Known bugs

Date	Reported by	Description
08/05/17	MS	Domain-averaged velocity variances over surface roughness heterogeneities look strange
28/04/17	MS	Spectra: wavenumber do not match theoretical value (Pi/dx) at grid cut-off length
22/02/17	MS	Segmentation Fault in netcdf_interface_mod in case of land surface model - Subscript 1 of the array SECTION has value 101 which is greater than the upper bound of 100
05/01/17	MS	Within revision r2092 (and former) there is still a bug in the LPM, particle speeds are getting too large so that particle exchange between cores does not work properly, possibly leading to segmentation faults. Remark: Fixed in revision r2100.
18/10/16	MS	Output of initial profile of passive scalar quantity #s is not correct.
12/08/16	MS	Only a 'latent' bug, however, in case of topography = .T. initialization of the PALM grid does not scale with the number of processors, as two global arrays required only (!) for NetCDF output are necessary (zu_s_inner, zw_w_inner). If you are short on internal memory, it is possible that the required memory exceeds the available memory in case of very large grids.
13/06/16	MS	Only if masking_method = .T.: Over-determination of Neumann boundary condition of perturbation pressure in poismg_noopt.f90 if a grid point within topography adjoins to two or more 'flow' grid points. This case, Neumann boundary conditions are not correctly set for the 'flow' grid points as the perturbation pressure inside topography is an average over all the adjoining 'flow' grid points. At this point it should be noted that is is not the reason for the velocity blow-up in case of complex topography.
13/06/16	MS	Less divergence reduction with masking_method = .T. compared to masking_method = .F. . Please see remarks below.
09/05/16	MS	For very complex topography, e.g. very narrow street canyons or corners that are resolved by only one or two grid points, the use of psolver = 'multigrid_noopt' and masking_method = .F. can lead to a blow up of wind velocity. Please see remarks below.
09/05/16	MS	In case of topography = .T., different processor topologies do not yield identical results.
18/04/12	MS	The use of the cache-optimized and the vector-optimized Palm versions do not yield identical results with topography. Reason: Different treatment of array rif_wall in cache and vector version in wall_fluxes. In vector version rif_wall is update after each component in prognostic equations, whereas in cache version old and updated values are mixed, leading to slightly different results in wall_fluxes_e.

Remarks with respect to usage of psolver = 'multigrid_noopt'

In case of very complex topography, including narrow street canyons or complex shapes resolved by only a few grid points, the use of psolver = 'multigrid_noopt' in combination with masking_method = .F. can lead to a blow up of wind velocity. So far, this never happened if masking_method = .T. or psolver = 'multigrid' (where topography is always masked). A possible workaround to prevent such velocity blow-up is to preprocess the topography by filling holes and removing complex shapes on the small scale before running the LES. (This approach should have no significant effect on the flow field, as in LES the flow in such regions is poorly resolved and do not yield to reliable physical information.) The reason for this velocity blow-up is unclear so far.

Remarks with respect to divergence reduction

If topography is prescribed and psolver = 'multigrid_noopt', the mean velocity-divergence reduction (see RUN_CONTROL) is smaller in case of masking_method = .T. compared to masking_method = .F. (about one-order of magnitude). This smaller reduction can be attributed to wall-bounded grid points, where the divergence after pressure correction remains significantly larger in case of masking, while the divergence reduction for non-wall-bounded grid points is similar in both cases. However, validation tests revealed that this smaller divergence reduction has no effect on the resulting flow.