Improving the multiple linear regression method of biomass estimation using plant water-based spectrum correction

Accurate observation of plant biomass is crucial for estimation of global carbon stocks and ecosystem productivity, and optical spectroscopy technology represents a potential solution. However, the plant water content limits biomass estimation accuracy, due to its contribution to the plant spectral characteristics. In this study, we used a plant water-based spectrum correction (PSC) method to decrease the influence of the plant water content, and biomass models were developed by combining the PSC and multiple linear regression (MLR) methods. The spectral data of 387 tree leaves were acquired using an ASD FieldSpec 3 spectrometer, and their water content values were determined by the oven-drying method. Based on the plant water values of the leaves, the optimal model for the PSC-MLR method was determined and was treated as the ultimate PSC-MLR model (coefficient of determination R (2) of validation = 0.6959, root-mean-square error of validation = 0.0220 kg m(-2), mean relative error of validation = 20.60%, ratio of performance to deviation of validation = 1.3492), which produced a better performance than the standard MLR model. This work shows the great potential of combining PSC with MLR for improved plant biomass estimation accuracy.