precipitation partitioning in CLM offline

[mizukami]

I am running off-line CLM with my own forcing data, and trying to check my input/output to make sure I am running correctly. I am checking precipitation partitioning (snow and rainfall).

Based on the CLM technical documents, precipitation is separated into snow and rain at each CLM dtime using

Psnow = P*(1-f)
Prain = P*f
where f = 0< 0.5*(Tatm-Tfreeze)

 

[slevis]

Dear Naoki,

I wonder whether you used the right temperature variable in your scatter plot. I think that Tatm corresponds to TBOT in the clm history file. If this does not explain the discrepancy, then consider using write statements in the code to investigate this behavior. Pls let us know what you find out.

Thanks,
Sam Levis

 

[mizukami]

Hi Sam,

Thank for the reply. I found out that I was looking at hourly average outputs (TBOT, SNOWATM and RAINATM) from 15min simulation. My data is hourly and the coupler takes hourly data and interpolate all the forcing at every 15min I believe. That was why I got some scatter around the line. I output every 15min and made the same plot, and I got all the points on 0.5 slope straight line.

I got further questions...

-I think I would like to run the model every hour (not 15min) to match up with forcing. I tried it but the simulation stopped after it ran for the first few years ( I was running 20years over 9x 15 grid with 1/16th degree resolution in Colorado). Is it some numerical stability issue (time step is too large)?

-Since I am running CLM over a particular river basin in Colorado, I would like to put basin mask over the grid if I can to save computation time (so running the model only inside the basin). I have not quite figured out this, but I am thinking of replacing land fraction grid in the domain grid (i.e., outside the basin is sea and inside is the land). I am wondering if there is better or appropriate way.

thanks,
Naoki

 

[slevis]

- Could be numerical instability. Not sure.

- This makes sense. I hope it works.

Sam Levis