论文部分内容阅读
大兴安岭是我国重点森林火灾区,准确预测该地区的森林可燃物含水率对于提高该地区林火发生预测的准确性意义重大。本研究采集典型林型的枯落物的光谱和含水率实测数据,通过一阶导数和去包络线的光谱分析方法识别森林枯落物含水率敏感波段。通过相关系数法从原始光谱、去包络线光谱、一阶导数光谱、去包络线之后的一阶导数光谱中筛选与枯落物含水率高度相关的波段作为含水率反演模型的备选自变量。利用逐步回归分析建立枯落物含水率反演模型,并对模型进行精度评价。结果表明,去包络线之后的一阶导数光谱对枯落物含水率变化存在显著响应,敏感波段位于398~668、768~1068、1098、1278、1388~1438、1458~1538、1868~1898、1988~2088、2198~2208、2228~2238 nm(P<0.05)。相关系数极值为-0.653、0.610,分别在波长2008、1888 nm处。通过多元逐步回归构建大兴安岭地区9种典型林型枯落物光谱和含水率的预测模型,模型决定系数R2=0.537,平均相对误差为0.303,均方根误差为0.499。本研究结果将为利用遥感技术快速测定森林枯落物含水率提供参考。
Daxinganling is a key forest fire area in China. It is of great significance to accurately predict the moisture content of forest fuel in this area to improve the prediction of forest fire in this area. In this study, the spectral data and moisture content data of typical litter were collected. The sensitive bands of moisture content of forest litters were identified by the first derivative and the spectral analysis of de-enveloping line. Correlation coefficient method was used to select the bands which are highly correlated with the moisture content of the litter from the original spectrum, the enveloping spectrum, the first derivative spectrum and the first derivative spectrum after enveloping, as an alternative to the moisture content inversion model Argument. The stepwise regression analysis was used to establish the model of the litter water content inversion and the accuracy of the model was evaluated. The results show that there is a significant response to the change of moisture content of litter after the first derivative of the envelope, the sensitive bands are located at 398 ~ 668,768 ~ 1068,1098,1278,1388 ~ 1438,1458 ~ 1538,1868 ~ 1898 , 1988 ~ 2088,2198 ~ 2208,2228 ~ 2238 nm (P <0.05). The correlation coefficient extremum is -0.653,0.610, respectively at the wavelength of 2008,1888 nm. Based on multivariate stepwise regression, the prediction models of spectral and water cut of nine typical forest litters in Daxing’anling were established. The coefficient of determination was R2 = 0.537, the average relative error was 0.303, and the root mean square error was 0.499. The results of this study will provide a reference for the rapid determination of water content of forest litter by using remote sensing technology.