LMAG 2025

 Overview

This is an informal workshop to be held at the Johns Hopkins University Applied Physics Laboratory (JHU/APL) from October 13-17, 2025 following the successful meetings of LMAG2020 and LMAG2023.
 

The idea of the 1st meeting (LMAG 2020) was inspired by impressive recent progress in three main disciplines, machine learning (L), data mining (M) and data assimilation (A) and the need to better understand the progress in concurrent research directions within different geospace disciplines to use each other’s methods, to combine the results (e.g., to advance Space Weather forecasts) and to find other ways of interaction, synergy and integration.

The main focus of the meeting is on:

Leveraging advanced tools to integrate observations with simulations, synthesize functional models, enable unbiased comparisons, and address the data sparsity challenge in Geospace.

Main topics :

• Understanding Geospace via LMA: How can LMA models be improved given more data? How do they improve our understanding of underlying physics (provide data discovery)? How do they provide nowcasting and forecasting of key Geospace parameters?

• Grey-box models: What are the pathways to combining LMA and first-principles approaches? In particular, can LMA be leveraged to facilitate data assimilation in and validation of geospace models?

• Explainable/interpretable methods for ML-driven geospace applications.

• Comparative LMA: How can LMA applications in other fields (e.g., solar physics, astrophysics, atmospheric physics, mission operations, and information technology) be used to develop and improve LMA methods in Geospace?

• LMA and (geo)space missions: How can LMA be applied to the formulation of future space missions, e.g., constellation-class missions, AI-enabled data selection, single-probe missions yielding constellation-class mission results? How can they be used to improve the present mission operations and data acquisition?

• LMA and information theory: How can the LMA analysis be improved using modern methods of information theory, complexity and uncertainty quantification?

• LMA tools, resources and infrastructure (e.g., data sets, facilities, codes, future planning)

Organizers: Savvas Raptis (JHU/APL), Anthony Sciola (JHU/APL), Harry Arnold (JHU/APL), Misha Sitnov (JHU/APL), Simon Wing (JHU/APL), Manolis Georgoulis (JHU/APL), Jon Vandegriff (JHU/APL)