Closed bascrezee closed 5 years ago
@lucap1979 This is not clear to me. In principle, I would expect my framework to work on any timeseries data. Although, currently, we use Iris to read in the data, therefore the data should be CF-compliant (which is a reasonable expection for datasets used in our project, I would say).
What exactly was the aim of the tests?
Hi Bas,
we manipulated the ERA-INTERIM dataset adding missing data in order to evaluate the impact in the trend. We found that the tool cannot handle missing data properly. So we should clarify in the guidance which kind of dataset the tool can ingest.
We discussed this at our Telecon, the tool has several options of dealing with missing data. For exampling resampling often effectively removes missing data. More advanced methods for missing data handling can be implemented in the future. Closing the issue for now.
• Test 1: taking out one decade, and reassembling the rest • Test 2: change to missing values • Test 3: as above, but updating metadata
This means that these tests are probably only useful for CF-compliant datasets, i.e. not ozone from satellites.