Putting computer science to work on biodiversity
01 June 2009
As concerns about global warming grow, scientists are turning to sophisticated computational models to better understand and ultimately predict the impact of climate change and human activity on biodiversity.
Policy makers are asking tough questions about how climate change may effect populations of species and their habitat. Scientists haven't been able to provide definitive answers because ecosystems, and even each plant or animal, are so complex, and much about them remains unknown. Now, scientists are looking more frequently to hardware and software tools to help them make sense of the huge amounts of data they are collecting on species, and to give them a solid base from which to predict what might happen in the future.
“It's very important for science to become better able to predict the behaviour of ecosystems,” says Rich Williams, a researcher with Microsoft Research, heading the Computational Ecology and Environmental Science group in the Cambridge (UK) lab. “Environmental policy making needs to move away from the triage behaviour it often has, and better scientific understanding of ecosystems is vital to that goal.”
To help push fundamental advances in ecosystem science, Microsoft Research and its partners are working on common methodologies, computer models and scientific workflow software that eventually could be used broadly by the scientific community. They hope their work will lead to new decision-support tools to better inform policymaking by governments and nongovernmental organisations worldwide.
Ecosystems pose large computational challenges. They are huge entities composed of plants and animals that rely on each another in many ways, such as predators or prey, and that vary greatly in lifespan, movement and physical size. “You have to come up with useful abstractions, and not know every detail,” Williams says.
The first steps are to get baseline data on species and then to standardise those data. Much of the data about the biological world cannot be shared and compared easily. They exist in scattered scientific papers, library archives, and in labs where scientists sometimes won't share their hard-toget field data.
Jorge Soberon knows first-hand the magnitude of the task of compiling reliable data to use for predictions. A senior scientist at the University of Kansas, Soberon is working with Williams and others to create a model of the geographical distribution of species in Mexico's cloud forests. This unique forest occurs worldwide on tropical mountains where there is frequent cloud cover or fog. They are rich in biodiversity, and contain many rare or endangered species, some of which occur only in the forests. In Mexico about 20 per cent of the plant species of the country live in the cloud forests, which occupy about 1 per cent of that country's total area. That's 10 times the number of plant species in Mexico's tropical rain forests, Soberon says.
“These are big simulations with thousands of species and hundreds of thousands of data points.”
Cloud forests strip moisture from the clouds and fogs to provide water for millions of people who live in villages below them. They are being threatened by farming, alien species and probably the biggest factor, climate change, according to a report by the UK-based United Nations Environment Programme World Conservation Monitoring Centre.
There are already indications of populations suffering heat-related declines. Some butterflies are leaving the mountains to head north, or moving higher into the mountains. “The butterflies require a certain temperature and precipitation,” Soberon says. “If the trend in warmer climate continues, they will run out of suitable climate in 50 to 100 years.”
He adds, “Since [climate] is going to change, we need to know where the species are going to be more endangered, and where the best possibilities are for doing something about it. We hope the results will be used to do policy and decision- making in Mexico.” Soberon is studying the locations where species have been observed in Mexico's cloud forests and their environment, such as temperature and rainfall, so he can understand what conditions the species need in order to live. There already are at least 20 competing methods for using these data to predict the geographic distribution of species. His work involves constructing virtual environments where all the relevant variables can be controlled, and testing and comparing the existing methods for predicting species distributions.
Soberon says his database of species in Mexico's cloud forests so far has nearly 6,000 valid names of vertebrate and plant species and 30,000 records of observations of them, with the aim of having twice that number of observations. Those records can be used with mathematical modelling algorithms to predict the distribution of each of the species.
“We will fit the species into the current and very recent past climates, and extrapolate given the future of climates that people like the IPCC (Intergovernmental Panel on Climate Change) provide you with,” Soberon explains. “These are big simulations with thousands of species and hundreds of thousands of data points.” The results of the project will be available publicly so other scientists can apply the methods to their own databases.
Tracking the invaders
The need for tools that can be broadly used prompted work on invasive species models by Elizaveta Pachepsky, a specialist scientist at the University of California at Santa Barbara. She began developing an invasion tool while doing postdoctoral research at Microsoft Research Cambridge (UK) in collaboration with Ed Baskerville, a visiting software developer. The tool aims to be easy for anyone to use with existing data to help develop models that predict which species are invasive and how fast they spread. The initial focus is on invasive plants, but the technology could be used for other species.
“My idea was to develop a software tool to develop models without having to have a math background,” she says. Most models now use software code that is not published or that consists of a command line and lots of mathematical symbols, which can scare off users. Pachepsky is basing her tools on Microsoft's Windows operating system and Silverlight graphical Web presentation.
The goal is to enable prediction and prevention strategies that could help stem invasive species' disruption to ecosystems and economic damage. By the end of the 20th century more than 120 marine species were transported to Europe by gardeners, on hiker's shoes, in ballast water of ships, and by other means, she says. DAISIE, the first comprehensive database of invasive and alien species in Europe, lists about 9,000 alien species. Only about 10 per cent are invasive, meaning they cause economic damage or threaten biodiversity. Among DAISIE's “100 of the worst” invaders is Acacia dealbata, a fast-growing tree also called the silver or blue wattle. In the United States, an estimated 50,000 invasive species of all kinds cause an estimated $137 billion in damage per year.
Pachepsky says there are two components to the tool: one works with existing data (the spread of invasive species over several years), and another predicts what will happen in the coming years. Pachepsky expects a basic Web-based version and standalone version of the tool to be released within Microsoft this year and then to the scientific community as a free download.
The work done by Pachepsky and Soberon requires collaboration among scientists in different disciplines. “For invasive species, the tool uses math, software development, computation, statistics and ecology. These are five major fields of science and inquiry,” says Pachepsky. “Bringing that together in a way that is coherent is a big challenge in science and progress in general.”
Microsoft Research's Williams, whose group comprises scientists who are knowledgeable about technology, sees such multidisciplinary collaborations as fertile ground for problem solving. “It's a two-way street,” says Williams. “Computational tools and techniques open up new ways of tackling ecological problems. And biological ideas and the needs of scientists inspire the development of novel computational methods. It's really fantastic.”
He adds that the results can be valuable in other areas as well. “There's already some interest in the potential for tech transfer from the tools that are being developed into product groups at Microsoft.”