2015 AGU Fall Meeting

December 14 - 18, 2015

Unprecedented, catastrophic events like Hurricane Sandy and the ongoing California drought are forcing regional and local-scale engineers, planners, and decision makers to formulate resilience and adaptation decisions for the hazards of today and under climate change. But these stakeholders are generally ill equipped in terms of time and quantitative tools. The contemporary explosion of in situ, remotely sensed and climate model-output data provides enormous opportunity for statistics, data mining, and machine learning communities to develop methods and solutions to address this gap. But “out of the box” data science methods will not suffice. The dependencies inherent in spatio-temporal data, as well as the paradoxical “small data” extreme events residing within “big data” that are the most crucial, necessitate tailored methods. Advances are needed in areas including uncertainty quantification, predictive modeling, and physics guided data mining for space-time hazards and extremes, especially those aimed toward addressing stakeholder needs at relevant scales.