Estimating dead fine fuel moisture content of forest surface, based on wireless sensor network and back-propagation neural network

Abstract

The dead fine fuel moisture content (FFMC) in forest affects the occurrence and spread of forest fires. Therefore, the estimation of surface dead FFMC plays an important role in forest fire behaviour, fire management and fire danger assessment. However, there are some challenges with current FFMC measurement methods. Modelling using meteorological variables may be a very potential method to achieve remote real-time FFMC estimation. A surface dead FFMC estimation method based on wireless sensor network (WSN) and back-propagation (BP) neural network was proposed. The WSN can realise the acquisition of microclimate data. The BP neural network can use these data to establish multiple FFMC estimation models for different terrain conditions and dead fuel types. The ability of these models to estimate FFMC at four different terrain sampling sites was evaluated. The results suggested that the dead FFMC can be estimated with some degree of accuracy. The correlation coefficients of the estimation results at the four sampling sites were all greater than 0.9, and the mean square errors were all less than 1. The method can be well applied to forest surface dead FFMC estimation and early fire danger assessment, which has practical significance for the rational allocation of fire fighting resources.