Dual NDVI ratio vegetation index: a kind of vegetation index assessing leaf carotenoid content based on leaf optical properties model

With characteristics of rapidness, non-destructiveness and high precision in detecting plant leaves, hyperspectral technology is promising in assessing the contents of leaf pigments and other biochemical components. Because the spectral absorption features of carotenoid and chlorophyll are overlapped in visible light region and that foliar carotenoid content is far lower than chlorophyll content, studies about constructing vegetation indices (VIs) for carotenoid is rare at home and abroad though carotenoid is one of the most important photosynthetic pigments. Hyperspectral data has abundant spectral information, so this paper proposed a multiple spectral indices collaborative algorithm to construct VIs on the basis of band-combination traversal and correlation analysis. Through a large number of simulated leaf reflectance spectra under different biochemical components contents run on PROSPECT model, a radiative transfer model, we successfully constructed a new kind of stable vegetation index (VI) for assessing carotenoid content at leaf level: RVI_DNDVI. Our results indicate that RVI_DNDVI is composed of two parts: (1)Narrow band NDVI constructed with 532 and 405 nm is high correlated with both carotenoid content and chlorophyll content while narrow band NDVI constructed with 548 and 498 nm is highly correlated with carotenoid content. The influence of chlorophyll content on RVI_DNDVI can be eliminated with the ratio combination of these two indices. (2) The influence of mesophyll structure parameter can be weakened by subtracting the reflectance at 916 nm, which has strong correlation with mesophyll structure parameter. RVI_DNDVIonly has high sensitivity to carotenoid content (the correlation coefficient is -0.94) at leaf level and R² of its exponential fit is 0.834 4. The estimation of RVI_DNDVIto carotenoid content can be verified with the validations of both simulated data and measured data.