embrs.tools.fire_predictor#

Module for running a prediction on a fire to project its state in a set amount of time.

Runs a simplified version of the fire simulator to provide predictions.

Utilizes a homogenous fuel map, adds noise to the wind forecast, and has the ability to be biased towards a more or less conservative prediction.

class embrs.tools.fire_predictor.FirePredictor(orig_fire: FireSim, time_horizon_hr: float, fuel_type: int = - 1, bias: float = 1, time_step_s: float = None, cell_size_m: float = None)#

Predictor class responsible for running predictions over a fixed time horizon.

Parameters:

  • orig_fire (FireSim) – FireSim object to predict the future spread for.

  • time_horizon_hr (float) – Time horizon in hours for which the model should predict over.

  • fuel_type (int, optional) – Fuel type to be used for the predictor’s homogenous fuel map. If -1, the dominant fuel type in the original map will be used, defaults to -1

  • bias (float, optional) – Bias term that controls how conservative the model is, will tend to over-predict if >1, under-predict if <1, defaults to 1

  • time_step_s (float) – Time step in seconds used for the prediction. This will be the temporal granularity of prediction result. If None, the time step of orig_fire will be used.

  • cell_size_m (float, defaults to None) – Cell size in meters used for the prediction. This will be the spatial granularity of the prediction result. If None, the cell size of orig_fire will be used.

generate_noisy_wind(wind_forecast: list) → list#

Adds noise to the true wind forecast using a auto-regressive model.

Parameters:

wind_forecast (list) – Wind forecast being used by the original fire simulation

Returns:

New forecast with noise added

Return type:

list

iterate()#

Iterate the fire prediction one time-step forward.

perform_actions()#

Perform any actions in specified action sequence that occur at the current simulation time.

property reduced_fuel_prediction: dict#

Get the regions predicted to be partially burnt by prescribed burning

Returns:

Dictionary where each key is a time-step and each value is a list of predicted reduced fuel locations (x, y) at that time-step. Time-steps start at the time-step the original fire when input to the prediction model.

Return type:

dict

run_prediction(action_sequence: list = None) → dict#

Run a prediction

Parameters:

action_sequence – Action sequence that should be completed during the course of the prediction run. Specify as a list of Action objects.

Returns:

Dictionary where each key is a time-step and each value is a list of predicted ignition locations (x, y) at that time-step. Time-steps start at the time-step the original fire when input to the prediction model.

Return type:

dict

Classes

FirePredictor(orig_fire, time_horizon_hr[, ...])

Predictor class responsible for running predictions over a fixed time horizon.