METEOR: Multivariate Emulation of Time-Evolving and Overlapping Responses

METEOR is a fast spatial climate emulator that generates large ensembles of monthly climate projections with realistic variability. Perfect for impact assessment, uncertainty quantification, and exploring climate scenarios.

Quick Start

Installation

git clone https://github.com/benmsanderson/METEOR.git
cd METEOR
make first-venv
make clean
make virtual-environment
source venv/bin/activate

Simple Example

Generate 100 realizations of global temperature and precipitation for SSP2-4.5:

from meteor import MeteorInterface

# Create and train emulator (uses cached models if available)
emulator = MeteorInterface(
    model='NorESM2-MM',
    variables=['tas', 'pr'],
    cache_dir='./cache'
)
emulator.train(verbose=True)

# Generate ensemble
ensemble = emulator.generate_ensemble_outputs(
    scenario='ssp245',
    start_year=1850,
    end_year=2100,
    n_realizations=100,
    timeseries=['global']  # Global mean
)

# Access results
global_temp = ensemble['tas'].timeseries['global']  # Shape: (100, 3012 months)
print(f"Ensemble shape: {global_temp.shape}")
print(f"2100 warming: {global_temp[:, -1].mean():.2f} ± {global_temp[:, -1].std():.2f} K")

📓 More Examples

For comprehensive examples including regional projections, impact metrics, gridded output, multi-scenario comparisons, and advanced customization, see:

METEOR Interface Examples Notebook

This interactive notebook covers:

• Regional and city-scale projections
• Climate impact metrics (heating/cooling degree days)
• Full gridded output generation
• Climatology without noise
• Multi-scenario comparisons
• Custom model configurations

What METEOR Does

• ✅ Fast Ensemble Generation: Create 100+ realizations faster than running a single GCM simulation
• ✅ Monthly Resolution: Realistic monthly variability with proper seasonality and persistence
• ✅ Spatial Detail: Global, regional (AR6 regions), and point-based projections
• ✅ Multiple Variables: Temperature, precipitation, and more with variable-specific treatment
• ✅ Impact Metrics: Built-in calculation of degree days and custom impact assessments
• ✅ Smart Caching: Automatic caching of trained models and downloaded data
• ✅ CMIP6 Integration: Direct access to cloud-based CMIP6 data

Architecture Overview

Core Components

Component	Purpose	Key Classes
High-Level Interface	Simple ensemble generation	MeteorInterface
Pattern Scaling	Annual climate projections from forcings	MeteorPatternScaling
Monthly Generation	Convert annual to monthly + seasonality	MeteorNoiseGenerator
Data Access	CMIP6 cloud data integration	Cmip6MeteorDataGetter
Impact Assessment	Climate impact calculations	Degree days, custom impacts
Variable Transforms	Variable-specific data processing	Gamma transform for precipitation
Ensemble Output	Structured output containers	EnsembleOutput, VariableOutput

Workflow

Emissions/Concentrations → Pattern Scaling → Annual Climate → Monthly Base
                                                                      ↓
                                                          + Stochastic Noise
                                                                      ↓
                                                          Monthly Ensemble
                                                                      ↓
                                                          Impact Calculations

Additional Resources

Documentation

• METEOR Interface Examples: Complete tutorial with visualizations

Other Notebooks

• Climate_Bench_METEOR.ipynb: ClimateBench metrics

Module Reference

Module	Purpose
meteor_interface.py	High-level MeteorInterface API
meteor.py	MeteorPatternScaling for annual projections
noise_generator.py	MeteorNoiseGenerator for monthly variability
cmip6_meteor_data_getter.py	CMIP6 cloud data access
ensemble_output.py	Output container classes
variable_transforms.py	Variable-specific transformations
impacts/	Climate impact assessment
prpatt.py	Pattern scaling algorithms
scm_forcer_engine.py	Simple climate model integration

Development & Testing

Run tests and checks:

# Run all tests
make checks

# Run specific test suite
pytest tests/unit
pytest tests/integration

#other make options
make format-checks        # run all the checks
make format               # re-format files
make format-notebooks     # format the notebooks
make black                # apply black formatter to source and tests
make isort                # format the code
make docs                 # build the docs
make test                 # run the full testsuite
make test-pypi-install    # test whether installing from PyPI works
make test-install         # test installing works
make virtual-environment  # update venv, create a new venv if it doesn't exist make
make first-venv           # create a new virtual environment for the very first repo

Contributing

1. Create feature branch from main
2. Add tests for new functionality
3. Ensure all tests pass (make checks)
4. Update documentation as needed
5. Submit pull request

---

Citation

If you use METEOR in your research, please cite:

@Article{sandstad_meteor,
AUTHOR = {Sandstad, M. and Steinert, N. J. and Baur, S. and Sanderson, B. M.},
TITLE = {METEORv1.0.1: A novel framework for emulating multi-timescale regional climate responses},
JOURNAL = {EGUsphere},
VOLUME = {2025},
YEAR = {2025},
PAGES = {1--49},
URL = {https://egusphere.copernicus.org/preprints/2025/egusphere-2025-1038/},
DOI = {10.5194/egusphere-2025-1038}
}

Support

• Examples: Explore notebooks/ directory
• Issues: GitHub Issues
• Updates: git pull && make virtual-environment

Note: Some research notebooks (METEOR_paper_figures_*.ipynb) require large datasets and are excluded from automated testing.