SPAD values estimation in jujube leaves based on lasso-based optimized spectral indices and a Stacking Algorithm

Jujube is an important economic forestry product, with Xinjiang being the leading producer in China, known for its high-quality fruits. Monitoring jujube growth is crucial for agroforestry management and Xinjiang's ecological economy. Chlorophyll content, a vital pigment associated with plant physiological activity and photosynthesis, can be non-destructively measured using the portable chlorophyll meter SPAD-502 Plus. However, its limitations in measuring SPAD values over a large area necessitate efficient estimation methods. Hyperspectral data and optimal band combination techniques are commonly employed to estimate SPAD values. However, the extensive hyperspectral data and time-consuming band calculations call for data dimensionality reduction methods. The LASSO method has shown promise in estimating SPAD values by selecting important hyperspectral bands. Nevertheless, its application in constructing vegetation indices for SPAD prediction in trees or vegetation leaves remains limited. To address this gap, this study utilizes LASSO-based optimized spectral indices and a Stacking Algorithm to estimate SPAD values in jujube leaves. The results demonstrate that LASSONDI and LASSO-RI are effective predictors of SPAD values in jujube trees. Additionally, the Stacking Algorithm combining SVR, GBDT, and Ridge models yields the best prediction performance with an R$^{2}$ of 0.98 and RMSE of 0.38.