P18     Fuel moisture model in WRF-Fire and assimilation of RAWS data.

Mandel, Jan, University of Colorado Denver, Adam Kochanski, University of Utah, and Martin Vejmelka, CEAI, Czech Republic

The behavior of wildland fire is highly sensitive to the fuel moisture content. With increasing fuel moisture content, the fire rate of spread decreases, and eventually, at the extinction moisture level, the fire does not propagate at all.
We have coupled a dead fuel moisture model with WRF-Fire. The moisture content in each idealized fuel species, identified by their characteristic response time as 1-hour, 10-hour, and 100-hour fuels, is simulated on a coarse grid, while the actual fuel used in the fire propagation is a mixture of these species on a much finer grid, where the fire simulation takes place.  At each point of the coarse grid, the moisture content of each fuel species is simulated independently by a first-order time-lag differential equation, whose solution approaches asymptotically an equilibrium fuel moisture content, estimated from WRF atmospheric conditions.
The model resolves automatically diurnal variability of wildland fire and the fire sensitivity to changing temperature and relative humidity.
The model can be steered by assimilating spatially sparse fuel moisture observations from remote automatic weather stations (RAWS). A trend surface model is used to extend the spatially sparse RAWS observations and their uncertainty estimate to the whole domain. At each grid point, this information is then combined with the model forecast using a nonlinear Kalman filter, which modifies the model state and parameters.
The model is also run as continuously as a part of the online wildland fire simulation system WRFx in standalone nowcasting mode, driven by real time weather forecast, and its results are available online.