Re: Problems with the gradient function

From: Dennis Shea <shea_at_nyahnyahspammersnyahnyah>
Date: Tue Dec 01 2009 - 15:10:13 MST


I speculate that you are using ?CAM? data that has been interpolated to
pressure levels.
If NCL's "vinth2p" or "vinth2p_ecmwf" has been used to perform the
interpolation, then the
argument "kxtrp" was likely set to False.

Logical. False => no extrapolation when the pressure level is outside of
the range of psfc.
As noted in the documentation, the gradient function, which uses
spherical harmonics,
requires: (1) the data are global and (2) no missing values be present.

[1] Regenerate the pressure level data with kxtrp=True
[2] Use

*poisson_grid_fill*( x, True, 1, 1500, 1e-2, 0.6, 0)

on each variable prior to invoking the gradient function.

No matter what, always remember that these are extrapolated or
interpolated values.

Did you try the gradient function on the original hybrid level data and
*then* interpolate the budget quantities to pressure levels via vinth2p
or vinth2p_ecmwf??

Dood lock

Katie Holman wrote:
> Dear all,
> I am running into a problem with the gradient function. The gradient
> function is outputting missing values at pressure levels that are very
> close to the surface (1000mb, 925mb, and 850mb), which is negatively
> impacting a moisture budget I am trying to perform at one specific
> location. I know this is an artifact of the function because the
> specific humidity data I am using has data at each level except the
> surface for the specific lat/lon location. The problem is the missing
> values that are arising at levels above 1000mb. Is there a way to
> mitigate this somehow?
> I have tried to shrink the spatial domain of my data to remove high
> elevation areas. This has helped slightly. However, I am still running
> into missing values above 1000mb. The main goal is to obtain the
> gradient of specific humidity at one point only, so the size of the
> domain isn't a problem, assuming that there are enough data points to
> perform the differencing scheme. Is there a way to perform the
> gradient around missing values so as to not eliminate an entire level?
> Has anyone had this problem before or worked with the gradient function?
> Thank you,
> Katie Holman
> _______________________________________________
> ncl-talk mailing list
> List instructions, subscriber options, unsubscribe:

Dennis J. Shea                  tel: 303-497-1361    |
P.O. Box 3000                   fax: 303-497-1333    |
Climate Analysis Section                             |
Climate & Global Dynamics Div.                       |
National Center for Atmospheric Research             |
Boulder, CO  80307                                   |
USA                        email: shea 'at' |
ncl-talk mailing list
List instructions, subscriber options, unsubscribe:
Received on Tue Dec 1 15:11:03 2009

This archive was generated by hypermail 2.1.8 : Thu Dec 03 2009 - 09:53:22 MST