The Helmholtz Representation Model for Climate Science (HClimRep) project seeks to leverage the success of foundation modeling in weather forecasting and apply it to climate timescales. This pioneering initiative focuses on developing a comprehensive AI‑based climate model that captures the complex interactions among the atmosphere, ocean, and sea ice across seasonal to decadal time scales.
The HClimRep prototype is based on the WeatherGenerator framework and aims to extend it by incorporating additional vertical levels of the atmosphere and additional Earth system compartments (e.g., oceanic fields).
Our goal is to build a Foundation Model for Climate Science that:
1. Delivers reliable AI‑based simulations across extended timescales
Figure 1. Autoregressive rollout with the WeatherGenerator/HClimRep prototype. Here, we show that the model remains stable for up to a year.
2. Captures atmospheric dynamics from the lower troposphere through the stratosphere
Figure 2. Temperature profile during a sudden stratospheric warming event. We observe that our prototype correctly captures both the target temperature profile and the temperature anomalies.
3. Represents ocean and sea‑ice interactions with high fidelity
Figure 3. Salinity (left) and sea surface temperature (SST) (right) in the tropical Atlantic. The model was trained on IFS-FESOM data and captures meaningful dynamics.
4. Adapts seamlessly to diverse downstream applications.
Figure 4. Sea surface temperature (top left) and ocean height (bottom left) downscaling results. Here, we fine‑tuned our prototype on downscaling. The training was completed in less than a day on 8 GPUs, demonstrating efficient adaptation to downstream tasks.
The HClimRep project is organized into 10 work packages (WPs). The project spans two phases, with foundational model extensions in Phase 1 and the construction of a novel, multi-scale and multi-compartment model in Phase 2.
WP 1
Onboaring
Objective
Build a community, provide training, support project partners, and organize a hackathon.
WP 2
AtmoRep Extension to the stratosphere (StratoRep)
Objective:
Extend AtmoRep to include stratospheric data and tracers, develop training concepts, and evaluate model performance.
WP 3
OceanRep Prototype Development
Objective
Develop processing strategies for ocean model data, pre-train OceanRep, build a sea ice component, and evaluate the model.
WP 4
Enable long-term Simulations
Objective
Develop strategies for long-term simulations of AtmoRep and OceanRep for seasonal to decadal timescales.
WP 5
HClimRep Concept Development
Objective
Develop the model architecture and training strategies for multi-resolution, multi-model integration.
WP 6
HClimRep Core Model Development
Objective
Implement HClimRep code, perform test runs, and optimize training performance.
WP 7
Pre-training of HClimRep
Objective
Supervise pre-training experiments, monitor performance, and evaluate model checkpoints.
WP 8
Exploration of Downstream Tasks
Objective
Develop embedding networks for downstream applications, fine-tune, and evaluate results.
The following two work packages span throughout the entire project duration:
WP 9
Data Management
Objective
Manage input and output data, prepare data publications, and ensure data accessibility.
WP 10
Project Coordination
Objective
Coordinate project activities, maintain the project website, organize meetings, and monitor progress.