Problem creating MPAS@3km input with ECMWF data

[thomasuh]

Hello,

I try to initialize a 3km mesh using ECMWF data on model levels. However, I get strange values for Theta, precipw, and rho in the input file and thus a crash of the atmosphere model:

For rho at the first layer in the input file:
Dry air density (kg m^{-3}) : min=-477.252 max=393.065
For Theta at first layer in the input file:
Potential temperature (K) : min=-147.332 max=342.869
for precipw in the input file:
precipitable water (kg m^{-2}) : min=-541.822 max=179.873

When checking the variables at the end of "init_atm_case_gfs" around line 4920, there are no theta values below 200 K. Am I looking at the right place in the code where "theta" values are actually written to the input file?

Any ideas what may go wrong here?

 

[mgduda]

Checking either theta_m ("t" in the code) or theta at the end of "init_atm_case_gfs" should be correct. There may be two tests worth trying:
(1) Using the same ECMWF intermediate file to initialize a simulation on a coarser mesh; and
(2) using data other than ECMWF model-level data to create initial conditions on the 3-km mesh.

For an easy alternative to the ECMWF data that you have, you could download the CFSR intermediate file valid 2010-10-23 0000 UTC from http://mpas-dev.github.io/atmosphere/real_data.html.

The outcome of these tests may help us to determine where to focus effort -- either on the use of model-level ECMWF data or on the 3-km static file.

 

[thomasuh]

Hello Michael,

In the meantime I found the reason for this strange behaviour. I ran init_atmopshere with single precision and obviously this caused the error. As soon as I switch to double precision during compilation, the error is gone (although writing out only with single precision).
Not sure why this happens...

Thomas

 

[mgduda]

That is interesting and good to note -- that producing the 3-km initial conditions with a double-precision build of the init_atmosphere core seems necessary. I can't think of a specific reason why this should be the case, but it's good that there is a solution in any case.