Inversion of Apple Leaf Moisture Content and SPAD Values Using Multispectral Vegetation Indices and Band Reflectance from UAVs

Accurate monitoring of leaf chlorophyll-related status (SPAD) and water status is important for evaluating apple growth and supporting orchard management. In this study, SPAD and leaf water content (LWC) were considered complementary indicators of chlorophyll-related and water-related physiological characteristics, respectively. Unmanned aerial vehicle (UAV) multispectral imagery and field measurements were collected during the flowering, fruit-setting, fruit-expansion, and fruit-coloring stages of Aksu red Fuji apples. Sensitive spectral features were screened using correlation analysis, and k-nearest neighbor (KNN), extreme learning machine (ELM), random forest (RF), and backpropagation neural network (BPNN) models were developed for SPAD and LWC estimation. The results showed clear relationships between multispectral features and both traits, with the RF model combining vegetation indices and band reflectance achieving the best overall performance. During the fruit-expansion stage, the optimal RF models achieved independent test-set R2 values of 0.767 for SPAD and 0.768 for LWC. These results demonstrate the potential of UAV multispectral remote sensing combined with machine learning for orchard-scale spatial monitoring of apple leaf chlorophyll and water status.