GODAE is sponsored by
Data definition
It is useful to adopt, with some slight modification, the nomenclature for the various stages of data preparation that was first used in FGGE (First GARP Global Experiment) and that were adopted by both TOGA and WOCE.
Level 0 data are raw instrument data, often in engineering units, as recorded by the instrument.
Level 1 data are related variables after the application of a raw signal to measurement algorithm,e.g. surface radiance or conductivity.
Level 2 data: Geophysical variables, e.g. salinity inferred from conductivity or SST inferred from surface radiance. This usually involves some form of empirical relationship. At each level we can distinguish between the degrees of screening (quality control, interpolation, etc.), from minimal actions through to sophisticated screening and interpretation. For example, at Level 2 often three sub-classes of data are distinguished:
→ Level 2a are data assembled and exchanged without significant levels of quality control, usually through an automated process in real-time. For example, data exchanged via the GTS and ingested by numerical prediction models can be classed as Level 2a data.
→ Level 2b are data that have been assembled and subjected to a significant level of quality control. The data set is changed from that telemetered from the platform, usually with the intention of adding value. The changes are subjective in the sense that a scientific or technical decision has been made in the screening process. We sometimes refer to such data as "research quality" or "climate quality" since there is an expectation that erroneous and/or biased data have been removed by "scientific" quality control.
→ Level 2c data are blended data, usually, but not always, from Level 2b. Data from different platforms, but for the same variable, may have been merged or different variables from the same or like platforms merged.
Level 3 data (products) usually take the form of a homogenous regularly gridded field.
→ Level 3a will derive from 2a, 3b from 2b, and 3c from 2c, and so on. It is not envisaged that any GODAE model and assimilation system will deliver pure Level 3a products since most systems will include at least some rudimentary form of data screening.
→ Level 3b data will usually be used to refer to model assimilation products where a more considered processing has taken place, often with some delay since the time of data collection (to allow the higher quality data set to be assembled).
→ Level 3c will be reserved for specialist, one-off products.
Level 4 data are products derived from further processing and consolidation of Level 3 data (value-adding). This may take the form of a local re-interpretation of a forecast, perhaps with the incorporation of independent local information. A seasonal climate outlook based (perhaps subjectively) on the NINO3 Index provides another example.
(Last Updated: 16-10-2007)