Keynote: Defensible Modeling of the Biosphere

Опубликовано 12 августа 2016, 2:59

To manage the planet on which we all depend, we need to predict the future outcome of various options. How would biofuel subsidies affect crop prices affect deforestation? CO2 emissions affect climate change affect fire? At present, we cannot make such predictions with any confidence. But, as I'll show in this talk, a computational approach to environmental science can change that. I'll explain how we built the first fully data-constrained model of the terrestrial carbon cycle, using Big Data, Cloud Computing, and Machine Learning. And I'll demo similar models for global food production, Amazon deforestation, and bird biodiversity. The prototype tools on which these models have been built - e.g. FetchClimate, Filzbach, World Wide Telescope - are freely available, and will hopefully allow other scientists to adopt a rigorous approach to modelling the complexities of the biosphere.