Undesirable non-indigenous species and emerging infectious diseases of wildlife and humans are urgent environmental concerns, creating health problems and costing society millions to billions of dollars annually to control. Considerable effort has been invested in understanding the ecology and evolution of non-indigenous species and diseases, but there has been relatively little success at forecasting outbreaks. Indeed, despite the fact that outbreaks of pest species and outbreaks of disease exhibit similar patterns, these phenomena are commonly studied by separate research communities for the purposes of risk analysis and intervention. My research aims to develop techniques for risk analysis of species introductions and diseases outbreak while exploiting this conceptual overlap.
One zooplankton species I have studied is the spiny water flea Bythotrephes longimanus. Spiny water fleas prey on native zooplankton and have caused declines in plankton biodiversity in some North American lakes. We found that seasonal fluctuations in the abundance of this species are driven by climate through its effect on water temperature. As a result there are "windows" of "invasion risk" during which introduction of spiny water flea is most likely to result in establishment of a new population. Unfortunately, because dispersal and population growth of spiny water flea is so unpredictable, knowing exactly where and when this invasive species will show up is impossible. Quantifying the limits of predictability given current ecological theory is an important component of risk analysis.
An analogous situation occurs when trying to predict disease outbreaks. Because so few people are infected early in an outbreak, predicting the final size of a potential epidemic is very difficult until a relatively large number of people have become infected. But, by then it's relatively clear how serious the outbreak is and forecasting is of little use. Again, understanding fundamental limits to potential forecasting precision is useful for making decisions about how to intervene. Using the 2003 outbreak of SARS in Singapore as an example, we took this problem of disease forecasting a step further and asked how large an outbreak is likely to be when the conditions for disease propagation are continuously changing, for instance as we improve our diagnoses and implement more effective controls. It turns out that this "societal learning" aspect is a crucial determinant of how big an outbreak is likely to be. It remains now to identify strategies that optimally allocate resources to control outbreaks when they first begin.
