GeoCAST

This repository will use test new foundation models for time series forecast to predict geoscientific time series.

It is not intended to build a new foundation model, but instead to test existing models.

We start with

• Chronos : https://github.com/amazon-science/chronos-forecasting (LLM)
• Moirai : https://github.com/SalesforceAIResearch/uni2ts (foundation TS) Both have been released in mid 2024 and then released last month a multi-scale / improved version.

Chronos and Moirai do not have dependency conflicts, so one conda environment can work:

conda create --name quake_chronos_moirai python=3.11
conda activate quake_chronos_moirai
pip install -r requirements_chronos_moirai.txt

Moiara can be installed as below

conda create --name quake_moirai python=3.11
conda activate quake_moirai
git clone https://github.com/SalesforceAIResearch/uni2ts.git
cd uni2ts
pip install -e '.[notebook]'
cd ..
pip install obspy pandas matplotlib

DATA

• Extreme events:

  • Earthquakes (earthquakes catalogs & synthetic catalogs)
  • Wildfire
  • Floods
  • landslides?

• Semi-periodic:

  • precipitation
  • hydrological loads on surface displacements
  • dv/v, temp, precips

• Transient waves:

  • earthquake ground motion
  • acoustics
  • bioacoustics