Find data in the catalogue

Data Explorer (choose it in the sidebar) is the catalogue: the datasets Orion hosts (boundaries, population, land cover, climate, transport, health and more), datasets OrionAI has found on publishers' own portals, and the datasets you have uploaded yourself, in one searchable place.

#Search

The search box has two modes, switched with the toggle beside it:

• Keyword — searches titles, descriptions, keywords and publishers. Results that also match by meaning are ranked in.
• ✦ AI — describe what you need in plain language ("rainfall over Meru county") and AI finds matching datasets by meaning.

Press Enter to see the results.

#Geographic coverage

A continent includes datasets for the continent itself, its regions and its countries. A region includes its countries; a country includes local datasets tagged with that country. For example, Africa includes East Africa, Kenya and country-tagged Nairobi datasets. A dataset appears once in each result list, even when several of its country tags match. Global lists explicitly worldwide datasets. Dataset page links keep their original place.

#Narrow the results

Use the filter chips above the results:

• Type — Points, Lines, Polygons (areas such as boundaries, districts or flood extents), Raster (gridded imagery and surfaces) or Table.
• Source — Hosted (data Orion holds) or Discovered (data OrionAI found on a publisher's own portal, marked "Discovered by OrionAI" with the publisher's name).

To browse by subject (boundaries, health and so on), choose a category card on Home; a Category chip for it then appears beside these filters.

Click a chip again to remove it. Switch between Grid, List and Map views with the view toggle, or choose Explore Catalogue (Explore Results when filters are on) to open everything on one interactive map. In ✦ AI mode the matches replace the keyword results and the view toggle is hidden.

Every card and list row has two buttons: View opens the dataset's page and Save bookmarks it to your data.

#Open a dataset

Each dataset has one page, at an address like /data/kenya/health/health-facilities-k7f3: the place it covers, its theme, and its name. The same page works for everyone, signed in or not, so you can send the link to anyone. It shows:

• a preview map of the data: its thumbnail, or until it has one, the data itself drawn live on a map you can pan and zoom;
• the description and documentation as the publisher wrote them;
• the columns and a few sample rows;
• details: publisher, licence, coverage, years, type, format, size and when it was last updated and checked, plus how to cite it;
• related datasets.

The actions are in the panel beside it:

• Open in map — a new map in Map Studio with the dataset on it.
• Add to project — adds it to one of your projects.
• Save — bookmarks it.
• Download — the whole dataset, where the licence allows: the publisher's own original file when they offer one, or a full export in the format you choose — GeoJSON, Shapefile (.zip), KML for Google Earth, GeoPackage, CSV (geometry as WKT) or FlatGeobuf. Hosted imagery downloads as GeoTIFF. A download is always the complete data from the source, never the sample rows shown on the page. Very large datasets (over 250,000 features) download as the publisher's original file or as an area you choose. After your first download we send a few short emails about that dataset (related open data for the same place, and how to see it on a map); turn them off any time under Email on your profile.
• Ask OrionAI — opens OrionAI with the dataset in context.
• Share — copy the link, or share it to WhatsApp, LinkedIn, X, Facebook or email. The link shows the dataset's map preview.

Viewing and sharing are open to everyone. The other actions need a free account: if you are not signed in, choosing one takes you to sign-up (or, if you already have an account, Sign in from there) and then straight back to the page with the action done.

Signed in, a hosted dataset also offers Explore on an interactive map and the full attribute table. The table keeps loading rows as you scroll, with its column names pinned at the top, so you can read the whole dataset without paging.

If something on a page is wrong, or you publish the data and want it removed, choose Report or request removal at the bottom of the panel. Our data team reviews every report.

#Browse without an account

Anyone can browse the open catalogue at /data, by place (for example /data/kenya) and by theme, without signing in. Every browse page has a search box for its own scope (all data, one country, one theme, or a theme in a country), which matches by keyword and by meaning. Lists keep loading as you scroll, so there are no pages to click through; this works the same in Data Explorer. Signed out, the browse pages and related datasets list only datasets that have passed review and are published; sign in to search everything in the index, including datasets still being reviewed.

#Add a dataset to a project

Choose Add to Project on a dataset, pick the project (or create a new one), and the dataset appears in that project's Data tab. You can then create a map from it. See Data in a project.

Datasets you upload appear in your own searches next to the others, but only you can see them until you make them public. Their pages are private too. See Upload your own data.

Old links to /dataset/<number> keep working: they take you to the dataset's page.

#Discovered datasets

Datasets marked ✦ Discovered by OrionAI live on their publisher's own portal (for example the Humanitarian Data Exchange, geoBoundaries, NASA, Copernicus, Digital Earth Africa or the Microsoft Planetary Computer). OrionGIS indexes their description, licence, coverage, columns and a few sample rows, and credits the publisher by name on the page. Each has an Access level:

• Open: the licence allows loading and copying.
• With your own account: the data loads through your own connection to the publisher.

The catalogue lists only data you can actually use. Data whose publisher allows viewing only, with no right to load or share it, isn't listed.

#When the publisher left something out

Some publishers give no description, or no coordinates at all (a table of inspections with only street addresses, for example). OrionGIS fills the gap and says so on the page:

• No description: OrionAI writes a short one from the dataset's title, columns and sample rows. It is marked Written by OrionAI under the text, and is replaced as soon as the publisher adds their own.
• A short description: when the publisher's description is under 120 characters, OrionAI adds two or three paragraphs under About this data, written only from the dataset's columns, sample rows, coverage, licence and publisher. They are marked ✦ Described by OrionAI, and the publisher's own text stays above them unchanged.
• No coordinates: OrionGIS finds the area the data covers from its address and place-name columns (street, town, county, state, country, or a site name with its admin names), or from the place named in its title or description, using OpenStreetMap. The coverage is marked Approximate area in the details.
• No columns or sample rows: OrionGIS reads a few rows from the publisher's own file or service (a feature service, GeoJSON, zipped GeoJSON or Shapefile, GeoParquet or CSV), trying each download the publisher offers until one works. It reads only the first rows, never the whole file, and tries again later if the source is unavailable.
• No area in the record: when the publisher's service knows its own extent (a PMTiles file, a WMS layer, an ArcGIS service or a Cloud-Optimised GeoTIFF), OrionGIS reads it from there. That area is exact, so it isn't marked approximate.

#Summary and questions on every dataset page

Every public dataset page opens with a one-paragraph summary under its title: what the data is, the area and years it covers, who publishes it, its licence and its size. Further down, Questions about this data answers what it contains, what area and years it covers, whether you can use it commercially, how to cite it and how to load it into a map. Both are built from the dataset's own details, so they always match the rest of the page.

#Load a discovered dataset into a map

Open and "With your own account" datasets have Open in map. Choose your area first (draw it, or use your project's extent). The area map shows the dataset itself, so you can see where the data is before you choose. Then:

• Small vector data is copied into your workspace as a private upload. Copies count toward your plan's storage.
• Larger data and all imagery are added as a live layer: streamed from the publisher and limited to your area. Live layers don't count toward storage. If an area holds more than a copy allows, OrionGIS offers a live layer instead.
• Save a snapshot of my area turns a live raster layer into a GeoTIFF of just your area, saved as a private copy.

The publisher, licence and citation travel with every layer you load.

#Satellite scenes

Imagery collections open a scene search: your area, a date range and a maximum cloud cover. Each scene shows its date, cloud cover and a quicklook. Add one in true colour or as an index computed by Orion's tiler:

• NDVI: vegetation
• NDWI: water
• NDBI: built-up land
• NBR: burn scars

You can also ask OrionAI, for example "NDVI for Nairobi from the clearest scene in September". See Ask OrionAI.

#Find places like this

Earth-embedding layers, such as the AlphaEarth Foundations Satellite Embeddings from Google and Google DeepMind, describe every 10 m of land with 64 numbers that summarise the satellite record for a year. On such a dataset, Open in map opens Find places like this:

1. Click a place on the map, for example a farm, a forest patch or a settlement.
2. Keep the map on the area to search (about 100 km across at most) and choose a year.
3. Press Find similar places. Every part of the area is coloured by how alike it is to your place.
4. Press Keep as a new map to save the result in Map Studio.

#Follow and publish

• Follow a place, theme, dataset or portal to hear about new and changed data. See Follow places, themes and datasets.
• Publishers can register their portal or dataset. See Publish your data on OrionGIS.
• Developers can read the index's public records as open STAC at /data/stac and DCAT at /data/dcat.jsonld.

#File previews and samples

Orion can read Shapefile, GeoPackage and KML/KMZ sources for map previews, thumbnails and loading, including supported files inside ZIP archives. The source must include a coordinate reference system. File reads have size and feature limits; a source that exceeds them stays unavailable rather than showing an empty preview.

GeoPackage and KML sources can supply sample columns and rows. Excel (XLSX) sources can supply table samples. A spreadsheet without geometry does not automatically get a map thumbnail.

Complex file reads run separately from the AI service. Preview processing is limited to 30 seconds, with a separate memory budget. If a source exceeds the budget, Orion reports that it could not read the preview and leaves the record for retry or review. This keeps other AI and data requests available.

#Every dataset is ready for a map

Everything you find in OrionGIS, in Data Explorer, on a search engine or through OrionAI, can be put on a map. A dataset that ships its own map file (GeoJSON, Shapefile, KML, GeoPackage, imagery or a map service) is published from that file; its CSV or spreadsheet, if it has one, stays available to download. A table with no map of its own appears only once OrionGIS has placed it:

• Coordinates become points.
• Admin areas at every level a country has (provinces, regions, counties, sub-counties, districts, wards and finer) are matched to OrionGIS boundaries by name or code, including ISO 3166-2 codes such as KE-43. A table covering many countries is placed on each country's own areas, so two places with the same name in different countries stay apart. When some rows give a ward and others only a county, the map shows the finest level and every row is joined to its own area in the full table and downloads. Each area shows its own latest figures.
• Countries are matched by name or ISO code. Regions such as East Africa, West Africa or Asia, and World Bank regions such as Sub-Saharan Africa or East Asia & Pacific, are drawn as the outline of their member countries. When a table lists countries and regions together, the map shows the countries and the region rows are joined in the full table and downloads.
• Addresses and place names are geocoded to points.
• National figures only are drawn on the country's outline.

Read the join summary on the page for the areas matched and any left out, and the breakdown label for the population or category shown. Maps show the selected breakdown, the latest value and a field per year; spatial exports contain every matched source row. OrionGIS stores only how to join the table: the rows are read from the publisher each time you open or download the data, so you always get the current figures.

#Premium Orion Datasets

When OrionGIS spatializes a statistics table — joins it to boundaries and adds the location component — the result is a derived product owned by Orion Terrametrics, marked on the page with a gold ★ Premium Orion Dataset badge and a Derived by row. The original publisher stays fully credited: their name, licence, citation and a description sentence naming the source. The marker identifies Orion's derived products, which will be offered under premium terms.

Tables come from statistics office APIs (SDMX, PX-Web, the KNBS Data Portal, ONS, Statistics Canada, IBGE and others), the World Bank, UNICEF and HDX APIs, and the CSV and Excel tables published on data portals. A table that lists the country, counties and sub-counties together is shown at the level that matches best. Tables that compare countries are drawn on country outlines.

Tables that cannot be reliably placed wait unpublished, and are retried. This includes unsupported formats, ambiguous names and tables beyond the current read limits (150 MB; 1 million statistics cells), which are being raised further. Coverage grows as each portal passes validation; an audited country is not necessarily available in the public catalogue yet.

Datasets from different organisations are all kept, even with the same title. Within one source, a second copy is only added when it is better (it works, is on the map, is official, has finer detail, is newer or more complete), and then alongside the original. Original pages and download links are preserved.