Data quality & validation
Keep masks, comparability and validation attached to every interpretation of change.

Could the apparent change be a difference in the observations?
Two images differ for many reasons. Before interpreting a change, consider registration, observation conditions, masks and processing choices. A comparison needs a clear account of what was included and excluded.
Keep the workflow reproducible. Preserve input identifiers, configuration and intermediate records so another person can follow the analysis. Separate a method demonstration on synthetic data from a finding about a real location.
Validation should be matched to the question. Review errors that could change the decision, not only the appearance of the output map. State what the evidence supports and where further observations are required.
What to take forward
- Comparable observations and explicit masks
- A reproducible processing record
- Decision-relevant validation
What should I bring to a conversation?
Bring the decision you need to make, the constraints you already know and a small, non-confidential example. We can then agree whether a resource, a focused review or a more detailed discussion is the right next step.
