The repo releases the data and code for the paper:
- OceanForecastBench: A Benchmark Dataset for Data-Driven Medium-Range Global Ocean Forecasting
Major Features
-
Deep Learning-Ready Dataset for Trainging. OceanForecastBench have compiled a comprehensive deep learning-ready dataset including diverse ocean variables necessary for medium-range global ocean forecasting.
-
Testing Method Based Observation. OceanForecastBench constructed a testing dataset by integrating multi-source observation data. A standardized model evaluation pipeline is also provided to align forecast data with discrete observation data and calculate evaluation metrics.
-
Baselines.
- ResNet
- SwinTransformer
- ClimaX
- FourCastNet
- XiHe
- PSY4
The data and pre-trained models used in the paper is available in the following links: Baidu Netdisk
Original Data Source
GLORYS12
ERA5
GHRSST
EN4
GDP
CMEMS L3 track
Traing Dataset
The training data files shall be organized as the following hierarchy:
├── 1.40625deg_npy
│ ├── train
│ │ ├── mra5_19930101.npy
│ │ ├── mra5_19930102.npy
│ │ ├── ...
│ ├── val
│ │ ├── mra5_20180101.npy
│ │ ├── mra5_20180102.npy
│ │ ├── ...
├── normalize_1.40625deg
│ ├── normalize_mean_23_1.40625deg.npz
│ ├── normalize_std_23_1.40625deg.npz
├── latlon_1.40625deg
│ ├── lat.npz
│ ├── lon.npz
The testing data files shall be organized as the following hierarchy:
Testing Dataset
├── ground_truth
│ ├── temperature_salinity
│ │ ├── EN.4.2.2.profiles.g10.2022/
│ │ │ ├── EN.4.2.2.f.profiles.g10.202201.nc
│ │ │ ├── EN.4.2.2.f.profiles.g10.202202.nc
│ │ │ ├── ...
│ ├── current_sst
│ │ ├── drifter_6hour_qc_c35b_4db8_1f03_U1718152782438.nc
│ ├── zos
│ │ ├── 202201
│ │ │ ├── dt_global_alg_phy_l3_20220101_20220701.nc
│ │ │ ├── dt_global_alg_phy_l3_20220102_20220701.nc
│ │ │ ├── ...
│ │ ├── 202202
│ │ │ ├── dt_global_alg_phy_l3_20220201_20220701.nc
│ │ │ ├── dt_global_alg_phy_l3_20220202_20220701.nc
│ │ │ ├── ...
Pre-Trained Models
├── ResNet
│ ├── fct_1d.ckpt
│ ├── fct_2d.ckpt
│ ├── ...
│ ├── fct_10d.ckpt
├── SwinTransformer
│ ├── fct_1d.ckpt
│ ├── fct_2d.ckpt
│ ├── ...
│ ├── fct_10d.ckpt
├── ClimaX
│ ├── fct_1d.ckpt
│ ├── fct_2d.ckpt
│ ├── ...
│ ├── fct_10d.ckpt
├── FourCastNet
│ ├── fct_1d.ckpt
│ ├── fct_2d.ckpt
│ ├── ...
│ ├── fct_10d.ckpt