“Spartan spatial random fields and stochastic local interaction (SLI) models”

1.    Introduction to Spartan spatial random fields (SSRFs): motivation and properties

2.    New SSRF covariance functions: properties and examples of applications to real data.

3.    Discretized SSRF models for data on regular grids and connection with Gauss-Markov random fields.

4.     Parameter estimation and interpolation, uncertainty estimation.

5.     Geometric anisotropy: an introduction and anisotropic SSRF models on regular grids.

6.    Introduction to the Stochastic Local Interaction (SLI) model for irregularly spaced data: kernel functions, adaptive selection of the kernel bandwidth, sparse precision matrices and applications to big spatial data.

7.    Applications to synthetic and real data, extensions to space-time models, topics for further research.

For the workshop's program see here

The abstracts of the presentations can be found here

Αίθουσα Συνεδριάσεων Σχολής ΜΗΧΟΠ, Δεύτερος Όροφος

This workshop will focus on applications of statistical physics in the modeling of environmental systems and the analysis of environmental data. Statistical physics has traditionally focused on the behavior of the microscopic systems. Environmental processes, on the other hand, typically involve macroscopic systems. In spite of the difference in physical scales, statistical physics and environmental modeling both investigate partially determined systems and require a stochastic approach, thus creating the potential for interdisciplinary transfer of knowledge. For example, statistical physics influenced subsurface hydrology which adapted and incorporated methods and ideas from statistical turbulence (structure functions, perturbation expansions, closure schemes), statistical field theory (Feynman diagrams, Renormalization Group theory, replica variational approach), and classical statistical mechanics (Liouville’s theorem, fractional Brownian motion). To date, statistical physics concepts are also used in statistical seismology and climate research. In addition, statistical and machine learning methods originating in statistical physics are used to analyze and process complex patterns in environmental data. This workshop aims to highlight such contributions and to present novel ideas and methods motivated by statistical physics that can lead to new environmental applications.

Contributions to this workshop should represent new theoretical, experimental, or computational approaches to environmental modeling inspired by statistical physics. Environmental modeling is widely construed to comprise mathematical and statistical models of physical, chemical and biological processes that affect the Earth’s environment and the global climate. A non-exclusive list of topics of interest includes novel computational and theoretical tools for the analysis of large spatiotemporal data sets, innovative approaches to complex environmental processes that combine nonlinear and stochastic components, methods that address the interaction of multiple scales, approaches for the reconstruction and simulation of non-Gaussian natural or artificial media, applications of stochastic differential equations to environmental processes, higher-order upscaling methods, applications of complex network theory, statistics and stochastic models of extreme events, and estimation of long-range correlations in environmental systems. Physical phenomena of interest include (but are not limited to) the flow and transport of pollutants in the atmosphere, the ocean and the subsurface, natural hazards (earthquakes, fires, avalanches, and landslides), precipitation, global circulation and the climate.