davidfrantz / force

Framework for Operational Radiometric Correction for Environmental monitoring
GNU General Public License v3.0
174 stars 50 forks source link

Apparent data loss when selecting EQUI7-EU as output #256

Closed SerRichard closed 2 years ago

SerRichard commented 2 years ago

Describe the bug Hello! Pretty sure this is a bug in FORCE, as my colleague and I have been able to replicate this anomaly across two FORCE deployments and with varying S2 file inputs. It appears that when defining the data-cube to re-project output to the equi7grid, we are consistently missing some outputs, see circled are of the image below. The S2 tile covers two other equi7 tiles which we would expect to see reflected in the product output.

S2 Grid as dashed line Equi7Grid as solid line

cover_missing excalidraw

Expected behavior We would expect 6 products to be output by FORCE, and not 4. This leads to orderly and repeated no data values across our ARD data. This doesn't seem to be down to the configuration of the prm file. If helpful, the input scene in this example is: S2B_MSIL1C_20200910T100559_N0209_R022_T32TPS_20200910T124936.

Parameterization In case you suspect this could be on our side, here is an example of our current parameters, with only any file-path options blank. If you see anything spurious on our side please let me know and I can correct & rerun.

example.prm

Setup

Though my colleague is using a different setup, I am not 100% what it is. Tia for any feedback and time spent replying to this, if we can provide anything else just let me know.

davidfrantz commented 2 years ago

Hi @SerRichard,

I am not convinced yet that it is a bug. Have you tested with MAX_CLOUD_COVER_TILE = 100?

Cheers, David

PS: you are using a nearly two year old release that should actually not work with newer S2 data anymore

SerRichard commented 2 years ago

Thanks for the speedy reply! Will change the parameter and also bump to 3.7.10! Will reply with findings!

SerRichard commented 2 years ago

Wonderful, this looks complete now! Many thanks! Will try do do a better job reading the parameterisation docs next time :wink: !

davidfrantz commented 1 year ago

I am glad it worked. Best, David