Research On Flammability Index Based On Multi-class Fuel Moisture Contents

In the context of increasing global warming and frequent wildfires,mega-fire appears in people’s vision.It burns a wider area,leading to more casualties and property losses.Moreover,it is more difficult to suppress it,requiring more fire fighters and increasing the risk of major casualties.At the same time,huge fires can worsen air quality and pose a serious public health hazard to nearby residents.According to the wildfire triangle model,Fuel Moisture Content(FMC)is the key parameter of wildfire risk assessment and early warning.According to the activity of combustible material,it can be divided into Live Fuel Moisture Content(LFMC)and Dead Fuel Moisture Content(DFMC).The advantages of remote sensing image such as large range,near real-time and multi-band provide convenience for parameter acquisition.The remote sensing data of fuel water content obtained by optical satellites are all used in the international fire warning system.California,the United States,has seen frequent fires in recent years.In 2018,there were two mega-fires that caused significant economic and human losses.At present,there is limited research on the role of water content of multiple types of fuels as indicators of large fire risk.Converting the FMC values into FI(Flammability Index)is an important step that contributes to its inclusion in an integrated fire risk assessment system.Yet further accuracy is needed because of the conversion of extinguishing humidity concepts and empirical statistics commonly used in previous studies.So far,no such assessment has been conducted in California,which has a diverse climate and ecology.Based on this,thesis analyzes the spatio-temporal distribution patterns of the2018 California wildfires by using four kinds of fuel water content parameters and MODIS(Moderate-Resolution Imaging Spectroradiometer)series of remote sensing products.Based on four kinds of fuel moisture content parameters,combustibility index FI is generated to further improve the warning accuracy.This study provides a reference for the correlation analysis between fuel moisture content and fire in fire-prone areas,and also integrates fuel moisture content into the fire risk assessment system.By studying the mega-fires in California,we can provide a prior knowledge analysis for preventing the occurrence of mega-fires in China,and reduce economic and social losses.The main work and achievements of thesis are as follows:(1)Comparison of the indicator effect of spatial and temporal distribution pattern of multi-class fuel moisture contents on mega-fire.Based on LFMC,FFMC(Fine Fuel Moisture Code),DMC(Duff Moisture Code)and DC(Drought Code)Moisture data sets,combined with the product MCD64A1.006,The actual combustion area of corresponding month is obtained.According to the IGBP classification standard of MCD12Q1.006,land cover products were divided into forest,shrub and grassland vegetation cover types.In this study,the temporal and spatial variation modes of combustion area and fuel moisture content were analyzed respectively.Meanwhile,the correlation between FMC and Fire Density(FD)was analyzed,and the indicator results of multiple types of fuel moisture content were further compared.The results showed that the four indicators had different meanings for the occurrence of large wildfires in the study area.In 2018,the performance of LFMC and DMC was better than that of FFMC and DC.For all types of fire,the correlation between LFMC and FFMC is higher,while the performance of DMC and DC is not good enough.(2)Study on the temporal correlation between water content of multi-class fuels and fire density.This study is based on LFMC,FFMC,DMC and DC,among which the latter three belong to dead fuel moisture content.Firstly,fire points and non-fire points were sampled based on MCD64A1.006 burnt ground products.The non-fire points were sampled by spherical model fitting semi-variogram values to determine the buffer zone.An equal number of sampling points were randomly selected outside the buffer zone to construct the fire data set.According to Pearson correlation coefficient and significance analysis method,the time-series correlation study of fuel moisture content and fire density was conducted to determine the optimal time window.The optimal time window is one month.(3)Productization of the Flammability Index(FI)based on the multiple types of fuel moisture contents.According to tsfresh time series feature extraction algorithm,the time dimension features in the optimal time window were extracted and the multidimensional feature data set was constructed.The fully connected neural network model is used to generate the combustibility index,and the regionalization research and typical case analysis are carried out.The results showed that the accuracy of the three quilt types was more than 0.8,and the AUC value of their models was more than 0.9.Compared with the traditional method of 0.7,there is a more obvious improvement.Compared with non-burning areas,the FI value of burning area is significantly higher,and has a significant increase trend in 2018 fire season,which proves the effectiveness of its burning index in 2018 California Fire,and provides support for FMC’s inclusion in the comprehensive fire risk assessment system.