Closed heikoklein closed 1 year ago
Fixed depositions now.
Refactoring the arrays or introducing a nx_out to make sure which are input and which are output are out of scope of this PR.
Results of SNAP with 100m output grid resolution, i.e. FIELD.OUTPUT_RESOLUTION_FACTOR= 25 Results of SNAP with standard 2.5km output grid resolution, i.e. FIELD.OUTPUT_RESOLUTION_FACTOR= 1
Hardly any runtime difference, memory consumption (inkluding 3d-fields) increased from 1G to 3.4G for high resolution.
Output writing is of course considerably slower in very high resolution, about 1min for this 5min run.
Coming projects required output in very high resolution, up to 100m. Lagrangian models carry model particles with exact coordinates and are have therefore not restricted to the grid of the meteorological driver. While it is already possible to interpolate the meteorology to high resolution using the FimexIO, this becomes very cost inefficient for such high resolutions. This PR allows therefore to just increase the output resolution to i.e. 25 times the input resolution, while working internally with the original meteorological resolution.