Consistent Bias in WRF Typhoon Track Simulations

[junius.Wang]

Dear WRF community,

I'm encountering an interesting phenomenon in my WRF simulations of a typhoon track and I'm hoping to get some insights from the community.

Setup:​

• WRF Version: [V4.6]
• Input Data: NCEP FNL (ds084.1)
• Simulation Period: [2021.07.21.12:00-2021.07.25.12:00]

Observations:​

I've run multiple simulations using different parameterization schemes, particularly varying the PBL & mu & cu & fdda schemes . I've uploaded images of the resulting typhoon tracks for each scheme, along with the JMA (Japan Meteorological Agency) observed track for comparison.

Key Findings:​

1. Consistent Deviation: All simulations, regardless of the PBL scheme used, show the typhoon stopping at a similar location in the final stages of the simulation.
2. Significant Bias: This common stopping point deviates significantly from the JMA observed track.

Possible Explanations:​

1. Model Limitations: There might be systematic limitations in WRF for this particular typhoon scenario.
2. Initial/Boundary Conditions: If all simulations use the same initial and boundary conditions, errors in these could lead to consistent biases.
3. Insufficient Physical Process Representation: Key physical processes affecting typhoon movement might not be adequately represented across different parameterization schemes.
4. Resolution Issues: The grid resolution might be insufficient to capture small-scale features crucial for accurate typhoon track prediction.
5. Large-scale Circulation Bias: There could be systematic biases in simulating large-scale circulation patterns that influence typhoon movement.
6. Observational Uncertainties: While less likely, uncertainties in the JMA observational data can't be completely ruled out.
7. Model Tuning: The model might be over-tuned to certain types of scenarios, leading to poor performance in this specific case.

Questions:​

1. Has anyone encountered similar consistent biases across different parameterization schemes?
2. Are there known issues with WRF in simulating typhoons in this particular region or under similar conditions?
3. What additional diagnostics would you recommend to further investigate this issue?
4. Are there specific modifications to the namelist.input that might help address this problem?

Attached Files:​

1. Images of simulated typhoon tracks for different PBL schemes(Due to the limitation of the number of uploaded files, only some images are available)）
2. A sample namelist.input file

Any insights, suggestions, or similar experiences would be greatly appreciated. Thank you for your time and expertise!

Junius

 

[William.Hatheway]

how are you making the simulated track?

that's something I've been trying to do in python for awhile now.

Sorry this is unrelated to your question.

 

[junius.Wang]

how are you making the simulated track?

that's something I've been trying to do in python for awhile now.

Sorry this is unrelated to your question.

This is not complicated. You can use wrf python to obtain the minimum slp for each time period and draw them out

 

[William.Hatheway]

This is not complicated. You can use wrf python to obtain the minimum slp for each time period and draw them out

could you provide a code example in wrf-python, to keep the issue away from this chat can you message me directly? @junius.Wang

 

[junius.Wang]

could you provide a code example in wrf-python, to keep the issue away from this chat can you message me directly? @junius.Wang

Sry for replying late. Sent already.

 

[junius.Wang]

I still need help.
I re simulated using data from gdas, and there were no major issues with the path. I need to clarify that only the initial data was replaced, and no other settings for the namelist were changed

 

[kwerner]

@junius.Wang
Do you mind attaching the domain plot, showing both domains? Thanks!

 

[junius.Wang]

Sorry to reply late!
Here is the domian plot by using NCL: