Exploiting Sensor Spatial Correlation for Dynamic Data Driven Simulation of Wildfire

Abstract

Dynamic data driven simulation based on Particle Filter (PF) has been shown to increase the accuracy of wildfire spread simulation by assimilating real time sensor data into the simulation. An important issue in dynamic data driven simulation is to utilize the sensor data in an efficient and effective manner. In our previous work, all sensor readings are treated as independent from each other, however, when sensors are randomly deployed, measurement data from nearby sensors could be correlated and thus biased observation could be incurred. This paper presents a spatial correlation model to exploit sensor correlations from sensor spatial locations and inter-distance, and integrate it as part of the PF measurement model. Experiment results show that with the information of sensor correlation simulation accuracy is further increased.