Fire behaviour model fails in debug mode

[DeniseWorthen]

Description

The FBH configuration fails in debug mode with unintialized variables

To Reproduce:

Compile and run the FB test in debug mode

### ATM-FBH test ###
COMPILE | atm_fbh | intel | -DAPP=ATMF -DCCPP_SUITES=FV3_HRRR -D32BIT=ON -DDEBUG=ON| - wcoss2 noaacloud acorn | fv3 |
RUN | cpld_regional_atm_fbh                             | - wcoss2 noaacloud acorn             | baseline |

It fails with

+ '[' NO = WHEN_RUNNING ']'
+ srun --label --distribution=block:block -n 31 ./fv3.exe
30: forrtl: error (182): floating invalid - possible uninitialized real/complex variable.
30: Image              PC                Routine            Line        Source
30: libc.so.6          0000146C87AE3D90  Unknown               Unknown  Unknown
30: fv3.exe            0000000009415E12  fire_behavior_nuo         797  fire_behavior_nuopc.F90

This change fixes at least this initial failure

diff --git a/state/state_mod.F90 b/state/state_mod.F90
index ce9fb9a..bee0e22 100644
--- a/state/state_mod.F90
+++ b/state/state_mod.F90
@@ -149,7 +149,7 @@
       allocate (this%fz0(ifms:ifme, jfms:jfme))
       allocate (this%fuel_time(ifms:ifme, jfms:jfme))
       allocate (this%fire_psfc(ifms:ifme, jfms:jfme))
-      allocate (this%fire_rain(ifms:ifme, jfms:jfme))
+      allocate (this%fire_rain(ifms:ifme, jfms:jfme), source=0.0)
       allocate (this%fire_t2(ifms:ifme, jfms:jfme))
       allocate (this%fire_q2(ifms:ifme, jfms:jfme))
       allocate (this%fire_rh_fire(ifms:ifme, jfms:jfme))

[mkavulich]

Thanks for this report @DeniseWorthen. I got a slightly different error message

30: forrtl: error (182): floating invalid - possible uninitialized real/complex variable.
30: Image              PC                Routine            Line        Source
30: libc.so.6          00001457C02ECD90  Unknown               Unknown  Unknown
30: fv3.exe            000000000945384F  fire_behavior_nuo         919  fire_behavior_nuopc.F90

but it points to the same problem: it seems that grid%fire_rain is used before being defined at the initial timestep.

 917     if (imp_rainrte) then
 918       ! convert m s-1 to m and accumulate
 919       grid%fire_rain(grid%ifps:grid%ifpe, grid%jfps:grid%jfpe) = grid%fire_rain(grid%ifps:grid%ifpe, grid%jfps:grid%jfpe) + &
 920           ( ptr_rainrte(clb(1):cub(1),clb(2):cub(2)) * ts )

I'll follow up with Pedro to see if this is the preferred fix or if grid%fire_rain should be initialized with zeros somewhere else.