of them are collaborating in a project DiGIR (Distributed Generic Information Retrieval), a common technology for the future of distributed biodiversity networks. In the next few months, the people from Mexico, Kansas, Berkeley, from Cria and Australia are concluding DiGIR, it will be possible to integrate the networks, and SpeciesLink will be able to see not only the 12 institutions of the state of São Paulo, but also the 80 from SpeciesAnalyst and the 30 from the Mexican network. The data will grow in richness in a remarkable way.
In your view, what are the priority issues for Brazil in this area?
There are some very interesting issues. One of them, which is a priority world-wide, is forecasting an analysis of basic species, which is closely associated with the movement of populations. Human beings are moving much more than 50 years ago. This movement makes species move as well, creating new problems. There is, for example, an insect called Homalodisca coagulata that transmits Pierce’s disease, caused by one of the varieties of Xylella fastidiosa. It comes originally from the south of the United States, but it has invaded California, to the west, and has become a dangerous pest not only for citrus trees but also for grapevines. To discover what the potential is for the Xylella that attacks the grapevines in California to invade South America, by means of these insects, we applied the models in the southeast of the United States.
And what happened?
On the native side, the tests were very precise. Applied to California, without overlapping points of infestation. The model indicates significant correspondence with the statistics. Not only can the native distribution be seen, but also the distribution of the invader. We projected the model for South America, worried about the citrus plantation in the state, but citruses do not appear to be very vulnerable to this insect. But regions like Salta, in Argentina, or even Rio Grande do Sul, seem to be very vulnerable.