Instruments you can trust, and how to know it

01A reading is a claim, and claims need evidence

Every instrument makes a claim: the air here is this warm, the wind this fast, the water this high. Instruments you can trust are the ones that can show why the claim should be believed.

Metrology has a precise language for that evidence. A measurement result is a value together with its measurement uncertainty, a parameter that describes how widely the true value may lie around the stated one¹.

Three further facts complete the picture. The calibration links the instrument to a reference, the siting says what the instrument could actually see, and the maintenance record says whether it was working as calibrated when the value was taken.

None of this is visible in the number itself. Two identical readings from two stations can deserve very different levels of confidence, and only the evidence beside them tells the user which is which.

This article follows that evidence from the laboratory to the archive: calibration and traceability, uncertainty and drift, siting, metadata, quality control and the shared rules that make many networks comparable.

02Calibration and the unbroken chain

Calibration is more than adjusting an instrument until it agrees. It establishes a relation between the values given by measurement standards and the instrument's indications, each with its uncertainty, and then uses that relation to obtain results¹.

The vocabulary warns against confusing calibration with adjustment, which is often mistakenly called self-calibration¹. An instrument that adjusts itself has not shown that it agrees with anything outside itself.

Metrological traceability is the property that lets a result be related to a reference through a documented, unbroken chain of calibrations, each contributing to the measurement uncertainty². The reference is usually a realisation of an SI unit held by a national laboratory.

National laboratories establish traceability for their own results, and other organisations are responsible for establishing the traceability of theirs³. A network operator cannot borrow traceability; it has to keep the chain.

For weather and climate instruments, regional instrument centres maintain calibration standards and help countries calibrate their national standards and monitoring instruments⁴. Several of these centres are accredited to the international standard for testing and calibration laboratories⁴.

Traceability does not on its own make a result fit for purpose. The vocabulary notes that it does not ensure the uncertainty is adequate, or that there were no mistakes².

03Uncertainty, drift and the calibration interval

Measurement uncertainty includes components from systematic effects, such as those tied to corrections and to the values assigned to measurement standards¹. It is a budget, and every link in the calibration chain spends part of it.

Instruments also change. Instrumental drift is a continuous or incremental change in an instrument's indication over time, caused by changes in its own metrological properties rather than in what it measures¹.

Drift is why calibrations expire. A field programme sets an interval for each sensor, based on how quickly that kind of sensor is known to change and how much uncertainty the users can accept.

Fire weather stations in the United States offer a worked example. Humidity and temperature sensors are recalibrated or replaced every year, and wind sensors every two years⁵.

Tipping-bucket rain gauges are checked in the field each year and calibrated or replaced at a depot every three years⁵. The depot's test equipment must itself follow a traceability protocol to national standards⁵.

Intervals like these are not universal constants. They are a programme's judgement, written down, so that every user can see how old a calibration may be.

04Siting: what the instrument actually sees

A perfectly calibrated sensor in the wrong place measures the wrong thing. A thermometer beside a wall, an anemometer in the lee of trees or a rain gauge under branches will all report faithfully on conditions nobody wanted to know.

The international guide sets siting recommendations for each variable, and a siting classification estimates how well a station meets them⁶. The classification was published jointly as an ISO standard in 2014 and as an annex to the guide⁶.

The class number is not a ranking of a station's value. A station may have very good reasons for its exposure, and its purpose decides whether that exposure is acceptable⁶.

Declaring the class is the point. Sharing each parameter's classification is more useful than implying that every station meets the guide, and it lets users judge quickly how representative the data may be⁶.

Programmes write the same idea into practical rules. The fire weather standard asks that a station stand at least as far from buildings, trees and dense vegetation as they are tall⁵. The comparison shows why.

05Metadata: the record that makes a value reusable

Metadata is the evidence written down. Without it, a long record cannot be trusted across a sensor swap, a station move or a change of method.

Good programmes are specific about what to keep. The fire weather standard requires the maintenance date, the specific sensors and their serial numbers, and a narrative of each site visit, with photographs and site access notes kept in the station file⁵.

It is just as specific about honesty. When a site is badly compromised, that fact belongs in the station metadata, with photographs, so that users can judge the data for themselves⁵.

Location needs the same care. Basic station metadata such as latitude, longitude, elevation and aspect is shared with many systems, so it has to be accurate⁵.

Open interfaces carry the metadata with the data. The OGC SensorThings API offers a standard way to manage and retrieve observations and metadata from different sensor systems over the web⁷.

06Quality control, from the edge to the archive

Quality control asks a simpler question than calibration: is this value plausible right now? It runs continuously, close to the instrument, and it leaves flags rather than deleting anything.

The fire weather standard describes two methods side by side. An automatic oversight system watches for out-of-range observations and non-functional sensors, and agency users periodically review and verify the data⁵.

Ocean observing offers a mature model. A long-running national effort writes authoritative procedures and manuals for real-time quality control of variables such as wind, water level and waves, with a common manual for quality flags⁸.

Lower-cost sensors need an extra step. Their accuracy is understood by collocation, running them side by side with reference monitors, and by evaluating them against performance targets⁹.

Every check that runs should travel with the value. A user who can see which tests passed, which failed and which were never run can decide how much weight a reading deserves.

07Shared rules for many owners

Evidence only travels if everyone describes it the same way. That is why measurement communities put so much effort into shared vocabularies, guides and procedures rather than into any single instrument.

The international observing programme for weather instruments sets technical standards, quality control procedures and guidance, with the aim of improving the quality and long-term stability of observations¹⁰.

Its vision is fit-for-purpose environmental measurement, in which users understand measurement quality and providers of essential climate variables are committed to measurement traceability¹⁰.

The current guide to instruments and methods of observation includes volumes on quality management, sampling, data reduction, and testing, calibration and intercomparison¹¹.

For a network of many owners, this is what makes federation possible. A value measured by one operator can be used by another, because both can read its calibration, siting, metadata and flags in the same terms.

Nexus Observatory Inc. designs its instruments and stations on this principle: every value leaves the site with the evidence needed to trust it, in open terms that any user can check.