基于Himawari-8静止卫星森林火灾识别技术研究

Approach to the fire detection technology and the algorithm from the geostationary Himawari-8

新一代静止气象卫星Himawari-8/9 AHI(Advanced Himawari Imager)传感器采用全圆盘扫描观测方式,常用极轨卫星Day/Night双模式火点算法不适用于AHI火点识别。本文以MODIS MOD14(MODIS Fire algorithm)火点识别算法为基础,通过分析静止卫星观测下的火点像元辐射特性和太阳中红外反射特征,确定适用于AHI的MIR/LIR火点识别因子及其动态检测阈值,并利用林业部门提供的大量实测历史火点数据校正识别因子和动态检测阈值,最终建立一种基于Himawari-8/9AHI传感器的采用中红外动态阈值的静止卫星火点识别算法模型。利用NPP VIIRS VNP14(NPP VIIRS Fire products)和JAXA EORC WLF (Wild Land Fire)火点数据,对一次兴安岭森林火灾案例中288个时次AHI全圆盘火点判识结果进行对比验证,结果表明该AHI火点识别算法准确率超过95%。根据新一代静止气象卫星Himawari-8/9AHI传感器提供的10分钟密度遥感观测数据,即可以实现森林火灾实时跟踪监测。

Based on the advanced theory of MODIS MOD14 fire algorithm,the present paper is engaged in a study on how to apply the fire detection algorithm to Himawari-8 AHI sensor. Due to the great contrast of the solar height angle of the pixel in the geostationary satellite observation, the general polar satellite day/night dual-mode fire algorithm cannot be directly applied to the AHI sensor. Therefore,we have to manage to analyze the specialty of the fire spot radiation and the solar infrared reflectance components by the geostationary satellites through the radiation transfer simulation to get the infrared reflection correction method of the surface pixel under the entire disk scanning observation mode of the geostationary satellite. At the same time,we have also kept tracing the thermal infrared radiation response of subpixel fire in the MIR/LIR band of AHI sensor,so as to confirm the MIR/LIR fire pixel identification factors suitable for AHI by determining its dynamic detection thresholds. And,then,the accuracy of the recognition factors and the dynamic detection thresholds can be regulated and corrected by the forestry administration departments with the help of the historical fire point data gained manually accumulated along with the Himawari-8 AHI historical data. By setting up the MIR/LIR combined factors,it would be possible to work out a fully automatic identification model of the fire pixel for geostationary satellite Himawari-8 AHI sensors. Thus,eventually,it would be possible to establish successfully a stationary satellite fire pixel identification algorithm based on the Himawari-8/9 AHI sensor via the mid-infrared dynamic threshold. And,at such a step,the identification results of 288 times of the AHI full disk fire pixel in the Xing’an Mountains forest fire case of North China can be compared with and verified through the NPP VIIRS VNP14 and JAXA EORC WLF fire point data. The results of the above examination and identification may demonstrate that the AHI fire pixel identification algorithm can be made to reach the accuracy of over 95%.Based on the application of the geostationary satellite fire algorithm,we have also developed a fire monitoring service real-time operation system known as Himawari-8 satellite remote forest sensing,which can continuously transmit the real-time fire monitoring data to the forest administrators all the 24 hours continuously for over one year. According to the response of the forestry administration,the given fire monitoring algorithm can be said qualified enough to meet the demands of the forest fire monitoring in terms of the recognition precision and the real-time response promptness. And,for the time being,the new generations of the geostationary meteorological satellite Himawari-8/9 can also be given a lot of improvements in the spectral observation performance through a 10 min. high temporal remote sensing,which has greatly heightened the forest fire monitoring effectiveness and efficiency in real time tracking and their recognition.