«Remote estimation of chlorophyll concentration in productive waters: Principals, algorithm development and validation A.A. Gitelson, Y.Z. Yacobi, D. ...»
Remote estimation of chlorophyll concentration in productive waters: Principals,
algorithm development and validation
A.A. Gitelson, Y.Z. Yacobi, D. C. Rundquist, R. Stark, L. Han, and D. Etzion
Anatoly A. Gitelson is a professor with both the School of Natural Resource Sciences, Institute of Agriculture and
Natural Resources, University of Nebraska-Lincoln, USA and the Institute of Desert Research, Ben-Gurion University
of the Negev, Israel. Gitelson holds a Ph.D. in Radio Physics (Russia, 1972). His general area of interest, beginning in the early 1980's, is physical aspects of remote sensing; radiative transfer; remote sensing of environment in visible, near infrared and microwave ranges of the spectrum.
Yosef Z. Yacobi is a senior scientist with Yigal Allon Kinneret Limnological Laboratory, Israel Oceanographic & Limnological Research, Israel. Yacobi holds a Ph.D. in Genetics (Israel, 1988). His general area of interest, beginning in 1986 is phytoplankton ecology in Lake Kinneret and in the Mediterranean Sea.. The central issue in his work is the relationship between light, photosynthetic activity of phytoplankton and algal biomass production. Since 1992 he is involved in research projects on phytoplankton remote sensing in various water bodies.
J.F. Schalles is professor and chair of the Biology Department at Creighton University in Omaha, Nebraska. He received his Ph.D. in Biology from Emory University in Atlanta Georgia. His current research focuses on aquatic remote sensing, using hyperspectral instruments and a combination of controlled mesocosm experiments and field observations. Current projects include spectroscopy of phytoplankton pigments, seston, and DOC in turbid inland and coastal waters and characterization of coral reefs.
Donald C. Rundquist is a professor with both the School of Natural Resource Sciences and the Conservation and Survey Division, Institute of Agriculture and Natural Resources, University of Nebraska-Lincoln. He also serves as Director of the Center for Advanced Land Management Information Technologies, a research and development facility focused primarily on remote sensing and geographic information systems. Rundquist holds a Ph.D. in Geography (Nebraska, 1977). His general area of interest, beginning in the early 1970's, is the application of remote sensing to natural-resources issues.
Robert Stark is Ph.D. student with Department of Geological and Environmental Sciences, Ben-Gurion University of the Negev, Israel. He received his M.Sc. degree in Geological and Environmental Sciences in 1997 with a specialization in the remote sensing of surface waters.
Luoheng Han received his Ph.D. in Geography from the University of Nebraska in 1994 with a specialization in the remote sensing. He is Associate Professor of Geography at the University of Alabama, where he teaches courses in remote sensing and GIS. His research interests are in hyperspectral sensing and water quality.
D. Etzion is Ph.D. student with Department of Geophysics and Planetary Sciences, Tel Aviv University, Ramat Aviv, Israel. He received his M.Sc. degree in Geophysics and Planetary Sciences in 1998 with a specialization in the remote sensing.
Abstract Most of the information pertaining to remote sensing of phytoplankton was developed for oligotrophic waters, where detritus and inorganic particles are scarce or their concentrations correlate with phytoplankton density. In our study we concentrated on inland and productive coastal waters, with the initial work done in Lake Kinneret, Israel. The primary objectives were: (1) to study the spectral features of reflectance of different water bodies during different seasons of the year, in order to find spectral features which are closely related to phytoplankton density; and, (2) to devise and validate algorithms for chlorophyll estimation using reflectance data as the measured variables. We found that the information gained from several spectral bands in the red and near-infra-red ranges of the spectrum were sufficient for the construction of algorithms for phytoplankton density estimation. These algorithms were validated in Lake Kinneret, as well as in other environments, with slight modification of the coefficients: the polluted water of Haifa Bay (Mediterranean Sea), fish ponds and wastewater reservoirs in Israel, and lakes with diverse trophic status in northwestern Iowa and eastern Nebraska (USA). Within the context of information essential for the estimation of chlorophyll concentration by remotely operated instruments, we discuss the requirements for satellite sensors to make them expedient tools for monitoring quality of productive aquatic ecosystems.
Introduction The patchy nature of phytoplankton distribution is a challenge to any survey effort on larger water bodies. Therefore, remote sensing techniques, which offer a synoptic view of most or all the surface zones of water bodies in question, are a promising solution. The initial efforts to use remote sensing for phytoplankton monitoring were done in marine pelagic environments, where phytoplankton is often a dominant component of the suspended matter, and the total concentration of chlorophyll a (Chl), a signature substance of phytoplankton is mostly below 1 mg/m3 (Morel and Prieur, 1977;
Gordon and Morel, 1983; Kirk, 1994, Bukata et al. 1995). Most inland waters are classified as Case II waters (Morel and Prieur, 1977); in productive waters, the Chl concentration is often moderate to high and phytoplankton concentration is not tightly coupled to the density of total seston.
Chl concentrations in Lake Kinneret, Israel, range from less than 5 mg/m3 to hundreds of mg/m3 (Berman et al.
1992). The spatial distribution of Lake Kinneret phytoplankton is very heterogeneous, particularly when the dinoflagellate Peridinium gatunense in almost all years forms a dense bloom from February through May (Pollingher, 1986; Berman, et al. 1995). Funding considerations limit the sampling schedule to a few stations, and routinely to a single station, located above the deepest point of the lake. It is questionable whether this single station (or even a few stations) is truly representative the overall condition. To overcome this problem, we have initiated a program for remote sensing of chlorophyll in Lake Kinneret in October 1992. The primary aims were (1) to study the characteristics of the reflectance spectra during different seasons of the year, (2) to evaluate concepts developed for estimation of Chl concentration in inland waters (Gitelson, 1986, 1993a,b), and (3) to devise algorithms for Chl estimation in Lake Kinneret from reflectance data. Following the establishment of algorithms for Chl detection in Lake Kinneret, an effort was undertaken to assess their validity in other productive and turbid water bodies in Israel and in the USA. In this paper, we present the summary of our work in those diverse water bodies, where remotely sensed data were used for the estimation of Chl concentrations.
Materials and Methods
Data were collected in several locations in Israel and in the north central United States. A list of water bodies examined is given in Table 1. Descriptions of optical properties of these water bodies have been published previously: Lake Kinneret (Gitelson et al. 1994a,b; Yacobi et al. 1995), Haifa Bay (Gitelson et al. 1996), Carter Lake (Schalles et al.
1998), northwestern Iowa lakes (Jones & Bachman, 1974), and wastewater treatment ponds (Oron and Gitelson, 1996;
Stark et al. 1996; Gitelson et al. 1997). In each experiment, upwelling radiance of water (Lw) and reference plate (Lref) were measured using high spectral resolution spectroradiometers: LI-1800 (aquatic systems in Israel), Ocean Optics ST1000 (Carter Lake), and ASD (Iowa lakes). Reflectance spectra were then calculated as R = Lw/Lref. Water samples were collected for analytical determination of Chl a concentration in the laboratory. Detailed descriptions of the analytical methods and data processing used for the treatment of the measured variables were presented in the mentioned above publications.
Table 1. Water bodies studied, period of observations, range of chlorophyll variation and dominant phytoplankton.
Reflectance spectra Upwelling radiance of water harbors information on the concentrations and composition of dissolved and suspended substances in the water, and is “the raw material” for remote estimation of concentrations of water constituents. Here we focused on spectral features determined by chlorophyll a absorbance and scattering by phytoplankton cells. All water constituents have significant optical activity in spectral range 400 to 500 nm. It includes absorption by dissolved organic matter and scattering by particular matter, those decrease toward longer wavelengths, and absorption by chlorophylls and carotenoids (e.g., Kirk, 1994). Of special importance is a reflectance minimum near 440 nm (Figs. 1a and 2b), caused by Chl absorption; this feature is used in oligothrophic waters, in a reflectance (R) ratio at 440 nm and 550 nm (R440/R550), to estimate Chl concentration (Gordon and Morel, 1983). The minimum near 440 nm is often indistinct in reflectance spectra of productive waters, due to strong absorption by dissolved organic matter and scattering by particulate matter (Figs. 1b and 2a). In such waters, the reflectance near 440 nm becomes less sensitive (if any) to Chl concentration (Gitelson et al, 1994a; Rundquist et al. 1995; Schalles et al. 1997). In the range near 490 nm another trough of reflectance is seen (Fig. 1a and 2b), caused by carotenoids absorption (Gitelson et al. 1995; Yacobi et al. 1995). In the case of Anabaena the most abundant of this substances of this group were zeaxanthin, myxoxanthophyll, ochinenone and β-carotene (Schalles et al. 1998). In productive waters, as in Haifa Bay (Fig. 1a), Iowa lakes (Fig. 1b), lake Kinneret during Peridinium period (Fig. 2a) and fish ponds (Fig. 2b), reflectance in the range 400 to 500 nm became low with no pronounced spectral features within a broad range of total seston and phytoplankton densities. Thus, all optically active constituents (dissolved and suspended materials) contribute to reflectance in the range 400 to 500 nm, and a common characteristic of reflectance spectra in this range is low sensitivity of reflectance to phytoplankton density. Absorption by pigments is masked by absorption of dissolved organic matter and scattering by suspended matter.
In the range 500 nm to near infra-red (NIR), the reflectance showed several distinct features: (a) a peak in the green range near 550-570 nm, (b) a trough near 625 nm, (c) a trough at 670-680 nm, and (d) a distinctive peak in the red - NIR boundary near 700 nm.
A prominent peak of reflectance in the green range represents the minimal absorption of all algal pigments.
Scattering by non-organic suspended matter and phytoplankton cell walls are factors, governing magnitude of reflectance in the green range (550 to 570 nm). Increase in concentration of particulate matter leads to increase in scattering and reflectance as well. Exceptions were Haifa Bay (Fig. 1a) and Lake Kinneret during Peridinium bloom (Fig. 2a), where non-organic suspended matter load was very low (phytoplankton was responsible for more than 90 percent of dry weight), and phytoplankton was the only constituent determined optical properties of water. Increase in phytoplankton density led to increase of scattering by phytoplankton cells and pigment absorption. Latter one was so strong that reflectance decreased, even scattering by phytoplankton cell increased. This decrease in reflectance appeared not only in the blue range but also in the green at wavelength located far away from main absorption bands of chlorophylls and carotenoids. It was caused by strong absorption by carotenoids, mainly those harbored by dinoflagellates, i.e., peridinin and diadinoxanthin. The position of the peak was dependent on carotenoid concentration as it was found in Haifa Bay (Fig. 1a) and in Carter Lake (Schalles et al. 1998). With increase in carotenoid concentration, absorption increased and manifestation of it was decrease in reflectance and shift of the peak position toward longer wavelength (from 550 to 570 nm, Fig. 1a). In Lake Kinneret during Peridinium bloom, absorption by carotenoids was so strong, that peak position was found at 570 even for Chl concentration as low as 5-15 mg/m3 (Fig.
In waters dominated by blue-green algae in the range of 620-630 nm, a decline of reflectance is noticed (Figs.
2b, and 3), caused by the absorption of the cyanobacteria phycobillins (Dekker, 1993; Gitelson et al., 1995b). The depth of this feature varied seasonally, in accordance with cyanobacter abundance and seasonality (Fig. 3, see also Schalles et al. 1998). Increase in phycocyanin concentration led to increase in depth of the trough and, as a result, to shift of green peak position toward shorter wavelength (Fig. 3). Thus, in waters with blue-green algae, green peak position depends upon at least two factors: carotenoid and phycocyanin concentrations.
Fig. 1: Reflectance spectra of aquatic systems with low to moderate Chl concentrations.
(a) Haifa Bay, June 1995. With increase in Chl concentration to 20 mg/m3, reflectance in the range 400 to 550 nm decreased; for Chl 20 mg/m3, the sensitivity of reflectance to Chl dropped significantly. In the range near 700 nm, reflectance was sensitive to Chl in a wide range of its variation.
(b) Lakes in northwestern Iowa, USA, June 1996. Peak near 700 nm was found to be the only spectral feature sensitive to Chl concentration.
Fig. 2: Reflectance spectra of aquatic systems with moderate to high Chl concentrations.
(a) Lake Kinneret, February 1994. In the range 400 to 550 nm, absorption by pigments was so strong that it was hardly to distinguish between spectra with Chl ranged between 5 and 150 mg/m3. Magnitude and position of the NIR peak depend strongly upon Chl concentration.
(b) Fishponds in the Jordan Valley, Israel, February 1998. Blue green algae were dominant in the ponds; thus, specific spectral feature of this algae can be seen as a gap near 625 nm.
The trough near 670 nm is due to maximum absorbance by chlorophyll a in the red range of the spectrum. For a Chl concentration of more than 20 mg/m3, the reflectance at 670 nm (R670) almost does not depend upon Chl (Fig. 2a) and primarily depends on concentration of non-organic suspended matter (Gitelson et al., 1993a,b; Dekker, 1993;