森林生物化学与CASI高光谱分辨率遥感数据的相关分析

Relationships between Forest Biochemical Concentrations and CASI Data along the Oregon Transect

该文介绍了利用光谱范围约 417— 80 0nm的航空高光谱分辨率数据估计森叶族叶化学成分浓度的分析方法。沿横跨美国俄勒冈州中西部收集 6个研究立地上的小型机载成象光谱仪 (CASI)数据。 3个族叶化学成分 [总叶绿素 (TC)、全氮 (TN)和全磷 (TP) ]从相同的研究立地取样并在实验室测定。使用多元统计和光谱微分技术评价CASI数据用于估计冠层生化浓度的潜力和效率。 12个族叶化学成分样本被测定 ,并同时在相同立地上提取CASI图象数据。 7个单回归模型被用来探索单波段和植被指数与 3个化学成分的线性与非线性相关关系及预测效果。光谱微分技术被用来压缩背景噪音对目标光谱的影响。利用CASI原始光谱、一阶和二阶微分光谱数据和逐步回归分析来预测TC、TN和TP。结果表明采用光谱微分技术能显著地改善由拟合度 (R2 )和均方根差(SE)描述的森林冠层化学成分浓度的估计精度。由单波段分析的结果说明族叶化学与CASI数据间的相关性很低。事实上 ,对于TP 。

In this paper,correlation between airborne hyperspectral data obtained with in the spectral range of approximately 417nm—800nm and three foliar biochemical constituents was studied.The hyperspectral data were acquired using Compact Airborne Spectrographic Imager(CASI) over six Study sites along an Oregon transect.Foliar biochemical constituents(expressed in concentration,mg/g of dry weight):total chlorophyll(TC),total nitrogen(TN),total phosphorus(TP),were measured from the same transect.The potential of CASI data for estimating foliar biochemical concentrations was evaluated,using multivariate statistical analysis and spectral derivative techniques.12 spectra corresponding to sample locations of foliar chemistry measurements were extracted from CASI images acquired within three weeks of the field sample collection.Univariate regression using 7 regression models was applied to explore both the linear and non-linear relations between individual channel spectral reflectances and vegetation indices derived from the CASI data and the three foliar chemical constituents.First and second order spectral derivatives were used to suppress the effects of low frequency spectra on those of the tree species.A piece-wise multiple regression procedure was used to generate multivariate linear equations for predicting TC,TN and TP using the original,and the derivative spectra of CASI data. Results show that with the spectral derivative technique,the estimation accuracies measured by Goodness-of-Fit(R 2)values and Root-Mean-Square Errors(SEs)of the three chemical constituents,can be greatly improved.The results obtained from the univariate analysis indicate that the level of correlation between foliar chemistry and CASI data is rather low.In fact,for TP no correlation exists between any individual band of CASI data and the measurements from laboratory.