coarse-grained-forecasting-benchmark/passive-scalar

Passive Scalar Conditional Forecasting

This dataset contains conditional ensembles of trajectories of a passive scalar advected by a steady cellular flow with weak diffusion. Each ensemble shares the same observed coarse initial state but differs in its hidden fine-scale initial conditions, producing multiple possible future trajectories.

The dataset accompanies an anonymous research submission.

Dataset

The observed state consists of 196 real-valued Fourier coefficients of the scalar field, recorded at 41 time points between (t=0) and (t=1).

The main file, data/trajectories_fine.npz, contains:

Array	Shape	Description
vecs_train	(80, 400, 41, 196)	Training ensembles
vecs_new	(1520, 20, 41, 196)	Additional training ensembles
vecs_eval	(20, 4096, 41, 196)	Validation and test ensembles

The main experiments used only the 80 initial conditions in vecs_train, with the first 200 of their 400 realizations each. The additional 1,520 conditions in vecs_new were not used in these experiments.

The file also contains group identifiers and time coordinates. All trajectory arrays are stored in float64.

Train, validation, and test conditions are specified in data/split.json and are disjoint. The dataset contains 1,600 training, 10 validation, and 10 test conditions.

Test reference ensembles

truth_bank/ contains 10,000 reference trajectories for each of the 10 test conditions. These are used to evaluate the predicted conditional distributions.

Files and usage

• data/trajectories_fine.npz — coarse trajectories
• data/split.json — group-level data splits
• data/normalization.npz — training-derived normalization parameters
• truth_bank/group_<gid>.npz — test reference ensembles
• SHA256SUMS — file checksums

Download using:

hf download coarse-grained-forecasting-benchmark/passive-scalar --repo-type dataset --local-dir passive-scalar

For the accompanying evaluation code, set PASSIVE_SCALAR_DATA to passive-scalar/data and PASSIVE_SCALAR_TRUTH_BANK to passive-scalar/truth_bank.

License

Creative Commons Attribution 4.0 International (CC BY 4.0).