When we run this algorithm in production, we'll be wanting to produces a per-scene product.
I wonder if we can frame up the processing to take a dataset ID and from that, go into the past or future of observations and build a new QA band that matches the original data footprint.
Also, output should be Cloud Optimised GeoTIFF and be accompanied with at least YAML metadata, including lineage, and preferably also a STAC conversion of the YAML.
When we run this algorithm in production, we'll be wanting to produces a per-scene product.
I wonder if we can frame up the processing to take a dataset ID and from that, go into the past or future of observations and build a new QA band that matches the original data footprint.
Also, output should be Cloud Optimised GeoTIFF and be accompanied with at least YAML metadata, including lineage, and preferably also a STAC conversion of the YAML.