Geospatial data catalog: which one should you choose?

What a geospatial data catalog is for

A geospatial data catalog brings together records for your datasets: what they contain, who produces them, which area they cover, when, and under what license. Depending on the tool, it either stops at that metadata or also hosts the data, shows it on a map and publishes it as OGC services.

Before comparing software, ask three questions: do you only want to describe your data, or also distribute it? Must you follow a mandatory metadata standard, such as INSPIRE in Europe? And is your data mostly geographic, or mixed with tables and documents?

The four tools at a glance

Which catalog for which need

• You must comply with INSPIRE or a national spatial data infrastructure: GeoNetwork is the reference for standardized metadata and harvesting.
• You publish open data of every kind (budgets, statistics, maps): CKAN, with its spatial extension, is the most widespread choice.
• You want to host, view and distribute your layers in one place: GeoNode combines catalog, maps and OGC services.
• Your whole organization already works in ArcGIS: ArcGIS Hub fits in naturally, and it is often the simplest path.

These tools can also be combined: GeoNetwork can harvest GeoNode records, and CKAN can reference services published elsewhere. The right choice depends on who needs to find your data and what they need to do with it.

Finally, think about the way out: how you would leave a tool if your choice turns out to be wrong. All three open source projects expose their records in open formats and through documented APIs, and your layers remain standard files or PostGIS tables. With a proprietary service such as ArcGIS Hub, migration is possible, but it goes through the platform's own export tools. For a public agency, this reversibility often weighs as much as features, because a catalog lives for years and usually outlasts the team that chose it. So ask the question at the tender stage: in which formats can records and data be retrieved, and at what cost? It is also worth asking how users and permissions are exported, since rebuilding access rules by hand is often the slowest part of a migration.

Writing useful metadata

Software does not make a good geospatial data catalog: records do. A vague record ("Roads layer 2") will neither be found by a search engine nor understood by a user. Whatever tool you choose, a good record answers the questions a stranger would ask before downloading the data.

• A title that stands on its own: subject, area and year, for example "Road network of the Central Region, 2025", rather than a file name.
• An abstract of three to five sentences: what the data contains, how it was produced, what it is for, and what it does not cover.
• An accurate extent: it drives search by area in GeoNode, in GeoNetwork, and in CKAN with its spatial extension.
• The date and update frequency: 2019 data can still be valid, as long as users know it.
• The coordinate system and scale or resolution: a 1:2,000 cadastral plan and a 1:200,000 map do not serve the same purposes.
• The license: open, restricted or paid; with no license shown, a careful user will not use the data.
• A contact: a department rather than a person, so the address stays valid when staff change roles.
• Keywords from a thesaurus: a shared vocabulary, such as the INSPIRE themes in Europe, makes your records comparable with those of other producers.

If you must comply with INSPIRE, the GeoNetwork documentation covers editing metadata that follows the technical guidelines based on ISO 19139, and the tool can validate records. Otherwise, start with a short internal record template that every producer fills in the same way. Ten complete records are worth more than a hundred half-empty ones.

Setting up a catalog, step by step

A successful catalog is an organizational project first and a technical project second. The following steps apply to all four tools compared here.

1. Take inventory: list existing datasets, their producers, their formats and where they are stored. The inventory often reveals duplicates and orphaned data.
2. Decide on the audience: internal staff, partners, the general public, buyers. The audience determines permissions, record language and the choice of tool.
3. Choose the tool based on the criteria above, then the hosting: your own servers, a hosting provider, or a service such as ArcGIS Hub.
4. Set a record template and a keyword vocabulary, with one completed example as a reference.
5. Load a first batch of fifteen to thirty of the most requested datasets, with complete records.
6. Open to a pilot group of users, collect their feedback on search and records, and fix what needs fixing.
7. Name an owner for each dataset and set an update rule, then open to the intended audience.
8. Measure: datasets viewed, searches with no results, downloads. Searches with no results show which data is missing.

Plan for harvesting from the start if your country has a national spatial data infrastructure. GeoNetwork can harvest on a schedule from CSW catalogs, WFS services, OAI-PMH sources or other GeoNetwork instances; in the other direction, the CKAN spatial extension relies on pycsw to expose a CSW service. A catalog connected to others is found far more often than an isolated one.

Common mistakes

The same mistakes come up in most geospatial data catalog projects. Knowing them helps you avoid most of them.

• Choosing the tool before the need. Installing GeoNode because it is comprehensive, when only a standards-compliant metadata list was required, adds components to maintain with no benefit.
• Underestimating memory. GeoNode relies on GeoServer, and its documentation recommends reserving at least 4 GB of memory for GeoServer. A small shared server is not enough.
• Launching without content. A catalog launched with three records makes a poor first impression that lasts. It is better to open later with a solid first batch.
• Forgetting permissions in the services. In GeoNode, permissions are enforced all the way into GeoServer through the GeoFence module. With other setups, check that a layer restricted on the portal is not still readable through its WMS or WFS address.
• Letting records go stale. A record with a broken download link undermines the whole catalog. A quarterly check of links and dates is usually enough.
• Neglecting software updates. A catalog exposed on the Internet receives security fixes that must be applied, together with the components around it.
• Relying on a single person. When the only administrator leaves, the catalog stops. Document the installation, the backups and the update procedure.

None of these mistakes is specific to one product. They almost always come from a project launched as a technical installation, with no one in charge of content and no maintenance schedule.

Six criteria for comparing geospatial data catalogs

Beyond the feature sheets, these criteria help you decide between the tools for your specific case. Rate each one for your context before looking at secondary features.

• Metadata standards. GeoNetwork natively handles ISO 19115 and ISO 19139, and its documentation covers editing INSPIRE metadata. CKAN has its own record model; its spatial extension can harvest ISO 19139 metadata. GeoNode produces records exposed through CSW.
• Data hosting. GeoNode stores layers and publishes them through GeoServer. CKAN stores files of any kind. GeoNetwork usually points to data hosted elsewhere.
• Geographic search. All three open source tools support search by extent; in CKAN, the ckanext-spatial extension adds it, and map previews come from another extension, ckanext-geoview.
• Harvesting. GeoNetwork harvests other CSW catalogs, WFS services or other GeoNetwork instances, which makes it a good node in a national infrastructure.
• Permissions. GeoNode manages per-resource permissions, enforced all the way into GeoServer. CKAN works by organization and by dataset.
• Running cost. The more a tool does, the more components it has to maintain: GeoNode needs more memory and care than CKAN or GeoNetwork alone.

One last point often matters more than features: the language and habits of your users. A catalog nobody fills in is useless. Plan who will enter records, how often, and choose the tool whose interface feels simplest to them.

Four typical situations

The criteria make more sense when applied to real organizations. Here are four common situations and the reasoning that leads each one to one tool rather than another.

Large organizations often combine two of them: GeoNode for the GIS teams' daily work, and GeoNetwork harvesting its records to meet publication obligations. An organization already equipped with ArcGIS Online, on the other hand, rarely gains from running a second portal alongside it: ArcGIS Hub, offered at two license levels, Basic and Premium, is then the most direct route. Whatever the case, write down the reasoning behind the choice, so that the next team understands why the catalog looks the way it does.

What if the catalog must also sell?

None of these catalogs is designed for selling first. They distribute, describe and control access, but they cannot take a payment, open access for a set period or clip data to the purchased area. Yet that is a common need for data producers, public and private alike.

For that case, LizPack Catalogue starts from GeoNode and GeoServer, two open source projects, and adds a store: full dataset download, extract for a chosen area, WMS and WFS access for a period, printable PDF map, and payment by bank card and Mobile Money. Data on sale shows only its preview; its attributes and files are reserved for buyers. The full approach is in our guide how to sell geospatial data online.

Hosting your geospatial data catalog

A catalog is a web service that must stay online, up to date and backed up. GeoNode, for example, combines several components (web application, PostgreSQL database, GeoServer) that must be upgraded together. That is real administration work, similar to what is described in hosting GeoServer online.

LizPack Catalogue can be installed on your own servers, if your data must not leave your premises, or hosted and maintained by LizPack. Pricing is by quote. For organizations that want their whole GIS platform in house, the LizPack Entreprise edition is covered in hosting your GIS on your own servers.

Frequently asked questions