Estimation Of Wildfire Spread Rate From Geosynchronous Orbit Satellite Remote Sensing Data

Wildfire spread rate(FSR),defined as the distance of wildfire spreading in a certain direction in a unit time,is an important parameter for estimating other wildfire behavior characteristics(wildfire flame height,wildfire line density,etc.).Near real-time wildfire spread rate data provide timely and effective monitoring information for wildfire prevention and control,and improve the efficiency of wildfire rescue for relevant departments.However,the traditional wildfire spread rate estimation methods are difficult to estimate wildfire spread rate in large area with near real-time,and can not provide scientific and timely decision-making information and support for relevant departments and the public.With the launch of new generation geosynchronous satellites,it is possible to acquire remote sensing data with high spatial and temporal resolution,which is helpful to improve the timeliness and accuracy of wildfire spread rate extraction.In this study,based on the date of Himawari-8,a Japanese new generation of geosynchronous satellite,a grassland wildfire broken in Australia on November 17,2015was taken as an example to carry out the research of the active wildfire detection method and wildfire spread rate extraction method for the geosynchronous satellite data.The main achievements are described as follows:(1)Four commonly used wildfire detection algorithms were applied to Himawari-8data to detect near real-time active wildfire in the study area.Four kinds of wildfire detection algorithms include three kinds of wildfire detection methods based on spatial information(fixed threshold,window context and global context)and one kind of wildfire detection method based on temporal information(time-series method).The parameters of different algorithms are set according to the information of the research region.Finally,based on the Moderate Resolution Imaging Spectroradiometer(MODIS)active wildfire products(MYD14),the quality of different wildfire detection algorithm results was evaluated.At the same time,the beta version near real-time active wildfire(WLF)products of Himawari-8 were also evaluated.The results show that the global context detection algorithm performs best in the research area,followed by the time series method,the fixed threshold method and the window context method have the same performance,and the WLF data quality is the worst.(2)A new near real-time wildfire rate extraction method(H8-FSR)based on geosynchronous satellite data is proposed.The method is based on Huygens theory and wildfire centroid theory.The implementation of the method is divided into three steps:time-series wildfire burned area extraction,time-series wildfire center extraction and near real-time wildfire spread rate calculation.The wildfire burned area data accumulate from the active wildfire data,and the center of wildfire extract from the centroid of wildfire burned area data.The method was applied to extract the near real-time wildfire spread rate in the study area.Compared with the FSR estimated from the Commonwealth Scientific and Industrial Research Organization(CSIRO)Grassland Fire Spread(GFS)model,H8-FSR achieved favorable performance with a coefficient of 0.944,determination coefficient(R~2)of 0.552,mean bias error(MBE)of-0.15 m/s,mean absolute percent error(MAPE)of 31.25%,and root mean square error(RMSE)of 0.930m/s.(3)The influence of wildfire quality and accumulated time of wildfire burned area on the performance of H8-FSR method was discussed,and the relationship between wildfire spread rate and vegetation/topography/weather data was analyzed.The analysis shows that the performance of H8-FSR is different under different accumulative time of active wildfire data.H8-FSR performs best when the accumulative time of active wildfire data detected based on global context algorithm is 1 hour and 2 hours.Under the same cumulative time,the higher the quality of active wildfire data corresponding the higher the performance of H8-FSR to extract wildfire.The correlation analysis between wildfire spread rate and vegetation/topography/weather information showed that wildfire spread rate was significantly correlated with combustible data,vegetation growth data and windspeed,but not with slope.