VWETKF for wildfire propagation prediction employing observations with missing values and/or spatial variations of error variance

Occurrence of wildfires is common in all continents with wildland vegetation. In general, complete observation of fire perimeter is carried out for a data assimilation framework. Moreover, the common practice assumes that observation data have a constant error. However, the airborne or spaceborne observations may be incomplete and/or have the spatial variations of error variance due to instrument failures or the presence of fire-induced smoke plume or cloud cover. To overcome these issues, a vertex weight was introduced for revising the nudge term to offset the limitation of the traditional implementation of ensemble transform Kalman filter (ETKF) in this study. Herein, the extent of nudging the prediction simulations toward the observation in the vertexes, where observation data were missing or exhibited lower fidelity, was adjusted. The merits of the flexible spatially distributed state estimation approach were demonstrated through a series of observing system simulation experiments (OSSEs). The performance of the vertex weight-based ETKF (VWETKF) was compared with that of the ETKF. The results have showed that the proposed VWETKF scheme is superior to ETKF and leads to improvement of the prediction accuracy.