This work, accepted for publication in Statistical Science, leverages the Gaussian mixture perspective to propose extensions covering new classes of covariance functions for nonstationary (univariate or multivariate) spatio-temporal GRFs, as well as simulation algorithms for those that are currently missing in the framework of spectral simulation.
Geostatistics, extreme events and Machine Learning for the climate transition
A project at heart of two transitions
A digital transition
Massive and heterogeneous environmental data, making new geostatistical and Machine Learning methods necessary.
A climate transition
In response to changes affecting air, water, soil and biodiversity, unprecedented in their amplitude, speed and simultaneous nature.
The objectives of the chair in three lines of research
Develop effective methods and tools to process spatialized and temporal data in order to assess the impacts and quantify the risks associated with ongoing climate change.
Predictive methods
Develop predictive methods for spatial and spatio-temporal phenomena, capable of processing large datasets.
A toolbox
Develop innovative simulation methods for extreme events and risk assessment and distribute them freely.
Hybrid approaches
Hybridize the ability of geostatistics to interpolate in space and time and that of Machine Learning to extract links and knowledge.
Publications
To predict smooth physical phenomena from observations, spline interpolation provides an interpretable framework by minimizing an energy functional associated with the Laplacian operator. This work proposes a methodology to construct a spline predictor on a compact Riemannian manifold, while quantifying the uncertainty inherent in the classical deterministic solution.
Events
Workshop on "combined extreme events" at the actuaries' congress
At the 25th actuaries' congress, the Geolearning chair will host a workshop dedicated to extreme weather events
Geolearning Chair Scientific Day
The Geolearning chair held a scientific day at Mines Paris – PSL on April 8, 2026. All the supported projects were presented there, as well as emerging challenges.