The application of cellular automata to weather radar
(D Panton with A Fitt and Y Wang)
Numerical modelling of geographically large scale weather systems has been
very successful in extending weather forecasts to periods of time of up to 5
days or more.
The ill-posed nature of the governing Navier-Stokes equations mitigates against reliable forecasts for longer periods than this. For geographically small scale systems, such as localised thunderstorms, whose lifetime is on a much shorter timescale however, numerical modelling is not so effective since it appears that the dynamics of such systems are a great deal more chaotic.
For such localised phenomena, weather forecasters rely heavily on radar observations to monitor the storm development. To do this, weather radar is normally used to measure the spatial distribution of raindrops suspended in air by detecting the returned signals of the transmitted electromagnetic waves. The returned signals are called weather echoes.
In this work a possible cellular automaton approach to weather (and in particular rainfall) modelling is considered. After posing a paradigm problem in a manner reminiscent of a numerical PDE solver and showing that the general approach appears to be valid, some more detailed modelling is done and it is commented on how this could be used to construct a genuine finite-state cellular automaton.