i.hyper.lib_zenodo harvests curated spectral libraries published on Zenodo into the same shared, source-tagged, partitioned local spectral database used by i.hyper.lib_ecosis, i.hyper.lib_gsti, i.hyper.lib_relab and i.hyper.lib_usgs. The default store is $HOME/grassdata/hyperspeclib, written as Hive-partitioned Parquet under source_database=zenodo, with a _manifest.json so a second run is incremental.
The five libraries already harvested are global or North American. Between them they hold almost no Mediterranean vegetation. A search of the local store for the species of a Provence site returned 4 spectra of Quercus ilex, 20 of Quercus pubescens and 10 of Laurus nobilis, and nothing at all for Pinus halepensis, Olea europaea, Cupressus, Pistacia, Arbutus unedo, Cistus or Rosmarinus.
That matters more than a gap usually does. A spectral match always returns a nearest neighbour, so matching a Mediterranean canopy against a North American library does not return no answer: it returns a confident wrong one. The regional libraries that fill the gap are published as individual Zenodo records rather than as a single database, which is what this module is for.
Zenodo has no common spectral schema, so each dataset is read by its own reader and declared in a registry carrying its record id, licence, site, instrument and the species it actually contains, verified against the dataset's own metadata rather than its abstract. Adding a library means adding a registry entry and a reader. Run with -l to print the registry.
| Key | Record | Licence | Level | Species |
|---|---|---|---|---|
| mediterranean_woodlands | 5176824 | CC-BY-4.0 | canopy and leaf | Quercus suber, Cistus salviifolius, Ulex europaeus, C3 grass mixture |
| montesinho | 10798148 | CC-BY-4.0 | leaf | Quercus rotundifolia, Quercus pyrenaica, Arbutus unedo, Cistus ladanifer, Erica arborea, Castanea sativa, Fraxinus angustifolia, Crataegus monogyna, Alnus glutinosa, Frangula alnus, Cytisus multiflorus, Genista falcata, Acacia dealbata, Dactylis glomerata, Trifolium repens |
Two points about what these contain. 974 of the 1216 observations in mediterranean_woodlands are canopy rather than leaf measurements, which is the reason to prefer it: a drone crown is a canopy, and a laboratory leaf spectrum is not comparable to one. And Quercus rotundifolia in montesinho is the Iberian holm oak of the Quercus ilex complex, the nearest reference this harvest can offer to a Provence holm oak.
Each spectrum carries its species in extra_metadata under the key species. That is the key the openspeclib loader reads to give a spectrum its material name, and writing it anywhere else loses the species between the store and the index. This is not hypothetical: the LOPEX93 and ANGERS records already in the store hold their species under Latin Name, which the loader does not read, so all 1212 of those named spectra reach the index carrying their dataset title instead and cannot be found by species.
Downloads are cached in _cache/zenodo under the store root and reused on a later run. Each download is checked against the size Zenodo reports, because a truncated file is otherwise parsed into a silently incomplete library.
A dataset already recorded in the manifest is skipped; pass -f to re-harvest it. Use -n to see what would be fetched without writing.
As harvested, both datasets give a species to every spectrum: 538 of 538 for montesinho and 1216 of 1216 for mediterranean_woodlands. Six of the latter need recovering, because the 2012-03-10 session is missing from the dataset's metadata table although its spectra are in the data file. Their species comes from the target ID instead, which is sound because species and measurement level are constant across all 27 targets, and each such row says so in species_source and is reported as a warning rather than passed off silently.
One value in that dataset is not a taxon: c3mixture, a mixed C3 herbaceous sward, carried in the genus column. It is renamed to C3 grass mixture and marked is_taxon: false on its 251 rows, so a mixture is not matched against as though it named a plant.
format=parquet needs pyarrow; format=sqlite has no extra dependency.
These are leaf and canopy reference spectra of Iberian and Mediterranean flora, not of Provence itself. They narrow the gap; they do not close it. A match against them is still a match against another site's plants, and the only thing that settles a species on the ground is identification in the field.
List the registry without downloading anything:
i.hyper.lib_zenodo -l
Harvest everything into the default shared store:
i.hyper.lib_zenodo
Harvest one dataset into a chosen store, as SQLite:
i.hyper.lib_zenodo dataset=montesinho output=$HOME/grassdata/hyperspeclib \
format=sqlite
Check what a record would cost before fetching it:
i.hyper.lib_zenodo record=5176824 -n
i.hyper.lib_ecosis, i.hyper.lib_gsti, i.hyper.lib_relab, i.hyper.lib_usgs, i.hyper.speclookup, i.hyper.spectroscopy
Spectral Feature Extraction and Interpretation Engine
Available at: i.hyper.lib_zenodo source code (history)
Accessed: Sunday Nov 15 10:56:53 2026
Main index | Imagery index | Topics index | Keywords index | Graphical index | Full index
© 2003-2026 GRASS Development Team, GRASS 8.5.1dev Reference Manual