Remote sensing of chlorophyll in Lake Kinneret using highspectral- resolution radiometer and Landsat TM: spectral features of reflectance and algorithm development

Authors

Yosef Z. Yacobi, Yigal Allon Kinneret Limnological Laboratory
Anatoly Giltelson, Ben-Gurion University of the Negev
Meir Mayo, Tel Aviv University

Document Type

Article

Date of this Version

1995

Comments

Published in Yacobi, Gitelson and Mayo in Journal of Plankton Research (1995) 17(11). Copyright 1995, Oxford University Press. Used by permission.

Abstract

High-resolution reflectance spectra in the range of 400-850 nm were obtained from Lake Kinneret during a period when dense populations of the dinoflagellate Peridinium gatunense dominated the phytoplankton. Chlorophyll (ChI) concentrations ranged from 5.1 to 185 mg m^-3 and from 2.4 to 187.5 mg m^-3 in the samples of two independent experiments. The most prominent features of the reflectance spectra were: (i) a wide minimum from 400 to 500 nm; (ii) a maximum at 550-570 nm, which did not surpass 3% in samples with high ChI concentration (>20 mg m^-3), indicating a strong absorption by pigments in the green range of the spectrum; (iii) a minimum at 676 nm; this was ~ 1 % and was almost insensitive to variation in ChI concentration >10 mg m^-3; (iv) a maximum reflectance showed near 700 nm; its magnitude and position were highly dependent on chlorophyll concentration. Highspectral- resolution data were used as a guideline for selection of the most suitable spectral bands for chlorophyll remote sensing. Models were devised, based on the calculation of the integrated area above the baseline from 670 to 850 nm and the reflectance maximal height within this range. Some algorithms already used in previous studies were tested and showed a plausible degree of accuracy when applied to the current data base. However, novel models devised in this study improved substantially the accuracy of ChI estimation by remotely sensed data, by reducing the estimation error from >11 to 6.5 mg m^-3. Those models were validated by an independent data set where ChI concentration ranged over two orders of magnitude. The use of three relatively narrow spectral bands was sufficient for ChI mapping in Lake Kinneret. Therefore, a relatively simple sensor, measuring only a few bands will be employed in future applications for ChI monitoring in inland waters. Radiometric data were also used to simulate radiances in the channels of TM Landsat and to find the algorithm for ChI assessment. The ratio of channel 4 to channel 3 was used and enabled ChI estimation with an error of <15 mg>m^-3. This algorithm was employed to map ChI in the entire area of Lake Kinneret with 10 gradations.