ResCu-en: Training and Testing Data for Moist Physics Ensemble

The ResCu-en project provides data and code for training, testing, and prognostic validation of a ResNet ensemble for moist physics parameterization in climate models. The repository includes datasets for both baseline and +4K sea surface temperature climates, sourced from NCAM, SPCAM, and CAM5 model outputs. This project is authored by Yilun Han from Tsinghua University and is built on Python 3.7 and TensorFlow-GPU 2.3.0.

Use Cases

• Training deep learning emulators for moist physics based on climate model outputs.
• Testing the performance of neural network ensembles under different climate scenarios.
• Prognostic validation of machine-learned parameterizations within climate models.
• Reproducing analysis and plots from the associated research project.

Strengths

• Includes data for two distinct climate scenarios (baseline and +4K SST).
• Provides associated Python scripts and Jupyter notebooks for analysis and plotting.
• Data is sourced from established climate models (NCAM, SPCAM, CAM5).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Last update date is unknown; freshness unverified.

Provenance

Source

Climate model outputs (NCAM, SPCAM, CAM5) processed by Tsinghua University.

Collection Method

Generated from climate model simulations.