Scheduled Downtime
On Friday 21 April 2023 @ 5pm MT, this website will be down for maintenance and expected to return online the morning of 24 April 2023 at the latest

Surface upwelling long wave radiation

This post was from a previous version of the WRF&MPAS-A Support Forum. New replies have been disabled and if you have follow up questions related to this post, then please start a new thread from the forum home page.

hongcheq

New member
Dear all,

Here I'm seeking some help with respect to surface upwelling longwave radiation in WRF v4.1.4. It was run with convection-permitting capabilities (4-km horizontal resolution, 65 vertical layers) during 2011 May-Aug, with the following configurations:
Microphysics: Morrison double-moment scheme.
Longwave radiation: RRTMG. Iacono et al. (2008, JGR)
Shortwave radiation: RRTMG. Iacono et al. (2008, JGR)
Surface Layer (sf_sfclay_physics): MYNN surface layer. Nakanishi and Niino PBL’s surface layer scheme.
Land-surface option: Noah-MP (multi-physics) Land Surface Model.
Planetary Boundary layer option: MYNN 3rd level TKE. Nakanishi and Niino (2006, BLM)

In the WRF simulations, supposedly, surface longwave upwelling radiation (variable LWUPB) should equal to EMISS*sigma*TSK^4, where variable EMISS is surface emissivity, variable TSK is the surface skin temperature, sigma is the Stefan-Boltzmann constant. However, when checked (see in figures), they do not equal to each other. The differences can go as high as 20 W/m2.

In the attached figures, first row shows LWUPB from WRF model outputs, the 2nd row shows the results from EMISS*sigma*TSK^4, the last two rows show differences between LWUPB and (EMISS*sigma*TSK^4) with single and double precisions. The value of sigma uses 5.67 * 10.0**(-8). Two figures show results at two different timestamps.

Has anyone seen similar symptoms in the WRF outputs? Are there ways to explain the differences between LWUPB in the model output and the re-constructed surface longwave upwelling radiation by using (EMISS*sigma*TSK^4)?

Any comments are greatly appreciated!

-----
Updates:
In module_sf_noahdrv.F file, found this line of code:
rl_up_rural=-emiss_rural(i,j)*sigma_sb*(tsk_rural(i,j)**4.)-(1.-emiss_rural(i,j))*glw(i,j)

The negative sign seems to indicate directions and the surface longwave upward radiation seems equal to emissivity * sigma * TSK^4 + (1-EMISS)*GLW, where GLW is downward longwave at ground surface. Tried this new method and found the differences between LWUPB and [emissivity * sigma * TSK^4 + (1-EMISS)*GLW] to be smaller than the previous method. However, the differences are still not zero (see newly attached figures). Scratching my head to figure out why...


Kind regards,
HC
 

Attachments

  • sfc_LW_1.png
    sfc_LW_1.png
    270.7 KB · Views: 649
  • sfc_LW_2.png
    sfc_LW_2.png
    234 KB · Views: 649
  • 03_EMISS_LWUPB_check14.png
    03_EMISS_LWUPB_check14.png
    174.6 KB · Views: 644
  • 03_EMISS_LWUPB_check23.png
    03_EMISS_LWUPB_check23.png
    175.1 KB · Views: 646
Hi,
Would you please do the same calculation using TSK at the time steps right before and after the model output time, then compare with LWUPB at the model output time? Please let me know whether you still see large difference.
 
Dear Ming,

Thanks for your reply.

I realized previous attached plots may involve a bit regridding errors. To get rid of this type of error, I test the variables on raw WRF curvilinear grid at particular times. The output frequency is hourly. Let TSK_before be at 20110501 01:00 UTC, TSK_after be at 20110501 03:00 UTC. All other variables be at 20110502 02:00 UTC.

error1 = LWUPB - EMISS * sigma_sb * TSK^4,
statistics of error1:
TSK_0501_0200.png

error2 = LWUPB - EMISS * sigma_sb * TSK_before^4,
statistics of error2:
TSK_Before.png

error3 = LWUPB - EMISS * sigma_sb * TSK_after^4,
statistics of error3:
TSK_after.png

Error values are not zero. I think radiation timestep is finer than 1 hour (15-min?) and the test above might not be ideal though. I'll try to do a similar test on radiation timestep.


Kind regards,
HC
 
An update:

Testing variables on raw WRF curvilinear grid at particular times. The output frequency is at radiation time step (15-min). Let TSK_before be at 20110621 15:00 UTC, TSK_after be at 20110621 45:00 UTC. All other variables be at 20110621 00:30 UTC.

error1 = LWUPB - EMISS * sigma_sb * TSK^4,
statistics of error1:
LWUPB_diff_TSK.png

error2 = LWUPB - EMISS * sigma_sb * TSK_before^4,
statistics of error2:
LWUPB_diff_2_TSK_before.png

error3 = LWUPB - EMISS * sigma_sb * TSK_after^4,
statistics of error3:
LWUPB_diff_3_TSK_after.png

It suggests that on radiation time-step, no matter TSK being one time-step before or after, LWUPB and (EMISS * sigma_sb * TSK^4) are not exactly equal to each other. But why is it the case? Shouldn't their values be the same (and thus the error1,2,3 be zeros everywhere)?
 
Top