[Wrf-users] Accumulating artefact at lateral boundary in long-term simulations

[Sebastian Knist]

Dear all,
  
in several long-term transient regional climate simulations with WRF v3.3.1, 
v3.5.1 and v3.6.0 (those we tested) using totally different setups, machines, 
forcing data, compilers we see accumulating unphysical artefacts along the 
lateral boundaries after some time (~100000 time steps, a few months for a 
60-seconds time step).
  
See these two figures as an example:

https://www.dropbox.com/l/8LRXt2pQW0hob6p4tn2bxr
  
The artefact is most pronounced at the outermost grid points (specified zone), 
but it is also seen in the relaxation zone and it is advected into the 
interior of the model domain. So there are stripe-like features (artefacts) in 
the interior of the domain that ultimately results from the artefacts along 
the boundaries. It is seen in the uppermost model layers first, later on in 
the simulation more and more also in lower levels. Several variables (P, T, U, 
V, W, QV) are affected whereby the artefact is most clearly seen in pressure 
perturbation and wind velocities.

The artefact only occurs in the parent model domain, or in case of a one-way 
nesting with ndown. wrfbdy files look fine. Since it accumulates slowly but 
continuously, it more and more affects also the inner model domain and may 
even lead to CFL related abort of WRF.
  
We wonder if some of you have seen this artefact in long transient runs, too? 
And you might already found a way to solve this problem?

Many thanks in advance for comments and suggestions.

Best regards
Sebastian

---
Sebastian Knist
Juelich Supercomputing Centre, Juelich Research Centre
Meteorological Institute, University of Bonn
Germany
Email: sknist at uni-bonn.de