Tutorial: Climate Change: Challenges for Machine Learning


December 8, 2014


Arindam Banerjee and Claire Monteleoni


Despite the scientific consensus on climate change, drastic uncertainties remain. Crucial questions about changes in regional climate, trends of extreme events such as heat waves, heavy precipitation, and mega-storms, and understanding how climate varied in the distant past, must be answered in order to improve predictions, assess impacts and vulnerability, and aid mitigation and adaptation efforts. Machine learning can help answer such questions and shed light on climate change. Similar to the case of bioinformatics, the study of climate change provides a data-rich scientific domain in which cutting-edge tools from machine learning can make a major impact. Further, such questions give rise to new challenges for the design of machine learning algorithms.

This tutorial will give an overview of impactful open questions about climate change, highlight recent successes of machine learning in this domain, and outline significant remaining challenges. Machine learning problems in climate change include prediction, reconstruction, causal attribution, analysis of quantiles and extremes, and exploratory data analysis. Challenges arise because the climate system is extremely complex, comprised of physical processes and their interactions, and the data is massive, high-dimensional, and spatiotemporal, with non-stationarity and potential long-range dependencies over time and space.