Water deficit-induced biochemical alterations and Vis/NIR spectral signatures in shallots

• Four water-deficit stress (WDS) levels were applied to shallot plants. • Bioactive compounds and Vis/NIR spectral response were evaluated under WDS conditions. • WDS induced distinct spectral features corresponding to biochemical changes. • Chemometric models revealed stress-dependent spectral–biochemical relationships. Water-deficit stress (WDS) is known as a condition in which a plant experiences limited water availability. This condition can trigger changes in metabolites and physiology in plants. Therefore, the study aimed to investigate water-deficit-induced biochemical changes in shallots to their spectral data, specifically in the visible/near infrared (Vis/NIR) region. In this study, three modeling approaches were evaluated separately. Subsequently, we assigned partial least squares (PLS), partial least squares–support vector machine (PLS-SVM), and partial least squares–feed-forward backpropagation neural networks (PLS-FFBPN) to develop regression models. Spectral and biochemical data were analyzed to explore how stress-induced variations within bioactive compounds are expressed in spectral data. The results showed that WDS level influenced the spectral patterns and model behavior, which was reflected by statistical indicators like root mean squares error of calibration (RMSEC), root mean squares error of cross-validation (RMSECV), and error difference (ED) values. Spectral-biochemical relationships became stronger for ascorbic acid equivalent antioxidant capacity (AEAC), total phenolic content (TPC), and total flavonoid content (TFC) with increasing WDS levels. But the opposite trend was observed for proline and IC 50 . On the other hand, other bioactive compounds like kaempferol, quercetin, and rutin did not exhibit a clear pattern with increasing stress levels. Among the three methods we tested, we found that PLS-FFBPN outperformed PLS and PLS-SVM because it could better capture the complex relationships between spectral data and bioactive compounds. The findings demonstrated that agronomic conditions such as WDS alter both the biochemical composition and its spectral signatures in shallots, especially for AEAC, TPC, TFC, proline, and IC 50 . In addition, hybrid regression methods (PLS-FFBPN) provided a robust framework for interpreting Vis/NIR responses associated with biochemical variation in shallots under varying water stress conditions.