Estimation Of Carotenoid Information On The Bifacial Leaves

As a fast and non-destructive method for estimating plant pigments, hyperspectral vegetation index is of great significance for the study of vegetation growth. Carotenoids are the second major types of pigments in plant pigments, and their concentrations can provide much information about the plant physiological state. However, most of the vegetation indices for estimating the carotenoid concentration was developed by the adaxial reflectance. In fact, while the remotely sensed data are being acquired, multiple scattering of higher-order canopy causes the incoming solar radiation to be reflected from understory and other leaves and enter the abaxial side of leaves. Furthermore, some foliage may change its orientation, turning the adaxial leaf surface away from the sun and exposing the abaxial leaf surface. This would make the remote sensing data contain spectral information from both adaxial Therefore, it is necessary for to create a new vegetation index based on the spectral information of both sides of the leaves to estimate the carotenoid concentration.In this thesis, we investigated the quantitative relationship between Car/Chi and different vegetation indices (VIs) on both adaxial and abaxial surfaces. The results showed that most of the published VIs did not have strong relationships with Car/Chl on either the one-surface dataset or the both adaxial and abaxial surface dataset. Among the reflectance indices tested, the modified Datt index performed best and is proposed as a new index for remote estimation of Car/Chl in plants with varying leaf surface structures. The best model of Narrow-leaved oleaster is (R720-Rs2o)/(R720-R73o), and the coefficient of determination (R2) is 0.91, the root mean square error (RMSE) is 0.14; the best model of White poplars is (R72o-R7io)/(R72O-R96o), R2 is0.75, RMSE is 0.12; the best model for Chinese elm is (R730-R500)/(R730-R760) and the R2 is 0.84, RMSE is 0.28; the best model for Virginia creeper is (R740-R4oo)/(R740-R750) and the f R2 is 0.80, RMSE is 0.23; the best model for Lilac is (R720-R510)/(R720-R590)and the R2 is 0.81, the RMSE is 0.38; the best model for the Grapes is (R710-R67o)/(R710-R4io) and the R2 is 0.66, RMSE is 0.08.;the best model for the torch is (Riooo-R64o)/(R1000-R500), and the R2 is 0.94, RMSE is 0.05. These results showed that the MDATT index is better than the other indices when it is applied on any individual plant species because it can remove the structure difference between different leaf surfaces which may have effect on the estimation of Car/Chl by reflectance.However, the estimation accuracy was obviously decreased when all the leaf samples of different species with the two surfaces. The best model for the this dataset is (R760-R640)/(R760-R480) and the R2 is only 0.52 and the RMSE is 0.37. It is considered that the MDATT model may be better in the removal of structure difference between the adaxial and abaxial surfaces in a plant species, but had some limitations in the removal of the structural differences between the species.