What is the best approach to simulate one year, for example, in WRF? I divided the simulations into 12 months, 52 weeks, or 5 days segments, as some papers have applied (but they did not justified this choice)? There is a best practice?
I'm not sure that a run for more two weeks is a best practice (see papers of Lorenz about predictability). On the other hand, the WRF usually provides best forecasts for 24 or 48 hours. So, I offer using the 365 runs for 1 day.
I agree with previous comment because without re-initialization wrf will diverge too much from the upper-level global meteorological fields which provide boundary conditions. The special 'spectral nudging' technique was developed specially to overcome this problem and to avoid re-initialization. This option is implemented in wrf, however I personally didn't use it yet. There are several papers which provide justification for re-initialization, such as this one Article Assessment of Three Dynamical Climate Downscaling Methods Us...
Thank you Valeriy Khokhlov and Ivan Kovalets for the answers. I am doing some testes for one month, simulating different run segments and using analysis nudging. My intention is to apply the WRF in an air quality model (CMAQ) for one year. The Meteorology-Chemistry Interface Processor (MCIP) for the CMAQ accepts up to 367 meteorology files (default). So, divide the WRF simulation into small periods may not be a problem to perform CMAQ.
I consider that it would be good, instead of 365 runs of 24 hours, to make the necessary runs, but for 96 hours. Here we are taking into account the spin-up. Certainly long runs go wrong. it would be necessary to run from day 1 to the day fourth, from the third to the day seventh, and so on, otherwise, each run will start at zero, and continuity would be lost.
- Badges
- Science topic
- Similar topics
- Papers