Floating point exception on output runoff & base flow within HyMAP2

[jnattala]

Hello,

This is regarding an overflow error on output runoff & base flow within HyMAP2.

Undefined values passed to HYMAP2_model_core seem to rely on a type conversion to real, of integer components 'imis', of HYMAP2_routing_struc array variables of derived type HYMAP2_routing_dec (i.e. actual arg. real(HYMAP2_routing_struc(:)%imis) within call to HYMAP2_model, from HYMAP2_routing_run).

For my configuration, HYMAP2_routing_struc(:)%imis = -9999 & HYMAP2_routing_struc(n)%evapflag is zero. This causes a overflow on output runoff & baseflow (runoff1 & basflw1), when dwiflag is zero, for a river sequence.

Below is a portion of a slightly modified version of lislog.0000 output, from a call to HYMAP2_grid2vector, from HYMAP2_routing_run (for imported surface runoff & baseflow), for some portions of LIS cycle times. The last six columns within modified log output from HYMAP2_grid2vector, show real(LIS_rc%udef), HYMAP2_routing_struc(n)%evapflag, ix, iy & nz. These are followed by corresponding modified log output, from HYMAP2_model_core.

[INFO] LIS cycle time: 07/30/2023 23:45:00
[JN] HYMAP2_grid2vector LIS cycle time: 07/30/2023 23:45:00 306321 surface_runoff(i,:): 0.00000E+00 -9999.00 0 729 799 1
[JN] HYMAP2_grid2vector LIS cycle time: 07/30/2023 23:45:00 306321 baseflow(i,:) : 0.17540E-09 -9999.00 0 729 799 1
[JN] HYMAP2_model_core LIS cycle time: 07/30/2023 23:45:00 306321 basflw0(ic): 0.17540E-09 runoff0(ic): 0.00000E+00 grarea(ic): 208000000.00000

[INFO] LIS cycle time: 07/31/2023 23:30:00
[JN] HYMAP2_grid2vector LIS cycle time: 07/31/2023 23:30:00 306321 surface_runoff(i,:): 0.99692E+37 -9999.00 0 729 799 1
[JN] HYMAP2_grid2vector LIS cycle time: 07/31/2023 23:30:00 306321 baseflow(i,:) : 0.99692E+37 -9999.00 0 729 799 1
[JN] HYMAP2_model_core LIS cycle time: 07/31/2023 23:30:00 306321 basflw0(ic): 0.99692E+37 runoff0(ic): 0.99692E+37 grarea(ic): 208000000.00000

When fill values on input runoff (runoff0) & input baseflow (basflw0) are not identical to real(LIS_rc%udef), as seen above for LIS cycle time 07/31/2023 23:30:00, there is a FPE on output runoff (runoff1) & baseflow (basflw1) (copied below).

Unsure if this is a known issue, if not, I have implemented a minor fix based on an absolute difference magnitude comparison, and can furnish it if needed. Please let me know if it would be useful.

thanks
Jyothi

FPE:

forrtl: error (75): floating point exception
LIS_I19_SL12 0000000000499BA1 hymap2_model_core 140 HYMAP2_model_core.F90
LIS_I19_SL12 0000000000498263 hymap2_model_ 200 HYMAP2_model.F90
LIS_I19_SL12 00000000004BD466 hymap2_routing_ru 676 HYMAP2_routing_run.F90
LIS_I19_SL12 000000000066083F lis_routingmod_mp 131 LIS_routingMod.F90
LIS_I19_SL12 0000000000F911D3 retrospective_run 206 retrospective_runMod.F90
LIS_I19_SL12 0000000000C01044 MAIN__ 68 lisdrv.F90

[jvgeiger]

Thanks for this issue.

@agetiran, would you please review this?

[agetiran]

I had a meeting with @jnattala earlier this week to talk about this issue. It looks like the floating point exception is occurring due to questionable values (i.e., 0.99692E+37) found in Qs and Qsb fields used as inputs in HyMAP. The problem seems to be related to either Noah-MP or LIS output file writing, not HyMAP.

[jvgeiger]

Thanks. I will close this discussion.