Pressure Gradient Errors in CAM-SE

Abstract

In models like CAM which use terrain following coordinates, there can
be large numerical artifacts in regions of rough topography related to
how one numerically computes horizontal gradients. The "pressure
gradient" term, in terrain following coordinates decomposes into two
large terms which ideally would cancel, but instead when small
relative errors in each term fail to exactly cancel, the result is
large errors creating spurious sources of divergence leading to
anomalously large vertical velocities over steep terrain.

It is conjectured that these anomalous velocities are the main contribution
to the precipitation biases seen in CAM over steep terrain.
Hence one of our tasks in ACME is to improve the numerics used to
compute the horizontal pressure gradient.
This subject has been extensively studied and there are several techniques
in the literature which will be tested in ACME. Before embarking on
this work, we will first establish in ACME how much of the precipitation
biases seen over steep terrain is related to anomalous vertical velocity,
as opposed to other effects such as the physical parameterizations
triggered by surface roughness. This poster will our experimental
setup and preliminary results looking at the relationship between
precipitation and vertical velocity as a function of topography and
surface roughness.