Earth Systems Lab 2026

Applied AI for Planetary Stewardship

Earth Systems Lab (ESL) brings together leading researchers in machine learning, Earth science, space exploration, climate science, disaster response, and related disciplines to tackle high-impact global challenges through applied AI.

Hosted in collaboration with the European Space Agency (ESA), ESL combines interdisciplinary research, rapid experimentation, and world-class mentorship in an intensive summer research sprint designed to create meaningful scientific and engineering outcomes.

Welcome to ESL 2026

The ESL Bootcamp is the launchpad for the research sprint, bringing participants together in Frascati, Italy, home to ESA’s Earth Observation Centre.

Over the course of ESL, researchers collaborate in interdisciplinary teams to explore ambitious ideas, develop AI pipelines, and create reproducible research outcomes with real-world applications.

Participants are selected not only for technical excellence, but also for their ability to thrive in collaborative, high-performance research environments.

What is Earth Systems Lab?

Earth Systems Lab is an applied artificial intelligence research cycle focused on advancing machine learning, data science, and high-performance computing for problems of material importance to humanity.

ESL tackles challenges across:

Earth science
Climate research
Disaster response
Space exploration
Planetary stewardship
Energy transition

The ESL’s core approach is pairing machine learning researchers with domain experts in intensive interdisciplinary teams.

While the research sprint itself runs for eight weeks, the full ESL research cycle extends over approximately twelve months, from challenge definition through to technical memos, scientific outputs, and deployable data products.

How ESL Works

Interdisciplinary Research

Each challenge team combines expertise from:

Machine learning and AI
Earth and climate science
Space science
Cyber-physical systems
Planetary science
Astrobiology and related domains

Teams are supported by faculty leads, technical mentors, partner organisations, and expert reviewers.

Rapid Iteration

ESL is structured around accelerated experimentation and continuous review. Teams rapidly prototype ideas, evaluate approaches, refine ML pipelines, and iterate toward reproducible outcomes.

2026 Research Challenges

MoEarths

Can we build on learnings from model switch out and uncertainty awareness to create optimal foundation model performance that can act as a unified estimator of truth?

There is no operational Earth-scale MoE (Mixture of Experts) system deployed today. The MoEarths (Mixture of Earths) challenge proposes a shift from selecting a single "best" model to a new paradigm of adaptive switch-out, prototyped during FDL 2025 (SHRUG) and FDL 2024 (MAESTRO). The applied aspect of this challenge is to look at an application that requires a transparent and uncertainty-aware Earth state estimator. MoEarths could provide the transparent third-party assessment required for forestry management or next generation carbon reporting and certification, addressing conflicting representations of biomass, land use, and carbon sequestration create a trust gap that complicates regulatory compliance. This project would be the first attempt to make a transparently arbitrated, uncertainty-quantified Earth state estimate.

Living Forests

Can we build a precursor foundation model of a dynamic, living system using ESA’s Biomass and Sentinel missions?

This challenge would develop the first AI Forest Foundation Model designed to learn the underlying physics of forests built on Biomass, Sentinel 3 and FLEX data. This effort would explore foundation model approaches that make predictions based on the latent structure in the embedding space. The approach effectively would produce a high-dimensional representation of the global forest, capable of simulating the outcomes of queries, its carbon carrying capacity and assessing resilience. A living forest model would work in concert with planned and existing SAR and optical foundation models in detecting biomass variation across diverse biomes, from boreal to tropical, by maintaining an internal representation of the forest in the latent space, helping to understand how the Earth’s living systems function and how they respond.

Tropical Cyclone Dynamics

Can we predict tropical storm intensity?

While recent work has successfully reconstructed 3D cloud structures at a three-kilometer scale, these models remain structural snapshots that lack the integrated thermodynamics and kinematics necessary to anticipate the destructive potential of rapid intensification events. This challenge invites the development of a novel multimodal AI framework that fuses 3D cloud morphology with temperature, precipitation and synthetic aperture radar (SAR) wind speeds. Wind speed fields derived from SAR will be paired with geostationary cloud-top observations and precipitation profiles from polar-orbiting sensors. These sources have different spatial/temporal resolutions and observation frequencies, which make this a bold challenge. Handling irregular grids will require sophisticated data fusion (for instance with graph neural networks or transformers and techniques for geospatial information encoding).

Research Outcomes

ESL outcomes are typically developed to mid-Technology Readiness Level (TRL), meaning:

AI and ML pipelines are validated in realistic scenarios
Results are reproducible
Scientific or engineering findings can be documented in technical memos or publications
Workflows follow best practices in software and data engineering

Community-Driven Excellence

A defining feature of ESL is its collaborative culture. Teams are encouraged to combine ambitious thinking with supportive working practices that enable creativity, resilience, and high-performance collaboration.

The ESL Research Cycle

Countdown Phase

1–15 June 2026

The Countdown Phase serves as the onboarding and orientation period for ESL.

Key goals

Meet fellow researchers and faculty
Establish team dynamics and communication practices
Build shared understanding of tools, terminology, and workflows
Gain access to compute environments and datasets
Prepare for the in-person Bootcamp

Key Sessions

Culture Session: 4 June

An introduction to the ESL community, collaboration principles, and team culture.

Tool Onboarding: 9 June

Introduction to communication platforms, project management tools, cloud compute resources, and best practices.

Initial Team Meetings: 1 - 15 June

Early exploratory meetings between challenge teams and faculty to align on expectations and begin preparing for the sprint.

Bootcamp Week

15–19 June 2026

Frascati, Italy

Bootcamp is the official launch of ESL.

Held in person in Frascati, Bootcamp brings together researchers, faculty, ESA stakeholders, and partners for an immersive week focused on:

Understanding challenge context
Exploring AI and ML methodologies
Building team cohesion
Establishing research directions
Learning collaboration frameworks and working practices

Throughout the week, participants take part in:

Science talks
Technical workshops
Practical sessions
Team exercises
Collaborative planning activities

Teams also begin developing their “Big Why”, the overarching motivation and ambition behind their research challenge.

Sprint Phase

15 June – 7 August 2026

ESL involves an intensive eight-week research cycle designed around rapid experimentation, prototyping, evaluation, and iteration.

Weekly Reviews

Teams participate in regular reviews with:

Faculty leads
Stakeholders
Industry experts
Scientific reviewers

These reviews provide:

Technical feedback
Research guidance
Evaluation of progress
Opportunities to refine direction
Exposure to external expertise

Research Structure

Week 1 — Bootcamp

Introduction to ESL, collaboration practices, and challenge exploration.

Week 2 — Exploration

Teams define and evaluate multiple research directions:

Safe Direction
Stretch Direction
Bold Direction

The goal is to identify both practical and breakthrough opportunities.

Week 3 — Development

Initial prototypes are tested and refined. Teams narrow toward their strongest concepts and begin developing formal technical plans.

Week 4 — Pipeline Development (“Max Q”)

Teams develop and stress-test their machine learning pipelines while receiving expert critique through focused workshops.

Week 5 — Calibration

Research directions are refined based on review feedback and experimental results.

Week 6 — Improvement

Teams strengthen evaluation methods, improve model performance, and prepare demonstration outputs.

Week 7 — Write-Up

Researchers prepare:

Technical memos
Scientific posters
Showcase presentations

Week 8 — Showcase

Teams present polished TED-style presentations during the ESL Live Showcase.

Live Showcase

7 August

The Live Showcase marks the culmination of the sprint phase.

Researchers present their work to:

ESA stakeholders
Scientific experts
Industry leaders
The wider ESL community

The showcase celebrates cutting-edge AI applications for planetary stewardship, climate resilience, and disaster response.

Technical Showcase

23 October

During the technical showcase with ESA stakeholders, the teams will present a deeper examination of the initial results of ESL. Researchers will share their results for approximately 30 minutes followed by 30 minutes of Q&A.

Sharing Phase

September – December 2026

Following the sprint, teams continue refining and publishing their work.

The Sharing Phase supports:

Technical memo completion
Scientific papers and posters
AI workflow documentation
Data product development
Conference presentations
Continued collaboration with partners

The goal is to ensure that outcomes are credible, reproducible, and ready for broader scientific and operational use.

At the end of Earth Systems Lab 2026, the teams have three key deliverables:

A scientific poster (using ESL templates)
A technical memo (using ESL templates)
A final technical showcase.

Faculty

Faculty are experts in either the scientific domain or machine learning (and sometimes both). Faculty form the core leadership team and are crucial to the successful development of ESL’s challenges.

Faculty support teams by:

Guiding research direction
Helping refine experimental approaches
Providing scientific and ML expertise
Supporting problem-solving and iteration
Encouraging ambitious thinking

Challenges are intentionally researcher-led and faculty-supported.

The objective is not to provide answers, but to create an environment where teams can discover breakthrough solutions together.

Collaboration and Culture

ESL is built on the belief that breakthrough applied AI research is fundamentally collaborative.

ESL encourages:

Interdisciplinary thinking
Constructive feedback
Curiosity and experimentation
Compassionate teamwork
Shared ownership of outcomes
Ambitious problem solving

A core principle of ESL is “co-opetition” - combining friendly competition with active collaboration and knowledge sharing.

The 7 Cs of ESL Culture

ESL encourages a collaborative mindset through seven guiding principles:

Community: Together we can do wonderful things.
Curiosity: Learn and Teach. Teach and Learn.
Compassion: Always be kind. Fear is the mind-killer!
Complexity: Your process together is not linear or predictable.
Courage: Aim for the stars.
Co-creation: Embrace diversity of thought to see a new possibility.
Comedy: Don’t forget to smile.

These principles help teams operate effectively under the pressure and uncertainty that often accompany high-impact research.

Deliverables

At the conclusion of ESL, teams produce a range of research outputs, including:

Scientific posters
Technical memos
AI workflows and pipelines
Data products
Showcase presentations
Conference and publication submissions

These outputs are designed to support reproducibility, scientific communication, and future deployment pathways.

Communication

Slack serves as the primary communication platform throughout ESL.

Participants are encouraged to:

Introduce themselves early
Engage actively with their teams
Maintain strong virtual working practices
Build rapport before the sprint begins

Strong communication and professionalism are considered foundational to successful collaboration throughout ESL.

Here are some of the previous researchers sharing a bit about them, and a few images, we highly encourage this to be done before bootcamp starts.

Why ESL Matters

ESL exists to accelerate the application of AI to some of the most urgent challenges facing humanity.

ESL combines:

Advanced machine learning
Scientific expertise
High-performance computing
Interdisciplinary collaboration

By bringing together researchers from around the world, ESL aims to create practical, ambitious, and scientifically rigorous outcomes that contribute to planetary stewardship and the future of human exploration.