Monitoring of arsenic uptake in Pteris vittata using short-wave infrared spectroscopy during multi-scale field trial

In this study an innovative multi-analytical approach to monitor arsenic (As) accumulation in Pteris vittata, using hyperspectral imaging (HSI) in the short-wave infrared range (SWIR: 1000–2500 nm), micro-X-ray fluorescence (micro-XRF) and spectroradiometric measurements obtained with a portable device, have been applied. The objective was to establish a reliable and non-invasive strategy for tracking As uptake dynamics and related physiological changes in plants under both controlled and field conditions. To this end, 28 ferns were planted in an As-contaminated soil and monitored for up to 120 days. The study was structured into three phases. First, micro-XRF was used to monitor As accumulation kinetics in the pinnae of a representative number of plants. In the second phase, the same pinnae were also analyzed using HSI to characterize spectral signatures related to As induced stress and to explore spectral variability through t-distributed Stochastic Neighbor Embedding (t-SNE). This analysis revealed specific spectral patterns linked to As accumulation. An ECOC-SVM-based classification model was then developed using HSI data to assess As amounts in laboratory scale. In the third phase, based on the spectral features and classification approach developed by HSI, a new ECOC-SVM classifier was produced using all the data acquired by a portable spectroradiometer in all field-grown plants. The results confirmed that micro-XRF efficiently tracked As accumulation, while HSI identified distinct spectral signatures associated with As induced stress. The t-SNE analysis demonstrated variability in spectral responses, which facilitated the development and classes set of a robust ECOC-SVM model. Importantly, applying the ECOC-SVM model to the portable spectroradiometer data demonstrated its effectiveness in real-time monitoring of As phytoextraction at field-scale. This multi-analytical approach provides an efficient and scalable tool for optimizing phytoremediation strategies and environmental monitoring, confirming its reliability in both laboratory and field settings.