sondes landing on ice (or "floaters") fail

[Geet-George]

Issue

For HALO-(AC)3, some sondes seem to have landed on ice. For example, 3 sondes in the 5-sonde circle flown close to the pole (last flight of the campaign, 20220412). Therefore, they kept transmitting data from the surface (outside of their design to sink, which they couldn't in this case). This causes problems with QC.

For example:

• the near surface coverage test does not work, because the alt is not calculated well by ASPEN for such sondes that keep transmitting surface values long after they have landed. Here, using gpsalt instead of alt from L1 files seems to work. The catch is that the GPS unit should be working for such sondes.
• the profile coverage also fails. From the D files, it seems that the sondes data was not received (maybe because HALO was outside range or some issue with the dropsonde instrument) for ~20 minutes, before data started being received again. My guess is that when HALO came within range again (HALO was flying circles), data started being recorded on AVAPS. In any case, the long time period without data makes the profile coverage very low and hence fails.

Suggested fix

I suggest creating another test detect_floater (I know "floater" is not an apt word here, but this is how the dropsonde community refers to it, so it is a quick reference to understand what is being said) which checks for long time periods (e.g. more than 1 minute) where the gpsalt is below 100 m and the values of pressure and gpsalt do not change significantly (this needs to be better defined).

With detect_floater, each sonde can be marked either as a floater or not. If it is one, then the near surface coverage will be calculated based on alt bounds on gpsalt and not alt. Also if it is one, the profile coverage will limit the profile measurements to a few seconds after the sonde has landed. This should check profile coverage correctly and also, there will be additional steps to exclude the "surface" values from L2 (because such measurements are no longer are a trajectory measurement) and other data levels. For example, if not excluded, the L3 regridding would otherwise be a temporal averaging along with spatial averaging/interpolation.

Implementing suggestions

• make detect_floater qc test
• add is_floater attribute
• if is_floater, the near surface qc test goes for alt bounds based on gpsalt, not alt
• if is_floater, the sonde profile should be restricted to a few seconds after landing
• the restricted profile should be used for profile coverage qc test
• the restricted profile should be used to create L2 data (and consequently, all subsequent data levels)

[Geet-George]

Solved with af7792c