Random-number seed in CESM-LE

[ryoung2]

The multiple simulations reported in a paper describing the CESM Large Ensemble Project [J. E. Kay et al., BAMS 96, 1333 (2015)] used a Monte-Carlo treatment of cloud cover. Was the same random-number seed used for each member of the ensemble? Or, did each member run through a different sequence of random numbers?

 

[hannay]

I assume you are referring to McICA code. I believe the seed depends on the column state, so it would be different between every ensemble.
I will ask around if I can get a better answer.

 

[ryoung2]

I assume that McICA means Monte-Carlo Independent Column Approximation, and that is the proper context. The seed would be set at the beginning of the simulation, and the input files probably allowed the user to set it for purposes of debugging (so that the computation would proceed through the same sequence of (pseudo)random numbers if the programmer needed to find a bug).

Thanks for taking up the cudgel!

 

[erik]

@nusbaume and @hannay since this is pretty specific to CAM I'm moving it to that forum.

 

[hannay]

I tried to ask around for more replies but I haven't found something more specify.
It looks to me the seed is defined in:
components/cam/src/physics/rrtmg/aer_src/mcica_subcol_gen_sw.f90
components/cam/src/physics/rrtmg/aer_src/mcica_subcol_gen_lw.f90
by the subroutine: kiss_seed

 

[ryoung2]

If kiss_seed is like a version that I found online, it can generate a random number seed by itself or accept a user-specified seed. Is it possible to find and examine the files that were used by whoever initiated the simulations? Or, is there an internal Technical Note that documents the simulations in Kay et al. at that level of detail?

If the answer to both questions is "No" or "It's not easy", it's time for me to give up and thank you for trying!

I wish you a pleasant holiday and a safe and satisfying new year.

Ralph