1 of 14

Earth System Models Pre-summit Intensive

Andreas Prein, Francina Dominguez, Ruby Leung, and Maria Molina

30. September 2024, University of Illinois Urbana-Champaign

X-SHiELD

Max Plank, DYAMOND

SCREAM

MPAS

NeuralGCM

DALL·E (OpenAI) visualization of a km-scale simulation of the coupled earth system

2 of 14

Outline

2

25.09.24

What we know and “agree” on

Potential Issues

Opportunities & open research questions

3 of 14

What we know and “agree” on

3

25.09.24

Our community would benefit from more organization

Daniel Klocke

4 of 14

What we know and “agree” on

4

25.09.24

Benefits of km-scale models

community

5 of 14

What we know and “agree” on

5

25.09.24

Model Requirements

community

benefits

Running on heterogenous architectures with accelerators

6 of 14

What we know and “agree” on

6

25.09.24

Code Structure

community

benefits

requirements

  • Software – Hardware co-design process
  • Collaborate with NWP community
  • Public-private partnerships for code development

7 of 14

What we know and “agree” on

7

25.09.24

community

benefits

requirements

code structure

8 of 14

Outline

8

25.09.24

What we know and “agree” on

Potential Issues

Opportunities & Open Research Questions

9 of 14

Opportunities & Open Research Questions�

9

25.09.24

Enhance speed and accuracy of model

Enhanced model resolution

Public-Private Partnerships

Community Building

Policy and Stakeholder Integration

Opportunities

10 of 14

Opportunities & Open Research Questions�

10

25.09.24

Opportunities

Research Questions

Balance between grid spacing and complexity

Validation of simulations and building trust with stakeholders

Coupling of individual components and new parameterizations

Ensemble size and uncertainty quantification

Computational efficiency and energy consumption

11 of 14

Outline

11

25.09.24

What we know and “agree” on

Potential Issues

Opportunities & Open Research Questions

12 of 14

Potential Issues

12

9/25/24

Energy consumption and HPC resources demand

Code rewrites are challenging and resource intensive

Balancing model complexity, grid spacing, and ensemble size

Model validation is difficult due to observational and methodological constraints

Inclusive framework that supports broad collaboration without compromising efficiency

13 of 14

Earth System Models Pre-Summit Intensive

13

25.09.24

What we know and “agree” on

Potential Issues

Opportunities & Open Research Questions

14 of 14

Earth System Models

14

25.09.24

Practitioners and Stakeholders

Computational Systems

Artificial Intelligence and Machine Learning

Starting point for discussion and feedback during this summit