Construction of hyperspectral reflectance and spectral index inversion model for the water content of Catalpa bungei leaves

Real-time monitoring of leaf water content is an important indicator of drought resistance in plants. In this study, hyperspectral reflectance and derived data are used to build an inversion model for leaf water content of Catalpa bungei. Rapid, non-destructive and real-time monitoring of leaf water content provides a high-throughput method for assessing drought resistance in tree seedlings. The hyperspectral reflectance and water content of mature leaves were determined and several models were built to evaluate the optimal combination using different variable selection and model construction methods. The results show that the PLS regression model constructed with reflectance as the input variable is the best for the test series. The MC-UVE method is the best for all models. With the PLS regression method, the model approach is optimal. MC-UVE-PLS model optimal test set regression coefficient (R2) maximum (0.7903), mean square root error (RMSE) minimum (1.7352). SR (1466 nm, 2128 nm) is the spectral index with the highest water correlation. First order differencing can effectively improve the correlation between the spectral data and water content, but the model cannot be optimised. Using MC-UVE as a variable screening method, PLS regression was used to build an inversion model for leaf water content, which provides technical support for real-time monitoring of leaf water content of Catalpa bungei.