基于归一化燃烧率火烧迹地遥感提取识别规则研究

Extraction and Recognition of Burned Area Based on Normalized Burn Ratio by Remote Sensing

为探讨基于归一化燃烧率(Normalized burn ratio,NBR)的森林火灾火烧迹地识别规则在我国森林火灾火烧迹地识别中的可用性和适用性,以2006年云南省安宁市"3·29"重大森林火灾场景为火烧迹地识别规则试验区,以2013年云南省大理市下关镇和2015年云南省昆明市石林县2起森林火灾场景为火烧迹地识别规则检验区,对基于森林火灾前后不同Landsat遥感影像像对NBR的9条火烧迹地识别规则进行火烧迹地识别对比研究。结果表明:1)阈值为0.2~1.2的识别规则适用于安宁和大理的森林火灾研究;2)美国地质勘探局和美国农业部林业局提供的差值归一化燃烧率(difference Normalized Burn Ratio,dNBR)参考阈值范围,应用于云南的森林火灾研究需要根据实际情况调整;3)以火灾发生的时间为基准,遥感影像成像时间越接近发生火灾时间的像对获得的dNBR,用阈值规则识别为火烧迹地的像元数越多。4)安宁和石林森林火灾识别的像元数量出现大幅变化与气温、降水量、植被长势等因素相关,植被长势越好识别的像元数越多。5)在固定识别规则阈值最小范围情况下,随着最大阈值范围的增加,识别像元数量变化较小。

In order to investigate the availability and applicability of recognition rules of forest fire burned area based on the NBR （Normalized Burn Ratio, NBR） in China,the scene of the burned area resulted from a fatal forest fire occurred on 29 March,2016 in Anning,Yunnan Province was used as experimental area to testify the recognition rules of burned area. The scenes of other two forest {ires occurred in Xiaguan Town- ship of Dali in 2013 and in Shilin County of Kunming in 2015 were used as checking areas. Based on the Landsat remote sensing images before and after the forest fire,a contrastive study was conducted on the 9 rules of NBR. The results showed that 1） rules with the threshold values of 0.2-1.2 were applicable for the forest fires occurred in Anning and Dali. 2） Ranges of reference threshold value o{ difference normalized burn ratio （dNBR） proposed by the United States Geological Survey and Department of Agriculture-Forest Service were suitable for the forest fire occurred in Yunnan after modification according to the actual situation. 3） More pixel numbers that were recognized by threshold rules were observed in the dNBR of remote sensing image pairs that were closer to the time of forest fire occurrence. 4） The number of pixels in forest fires in Anning and Shilin changed greatly, which was related to the temperature, precipitation, vegetation growth and so on. The more vegetation grew, the more pixels were identified. 5） The minimum range of the threshold of the fixed recognition rule, the number of pixels varied little with the increase of the maximum threshold range.