I want to test some 3D weather radar software by simulating rainy weather. I want some degree of realism in the sense that it simulates rain systems and their boundaries and also simulates rain rates over their volumes. Weather systems should be dynamic in the sense that they can evolve and change at typical rates.

I would be satisfied with a statistical model if that's possible, rather than one driven by solving differential equations, based on my assumption that generating random numbers is easier than solving large systems of equations.

The model should be capable of running for long periods of time (a few hours of simulated real time) while generating an assorted set of weather patterns (clear weather and various levels of rain intensity). It should be suitable for real time simulation of the radar software, so weather development over time is an important requirement.

Being able to support real-time simulation is another reason I am leaning toward any more simplistic (by arguably reasonable) model.

I don't know where to begin finding a model like this. The models I have seen appear to be more complex than I would like, and probably not really suitable for my limited requirement.

What would help me is to identify any existing models that might be used for this purpose or any literature that describes how to create a model like this.


1 Answer 1


The problem with a statistical model is the reliance on dynamical model input, plus the requirement of climatology. A statistical model will also only give data for a solitary point, not an area. The simplest statistical model is persistence, which is the current weather, followed by climatology. Both are the lowest in forecast skill.

For a dynamic model, try using the WRF model (http://www2.mmm.ucar.edu/wrf/users/download/get_source.html).

It is a commonly used model. The runtime will range, depending on your computational power and needs. A larger domain will take longer, and require more. I know people who simulate hurricanes with it, and people who run fine scale modeling.

I can run WRF with a 200 km domain size for a 36 hour forecast in 1-2 hours.

Again it will really depend on your resources. Running WRF may require compilers you do not have access to. You can develop the visualization software after you ran the model, with whatever tools you find. Python and NCL are good languages, but MATLAB and a variety of other languages can be used.

  • $\begingroup$ I included persistence and climatology as statistical models. $\endgroup$ Mar 13, 2017 at 17:39
  • $\begingroup$ Thanks. I am looking this over now. It appears not to be compatible with Windows. Is that correct? (That may not be a problem, but I will need other resources, other than my desktop, if it is.) $\endgroup$
    – Jim
    Mar 13, 2017 at 18:14
  • $\begingroup$ Well, theoretically you can make it be compatible, but I have not been able to do it. The idea would be to use Cygwin (cygwin.com) to emulate Linux, and run it like that. It would likely be much slower than on a Linux server or computer, plus you would likely need an intel or PGI compiler. $\endgroup$ Mar 13, 2017 at 18:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.