Quality control of metocean data: what the flags mean
Geomarins Integrated Limited ·
Why raw metocean data needs quality control, the common errors in wave, wind and current records, and how good, suspect and bad flags keep data trustworthy.
Instruments at sea work in one of the harshest environments there is. Salt, waves, marine growth, power limits and the occasional passing vessel all affect what they record. Raw metocean data always contains some errors. Quality control finds them, so that the data people use is data they can trust.
Common problems in raw data
- Spikes: single values far from their neighbours, often caused by interference or a momentary fault
- Flat lines: a sensor stuck on one value because it has failed or frozen
- Drift: a slow change in a sensor's reading over weeks, often from fouling or ageing
- Out of range values: numbers that are physically impossible or beyond the sensor's range
- Timing errors: gaps, repeated records and clocks that have drifted
- Position errors: a buoy that has dragged or broken free is no longer measuring at the intended site
How checks work
Automatic checks run on every record. Range checks compare each value with physical and instrument limits. Spike and step tests compare each value with those around it. Rate of change tests look for changes faster than nature allows. Consistency checks compare related parameters: strong winds with no change in waves, for example, may point to a problem. Position checks confirm the buoy is still on station.
Automatic checks are followed by expert review, because some problems only show up when a person looks at the data in context, alongside the weather that was happening at the time.
Flags, not deletions
Good practice is to flag data rather than delete it. Each value is marked as good, suspect or bad. Suspect values may be real but unusual, such as a genuine storm peak, and are kept for review. Bad values are excluded from statistics but remain in the archive. The original raw files are always kept, so the record can be reviewed again if needed.
Programmes such as the US Integrated Ocean Observing System's QARTOD manuals describe widely used real-time checks for ocean data, and many projects follow similar approaches.
Why it matters
A single spike in a wave record can become a false extreme value in a design report. A drifting current meter can bias a year of statistics. Quality control protects against both, and the flags and metadata that come with it let anyone using the data later understand exactly what they are working with.