The copy of field parameters from geogrid to ERTBOX grid may give ensembles with 0 standard deviation for some grid cells

[oddvarlia]

ERT update will get problems if field parameter have 0 standard deviation. The help function to copy from geogrid to ERTBOX grid and which is used to do that for typically petrophysics field parameters does not guarantee that the whole ensemble of realizations can have grid cells with the same value (0 standard deviation) for some of the grid cells that exists in the ERTBOX grid but not in the geogrid. The extrapolation methods can for instance assign a constant value and the same value to all realizations in some cases. This will create 0 standard deviation for the ensemble for those grid cells and will create warnings and errors from update algorithms in ERT and probably also slow down the runtime for ERT in adaptive localisation.

To avoid this, a workaround today is to add some small noise to the values of the extrapolated field in ERTBOX.
This has no consequences for APS since in APS we always simulate the GRF's in the ERTBOX (and always fill the whole ertbox). It has only effect for application of the geo to ertbox copying which used extrapolation when applied to petrophysical field parameters in FMU workflows.

The effect of adding noise will only be for the grid cells that are within the ERTBOX, but not within the geogrid. So, for most of the grid cells that are physical grid cells, there are no effect of this, but for for grid cells at the boundary of active and inactive grid cells in the geogrid, the update may be have a very minor effect which in practise should be ignorable.