single mask for all inputs is wrong

[RolfHut]

The current suite setup assumes that both the GFS, GEFS and HSAF data needs to be interpolated to identical masks. This is not correct. GFS and GEFS need to be interpolated to a mask representing the grid of PCRGlobWB. HSAF doesn't need a mask, but the option for a mask is included in the hsaf preprocess script to select out regions that are of no interest (sahara?).

[nielsdrost]

I agree it should not be needed.

Mask is currently used to reduce number of observation points to something we can still handle in our DA. This is thus a performance problem.

[RolfHut]

but as I read it, all input have to use the same mask now, I do not see how we turns this off on a bigger computer in the future.