FireFly has been evaluated for both parameter estimation and state estimation approaches in

• academic tests using synthetically generated observations based on a reference run of the fire spread model with assumed true values of the control variables (this is a common practice in data assimilation to verify the consistent behavior of the data-driven system and to test different implementations of the data assimilation algorithms)

• reduced-scale controlled fire experiment (4 m x 4 m)

• FireFlux experiment (30 ha) with the extension of the EnKF algorithm to a spatially-distributed parameter estimation approach

emphasizing the potential of data assimilation to dramatically increase wildfire spread simulation accuracy. The current strategy adopted in FireFly is suitable for surface wildfire spread, but would require further developments to deal with a land surface-atmosphere modeling system.