version 21.10

# Changes since last svn-revision r4901

The following lists shortly describe the main differences between this release and the last frozen revision r4901 of the old svn-repository.

## Bugfixes:

- sky view factor restart MPI I/O: error message string was too long, initial setting of svf IO indices to zero was wrong (bugs were introduced with the last svn revision 4896 and 4898), total domain boundaries must not include the ghost layers, because SVF are not defined there, some variables extended to 64-bit INTEGERs, MPI error at end of run fixed (solves trac ticket #1269)
- ALLOCATE statement was missing for global_end_index (restart input of surface data)
- calculate correct maximum of plant canopy factors in case of calc_svf = .TRUE.
- avoid unrealistic RTM output of the incoming SW radiation fluxes due to angular discretization
- stop the model with error message when improper values for raytrace_discrete_azims and raytrace_discrete_elevs set, issue warning when few values of raytrace_discrete_azims and raytrace_discrete_elevs are set instead of the default values
- division-by-zero with mrt_nlevels = 4
- discretization of wall temperatures in the edges of layers modified. The approach shall have smaller discretization errors in such a situation. The change follows similar change made in LSM some time ago and it makes discretization approach in USM and LSM again consistent. Beside this, the update also improves pre-set parameters of the pavements in LSM and radiation modules and add a few minor bug fixes.
- at east/westward-facing walls the surface-parallel velocity was not computed correctly, the v-component needs to be taken instead of the u-component (see calc_uvw_abs_v_vgrid). The error propagated into the horizontal momentum fluxes and caused an asymmetry e.g. between north-south and east-west orientated canyons. The bug is in since the very first implementation of the topography.
- at land- and building surfaces fluxes still contained the density. For horizontal surfaces this does not matter since the density cancels out in diffusion_s. At vertical surfaces the heat input is overestimated by about 20%. Tests for a medium-sized setup for Berlin has shown that the impact on boundary-layer structure is not significant.
- some restart MPI-I/O errors regarding specific output quantities and 3d-arrays of type LOGICAL, MPI I/O bug with cyclic-fill when using topography in the main run
- wrong vertical velocity removal if topography is used
- output of perturbation pressure corrected (so far, it was normalized by the near surface density)
- wrong shift of the coordinates of the virtual measurement (VM) locations within the output, wrong velocity interpolation in the virtual flight module
- height levels for some profile output quantities revised
- aerosol emission fluxes in SALSA: timestamp was not correctly selected if the simulation does not start from t=0 (i.e., restart runs)
- lpm stochastic model: wrong calculation of the CFL-condition when the stochastic velocity components are added to the interpolated resolved velocity components, wrong calculation of stochastic term of the Weil equation
- calculation of subsidence velocity at the first computational level (k=1)
- several minor bugfixes

## Changes:

- transpose-compute overlap feature of the FFT-pressure-solver removed from the code (initialization parameter transpose_compute_overlap removed)
- possible options and default settings for initialization parameters flux_input_mode and flux_output_mode changed, default now is 'application-specific', the old default 'approximation-specific' does not exist any more

## New features:

- pressure drag on vertical surfaces added as new 2d-output quantity (data_output = 'pressure_drag_x*', 'pressure_drag_y*'))
- true air temperature at 2m added as new 2d-output quantity (data_output = 'ta_2m*')
- in case of using the FFT-solver, the total number of gridpoints along x or y ( nx+1, ny+1 ) must not be multiples of npex / npey any more. In such cases, subdomains are not uniform in size any more (some of them have one grid point less than others).
- Lagrangian particle model: simple interpolation method extended to subgrid-scale velocites
- enable virtual flights that start with a descent
- FFT-solver accepts non-cyclic boundary conditions along x/y. ATTENTION: this is an experimental feature an has not been rigorously tested so far.
- direct radiative flux is taken into account for photolysis
- diagnosis and output of cold-air drainage flow quantities