Modeling the Future
In the 1920s, when prickly pear cacti were an invasive species in Australia, huge fields of the plant took over the native landscape. To combat the cacti, the Australian government shipped in cactus moths. What happened decades later illustrates the power of harnessing ecological information to predict change.
The cactus moth, Cactoblastis cactorum
, native to Argentina, had a taste for prickly pear and helped control the problem for many years.
However, by 1989, the cactus moth had island-hopped its way from Australia to Florida, where hundreds of species of endangered cacti live. Scientists worried that the moth would destroy cacti populations in Florida and then move on to Mexico, where cacti are an important part of the agricultural economy and an essential food source for cattle and people.
Another insect that Soberon intends to model: the butterfly Callophrys xami
“We had to know, ‘Where is it going to attack?,’” said Jorge Soberon, research scientist at the KU Biodiversity Institute. “We had to predict the distribution of this little moth ahead of time.”
The cacti of concern were spread throughout so many countries that scientists needed to know where to focus their efforts to protect them. Soberon used a type of research called predictive modeling to identify where the moth could travel and where it could flourish, so that the cacti in those areas could be protected.
Layers of information
Predictive modeling is a way of taking information from many different fields — ecology, botany, genetics, meteorology, even sociology — and layering the information together, via computers, to predict future outcomes. This idea may seem pretty sci-fi, but it’s extremely practical, said Soberon.
"Through predictive modeling, we can predict where diseases are likely to migrate by tracking the insects that spread them – diseases such as malaria, Ebola, bird flu, West Nile," he said. "We examine the patterns of pests of corn, oranges and grapes."
But how does predictive modeling work, exactly? Soberon offers another example: Say you were studying the plague to find places where a possible outbreak could occur. “To find out where the plague could survive, you’d look for places with rats and fleas,” Soberon said.
Next, you’d look at the climate of places where the plague was successful in the past. Then you would search for regions that were similar to those places in average temperature, the hottest and coldest months, precipitation and humidity. You’d also factor in human and rat migration: where might an infected rat or human travel? Where are there geographic barriers to both, such as mountains? When looking at such a question, Soberon’s research team takes into account nineteen of these different variables.
A graph that models the temperature, precipitation and seasonality in the Americas.
Finally, you would enter all of this data into a computer program, which layers them together over a topographical map. The computer program identifies the areas of highest risk.
At that point, you could begin to act with this information. If you were trying to prevent plague, you would work with the World Health Organization, non-government organizations and even local governments in the areas of concern to educate the public and health professionals about taking precautions, identifying symptoms and administering treatments.
In the case of the moths in Mexico and Florida, their departments of agriculture were able to distribute materials and information, and volunteers from non-government organizations even went into areas at risk and removed moth eggs from cacti by hand.
This kind of success story speaks to the usefulness and accuracy of predictive modeling.
“We’re not just filling in historical gaps, but interpolating and looking forward,” says A. Townsend Peterson, ornithology curator at the Biodiversity Institute and KU professor of ecology and evolutionary biology.
Diseases on the move
Peterson focuses on predicting outbreaks of diseases. He has studied the Marburg virus, a deadly disease with symptoms similar to the Ebola virus. Named for Marburg, Germany, the site of the first outbreak in 1967, the disease infected 31 people that year who were mostly staff in a laboratory studying monkeys. The next infections were scattered across space and time: Johannesburg, South Africa, 1975; Kenya, 1980 and 1981; and the Democratic Republic of Congo, from 1998 to 2000.
By looking at climate data and paths of transmission, it is possible to predict the path of a disease worldwide. Scientists examine where a disease could travel, and where it could actually thrive. For example, a virus could infect a human who travels to Antarctica, but the virus wouldn’t succeed in the climate or have the population density to spread.
Peterson used predictive modeling techniques to examine the climate in Marburg and to locate similar climates. He was able to find some areas that had the potential for outbreaks, including Angola.
By the end of 2004, as the computer models had shown the possibility, there was a major outbreak in Angola. Since then, Peterson's research team has been able to estimate where each of the outbreaks would be, ahead of time.
“It has the potential for telling us not only where, but when,” Peterson said.
A tool for preservation
Cloud forests are a rare ecological formation found in the Sierra Mountains in Mexico. The Sierras are so tall that clouds form around the mountains only at certain elevations. The forests correspond to the rings of fog, rainclouds, and moisture that circle the mountains, “like a monk’s haircut,” as Soberon described them.
These forests are endangered by climate change and deforestation. Many non-government organizations are working to protect the forests, but this work is complicated. Because the cloud forests are so dependent on their rain clouds, and global climate is changing so quickly, the forests’ locations may be different in the future. Changes in temperature on the mountain could cause the cloud forest to shift up or down the mountain, or disappear altogether, for example.
Soberon uses modeling to predict these future scenarios so that non-government organizations can put energy and funding into the areas with the best chance of survival. He noted that the good news is that predictive modeling finds not one solution, but many; his goal is to find “the optimal solution,” he said.