Support for "no-data" values in a deformation model

[ccrook]

This issue discusses how the deformation model should handle "no-data" values, where the deformation model is not defined by a producer. This could be because the producer does not have data for a region, for example offshore locations where the deformation is not measurable, or because they do not have jurisdication to define deformation in an area. In both cases it is likely that a grid covering the area in which the deformation is defined will also overlap the area in which it is not. On proposal is that the model should support some form of "no-data" value to indicate regions where the deformation is not defined.

The following notes were hastily taken during the 30-Nov-2020 meeting.

RL: define area where defined MD: need a value, add large uncertainty. Could flag ER: Defining complex area of validity would be complex to implement. MD: Issue transforming geometries spanning region MD: No-data value vs high uncertainty Complex shape of validity vs no data value - and much more efficient Choice for software handling transformation extending beyond area ER: Could be software implementation issue - not specified in standard VM: Any coordinate transformation outside grid or in no data value should fail MD: Should we recommend to implementer how to handle outside data CC: Could be handled by implementer ER: Could require implementer to evaluate to zero at edge of boundary of model RL: MD points out we need to give guidance to implementers/producers on handling extent of model.

[ccrook]

I have amended the functional model strawman with a specification and discussion to support including a "no-data" value.

[ccrook]

The specification currently does not define how no-data is represented. However this may be defined by the carrier for the model, eg GGXF. For example it could be a specific numeric value, or a NaN (not a number) value.

[ccrook]

Closed at Jan meeting with acceptance of "no-data" values

RL: Note relationship with GGXF