1-01-01 Atmospheric variables

[joergklausen]

From discussions with colleagues at NILU:

Especially for variables on aerosol particle properties, variable names often need further modification. An example: one such property is the particle light scattering coefficient. This variable is measured for various particle size fractions: PM1 (aerodynamic diameter < 1µm), PM2.5, PM10, TSP (total suspended matter). It can be measured for various conditions of temperature and pressure: standard, ambient, lab. It can be measured with several types of sample conditioning, here RH: dry, dried, ambient. In this example, this leads to 36 different combinations of modifiers for one variable. Depending on the use case, these modifiers can be regarded as discovery or use metadata. An expert user wants to choose exactly matching time series to compare. Other users might regard these modifiers as additional information, i.e. use metadata.

A well defined metadata schema should be usable for both use cases.

In the current WMDR specification, these modifiers are added to the variable name, resulting in numerous versions of the same variable. This approach is certainly possible if such modifiers occur only occasionally. However, this approach has 2 drawbacks: 1) in case the modifiers are used more than occasionally, the list of variables is blown up considerably; 2) variables can’t be served according to use case, the modifiers are always present, needed or not. If we use this approach of appending the modifiers to the variable names we need to define roughly 1700 new variable names.

This would be unreasonable. We need to develop a consistent approach to this.

Related issues are #168, #169, #170.

[markusfiebig]

The solution to this issue could be rather simple by introducing a key/value pair into the WMDR schema, e.g.

• Key: particle size fraction, values: PM1, PM2.5, PM10, TSP
• Key: wavelength, values: 355 nm, 470 nm, …
• Key: reference pressure, values: 1013.25 hPa, ambient, instrument internal, ...

[gaochen-larc]

We need a systematic approach for the aerosol variables. It is not practical to add modifiers to the variable names. Are we allowed to add attributes or variable descriptions? Fundamentally. there are many a dozen or so aerosol measurements. These variables, however, need additional attributes to define: size range, size type (for microphysical and composition measurements), sampling RH, reporting temperature and pressure, and wavelength (for optical properties). Furthermore, if size ranges are limited to PM1, PM2.5, PM10, and total, the data providers will have to look for a close match. The size range for measurements I deal with is less 5 um.

[fstuerzl]

Proposal of a systematic approach for modifiers (cited from Wiki - WG ACV):

Separate modifiers from variable names (see issue #173). Review elements proposed for new table "Sampling procedure" (Issue #111) that currently combines size cut off + sampling condition. Review elements proposed for new table "Sample treatment" (Issue #112). Establish new elements for reference conditions (referencePressure, referenceTemperature, referenceHumidity) for use in DataTypes ‘Sampling’ and ‘Reporting’. This involves an extension of the WMDR model (!).

[gaochen-larc]

Size cut off should indicate both lower and upper limits. Propose to add sizing technique attribute.

[amilan17]

@joergklausen - I just noticed that this issue doesn't have a branch. Can you confirm that this issue was fully addressed in in other branches and can be closed with FT2021-2 is released?

[joergklausen]

Issue should be treated as part of the I-ADOPT approach to variable name construction (#464).