WRF stop without any error

[caoyer]

Without any mistake being visible in rsl.out or rsl.error, my WRF stopped running.
The WRF stopped roughly in the third month of my 4-month experiment.
I discovered 2 solutions to this issue: first, minimize the domain area, for instance, when my larger domain encompassed all of India, WRF would terminate without a problem, but when my smaller domain only covered South India, everything would work as it should; and second, change the land surface model. WRF stops while using Noah-MP to simulate, however other LSMs don't run into any problems, whether big or small domain.
Is Noah-MP at fault here?

 

[Ming Chen]

Did you check all the rsl files when the case failed running to the end? Note that the error message may not necessarily be in rsl.error.0000.
Also, did you check whether you have enough storage space in your computer? I ask this because, if the case can run successfully with reduced domain size, it may indicate that there is sufficient space for a larger domain run.
NoahMP is more complicated than Noah, but if physics is wrong in NoahMP, you should receive errors....

 

[caoyer]

Did you check all the rsl files when the case failed running to the end? Note that the error message may not necessarily be in rsl.error.0000.
Also, did you check whether you have enough storage space in your computer? I ask this because, if the case can run successfully with reduced domain size, it may indicate that there is sufficient space for a larger domain run.
NoahMP is more complicated than Noah, but if physics is wrong in NoahMP, you should receive errors....

I appreciate your response. First, I'm certain I have enough storage, and second, I've reviewed all the rsl* files and found no problem messages—just regular output with an sudden stop.
In addition, the very huge outliers of the LAI (GPP, NPP) data in the first timestep of the wrfout file also puzzle me. There are locations with LAI data of -1e-36 after the first timestep. Only Noah-MP experiences these circumstances; in Noah, LAI data there are within typical bounds. wrfinput file did not have this circumstance, whose LAI was processed by real.exe from the met_em* files. So i think there are problems in Noah-MP.