Mapping 3D plant chlorophyll distribution from hyperspectral LiDAR by a leaf-canopyradiative transfer model

The three-dimensional (3D) plant chlorophyll distribution plays an essential role in plant physiological and ecological analysis. The hyperspectral LiDAR (HSL) technology, as a new active technology, has the advantage of directly and simultaneously obtaining plant hyperspectral and 3D spatial information. It provides an alternative opportunity for fast and nondestructive 3D plant chlorophyll estimation. Current estimation models are developed by statistical relationships between HSL spectra and plant chlorophyll content, which perform limited robustness and lack physical mechanisms. The leaf-canopy radiative transfer models (RTMs) could describe physical responses between spectra and chlorophyll content, and show great robustness from optical hyperspectral data, which have the potential to overcome the limitation of current estimation models. However, it is unclear whether leaf-canopy RTMs could be applied to HSL spectra for 3D chlorophyll mapping. Therefore, this study explored the potential of a leaf-canopy RTM with HSL spectra for 3D plant chlorophyll estimation. Firstly, an optimal leaf RTM was selected from a series of PROSEPCT models, and determined by estimating leaf-scale chlorophyll content across multi-species and multi-state leaves. Then, the 4SAIL model, a canopy RTM, was optimized to meet the active detection mechanism of the HSL technology. Next, coupling the best PROSPECT with optimized 4SAIL models was used as the leaf-canopy RTM. The 3D chlorophyll distribution was estimated by the leaf-canopy RTM with a lookup table (LUT) algorithm under six cost functions. Finally, the performance of 14 vegetation indice (VI) models was tested as a comparison. Results demonstrated that the PROSPECT-5 model was the best leaf RTM with the best R2 of 0.60. The combination of PROSPECT-5 and optimized 4SAIL model could successfully achieve 3D plant chlorophyll estimation. This leaf-canopy estimation RTM with the best cost function (Least square estimator) had the best accuracy with an R2 of 0.73. Its accuracy was the same as the median accuracy of the best VI model (MTCI model), and higher than that of other VI models. The estimated and actual 3D chlorophyll distributions were consistent. The application of the leaf-canopy RTM with HSL technology offers an alternative approach to plant physiological and ecological research.