T2 cold bias in WRF (other variables look reasonable)

[aba0327]

Hello,


I am running WRF using LDAPS data (KMA, 1.5 km) as the initial and boundary condition.
Most variables (U10/V10, pressure, winds, precipitation) look reasonable compared to observations, but near-surface temperature (T2) shows a strong cold bias.


My settings:

• Initial condition: LDAPS (1.5 km)
• Domains: 500 m resolution, 100 × 100 grid (very small domain)
• Start time: 03 KST, with 21 h spin-up, using the following 24 h for analysis
• Physics:
  • mp_physics = 8 (Thompson)
  • bl_pbl_physics = 2 (MYJ)
  • sf_surface_physics = 4 (Noah-MP)
  • ra_lw_physics = 4, ra_sw_physics = 4 (RRTMG)
  • cu_physics = 0
• sf_urban_physics = 0 (no UCM)
• num_land_cat = 21 (MODIS)

The issue:

• T2 is consistently colder than both LDAPS and observations (especially at night).
• Other variables do not show such a strong bias.

Questions:

1. What are the common causes when WRF produces a strong cold bias in T2 while other fields are okay?
2. Could the very small domain size (100×100, 500 m) or the start time (03 KST) be contributing to this?
3. Would you recommend specific PBL or land-surface scheme changes (e.g., MYNN, UCM) to reduce this bias?
4. Are there other diagnostics (HFX, LH, RNET, SWDOWN, LWDOWN, etc.) I should check to understand why T2 is too low?

Thank you in advance for your help!

 

[Ming Chen]

Would you please upload your entire namelist.input for me to take a look? Also, what is LDAPS?

Which version of WRF did you run? How long did you integrate?

100 x 100 grid numbe ris way too small. Can you enlarge the model domain? I won't expect a larger domain can greatly improve the model performance, but I expect that the cold bias can be less severe. Another option is to run with Noah LSM. Please let me know whether Noah performs better.

We got a few reports on cold bias issues related to NoahMP. I will talk to our expert and check whether he has any solutions.

 

[cenlinhe]

Do your simulation period and domain include snow? If so, you may want to increase the SCFFAC and/or MFSNO parameters in MPTABLE.TBL to reduce snow cover in Noah-MP. Activating sst_update and time-varying LAI & SST may also help. You could also test reducing CBIOM to a very small value (e.g., 0.0002) in MPTABLE.TBL.

 

[Ming Chen]

Please follow Cenlin's instruction and rerun your case. Let us know whether the above changes can give you better results. Thanks in advance.