Dear Sir,

Thank you for your suggestion and useful stuff. I would begin from these. I have no idea whether it can be integrated with the application of numerical models, such as MM5 and WRF, and radar which are widely used for weather forecast nowadays or not, sir.

Best regards,

Sompoke

I'm not sure exactly what you mean by micro-community scale, but at the best operational modeling scales in the U.S. (still of order ~1-3 km) there is enough uncertainty that a probabilistic method is best.  No single deterministic solution is likely to provide the answers you seek but you may be able to find a range of probabilities and solution methods to give some useful skill to warn the communities you seek, particularly if using data assimilation with radar and surface observations.  You might also be interested in this book or its reference list.  http://www.nap.edu/catalog.php?record_id=13328

Thank you for your comment, sir. Your suggested book and its references inside are useful. From the book, I think FAST3DCT simulation and urban meteorology are interesting. They are new and challenge for Bangkok, Thailand. Here, I only experience with the simulation of CAMx/MM5 in grid resolution of 2km for ozone air quality. According to your suggestion, a probabilistic method is interesting to apply, but the design of research methodology is really not easy. This topic is probably my future research topic. Thank you, sir.

In wrf there is "idealized case" to run for storms, maybe be suitable for the purpose :)