DINO dataset for nemo-spinup-bench

This dataset accompanies nemo-spinup-bench, a benchmarking suite for machine learning-based forecasting methods applied to ocean model spinup, comprising the evaluation, forecast and restart packages. 

The dataset contains grid outputs and restart files from runs of the nemo ocean model using the DINO idealised configuration, covering 50-year and 200-year simulations. It is intended for training and evaluating ML models that forecast ocean model state during spinup, with the goal of reducing the computational cost of reaching equilibrium.

Please refer to the included README.md for full details on file contents, data processing steps, and compression.

Description of the files:

Baseline example in nemo-spinup-bench:

• 50.zip: grid outputs (temperature, salinity, velocity, SSH) and a restart file from a 50-year NEMO / DINO simulation
• 200.zip:  as above, but for a 200-year simulation with outputs concatenated across four 50-year segments
• restart.zip: 200 annual restart files spanning the full 200-year simulation.
• README.md: full description of file contents, data processing steps, and compression details

Example 1-resampling-temporal-data in nemo-spinup-bench:

•  restart3.zip: Compressed grid files at various temporal granularities. See README.md in .tar for more details. 

Example 2-regrid-evaluate-gorce in nemo-spinup-bench:

• regrid-evaluate.zip: Contains reference files for example 2 in 

Testing data for nemo-spinup-evaluation testing data:

• evaluation-test.zip:  Data required for the CI.

Testing data for nemo-spinup-restart: 

• restart-test.zip: Data required for the CI. It contains duplicate 50 year restart file (also in 50-year.zip - DINO_00576000_restart.nc) and a reconstructed restart state at 50 year (30 year jump) prepended with NEW in the reconstructed_restarts directory -  NEW_DINO_00576000_restart.nc. As well as the raw numpy files in the simu_predicted/ directory.