forrtl: error (65): floating invalid

[dashline]

I run the 60km_3km variable mesh with MPAS tutorial initial data(GFS:2014-09-10_00) from https://www2.mmm.ucar.edu/projects/mpas/tutorial/Boulder2019/index.html. I can't get the log.atmosphere.0000.err.

The Script begin at 2021-11-01 14:28:44
forrtl: error (65): floating invalid
Image              PC                Routine            Line        Source
atmosphere_model   00000000029AAEEE  Unknown               Unknown  Unknown
libpthread-2.12.s  00000039C9E0F500  Unknown               Unknown  Unknown
atmosphere_model   0000000002A700CF  Unknown               Unknown  Unknown
atmosphere_model   0000000000D7C50B  atm_time_integrat        5975  mpas_atm_time_integration.F
atmosphere_model   0000000000BB315D  atm_core_mp_atm_m         390  mpas_atm_core.F
atmosphere_model   0000000000BADC9A  atm_core_mp_atm_c         163  mpas_atm_core.F
atmosphere_model   000000000041B041  mpas_subdriver_mp         331  mpas_subdriver.F
atmosphere_model   00000000004164E5  MAIN__                     14  mpas.F
atmosphere_model   000000000041646E  Unknown               Unknown  Unknown
libc.so.6          0000003BD661ECDD  __libc_start_main     Unknown  Unknown
atmosphere_model   00000000004163A9  Unknown               Unknown  Unknown

I see the forum and think maybe it's the config_dt problem. initally, the config_dt is 18. And I try 9. The error still happen.

 

[mgduda]

From your stack trace it looks like the code is stopping in the atm_init_coupled_diagnostics routine, which is called when the model starts up. In the v7.0 release, line 5975 in mpas_atm_time_integration.F is where the exner function is computed from d(zeta)/dz, rtheta_base, and rtheta_p, with the latter two variables being computed earlier from rho_base and theta_base.

Can you check the base-state potential temperature (theta_base) and base-state dry air density (rho_base) fields in your initial conditions file to see that they look reasonable?

 

[dashline]

From your stack trace it looks like the code is stopping in the atm_init_coupled_diagnostics routine, which is called when the model starts up. In the v7.0 release, line 5975 in mpas_atm_time_integration.F is where the exner function is computed from d(zeta)/dz, rtheta_base, and rtheta_p, with the latter two variables being computed earlier from rho_base and theta_base.

Can you check the base-state potential temperature (theta_base) and base-state dry air density (rho_base) fields in your initial conditions file to see that they look reasonable?

From the exner function, I notice the formula rtheta_base(k,iCell)))**rcv. Is it the error floating invalid comes from underflow or overflow? I just use ncl to check the minimum and maximum of the variables theta_base and rho_base from my SouthChina.init.nc.
(0) Base state potential temperature (K) : min=0 max=792.208
(0) Base state dry air density (kg m^{-3}) : min=0 max=3.56364
According to the formula of caculating rtheta_base, it doesn't seem to overflow or underflow?

 

[mgduda]

Based on the range of rho and theta values, it's not immediately clear to me why a floating-point exception is being generated on line 5975 of mpas_atm_time_integration.F. But note that there are other sorts of exceptions besides overflow and underflow -- for example, raising a negative number to a fractional power.

In any case, potential temperatures of zero (in K) aren't reasonable, nor are dry air densities of 3.56 kg/m^3. As I mentioned to you in this thread, I strongly suspect there are problems in your initial conditions, and examining the fields in your initial conditions file would be a good idea. It may also be worth trying to start over and create new ICs step-by-step.

 

[dashline]

Based on the range of rho and theta values, it's not immediately clear to me why a floating-point exception is being generated on line 5975 of mpas_atm_time_integration.F. But note that there are other sorts of exceptions besides overflow and underflow -- for example, raising a negative number to a fractional power.

In any case, potential temperatures of zero (in K) aren't reasonable, nor are dry air densities of 3.56 kg/m^3. As I mentioned to you in this thread, I strongly suspect there are problems in your initial conditions, and examining the fields in your initial conditions file would be a good idea. It may also be worth trying to start over and create new ICs step-by-step.

Thank you for your explanation. In the thread that you mention, the ICs is from ERA5. The data in this thread is tutorial GFS. You mean to check my original data or the intermediate data from ungrib? Is it possible that the problem comes from the static.nc that I create?