Discussion
A feature around 7 times larger than the grid spacing is the smallest feature
the model can simulate moderately well with a somewhat realistic-appearing
structure. For a model with a grid spacing of 50 km, this is a feature with
a crest-to-crest wavelength of 7x50= 350km. For a model with 20km grid spacing,
the smallest realistic-looking feature has a wavelength of 140km. Remember,
a coarse-resolution model tries to simulate things at an unrealistically large
scale, including MCCs, grid-point precipitation bulls-eyes, sea breezes, fronts,
etc. Topography and coastlines can force a model to create features at scales
down to around 4 times the grid spacing, but those features will be crude.
If the model makes a spurious feature toward the small end of its simulation
capability, obviously that feature will be smaller in a finer resolution
model. Therefore it will impact a smaller area at first. Additionally, the
smaller disturbance is dynamically smaller - it is smaller compared to the
Rossby radius. Therefore, it is more likely to dissipate faster. Thus,
the answer to the question is: spurious features occurring in higher-resolution
models tend to have less adverse impact on the large-scale forecast, and
even the mesoscale impact of features their size becomes shorter-lived as
model resolution improves.
