LoginSign Up
← Back to all sparks
N

n2khab

ANALYTICS
Velocity0.0

Providing Preprocessed Reference Data for Flemish Natura 2000 Habitat Analyses

github.com/inbo/n2khab ↗

n2khab keeps retracting interpretations of habitat data it can't actually support

natura 2000rhabitat mappingreproducible researchflandersspatial data
◆Current state
n2khab reads and prepares the standardised Flemish Natura 2000 habitat data sources — habitat maps, water surfaces, GRTS master grids — for reproducible analysis. Releases track the publication of new versioned data sources on Zenodo, but the more consequential ones change how the package interprets what it reads. The latest removes an argument outright after the reasoning behind it was found to be wrong.
◆Where it's heading
Two forces shape the package. The first is external: each new habitatmap or watersurfaces vintage needs a supported reader, and the package has absorbed a steady stream of them. The second is a willingness to break its own API when the ecology does not support what the code claimed — the interpreted argument removed because type 3130 occurrences cannot be resolved to a single subtype, the rbbvos+ type dropped as too loosely defined, the collapse default changed to match how users actually need the output shaped. Return structures are also being normalised so element names no longer vary with data source version.
◆Prediction
The package has said it expects future watersurfaces_hab versions to implement collapsing in the data source itself, so the corresponding argument is a candidate for removal once that lands — the same path the interpreted argument took.

◆Recent moves

  1. 1mo ago

    n2khab 0.15.1

    A hotfix correcting the name of the most recent watersurfaces version in read_watersurfaces(). Published a week after 0.15.0 and only meaningful alongside it.

    View source ↗
  2. 1mo ago

    n2khab drops an argument built on a wrong ecological assumption

    read_watersurfaces_hab() loses its interpreted argument entirely: the rationale for interpreted = TRUE rested on the assumption that occurrences coded as main type 3130 could be resolved to a specific subtype, which they cannot. The former FALSE default is what continues. The release also stops imitating old package behaviour for old data source versions, so returned list element names are now consistent regardless of vintage.

    View source ↗
  3. 5mo ago

    n2khab adds a reader for the watersurfaces reference points source

    New read_watersurfaces_refpoints() covers a newly published processed data source, with three archived versions supported alongside the latest. The notes distinguish interim versions, which skip the usual quality control, and advise against them — the kind of provenance labelling that makes this package usable for regulatory work.

    View source ↗
  4. 7mo ago

    n2khab collapses watersurfaces output to unique polygon-type pairs

    A breaking default: read_watersurfaces_hab() gains collapse = TRUE, reducing the watersurfaces_types element to unique polygon_id and type combinations. The release notes tell users to check existing code for changed results rather than assuming the new shape is safe — consistent with how this package handles its own breaking changes.

    View source ↗
  5. 1y ago

    Support for watersurfaces 2024 and watersurfaces_hab v6

    Two new data source versions wired into their existing readers. Routine vintage tracking, the package's most common release type.

    View source ↗
  6. 1y ago

    n2khab supports the 2023 habitat maps and drops the rbbvos+ type

    Five 2023-vintage data sources gain support at once. The removal is the more interesting half: the rbbvos+ type is dropped as too loosely defined — an intermediate between rbbvos and 6510_hua that was never mapped consistently. Same instinct as the later interpreted retraction.

    View source ↗