Adjacent pixel drill results have different time series dates

[NormanGA]

We recently ran the command line pixel drill on 5 sites of 9 pixels each (5 3x3 spots) to return NBAR and FC reading across all available bands between 1987 and 2015. We specified that the drill use PQA filtering. The command used was:

while read lat lon; do echo "Retrieving time series for $lat / $lon" ; retrieve_pixel_time_series.py --lon $lon --lat $lat --acq-min 1987-06 --acq-max 2015-10 --satellite LS5 LS7 LS8 --mask-pqa-apply --dataset-type ARG25 --hide-no-data --output-directory $PWD --overwrite ; done <pixels.txt

We have found that the pixel drill results for adjacent pixels have quite different dates in their time series. We expect some, but are finding that there is much more than we expected. Can we check that the drill is retrieving all results in the time series repeatably?

Norman

[NormanGA]

pixels.txt

[sixy6e]

I've had a quick look at the file list returns for the locations: (145.38525, -16.85219) (145.38525, -16.85469)

And the lists contain the same records. But the data that is output is different.

My guess is it has to do with what constitutes a "has_data". https://github.com/GeoscienceAustralia/agdc/blob/master/api/source/main/python/datacube/api/tool/retrieve_pixel_time_series.py#L228

This line will exclude pixels not satisfying the "has_data" condition.