---

name: rockhound description: Maintain or migrate legacy RockHound geoscience dataset downloads, caches and loaders such as PREM. Use when an existing project imports RockHound; prefer maintained Ensaio fetchers for supported new dataset workflows. license: MIT metadata: version: "1.0.0" author: Geoscience Skills skill_type: domain tags: '["Legacy", "Datasets", "Migration", "PREM"]' dependencies: '["rockhound==0.2.0"]' complements: '["ensaio", "pooch", "xarray", "disba"]' workflow_role: data-loading

RockHound maintenance and migration

RockHound's repository was archived in 2022 and recommends Ensaio. Preserve a working legacy environment while assessing supported replacements; there is no universal one-to-one mapping of fetcher names or return objects.

Inspect an existing workflow

Identify the exact fetcher, data version, source URL/hash, cache location, parser and scientific conventions. Set ROCKHOUND_DATA_DIR before importing RockHound if a separate cache is needed. Never delete a user's cache to test a migration.

Legacy PREM load

This fragment uses the upstream registry and may download from EarthScope/IRIS. It loads the 1 s PREM radial Earth model, not a local field observation. The registry URL returned HTTP 404 during the 2026-09-14 check. This example currently needs an existing hash-verified cache or a separately verified source; it is not a successful fresh-download recipe. The test suite exercises the real loader with a clearly synthetic cache fixture, without claiming PREM validation.

# example: legacy-prem
import rockhound

prem = rockhound.fetch_prem()
radius_km = prem["radius"].to_numpy()
depth_km = prem["depth"].to_numpy()
density_g_cm3 = prem["density"].to_numpy()
vpv_km_s = prem["Vpv"].to_numpy()
vsv_km_s = prem["Vsv"].to_numpy()

The table contains discontinuities and distinct vertical/horizontal velocities. Preserve repeated boundary radii and choose a side explicitly for interpolation. Zero shear velocity in a fluid region is meaningful, not a missing value. Do not send the full radial model directly to a plane-layer surface-wave solver.

Migration steps

1. Find a maintained dataset release, its redistribution license and immutable identifier. If Ensaio has no matching dataset, use a documented Pooch registry and the original provider instead of substituting unrelated sample data.
2. Compare downloaded hashes or document changed releases; inspect units, grid registration, coordinates, missingness and source preprocessing.
3. Adapt the return contract: an Ensaio path still needs a CSV/NetCDF reader, while RockHound may return a DataFrame or xarray object.
4. Verify representative values and downstream results before replacing imports. Diagnose failed old URLs; do not bypass TLS or checksum checks.

See the archived project notice, PREM loader contract and Ensaio catalog.