Characterization of the distribution of mineral elements in chromium-stressed rice (Oryza sativa L.) leaves based on laser-induced breakdown spectroscopy and data augmentation

The fast and precise visualization of mineral elements in contaminated plants is crucial for understanding nutrient dynamics, plant health, and environmental monitoring. In this study, a Wasserstein generative adversarial network (WGAN) is proposed to expand the data scale and combines laser-induced breakdown spectroscopy (LIBS) with three machine learning methods for mineral elements analysis in Cr-stressed rice (Oryza sativa L.) leaves. In the quantitative analysis of mineral elements, the proposed method improved the prediction performance for Cu, K, Mg, Mn, and Na, except for Ca and Fe, demonstrating the effectiveness of data augmentation in enhancing the quantitative models. Mapping the distribution of Ca, Fe, Mg, K, Mn, and Na in rice leaves shows higher concentrations towards the apical regions and approximately symmetrical distribution along the leaf vein. Additionally, Fe, K, and Mn concentrations are significantly lower in Cr-polluted leaves compared to uncontaminated leaves. These preliminary findings offer insights into the macro distribution of mineral elements in plants.