Proposal Description

Recent advances in data acquisition technologies have enabled the observation of environmental and geophysical systems at unprecedented high resolution, resulting in increasingly complex and unconventional data structures. These diverse data types include high-dimensional multivariate data, functional data, and circular data, often contaminated by atypical patterns, anomalies, or extreme values. Such characteristics pose significant challenges for traditional statistical methodologies, creating a growing need for robust, flexible, and modern approaches.
This session brings together recent developments in modern statistical theory designed to address the challenges posed by complex environmental and geophysical data, focusing on methods for anomaly detection, extreme-value analysis, and robust modelling across diverse data types. The session focuses mainly on presenting a range of methods for anomaly detection, extreme-value modelling, and robust estimation in the presence of atypical observations across diverse data types. Leveraging such recent methodological developments in statistics, researchers can extract meaningful insights from unusual, extreme, and structurally complex data, ultimately improving the scientific understanding and predictive capabilities of an environmental phenomena like climate change.
The session will feature three presentations that highlight innovative contributions in this area. The first presentation introduces a tail-calibrated quantile treatment effect framework for extreme event attribution, enabling more reliable estimation of anthropogenic contributions to extreme weather events. The second presentation develops robust estimation methods for circular data models, ensuring stability and interpretability in the presence of atypical observations. The final presentation presents an uncertainty-aware anomaly detection approach for seismic waveforms using conformal functional prediction to improve the detection of unusual seismic events in high-resolution time-varying signals by means of functional data analysis.
Collectively, these contributions will showcase the potential of modern statistical methods to analyze diverse complex, high-dimensional, and unconventional environmental datasets, with a particular focus on extremes and anomalous behaviours. The session highlights how innovative statistical methodologies can enhance the understanding of extremes, unusual or structurally complex phenomena in environmental and geophysical sciences. By attending this session, researchers will gain insight into the latest statistical methods and tools for analysing extreme values and anomalies across diverse data types, learning approaches that combine theoretical rigour with practical applications in high-resolution environmental monitoring, prediction and risk assessment.
Prof. Marian Scott, Emeritus Professor of Statistics with a long record in environmental statistics and the application of statistical methods to environmental sciences, enriches the session as a discussant, highlighting the potential future areas of research and challenges in modelling diversified environmental data.