Spectral analysis of forest fire noise for early detection using wireless sensor networks

Abstract

Crown fires are extremely dangerous, very difficult to fight and often have a rate of spread over 100 times more than a surface fire. Therefore, it is important to determine the type of forest fire in the early detection based on wireless sensor networks (WSNs) to adopt the proper strategy to fight the fire. It is shown that this could be done analyzing the noise power spectrum of forest fires: surface fires noise spectrum can be modeled as the red noise (gradual increase of trend line amplitude toward lower frequencies), while for crown fires noise spectrum trend line has an almost bell-shaped (Gaussian) type. The noise frequency range is relatively narrow for crown fires and ranged from 250 to 450 Hz. The intermediate type of fires (strong surface fire and incipient crown fire) has a transient noise spectrum from broadband red to narrowband Gaussian. The article presents the spectrums of 9 different forest fires. The different trend line of the forest fire noise power spectrum is the parameter that can be used to determine the type of forest fire in WSNs.