Changjiang River Scientific Research Institute, Changjiang Water Resources Committee, Wuhan, 430010, China; Wuhan Center for Intelligent Drainage Engineering Technology Research, Wuhan, 430010, China.
School of Resource and Environmental Sciences, Wuhan University, Wuhan, 430079, China.
Environ Pollut. 2023 Dec 1;338:122618. doi: 10.1016/j.envpol.2023.122618. Epub 2023 Sep 25.
Neither cadmium (Cd) nor lead (Pb) is necessary for crop growth, but they both can accumulate in soil and crop tissues, resulting in land degradation and crop reduction. Few researchers have explored how to detect Cd-Pb co-accumulation in leaves using proximal sensing techniques, especially by low-cost, easy-to-use leaf clips that capture hyperspectral reflections at suitable foliar positions. In this study, a hyperspectral imager was employed to collect images of the rice canopy from a designed greenhouse experiment that included 16 pretreatments of Cd-Pb co-accumulation, followed by spectral extractions from 3 foliar positions: the blade root, the middle of the leaf, and the leaf apex. A support vector machine with leave-one-out cross-validation was performed to diagnose the contaminative levels based on the feature wavelengths selected by an improved successive projection algorithm. Partial least squares regression was used to predict Cd-Pb concentrations in rice blades. The results indicated that diagnostic accuracies were varied using spectra of different foliar positions. The blade root and leaf apex of rice blades were the optimal foliar position for detecting Cd and Pb contamination, respectively. At the optimal foliar positions, diagnostic accuracies exceeded 0.80 for distinguishing whether the rice is subject to Cd-Pb contamination. The Cd prediction performed 'very good' with a residual prediction deviation (RPD) of 2.21, a R of 0.79, and a root mean square error (RMSE)of 6.14, while that of Pb was 1.62, 0.61, and 186.54. Important wavelengths were identified at 659-694 nm and 667-694 nm to detect Cd and Pb contamination. In summary, our results verified the feasibility and clarified the optimal foliar positions of rice blades to detect Cd-Pb contamination. The wavelengths selecting have the great potential in the design of future leaf clips, and the optimal foliar position can provide suggestions to improve diagnostic performances in field applications.
镉（Cd）和铅（Pb）都不是作物生长所必需的，但它们都可以在土壤和作物组织中积累，导致土地退化和作物减产。很少有研究人员探索过如何使用近地感应技术检测叶片中的 Cd-Pb 共积累，特别是使用低成本、易于使用的叶片夹在合适的叶面位置捕获高光谱反射。在这项研究中，使用高光谱成像仪从一个设计好的温室实验中收集水稻冠层的图像，该实验包括 16 种 Cd-Pb 共积累的预处理，然后从 3 个叶面位置提取光谱：叶片根部、叶片中部和叶片顶端。采用带有留一法交叉验证的支持向量机，根据改进的连续投影算法选择的特征波长来诊断污染水平。使用偏最小二乘回归预测水稻叶片中的 Cd-Pb 浓度。结果表明，使用不同叶面位置的光谱，诊断准确率有所不同。叶片根部和叶片顶端是检测 Cd 和 Pb 污染的最佳叶面位置。在最佳叶面位置，区分水稻是否受到 Cd-Pb 污染的准确率超过 0.80。Cd 的预测表现“非常好”，剩余预测偏差（RPD）为 2.21，R 为 0.79，均方根误差（RMSE）为 6.14，而 Pb 的 RPD 为 1.62，R 为 0.61，RMSE 为 186.54。在 659-694nm 和 667-694nm 处鉴定出重要波长来检测 Cd 和 Pb 污染。总之，我们的结果验证了该方法的可行性，并明确了水稻叶片检测 Cd-Pb 污染的最佳叶面位置。所选波长在未来叶片夹的设计中具有很大的潜力，而最佳叶面位置可为提高田间应用中的诊断性能提供建议。
