Page - 7 - in options, Volume winter 2016/2017

research tools & methods 7winter 2016/2017 + optionswww.iiasa.ac.at iiasa research Cutting model uncertainty using natural selection P lants are not only responsible for the oxygen we need to live. They also provide our food, clean our air, and help mitigate climate change. Dynamic vegetation models track these vital services by giving estimates of the amount, type, and growth of vegetation around the world, and how it changes over time. This information underpins important estimates of carbon storage and climate change mitigation. However, the biosphere is an astoundingly complex place. To accurately estimate vegetation patterns, such models must incorporate a huge number of parameters— quantifying factors such as the growth rate of individual species, when they first make seeds, and how tall they grow to become. “Each of these parameters is associated with some uncertainty, and therein lies the problem,” says IIASA Evolution and Ecology Program Director Ulf Dieckmann. “You can make these models increasingly sophisticated, but with each dynamical element added, the uncertainty grows.” Ultimately, the uncertainty around what is put into the model dictates the confidence in the results that come out. Uncertainties that are too large can mean near‑worthless conclusions. To address this problem, IIASA has launched a new cross‑cutting project, Dynamic Vegetation Models: The Next Generation. Building on recent scientific advances by the project organizers, the initiative aims to predict plant dynamics and reduce uncertainty based on principles rooted in natural selection. For instance, consider how many seeds a tree produces. There is an optimal number for the tree— enough to effectively pass on its genes, but no so many that it overconsumes valuable resources. “Natural selection has a sweet spot for these physiological parameters, and that means that we can use optimization in the model itself to estimate them,” says  Dieckmann. Not all variables have a single optimum, however. The best height for a tree, for  instance, changes depending on the height of those surrounding it. For a tree encircled by short shrubs, investing less in growth and more in reproduction might be best. In work for the new project, the  rules governing these dynamics are explicitly incorporated into the model. This means that, rather than having to use uncertain estimates of the number of seeds a tree will produce, or of its maximum height, the  model itself can calculate these features. “Every plant you can see is the result of natural selection,” says IIASA researcher Oskar Franklin. “By building this ‘missing law’ into the models, we can substantially reduce the number of uncertain parameters and develop models that are both simpler and more accurate than current models.” As well as combining expertise from across the institute, the project will establish a working group made up of 20–30 experts from around the world, and will culminate in an international conference on the topic. DB Further info www.iiasa.ac.at/project/dynamicveg Oskar Franklin franklin@iiasa.ac.at Ulf Dieckmann dieckmann@iiasa.ac.at Elena Rovenskaya rovenska@iiasa.ac.at This new section of Options focuses on methodology—the  study of theory underpinning IIASA models and tools. Exploration and development of innovative mathematical methods and computational techniques help IIASA researchers understand the complex systems at the heart of global  challenges.