Uninitialized variables in NGEN/Snow-17

[drakest123]

The values of precip_scf(nh) and forcing%pa are initialized as huge(1.0) but never overwritten in the NGEN/Snow-17 executable.

Current behavior

When compiling Snow-17 with NGEN, the precip_scf output value in the cat-*.csv file is "inf".

Expected behavior

The precip_scf output value should not be "inf"

Nature of the problem

The value of precip_scf(nh) is initialized as huge(1.0) in forcingType.f90::initForcing() but never overwritten:

this%precip_scf(:)     = huge(1.0)

forcingType.f90::initForcing(): this%precip_scf(:):    3.40282347E+38

When this value is multiplied by the catchment area, it becomes "inf".

In contrast, when snow-17 is run without NGEN, it calls read_areal_forcing(). This is where precip_scf and forcing_pa(nh) are initialized:

      forcing%precip_scf(nh) = forcing%precip(nh) * parameters%scf(nh)   ! scale input precip by snow correction factor
                                                                         ! (note: this is for output; model input
                                                                         ! precip is scaled in pack19()

      call sfc_pressure(parameters%elev(nh), forcing%pa(nh))             ! fill in the surface pressure field

Solution

Minimally, precip_scf(nh) must be initialized. For completeness, the surface pressure field should also be initialized although It is not clear that it is used.

[andywood]

Hi Steve, can you check how this relates to PR#48? I think I corrected this issue. -Andy

…

On Sat, Sep 7, 2024 at 1:20 PM StephenDrake-NOAA ***@***.***> wrote: The values of precip_scf(nh) and forcing%pa are initialized as huge(1.0) but never overwritten in the NGEN/Snow-17 executable. Current behavior When compiling Snow-17 with NGEN, the precip_scf output value in the cat-*.csv file is "inf". Expected behavior The precip_scf output value should not be "inf" Nature of the problem The value of precip_scf(nh) is initialized as huge(1.0) in forcingType.f90::initForcing() but never overwritten: this%precip_scf(:) = huge(1.0) forcingType.f90::initForcing(): this%precip_scf(:): 3.40282347E+38 When this value is multiplied by the catchment area, it becomes "inf". In contrast, when snow-17 is run without NGEN, it calls read_areal_forcing(). This is where precip_scf and forcing_pa(nh) are initialized: forcing%precip_scf(nh) = forcing%precip(nh) * parameters%scf(nh) ! scale input precip by snow correction factor ! (note: this is for output; model input ! precip is scaled in pack19() call sfc_pressure(parameters%elev(nh), forcing%pa(nh)) ! fill in the surface pressure field Solution Minimally, precip_scf(nh) must be initialized. For completeness, the surface pressure field should also be initialized although It is not clear that it is used. — Reply to this email directly, view it on GitHub <#50>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABIKARLFAQVM2QKNIT7AILTZVNGZLAVCNFSM6AAAAABN2IPKZCVHI2DSMVQWIX3LMV43ASLTON2WKOZSGUYTEMBTGA4TCMA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[drakest123]

Yes, that is the same problem. I tested your file changes and they worked. I will close this comment. BTW, I'm working on another issue with units and I'll post that in case you've encountered it. I've seen some scuttlebutt on this issue but I don't know if it has been resolved.