I've worked for a research center which tries to predict forest fire spread based on "current weather" and "terrain type". They have moved on from the project since, and went into building new weather prediction methods, using simulations and heuristics and whatnot (I don't have details).

Suffice it to say this is very high-level stuff, applying physics to large masses of air at different pressures and speeds (wind), taking into account all parameters like humidity and temperature, to an uneven land with trees and houses and whatever, so never is the starting data set large enough to model the current reality and predict the future reality so far in the future... It's extreme.

I've worked for a research center which tries to predict forest fire spread based on "current weather" and "terrain type". They have moved on from the project since, and went into building new weather prediction methods, using simulations and heuristics and whatnot (I don't have details).

Suffice it to say this is very high-level stuff, applying physics to large masses of air at different pressures and speeds (wind), taking into account all parameters like humidity and temperature, to an uneven land with trees and houses and whatever, so never is the starting data set large enough to model reality and predict the future reality... It's extreme.

I've worked for a research center which tries to predict forest fire spread based on "current weather" and "terrain type". They have moved on from the project since, and went into building new weather prediction methods, using simulations and heuristics and whatnot (I don't have details).

Suffice it to say this is very high-level stuff, applying physics to large masses of air and wind, taking into account all parameters like humidity and temperature, to an uneven land with trees and houses and whatever, so never is the starting data set large enough to predict reality... It's extreme.