Forest fire hazard assessment by clustering and using neural networks under unstability and uncertainty

Abstract

The paper focuses on the data on forest fires and identification of key natural and anthropogenic factors that are crucial for forest management, especially, for developing and implementing the fire safety measures. In recent decades, there have been observed the increased environmental, social and economic losses from the forest fires on a global scale, which has required stepped-up fire-fighting surveillance, especially in the preventive forest fire risk assessment. In all the variety of modern approaches aimed at assessing the fire hazards to the forests and taking into account the effecting environmental factors, most of them are based on simplified calculations and do not take into account different factors, mainly anthropological ones. The purpose of the study is to assess the forest fire risk depending on the environmental factors by using cluster analysis in conditions of instability and uncertainty. It could help applying the integrated approach to forest fire risk assessing in order to take into account both natural and anthropogenic factors in difficult conditions. To assess the forest fire risk, there were used the data obtained by MODIS spectroradiometer from January 1, 2014 to November 24, 2019: latitude; longitude; acquisition time and date. The following parameters were used as additional: Fire Weather Index; minimum distance to an inhabited locality; minimum distance to the road (highway or railway); minimum distance to the water area; holiday / day off; potential value. According to the results of the spatial distribution of forest fires and taking into account the data on the environmental factors there have been formed three clusters; there has been revealed a key relationship between the probability of a forest fire and proximity to the inhabited locality. There has been submitted the index of forest fire risk assessment (the Fire Weather and Human Index (FWHI)) based on the natural and anthropogenic impacts. Identification of social and biophysical aspects of the community exposure to fires and the adaptation of the existing fire prevention strategy will improve the forest fire safety system.

Forest fire monitoring via uncrewed aerial vehicle image processing based on a modified machine learning algorithm

“…In order to obtain a more detailed flame image, the flame pixels can be segmented in YCbCr color space and RGB color space, and the decision conditions can be obtained according to the two color spaces ( Stankevich, 2020 ). In case of forest fire risk, the color of forest fire is quite different from the background color of the forest environment, and the characteristics are obvious.…”