Spanning many orders of magnitude in time and space scales, Earth science data, from point measurements to process-based Earth system model output, are increasingly large and complex, and often represent very long time series, making these data difficult to analyze, visualize, interpret, and understand. An “explosion” of heterogeneous, multi-disciplinary data–including observations and models of interacting natural, engineered, and human systems–have rendered traditional means of integration and analysis ineffective, necessitating the application of new analytical methods and the development of highly scalable software tools for synthesis, assimilation, comparison, and visualization. For complex, nonlinear feedbacks among chaotic processes, new methods and approaches for data mining and computational statistics are required for classification and change detection, model evaluation and benchmarking, uncertainty quantification, and incorporation of constraints from physics, chemistry, and biology into analysis.

This workshop explores various data mining approaches and algorithms for understanding nonlinear dynamics of weather and climate systems and their interactions with biogeochemical cycles, impacts of natural system responses and climate extremes on engineered systems and interdependent infrastructure networks, and mitigation and adaptation strategies for natural hazards and infrastructure and ecosystem resilience. Encouraged are original research papers describing applications of statistical and data mining methods that support analysis and discovery in climate predictability, attributions, weather extremes, water resources management, risk analysis and hazards assessment, ecosystem sustainability, infrastructure resilience, and geo-engineering. Rigorous review papers that either have the potential to expose data mining researchers to commonly used data-driven methods in the Earth sciences or discuss the applicability and caveats of such methods from a machine learning or statistical perspective, are also desired.

Methods may include, but are not limited to cluster analysis, empirical orthogonal functions (EOFs), extreme value and rare events analysis, genetic algorithms, neural networks and deep learning methods, physics-constrained data analytics, automated data assimilation, and other machine learning techniques. Novel approaches that bring new ideas from nonlinear dynamics and information theory, network science and graphical methods, and the state-of-the-art in computational statistics and econometrics, into data mining and machine learning, are particularly encouraged.