{"success":true,"data":{"author":"Polymathic AI (The Well); Michael McCabe","bucketAccessType":"public","bucketName":"polymathic-ai/planetswe","bucketPrefix":"","category":"Physics","cloudProvider":"huggingface","createdAt":"2026-11-14T01:01:19.579Z","description":"Forced, hyperviscous rotating shallow water equations on a sphere with Earth-like topography and daily and annual periodic forcing, a simplified stand-in for a single pressure level of an atmospheric model. The shallow water equations are a depth-integrated form of incompressible Navier-Stokes for flows whose horizontal scales far exceed vertical ones. Solved with Dedalus on an equiangular 256 by 512 grid; each trajectory runs 3,024 steps (24 steps per model day, 1,008 per model year) for 3 model years. Fields are height and velocity. The datasheet lists 40 trajectories from different initial conditions. Listing data/ (measured 2026-11-13) shows train/ (96 files, 154 GB, 1.6 GB per file, one per initial condition and year segment), valid/ and test/ (12 files, 19.2 GB each); about 192 GB in the repository. Files are HDF5 in the common The Well format; open them with the_well package (WellDataModule with base path hf://datasets/polymathic-ai/) or directly with h5py. Part of The Well (Ohana et al., NeurIPS 2024 Datasets and Benchmarks, arXiv:2412.00568); released under CC-BY-4.0, cite The Well and the simulation's source papers.","displayTitle":"The Well: PlanetSWE, Forced Rotating Shallow Water on a Sphere","id":"d63753b8-26ea-5dae-b43d-1c56df7904e8","license":"CC-BY-4.0","licenseEvidenceUrl":"https://arxiv.org/html/2412.00568","objectCountEstimate":137,"probeStatus":"ok","referenceUrl":"https://arxiv.org/abs/2306.10619","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"polymathic-ai/planetswe\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"polymathic-ai/planetswe\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=polymathic-ai, license:cc-by-4.0, downloads=2050"],"format":"mixed","generatedAt":"2026-11-14T01:01:20.692Z","layout":{"example":"data/test/planetswe_IC36_s1.hdf5","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"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","encoding":"hex","key":"data/test/planetswe_IC36_s1.hdf5","truncatedAt":512}],"sampleObjects":["data/stats.yaml","data/test/planetswe_IC36_s1.hdf5","data/test/planetswe_IC36_s2.hdf5","data/test/planetswe_IC36_s3.hdf5","data/test/planetswe_IC37_s1.hdf5"],"sidecars":[{"bytes":"---\nlanguage:\n- en\nlicense: cc-by-4.0\ntags:\n- physics\n\ntask_categories:\n- time-series-forecasting\n- other\ntask_ids:\n- multivariate-time-series-forecasting\n---\n\nThis Dataset is part of [The Well Collection](https://huggingface.co/papers/2412.00568).\n\n# How To Load from HuggingFace Hub\n\n1. Be sure to have `the_well` installed (`pip install the_well`)\n2. Use the `WellDataModule` to retrieve data as follows:\n\n```python\nfrom the_well.data import WellDataModule\n\n# The following line may take a couple of minutes to instantiate the datamodule\ndatamodule = WellDataModule(\n    \"hf://datasets/polymathic-ai/\",\n    \"planetswe\",\n)\ntrain_dataloader = datamodule.train_dataloader()\n\nfor batch in dataloader:\n    # Process training batch\n    ...\n```\n\n# PlanetSWE\n\n**One line description of the data:** Forced hyperviscous rotating shallow water on a sphere with earth-like topography and daily/annual periodic forcings.\n\n**Longer description of the data:** The shallow water equations are fundamentally a 2D approximation of a 3D flow in the case where horizontal length scales are significantly longer than vertical length scales. They are derived from depth-integrating the incompressible Navier-Stokes equations. The integrated dimension then only remains in the equation as a variable describing the height of the pressure surface above the flow. These equations have long been used as a simpler approximation of the primitive equations in atmospheric modeling of a single pressure level, most famously in the Williamson test problems. This scenario can be seen as similar to Williamson Problem 7 as we derive initial conditions from the hPa 500 pressure level in ERA5. These are then simulated with realistic topography and two levels of periodicity.\n\n**Associated paper**: [Paper](https://openreview.net/forum?id=RFfUUtKYOG).\n\n**Domain expert**: [Michael McCabe](https://mikemccabe210.github.io/), Polymathic AI.\n\n**Code or software used to generate the data**: [Dedalus](https://dedalus-project.readthedocs.io/en/latest/), adapted from [this example](https://dedalus-project.readthedocs.io/en/latest/pages/examples/ivp_sphere_shallow_water.html).\n\n**Equation**:\n\n$$\n\\begin{align*}\n\\frac{ \\partial \\vec{u}}{\\partial t} &= - \\vec{u} \\cdot \\nabla u - g \\nabla h - \\nu \\nabla^4 \\vec{u} - 2\\Omega \\times \\vec{u} \\\\\n\\frac{ \\partial h }{\\partial t} &= -H \\nabla \\cdot \\vec{u} - \\nabla \\cdot (h\\vec{u}) - \\nu \\nabla^4h + F\n\\end{align*}\n$$\n\nwith \\\\(h\\\\) the deviation of pressure surface height from the mean, \\\\(H\\\\) the mean height, \\\\(\\vec{u}\\\\) the 2D velocity, \\\\(\\Omega\\\\) the Coriolis parameter, and F the forcing which is defined:\n\n```python\ndef find_center(t):\n    time_of_day = t / day\n    time_of_year = t / year\n    max_declination = .4 # Truncated from estimate of earth's solar decline\n    lon_center = time_of_day*2*np.pi # Rescale sin to 0-1 then scale to np.pi\n    lat_center = np.sin(time_of_year*2*np.pi)*max_declination\n    lon_anti = np.pi + lon_center  #2*np.((np.sin(-time_of_day*2*np.pi)+1) / 2)*pi\n    return lon_center, lat_center, lon_anti, lat_center\n\ndef season_day_forcing(phi, theta, t, h_f0):\n    phi_c, theta_c, phi_a, theta_a = find_center(t)\n    sigma = np.pi/2\n    coefficients = np.cos(phi - phi_c) * np.exp(-(theta-theta_c)**2 / sigma**2)\n    forcing = h_f0 * coefficients\n    return forcing\n```\n\nVisualization:\n\n![Gif](https://users.flatironinstitute.org/~polymathic/data/the_well/datasets/planetswe/gif/planetswe.gif)\n\n| Dataset    | FNO | TFNO  | Unet | CNextU-net\n|:-:|:-:|:-:|:-:|:-:|\n| `planetswe`  | 0.1727| \\\\(\\mathbf{0.0853}\\\\) | 0.3620 | 0.3724|\n\nTable: VRMSE metrics on test sets (lower is better). Best results are shown in bold. VRMSE is scaled such that predicting the mean value of the target field results in a score of 1.\n\n## About the data\n\n**Dimension of discretized data:** 3024 timesteps of 256x512 images with \"day\" defined as 24 steps and \"year\" defined as 1008 in model time.\n\n**Fields available in the data:** height (scalar field), velocity (vector field).\n\n**Number of trajectories:** 40 trajectories of 3 model years.\n\n**Estimated size of th","encoding":"utf8","key":"README.md","name":"README.md","size":5841,"truncated":true},{"bytes":"---\nlanguage:\n- en\nlicense: cc-by-4.0\ntags:\n- physics\n\ntask_categories:\n- time-series-forecasting\n- other\ntask_ids:\n- multivariate-time-series-forecasting\n---\n\n# How To Load from HuggingFace Hub\n\n1. Be sure to have `the_well` installed (`pip install the_well`)\n2. Use the `WellDataModule` to retrieve data as follows:\n\n```python\nfrom the_well.data import WellDataModule\n\n# The following line may take a couple of minutes to instantiate the datamodule\ndatamodule = WellDataModule(\n    \"hf://datasets/polymathic-ai/\",\n    \"planetswe\",\n)\ntrain_dataloader = datamodule.train_dataloader()\n\nfor batch in dataloader:\n    # Process training batch\n    ...\n```\n\n","encoding":"utf8","key":"generation/README.md","name":"README.md","size":652,"truncated":false}],"sourceObjectKey":"data/test/planetswe_IC36_s1.hdf5","starterCell":"from huggingface_hub import hf_hub_download\npath = hf_hub_download(repo_id=\"polymathic-ai/planetswe\", filename=\"data/test/planetswe_IC36_s1.hdf5\", repo_type=\"dataset\")\nprint(path)","starterSamplePath":"data/test/planetswe_IC36_s1.hdf5","totalObjects":137},"summary":"Forced, hyperviscous rotating shallow water equations on a sphere with Earth-like topography and daily and annual periodic forcing, a simplified stand-in for a single pressure…","title":"The Well: PlanetSWE, Forced Rotating Shallow Water on a Sphere","updatedAt":"2026-11-14T02:32:56.752Z","visibility":"public","websiteUrl":"https://polymathic-ai.org/the_well/"}}