akerosenqvist
commented
2 months ago All good comments, Marko. While the current NRB spec works well for certain multi-source composite products (such as JAXA’s ALOS global mosaics), I do agree that an averaged product - in particular if generated from different orbits and look directions - poses some new considerations. The fact that the local inc angle and acquisition ID images aren’t defined for such a product shouldn’t render it non-compliant by default. Perhaps something we could discuss in the CARD4L SAR team
Comment from Marko Repse (Sinergise) regarding the averaged S1 mosaic products they produce at the SentienHub: - Ake, Clement, Peter, I had a look at the latest specs you posted and it seems to me that our mosaics cannot be compliant due to 2.4 Local Incident Angle. Since output pixels are combined from different orbits and look directions the incidence angle cannot be defined. For similar reasons, 2.8 Acquisition ID Image isn't useful even though it is required. 1.6 Source Data also balloons greatly without adding much value IMO. That said, in our internal QA of the output it is clear that the quality varies depending on the amount and properties (orbit) of the source data. Having an indicator of this in some future specs could be of interest but at the same time nontrivial to obtain (a candidate for mosaic "target" maybe?). Overall I have a feeling that for non-single-acquisition data the specs could be relaxed somewhat, the idea being that the individual source data becomes less relevant to the whole. Maybe for NRB it could be enough for the sources to be RTC? I don't know.