Island, archipelago, plate, mountain and point labeling dataset

Most accurate island shoreline vector dataset currently available, adjusted from Sayre et al. (2019) and completed with Sayre (2023), tectonic plates adjusted from Bird (2003), island point data from Weigelt et al. (2013), the complete geoNames point location dataset (2023) and the GMBA Mountain Inventory v2 from Snethlage et al. (2022), to be used as labeling datasets for identified polygons/features by the TABS R-package (https://github.com/jedgroev/tabs). All layers use the reference system WGS84, SRID 4326.
The dataset is provided in two formats with the same content:
1. GEOPACKAGE (labs.zip)
The zip includes two files (reference.gpkg, mountains.gpkg), where "reference" includes all island datasets and geoNames and "mountains" contains the GMBA mountain inventory. This is the file downloaded by setup() of the tabs R package, for offline use.
2. GEOPARQUET (one .parquet file per layer, new in version 6)
Every layer of the two GeoPackages as a separate GeoParquet file (GeoParquet 1.0.0, geometry as WKB), named after the layer (e.g. islands.parquet, mountains_standard.parquet). These files can be queried directly from Figshare without downloading them, e.g. with DuckDB; the tabs R package reads them this way when the GeoPackages are not on the disk. Spatial layers are sorted along a Hilbert curve and have a column bbox (xmin, ymin, xmax, ymax of each feature), so that a query on an area reads only the row groups (2,000 rows each, ZSTD compression) that overlap it. Attribute tables (lookup tables) are plain Parquet. The QGIS styles (layer_styles) are only in the GeoPackages, and the GeoParquet files have no fid column. Example (DuckDB): SELECT name, area_km2 FROM read_parquet('[download link of islands.parquet]') WHERE bbox.xmin < 24.2 AND bbox.xmax > 23.6 AND bbox.ymin < 39.5 AND bbox.ymax > 39.0
CHANGES IN VERSION 6

Islands that are new in the 2023 release of the global islands (Sayre 2023) are added; islands cut at the 180th meridian are joined into one feature.
Unnamed islands that contain exactly one geoNames island name are named after it (33,203 islands); the column name_source gives the source of each name (USGS, WCMC, GeoNames, OSM or manual).
Islands misplaced by 1-12 km in the source (127 islands, mainly in Antarctica and the Arctic) are flagged by comparison with OpenStreetMap land polygons: columns shifted, shift_east_m, shift_north_m and shift_method. The geometries themselves are not changed.
The setting of every island (column island_setting_id, lookup table lu_island_setting): sea island, mainland-associated island, water, noise or island-associated island (with associated_island_id), and land_share, the percentage of the island that is land in ESA WorldCover 2021. New layer deltas (river delta outlines from Edmonds et al. 2020), used for the setting.
Two duplicated islands were removed, two islands cut into pieces were joined and one outline was corrected after review.
The GeoParquet files (2).
Tables in reference.gpkg

Sayre et al. (2019), Sayre (2023): islands (polygons), continents (polygons), continents_metadata, lu_land_type, lu_country_code, island_size, lu_island_setting
Bird (2003): plates (polygons), orogens (polygons)
Weigelt et al. (2013): islanddata (points), archipelago (polygons), archipelago_country (polygons)
geoNames: reference_points (points), lu_feature_class, lu_feature_code
Edmonds et al. (2020): deltas (polygons)
Tables in mountains.gpkg

Snethlage et al. (2022): mountains_broad, mountains_standard
TABLE DESCRIPTIONS
islands
Global shoreline vector and integrates all named islands (All_BI_Named_2) and small islands (global_islands_outputs.gdb) of Sayre et al. (2019), completed with the islands that are new in the 2023 release (Sayre 2023). Islands cut at the 180th meridian are joined into one feature. The mainland geometry is split and stored in a separate table (continents). - name_source: the source of each name: USGS, WCMC (Sayre 2023, for new big islands without a USGS name), GeoNames (an unnamed island that contains exactly one geoNames island name; for islands with several names the name chosen in a review), OSM or manual. - shifted: TRUE when the island is misplaced by 1 km or more compared with OpenStreetMap land polygons, FALSE when its position agrees, NA when not assessable (e.g. islands in lakes or under ice). shift_east_m and shift_north_m give the displacement to the OpenStreetMap position, shift_method how the flag was obtained (osm, review on satellite imagery, or neighbours: taken over from the nearest measured island).
- island_setting_id: the setting of the island (lu_island_setting): 1 sea island; 2 mainland-associated island (separated from the mainland only by sheltered water: large deltaic systems, wetlands, rivers and lakes); 3 water (a river, lake or channel stored as an island); 4 noise (a sliver or unclear polygon); 5 island-associated island (in a lake, river or closed lagoon of a larger island, given in associated_island_id). The classes come from rules checked against reviews of about 19,000 islands, using OpenStreetMap land polygons, the river deltas of Edmonds et al. (2020), ESA WorldCover 2021 and Global Mangrove Watch v3.0 (Bunting et al. 2022). - land_share: the percentage of the island that is land in ESA WorldCover 2021.
continents / continents_metadata
Mainland geometries provided by the global shoreline vector (global_islands_outputs.gdb). The data were pre-processed to make the geometries easily queryable by subdividing the continents into smaller and manageable polygons. continents_metadata includes all relevant records associated to the original continent geometries (i.e., Australia, Africa, South-America, North-America, Eurasia).
lu_land_type
Lookup table for the land type including three classes (1, mainland; 2, big island; 3, small island) as differentiated by Sayre et al. (2019).
lu_country_code
Lookup table for country codes.
island_size
Number of islands per island size (in m², ha and km²).
lu_island_setting
Lookup table for the setting of the islands (1, sea island; 2, mainland-associated island; 3, water; 4, noise; 5, island-associated island).
plates
Geometries of different plates. The dataset that was used is a conversion of the dataset originally published in the paper Bird (2003). This source has been made available by Hugo Ahlenius, Nordpil and Peter Bird and is made available under the Open Data Commons Attribution License: http://opendatacommons.org/licenses/by/1.0/.
orogens
Geometries of different orogens. The dataset that was used is a conversion of the dataset originally published in the paper Bird (2003). This source has been made available by Hugo Ahlenius, Nordpil and Peter Bird and is made available under the Open Data Commons Attribution License: http://opendatacommons.org/licenses/by/1.0/.
islanddata
Island point data including bioclimatic and physical characterization of the world's islands from Weigelt et al. (2013) without any modifications.
archipelago
Archipelago geometries derived from Weigelt et al. (2013) by calculating the convex hull of points per archipelago.
archipelago_country
Archipelago-country geometries derived from Weigelt et al. (2013) by calculating the convex hull of points per archipelago-country.
reference_points
Point geometries which can be used for labeling features. The data has been downloaded from the allCountries.zip hosted by geoNames.
lu_feature_class
Feature classes identified in geoNames including: A: country, state, region,...; H: stream, lake,...; L: parks, area,...; P: city, village,...; R: road, railroad,...; S: spot, building, farm,...; T: mountain, hill, rock,...; U: undersea; V: forest, heath,...
lu_feature_code
Feature codes (in total 681) for each feature class identified in geoNames.
deltas (new in version 6)
Outlines of 2,174 river deltas (convex hull of five points per delta) from Edmonds et al. (2020), CC BY 4.0, used for the setting of the islands.
mountains_broad / mountains_standard
The GMBA Mountain Inventory v2 from Snethlage et al. (2022) including both the broad and standard hierarchical mountain range dataset. The dataset was imported in the Geopackage without further modifications, except for some invalid symbols.
ATTRIBUTION
The position flags (shifted, shift_*) and the island settings were derived with OpenStreetMap land polygons, © OpenStreetMap contributors (ODbL); no OpenStreetMap geometries are included.