Observing the planet as one network

01Many owners, one picture

The instruments that watch the weather, the water and the ground belong to many hands. National services, cities, utilities, ports, universities and volunteers each place sensors for their own purpose and on their own budget.

Each owner sees a piece of the picture. A flood crosses a border, a storm crosses a coast, and a heat wave crosses every network at once, so the questions people ask are bigger than any one owner's map.

The meteorological community has answered this with an overarching framework for all its observing systems, which promotes network integration and partnership so that members can build capability, widen coverage and improve economic efficiency¹.

Gaps remain. In some parts of the world observations are not made or not exchanged internationally, and in others they are not made or exchanged often enough².

That is why a basic surface network is now being designed, defined and monitored at the global level rather than country by country².

The stakes are public. The United Nations has set the aim that everyone on Earth is protected by early warning systems by the end of 2027³.

No single operator can meet a goal like that alone. A network of many owners can, if each piece is built to join the others and to keep working when one of them is missing.

02Open data models: one way to say what was measured

Joining starts with language. If one owner records a gauge as a row in a spreadsheet and another as a file named by date, nobody can merge the two without a translator.

Open data models remove the translator. OGC SensorThings gives an open, standard way to manage and retrieve observations and metadata from different sensor systems over the web⁴.

Its sensing model names the parts that every observation shares: the thing that carries the sensor, its location, the sensor, the property observed, the observation and the feature of interest it describes⁵.

An observation in that model can also carry a statement of result quality and the period in which the result may be used⁵.

Common models do not force owners to buy the same equipment. They only ask that every station describes itself in terms that others can read, which leaves each owner free to choose instruments for its own needs.

The same idea applies to web access. OGC API Features provides building blocks to create, modify and query features on the web, so stations and records can be shared with ordinary tools⁶.

A shared model also makes gaps visible. When every station is described the same way, a catalogue can show where a property is measured, where it is not, and where a record has gone quiet.

03Shared time and shared identity

Two more things must be shared before records can be merged: the moment and the source. A value is only comparable with another if both clocks point to the same reference.

Coordinated Universal Time is that reference. National metrology institutes maintain physical realisations of it and contribute their clock data to the international bureau that publishes the time scale⁷.

Shared time also exposes faults. When many stations from different owners report against one reference, a clock that drifts or a station that has been tampered with stands out against its neighbours.

Identity matters just as much. In a network of many owners, nobody can trust a station because of where it is plugged in or who bought it.

Zero trust architecture states this plainly: no implicit trust is granted to assets based solely on their network location or ownership, and authentication and authorisation happen before a session is established⁸.

Remote attestation supplies the evidence for that decision. An attester produces believable information about itself, a verifier appraises it, and a relying party decides whether to treat the attester as trustworthy⁹.

Together these let a station from one owner join another owner's network on evidence rather than on reputation. The same rules apply to a national agency and to a school with one weather mast.

04Federation: publish once, reach everyone

Federation is how many networks act as one without merging into one. Each owner keeps its own systems and publishes through shared services that no single owner controls.

The WMO Information System is a working example: a framework for data sharing across all WMO domains and disciplines, built on open standards and web technologies¹⁰.

Its technical specifications and implementation plan were approved in October 2022, and it will gradually replace the older global telecommunication system¹⁰.

In that design a publisher's node announces each new resource with a notification message over MQTT, and everyone subscribed to that dataset's topic receives it at once¹¹.

Global brokers republish the notifications for resilience, and consumers are advised to subscribe to more than one broker so that notifications keep arriving if one fails¹¹.

Global caches keep copies of core data for rapid access, so a small national centre does not have to serve the whole world from its own link¹¹.

The lesson travels beyond weather. A federation needs open message formats, more than one path for every message, and a catalogue that tells users what exists and where to find it.

Federation also spreads the effort. Each owner looks after its own stations and its own node, while shared services run in more than one place so that none of them becomes a single point of failure.

05Evidence that travels with the data

A shared network is only as trustworthy as the weakest record in it. Users need to know where each value came from, what was done to it and who was responsible.

Provenance is information about the entities, activities and people involved in producing data, and it can be used to assess its quality, reliability or trustworthiness¹².

The W3C PROV family defines a model, serialisations and supporting definitions so that this information can be exchanged between very different systems¹².

The FAIR principles ask the same of research data in general: metadata should be rich, carry a clear usage licence, be associated with detailed provenance and meet community standards¹³.

When evidence travels with each value, a user far downstream can judge it without calling the owner. When it is stripped away, every merged record becomes a matter of faith.

Corrections need the same path. If a calibration upstream turns out to be wrong, the record should show which values depended on it and which have since been reissued.

Evidence also protects owners. A station that publishes its quality flags and its calibration history can show that its data was sound, even when someone else's analysis went wrong.

06Built from open standards that already exist

None of this needs a new invention. The building blocks are published, open and in use, and the work lies in joining them so that owners can take part without giving up control.

Provenance has had a W3C model since 2013, and sensor observations have had an OGC web model since 2016¹²¹⁴.

The FAIR principles for data were published in 2016, and zero trust architecture and remote attestation give the vocabulary for admitting stations on evidence¹³⁸⁹.

What remains is discipline. Every owner publishes in open models, keeps time to a shared reference, proves the identity of its stations and sends the evidence with every value.

Start small. A single station that keeps good time, describes itself openly and publishes its evidence is already ready for the wider network, even before anyone else connects to it.

Nexus Observatory Inc. designs observing sites and software around these open standards, so that each station can join a wider network from the day it is installed.

The result is a network that no single party owns and every party can rely on. Many owners, one picture, and a record that anyone can check.