Opened 9 months ago

Closed 9 months ago

#636 closed defect (fixed)

crash in FG2 with lcchange

Reported by: jgipsl Owned by: luyssaert
Priority: major Milestone: IPSLCM6.v2
Component: Anthropogenic processes Version:
Keywords: Cc:

Description

Revision 6358, CN-CAN, crashes with setup OOL_SEC_STO_FG2 after 13 years of global simulation. It crashes with following message:

15 At the pixel          17  for element           1  pool_end is NaN
15  pool_end:    6253.98715759889       0.000000000000000E+000
15  0.000000000000000E+000   12.5167455770299                          NaN
15   17.0080726969985       0.000000000000000E+000  0.000000000000000E+000
15  0.000000000000000E+000   561.209171031187       0.000000000000000E+000
15   94.8813663420977        17.4155028846522       0.000000000000000E+000
15  0.000000000000000E+000
15
15FATAL ERROR FROM ROUTINE sapiens_lccchange
15 --> NaN problem

Pixel 17 at proc 15 corresponds to lat/lon: 41.0000000000 95.0000000000

If instead using MIN_VEGFRAC=0.001 instead of default value 1.00E-07, the model runs longer but it still crashes after 33 years. The error message is the same but for another pixel:

5 At the pixel         132  for element           1  pool_end is NaN
5  pool_end:    39.7716653405912       0.000000000000000E+000
5  0.000000000000000E+000  0.000000000000000E+000   570.519822412721
5  0.000000000000000E+000                     NaN   1160.37141536533
5   109.831348305882       0.000000000000000E+000  0.000000000000000E+000
5  0.000000000000000E+000  0.000000000000000E+000  0.000000000000000E+000
5   1797.88054918439
5
5FATAL ERROR FROM ROUTINE sapiens_lccchange
5 --> NaN problem

Pixel 132 at proc 5 corresponds to lat/lon: 65.0000000000, 43.0000000000

Change History (6)

comment:1 Changed 9 months ago by luyssaert

A global FG2 run with LCC ran for 110 years without a crash (r6332). This could be used as a starting point for the debugging.

Last edited 9 months ago by luyssaert (previous) (diff)

comment:2 Changed 9 months ago by luyssaert

The crash could be reproduced (r6358) after 90 years (1991) with FG2 for a smaller domain on irene
LIMIT_WEST=80
LIMIT_NORTH=60
LIMIT_SOUTH=30
LIMIT_EAST=110

comment:3 Changed 9 months ago by luyssaert

  • Owner changed from somebody to luyssaert
  • Status changed from new to assigned

comment:4 Changed 9 months ago by luyssaert

r6390 contains a solution for the underlying problem of this crash: in sapiens_kill and IF-statement tested for wehther circ_class_biomass was empty but circ_class_biomass should be changed in sapiens_kill (unless circ_class_n became zero). In that if-statement circ_class_biomass has been replaced by circ_class_n.

comment:5 Changed 9 months ago by jgipsl

Revision 6398 has been run successfully in global using the default set up for OOL_SEC_STO_FG2, starting without restart file. The full period of 110 years has been running with success.

comment:6 Changed 9 months ago by luyssaert

  • Resolution set to fixed
  • Status changed from assigned to closed
Note: See TracTickets for help on using tickets.