Key research themes
1. How can spatio-temporal distances incorporating causality and seasonality improve interpolation in sparse ecological datasets?
This research theme explores innovative interpolation techniques that address the entanglement of spatial and temporal variability in ecological data, which is often sparse and irregular. By defining a spatio-temporal distance metric that embeds causality constraints and allows for seasonal behavior, the aim is to improve the estimation of spatial fields over time. This has implications for ecological modeling where classical spatial or temporal methods fail to capture complex dynamics due to data limitations or non-stationarities.
2. What statistical frameworks, variogram estimators, and computational tools best support robust spatio-temporal data modeling and inference?
This theme focuses on the methodological advancements in spatio-temporal statistics for analyzing, modeling, and predicting complex environmental and ecological phenomena. It emphasizes the challenges posed by high dimensionality, data sparsity, and outliers, examining robust variogram estimation, covariance modeling, and computational frameworks—as well as the role of software such as R—in facilitating principled inference and prediction in spatio-temporal domains.
3. How can machine learning and dynamical systems methods address complex spatio-temporal dynamics for forecasting and interpretation in biological and physical systems?
This theme centers on the application of machine learning techniques, including convolutional neural networks, conditional random fields, and active learning, combined with physics-based and dynamical systems models to predict, estimate, and understand high-dimensional spatio-temporal phenomena. The focus lies on extracting meaningful dynamics from limited observations, enabling cross-variable estimation, and improving prediction accuracy in systems such as cardiac tissue excitation, neural dynamics, and general dynamical systems with complex spatio-temporal structures.