重金属铜胁迫下玉米的光谱特征及监测研究

农作物重金属污染监测是当今高光谱遥感研究的重要内容之一，旨在设计一种新的窄带植被指数，以实现不同培育期的两种玉米品种的重金属铜胁迫监测。研究设计了不同浓度的铜污染实验，采用SVCHR-1024I型高性能地物光谱仪测量不同浓度铜离子（Cu2+）胁迫下玉米叶片的光谱反射率，并同步获取了玉米叶片中Cu2+含量数据。首先，对玉米叶片原始光谱数据进行一阶差分处理，并计算一阶差分反射率与叶片中Cu2+含量的相关系数（r），筛选对铜胁迫敏感的波段。计算结果显示，489~497，632和677 nm波长附近的一阶差分反射率与叶片中Cu2+含量显著相关，可将其视为敏感波段。其次，根据以上3个敏感波段，建立基于一阶差分反射率的铜胁迫植被指数（dVI）。对所有可能的dVIs和Cu2+含量进行一元回归分析，并采用决定系数（R2）和均方根误差（RMSE）对回归结果进行评估，以筛选最佳指数。最后，采用不同生长年份的玉米实验数据对敏感波段的稳定性及dVI的适用性进行了验证评估；同时，通过与归一化植被指数（NDVI）、红边叶绿素指数（CI_red-edge）、红边位置（REP）、光化学反射指数(PRI)等常规重金属胁迫植被指数进行应用比较，证明dVI更具有优越性。结果表明：一阶差分处理后，在450~500，630~680和677 nm波长处的叶片反射率与Cu2+含量的相关系数明显增大。基于一阶差分反射率的特征波段具有稳定性，对于不同生长年份的玉米叶片数据，特征波段的波长位置不变。一元回归分析结果表明，结合497，632和677 nm波长的一阶差分反射率的指数与Cu2+含量具有显著的相关性，对于不同生长年份的2种玉米品种数据集，R2都高达0.75以上。另外，与常规植被指数比较结果表明，该研究所提出的dVI具有更好的鲁棒性及有效性，可为冠层尺度的重金属胁迫监测提供理论基础。

Spectral Characteristics and Identification Research of Corn under Copper Stress

The monitoring of heavy metal pollution in crops is one important application of hyperspectral remote sensing study. The objective of this work was to develop a new narrow-band vegetation index to characterize the Cu (copper) stress degree in two corn species at two growing years. The experiment on the copper pollution was designed based on its different concentrations, meanwhile, the hyperspectral reflectance of corn leaves stressed by different Cu2+ concentrations were measured using hand-held spectrometer(SVC, USA) and leaf Cu2+ contents were also measured. The first difference reflectance and biochemical data of corn were analyzed using Pearson correlation coefficient (r) to select wavelengths sensitive to Cu stress. The calculated Pearson correlation coefficients suggested that the first difference reflectance near 489~497, 632 and 677 nm wavelengths was significantly correlated with Cu2+ contents in leaves. The selected wavelengths of 489~497, 632 and 677 nm were used to establish the Cu stress vegetation index based on the first difference reflectance (dVI). To select index with the highest possible correlation to Cu stress, all possible dVIs were related through simple regression models with Cu2+ contents andthe predictive abilities of those models were evaluated through the R2 values and the root mean square error (RMSE). The stability of the sensitive bands and the applicability of dVI were assessed using corn data from different growth years. Meanwhile, the performance of dVI was compared with that of existing popular vegetation index (VIs) related to heavy metal stress, such asnormalized difference vegetation index (NDVI), red-edge chlorophyll index (CI_red-edge), red-edge position (REP), photochemical reflectance index (PRI). The results suggest that the corn spectral characteristics in response to copper stress are enhanced with the first-order difference treatment. Compared with the original reflectance, the correlation coefficient between first difference reflectance at wavelengths of 450~500, 630~680 and 677 nm and Cu2+ content increases. The wavelength position of copper stress sensitive band based on the first-order differential reflectance is stable for the data sets of different growth years. The index that combined the first difference reflectance in 497, 632 and 677 nm wavelengths is found to be a potential useful index to predict leave Cu concentration for different data sets. And the correlation of dVI was much stronger than that of other VIs for all the tested data sets from two corn species at two growing years. The proposed dVI characterizes the Cu stress degree on vegetation with advantages of better effectiveness and robustness. This study focuses on the spectral reflectance at the leaf scale, so it is expected that future work will extend it to canopy scale.