IAMG2025
The 23. annual conference of the IAMG
October 08 - 13, 2025, Zhuhai, China

Short Courses

Comming soon

Data-Driven Methods for Spatial (primarily) and Spatiotemporal Regression↑

  • organized by: Dionissios T. Hristopulos, Technical University of Crete, Greece
  • Venue: coming soon
  • Date: coming soon
  • Time: coming soon
  • Min. number of participants: 5
  • Max. number of participants: 15
  • costs per participant: coming soon
  • Attendance Requirements: -
  • Recommended prerequisites: Knowledge of probability and statistics.

General course description

This one-day course will focus on probabilistic spatial regression methods. Such methods help to determine relations between spatially distributed variables and exploit them for spatial prediction. The focus will be on data-driven methods of regression in which the underlying relations are primarily inferred from the data (i.e., without assuming underlying parametric models). More precisely, we are going to discuss covariance-based methods such as kriging and Gaussian process regression (GPR) as well as the covariance-free method of stochastic local interactions (SLI) which employs the precision (interaction) matrix. The course will include introductory-level topics in spatial statistics and Gaussian processes as well as some more advanced topics.

Course objectives and learning outcomes

To gain familiarity with data-driven spatial regression methods. To develop an understanding of fundamental theoretical concepts such as the covariance kernel, the variogram function, stationarity, precision matrix, cross-validation, prediction and uncertainty quantification. To familiarize with the “zoo” of covariance kernels and their properties. To become acquainted with model estimation methods such as variogram fitting and maximum likelihood. To understand the principles of kriging, GPR and SLI and the computational requirements of each method. To highlight connections between geostatistics and machine learning methods of spatial regression. To develop the ability to use the data-driven regression methods in practical studies.

Course content

Spatial model for regression, Trend Models, Kernel Smoothing. Positive-definite functions (covariance kernels), Isotropy and types of anisotropy, variogram function Model Estimation (variogram fitting and maximum likelihood) Spatial Prediction with Kriging, uncertainty estimates, computational requirements Introduction of Gaussian process regression and relation to geostatistical framework The precision (interaction) matrix and Stochastic local interaction models Warped Gaussian processes and nonlinear regression What are the complications in the case of spatiotemporal datasets? Examples

Quantitative Causality Analysis with Information Flow: A Tool for Data-Driven Discovery in Earth and Geosciences↑

  • organized by: X. San Liang, Dept. of Atmosphere-Ocean Science, Fudan University, Shanghai, China; ivision of Frontier Research, Southern Marine Laboratory, Zhuhai, China; School of Artificial Intelligence, Sun Yat-Sen University, Zhuhai, China
  • Venue: coming soon
  • Date: coming soon
  • Time: coming soon
  • Min. number of participants:10
  • Max. number of participants: unlimited
  • costs per participant: coming soon
  • Attendance Requirements: Familiar with a programming language such as Matlab, Python, R, and should have it installed on their laptops.
  • Recommended prerequisites: -

Summary

Data-based causality analysis is a great help for scientific discovery. In this half-day course, we will introduce, in an easy way, a recently developed quantitative causality analysis, which has been widely applied in different disciplines, Earth and geosciences in particular, with interesting discoveries which would otherwise be difficult, if not impossible, to discover. We will take you for a brief stroll through the theory, with more emphasis on real applications. There also will be a lab lesson for you to gain hands-on experience in using the provided software. Causality analysis is an important and old problem lying at the heart of scientific research. Causality analysis based on data, however, is a relatively recent development. Traditionally causal inference has been classified as a field in statistics. Motivated by the predictability problem in physical science, it is found that causality in terms of information flow/transfer is actually a real notion in physics that can be derived ab initio, rather than axiomatically proposed as an ansatz, and, moreover, can be quantified. A comprehensive study with generic systems (both deterministic and stochastic) has just been fulfilled, with explicit formulas attained in closed form (Liang, 2016). In the linear limit, its maximum likelihood estimator is concise in form, involving only the commonly used statistics, i.e., sample covariances. An immediate corollary is that causation implies correlation, but the converse does not hold, expressing the long standing philosophical debate ever since Berkeley (1710) in a transparent mathematical expression. The above rigorous formalism has been validated with benchmark systems like baker transformation, Hénon map, stochastic gradient system, and with causal networks in extreme situations such as those buried in heavy noises and those with nodes almost synchronized (e.g., Liang, 2021), to name a few. They have also been applied to real world problems in the diverse disciplines such as climate science, meteorology, hydrology, turbulence, artificial intelligence, neuroscience, financial economics, quantum mechanics, etc.

Learning outcomes

  • The participants will have a basic understanding of information flow (IF), a quantitative causality measure rigorously derived from first principles, and its properties such as invariance upon nonlinear transformation, capability of handling latent confounding, self-loop, etc.
  • The participants will learn how to estimate IF from data, and use the resulting estimator to easily infer the causal relations, in a quantitative way, between two time series.
  • The participants will also learn how to deal with the causal inference with panel data.
  • The participants will be convinced of the easiness in making scientific discoveries through a variety of realistic applications.
  • The participants will gain hands-on experience in using the software as provided.

Prerequisites and materials

  • No coding experience required.
  • Some basic knowledge of calculus and statistics.
  • Familiar with one programming language such as Matlab, Python, R.
  • All datasets will be provided, together with open source codes.

Density data analysis for big data processing in geosciences↑

  • organized by: Karel Hron, Palacky University Olomouc, Czech Republic (onsite); Alessandra Menafoglio, Politecnico di Milano, Italy
  • Venue: coming soon
  • Date:
  • Time: coming soon
  • Min. number of participants: 5
  • Max. number of participants: coming soon
  • costs per participant: coming soon
  • Attendance Requirements: Attendees should bring their own laptops with R installed.
  • Recommended prerequisites: Multivariate statistics, Introductory courses in algebra and calculus, Basic knowledge of R.

Course description

The analysis of distributional data (probability density functions or histogram data) has recently gained increasing attention in the geosciences due to the possibility of automated data collection. Distributional data, expressed as probability density functions, are often observed by themselves (particle size distributions) or as a result of aggregation of large data streams (e.g., distributions of concentration values). This course provides an introduction to Density Data Analysis (DDA) from the perspective of Bayes spaces using methods of functional data analysis. Bayes spaces are mathematical spaces whose points are densities, generalizing to the functional setting the Aitchison simplex for multivariate compositional data. The course will give a brief overview of the concise theory of Bayes spaces as well as statistical methods developed in this setting. All methods will be illustrated with examples from real case studies in the geosciences and beyond.

Course objectives and learning outcomes

The course aims to introduce attendees to the principles and basic methods of density data analysis, how to apply them with R and how to interpret the results obtained. The course combines theoretical classes with practical data analysis.

Content

  • DDA and the geometry of Bayes spaces
  • Exploratory DDA and dimensionality reduction
  • Density-on-scalar, scalar-on-density and density-on-density functional regression
  • Bivariate densities and their orthogonal decomposition
  • Introduction to multivariate Bayes spaces

Free software

R Project for Statistical Computing: www.r-project.org