Differences caused by different running times

[jc sun]

Hi all,
Recently, I used WRFV4.2 simulating Typhoon Lekima (2019).I have found that the timing of the namelist setting makes a huge difference in the results. As shown in the figure, the simulation start time of typhoon is 2019-08-01, and the simulation start time of typhoon 2 is 2019-08-07. The simulation duration of typhoon 1 is 13 days, and the simulation duration of typhoon 2 is 6 days. The parts where the time overlaps are shown in the figure, and the difference is huge. The simulation of Typhoon 2 is close to the actual situation. What is the reason for this? If the simulations start and end at different times, would the partial results of the time crossing of 2 simulations differ so drastically?How to improve to solve this problem?

 

[Ming Chen]

This question involves 2 issues:
(1) The initial time of WRF run is important for short-term weather forecast. e.g., 2-5-day forecasts, while lateral forcing is more imortant for longer term simulation. For your cases, I suppose the initial condition of case 2 is more consistent with observations than that of case 1, which explains why case 2 gave better results,
(2) SST is fixed in WRF salutation unless sst_update is turned on. This option is important for longer simulations. Your case 1 covered 13 days and I would suggest to turn on this option.

 

[jc sun]

Hello, chen
Thanks a lot for your answer.
(1) Regarding your first reply, what I want to ask is what determines and changes the condition of lateral forcing, the parameters in the namelist?
(2) The second is that I tried the simulation from August 5th, and the results were almost the same as those of typhoon 1, even 2 days before the typhoon appeared, there was no improvement. Is your suggestion to start the simulation at the moment when the typhoon starts to appear? If nesting is carried out, the large area is simulated a few days before the start, and the small area is simulated when the typhoon appears, will the results be different? For the sst_upstate you proposed, I have used the coupling of roms and wrf, and the result has not changed significantly. I will still open if needed.
(3) Another question is if there are 2-3 typhoons in a short period of time in this area, then your suggestion is to conduct a separate simulation for each typhoon, or conduct a complete simulation within a specific time? Because I'm thinking about coupling it with roms to see how the ocean responds to this phenomenon. Looking forward to your reply!

 

[Ming Chen]

Please see my answers below:
(1) Lateral forcing condition is derived from large-scale forcing data, i..e, the data you ungrib. Nothing in namlelist can change the lateral forcing condition.
(2) The model results are dependent on many issues, including the initial condition. However, this doesn't mean that different initial time will definitely lead to significantly different results. For short-term simulation, if the initial and boundary forcing are reasonable and if they are from the same large-scale forcing data, then model results could be quite similar even if the initiation time is different.
(3) if you run the case with a coupled ROM, then sst should be predicted by ROM. In this case you don't need to turn on sst_update.
(4) I am not sure whether the 2-3 typhoons appeared simultaneously? if so, it could be a better idea to run a single simulation for all the typhoons because they might intercact with east other. Otherwise, it could be a better idea to study them separately. This is my personal opinion. Please determine what to do based on your specific cases.

 

[jc sun]

Thanks a lot for your answer. For the consecutive typhoons that appear in the region, your suggestion is also a complete simulation from the appearance of the first typhoon to the disappearance of the last typhoon?