Life is found almost everywhere on Earth, but each species is limited in the range of places and environments within which it can live. Understanding the distribution limits of a species is an old and fundamental problem in ecology. It is also an important practical problem.
We need computational tools to predict how the potential distributions of pest species, disease vectors and threatened species may change with the climate if we are to manage them properly.
One of the classic early texts in ecology is The Distribution and Abundance of Animals, written by Australian scientists Herbert Andrewartha and Charles Birch in 1954.
At the time it was written, tolerances and responses of animals to different environments were measured directly in the laboratory. These were then compared with weather station observations at particular sites. Distribution predictions involved hand-drawn contour maps based on this information.
Computers can help
In the 1980s, Professor Mike Hutchinson, from the Australian National University, revolutionised the field by developing computational methods to make continent-scale gridded climate layers from weather data.
Methods soon developed to mathematically describe the suitable environmental space of a species by querying those gridded layers at places where a species was known to occur. And the computed environmental spaces could then be projected back onto the landscape to predict the distribution of species.
This statistical approach to modelling the distribution of species has become one of the biggest fields in ecology today. A wide range of powerful computational methods are routinely used to understand where different species could occur under present climatic conditions.
These models are also being combined with the outputs of general circulation models to predict where a species might occur in the future.
Correlation vs causation
But care should be taken when using these correlative modelling approaches. They are statistical descriptions and are thus only reliable within the range of environmental conditions under which they were developed.
When correlative models are projected to novel environments, such as future climate change scenarios, they can be misleading.
This extrapolation problem has encouraged the development of mechanistic approaches to modelling the distribution of species. These approaches start, not with known distribution, but with measured tolerances and responses of organisms.
The field is now returning to biology-driven approaches from the days of Andrewartha and Birch, but with the more powerful tools and data now available.
My research group is focused on developing mechanistic species distribution models grounded in the physics of heat and mass exchange. We use these models to compute the inputs and outputs of heat to an organism at particular times and places in its habitat.
We develop algorithms of the behaviours and physiological responses that a species might use to buffer itself against harsh conditions. These include seeking shade, moving underground or changing colour.
We then compute outcomes, such as whether an animal could survive and, if so, how much time it would have to forage, how fast it could grow, how much energy and water it could obtain compared to what it lost and, ultimately, how many offspring it could have.
This brings us to helping save the koala. Koalas, being warm-blooded like us, keep a very constant body temperature despite changes in their environment.
But when it gets too cold, they need to expend extra energy to produce metabolic heat. And in hot weather, they need to lose extra water for evaporative cooling.
We can compute the energy and water costs imposed by the climate at a particular location, accounting for subtle responses koalas have. For example, koalas hug cool tree trunks to lose heat without having to spend water.
This cooling behaviour is something we discovered as part of our research. Knowing how much energy and water is in eucalyptus leaves, and how this converts physiologically into the production of offspring, we can estimate whether a koala could survive and reproduce at a particular place.
By repeating solutions of these kinds of calculations across grid of environmental conditions, we’ve produced maps of potential distribution.
The value of mechanistic predictions is twofold. First, we gain a greater understanding of what limits distributions. Second, we have a robust prediction that can be extended more confidently to novel conditions.
Our predictions for the koala involved solving the heat budget algorithm 40 billion times (115,144 locations × 20 years × 365 days × 24 hours × two climate scenarios) with the aid of the Victorian Life Sciences Computation Initiative super-computing facility.
These predictions indicate substantial range contractions towards the coast by 2070, especially in northern Australia. We can use this information to prioritise conservation reserves, and possibly translocation programs, to help koalas adapt to a changing climate.
The combined computational arsenal of correlative and mechanistic tools that ecologists now have at their disposal would amaze early ecologists such as Andrewartha and Birch, and will help humans adapt to future environmental change.