Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
6539722 | Computers and Electronics in Agriculture | 2018 | 9 Pages |
Abstract
Carotenoids play important roles regarding photoprotection as well as light harvesting during the process of photosynthesis, resulting in the opportunity of quantifying carotenoids content to evaluate the productivity of vegetation. The traditional approaches such as ultraviolet and visible (UV-vis) spectroscopy are destructive and hence do not allow to determine the temporal dynamics of carotenoids content over time. As a promising alternative, hyperspectral remote sensing provides a way to evaluate carotenoid content changes over time and at multiple scales. Furthermore, it is easier to expand such approaches for large scale monitoring. However, to identify a generally applicable hyperspectral index sensitive to carotenoids remains a big challenge. In this study, we have evaluated thirteen available hyperspectral indices to quantify carotenoids, based on four independent datasets including two field datasets from Japan and two publicly available datasets (LOPEX and ANGERS). We attempted to develop a new generally applicable hyperspectral index for broadleaved plant species using the original and first derivative reflected spectra of the four datasets. We found that dND (516,744), a normalized differences type index using reflectance derivatives at 516 and 744â¯nm ((D516-D744)/(D516+D744)), had the highest robustness among all datasets and also was the best index when all data were combined (R2â¯=â¯0.475, WAICâ¯=â¯2430.1, and RPDâ¯=â¯1.45 for all datasets), suggesting its potential for general applications. Further extensive evaluations of the proposed index in other types of plants is required to test whether it can also be applied in other than broadleaved species.
Related Topics
Physical Sciences and Engineering
Computer Science
Computer Science Applications
Authors
Rei Sonobe, Quan Wang,