Add Wind-Driven Forest Fire example

[liyuezha]

Summary

This PR adds a new model example: Wind-Driven Forest Fire.

Compared to forest fire model which focuses on critical thresholds, this model introduces anisotropy caused by wind. It demonstrates how to implement directional bias in grid-based models and explicitly measures the Rate of Spread (ROS) for both the fire head (downwind) and the flank (crosswind).

Motive

Modeling Anisotropy
The model implements a wind-biased ignition probability formula based on the angle between the wind vector and the neighbor direction:
$$p = p_{\text{spread}} \cdot (1 + \text{strength} \cdot \cos(\theta))$$

Implementation

Instead of just counting "burned trees," this model calculates spatial metrics:

• Head Distance: The furthest distance the fire has traveled downwind.
• Flank Width: The lateral width of the fire scar perpendicular to the wind.
• ROS (Rate of Spread): Real-time calculation of the fire front's speed.

Usage Examples

Additional Notes

[EwoutH]

Cool concept, thanks!

Instead of a complete new model, could we also add a wind argument(s) to the existing model? That can be optionally enabled?

[EwoutH]

Instead of a complete new model, could we also add a wind argument(s) to the existing model? That can be optionally enabled?

@liyuezha what do you think of this?

[liyuezha]

Hi, @EwoutH

Thanks for the suggestion! I agree these features could be integrated.

For context, the original forest_fire model:

• Randomly fills cells with trees (probability = density)
• Sets all trees in the left column (x=0) on fire
• Uses uniform propagation probability

The main differences in my wind demo are:

1. Center Ignition: Fire starts from a 3×3 block in the center (instead of the left edge), which better illustrates directional wind effects
2. Wind-weighted spread: Full grid of trees with fire spread governed by wind-influenced probabilities (rather than uniform probability)

I think these could be integrated as optional modes while keeping current behavior as default:

ignition_mode = "edge" | "center"
spread_mode = "uniform" | "wind_weighted"

Wind parameters would only be active when spread_mode="wind_weighted".

This way wind becomes an optional advanced extension. If that sounds aligned with the project's style, I'm happy to refactor accordingly!

[EwoutH]

Sounds good! It’s always great to have some options in a model to play with and see what it does.