MERIS satellite chlorophyll mapping of oligotrophic and eutrophic waters in the Laurentian Great Lakes

Abstract

Chlorophyll-a (Chla) concentrations and ‘water-leaving’ reflectance were assessed along transects in Keweenaw Bay (Lake Superior) and in Green Bay (Lake Michigan) (two of the Laurentian Great Lakes, USA), featuring oligotrophic (0.4–0.8 mg Chla m^− 3) and eutrophic to hyper-eutrophic waters (11–131 mg Chla m^− 3), respectively. A red-to-NIR band Chla retrieval algorithm proved to be applicable to Green Bay, but gave mostly negative values for Keweenaw Bay. An alternative algorithm could be based on Chla fluorescence, which in Keweenaw Bay was indicated by enhanced reflectance near 680 nm. Bands 7, 8 and 9 of the Medium Resolution Imaging Spectrometer (MERIS) have been specifically designed to detect phytoplankton fluorescence in coastal waters. A quite strong linear relationship was found between Chla concentration and fluorescence line height (FLH) computed with these MERIS bands. The same relationship held for observations on oligotrophic waters elsewhere, but not for Green Bay, where the FLH diminished to become negative as Chla increased. The remote sensing application of the algorithms could be tested because a MERIS scene was acquired coinciding with the day of the field observations in Keweenaw Bay and one day after those in Green Bay. For Green Bay the pixel values from the red-to-NIR band algorithm compared well to the steep Chla gradient in situ. This result is very positive from the perspective of satellite use in monitoring eutrophic inland and coastal waters in many parts of the world. Implementation of the FLH relationship in the scene of Keweenaw Bay produced highly variable pixel values. The FLH in oligotrophic inland waters like Lake Superior appears to be very close to or below the MERIS detection limit. An empirical algorithm incorporating three MERIS bands in the blue-to-green spectral region might be used as an alternative, but its applicability to other regions and seasons remains to be verified. Moreover, none of the algorithms will be suitable for mesotrophic water bodies. The results indicate that Chla mapping in oligotrophic and mesotrophic areas of the Great Lakes remains problematic for the current generation of satellite sensors.

Introduction

Containing one-fifth of the world's surface freshwater (Wetzel, 2001), the Laurentian Great Lakes provide drinking water, food, recreation and transportation for a growing population of about 35 million presently. On both sides of the US–Canadian border their ecological state is regarded as reflecting environmental health in the heart of the continent. The three largest of these ‘inland seas’, Lake Superior (surface area A = 83 10³ km²; mean depth z̄ = 147 m), L. Michigan (A = 58 10³ km²; z̄ = 85 m) and L. Huron (A = 60 10³ km²; z̄ = 59 m) are also deep, and their enormous volumes involve water renewal times of decades, even up to nearly two centuries in L. Superior (EPA, 1995). Large parts of these lakes therefore appear to be unaffected by regional environmental changes even long after significant demotechnic growth in the drainage basin.

In the vicinity of land the impacts of increased nutrient loading, soil erosion and invasive species may be apparent already in an early stage. The physical, chemical and biological gradients in such regions are highly interesting for mechanistic studies, but also because they are first in signaling environmental changes. These lake areas are typically receiving riverine inputs from hinterlands with cities, industries, forestry and agriculture. Prominent examples are Green Bay, L. Michigan (Auer et al., 1986, De Stasio and Richman, 1998, Millard and Sager, 1994), and Saginaw Bay, L. Huron (Bierman et al., 1984, Budd et al., 2001).

Monitoring of water quality in the Great Lakes of course is strongly facilitated by optical remote sensing, certainly in combination with the study of processes in situ and numerical modeling (Ji et al., 2002, Lathrop et al., 1990, Mortimer, 1988, Shuchman et al., 2006). One of the parameters most quickly responding to environmental change is the ubiquitous phytoplankton pigment chlorophyll a (Chla), which exhibits a unique spectral absorption signature with marked peaks in the blue and red wavebands. In oceanography great progress has been made in the retrieval of this pigment from blue-to-green ratios of remote sensing reflectance, paving the way to estimates of global carbon fixation (Campbell et al., 2002, O'Reilly et al., 1998). In contrast to the clear oceanic waters (case-1 waters), serious Chla retrieval problems have arisen for coastal and inland waters in which the optics are not closely related to phytoplankton (case-2 waters); see Morel et al. (2006) for a recent discussion of case-1 and case-2 marine waters, Bukata et al. (1991) for the example for Ladoga lake, the largest lake in Europe, and Budd and Warrington (2004) and Li et al. (2004) for L. Superior. Indeed, L. Superior, the clearest of the Great Lakes, can be classified as oligotrophic case-2 rather than case-1 (Budd and Warrington, 2004, Gons and Auer, 2004).

Hence, whereas Chla in case-1 waters can be accurately estimated on the basis of the pigment's absorption peak in the blue, the approach does not work for case-2 waters because of high and variable absorption by chromophoric dissolved organic matter (CDOM; otherwise known as Gelbstoff, dissolved humic substances or gilvin) and detritus particles in this spectral region. In oligotrophic case-2 waters, estimation on the basis of the Chla absorption peak in the red can be no alternative due to the overwhelming absorption by water of the red and near-infrared (NIR) wavelength bands. However, this spectral region also features the emission of Chla fluorescence, which may be detected by dedicated satellite sensors (Babin et al., 1996, Gower and Borstad, 2004). The same spectral region is important with regard to eutrophic case-2 waters, where the Chla absorption in the red becomes so strong that Chla can be estimated from reflectance in the vicinity of the Chla absorption peak in the red and in the NIR wavelength band (Dall'Olmo and Gitelson, 2005, Gons et al., 2002).

The Medium Resolution Imaging Spectrometer (MERIS) provides imagery for the spectral radiance in bands centred at 665, 681 and 709 nm. The aim here is to derive the enhancement of reflectance associated with Chla fluorescence in seawater. Fluorescence could be detected with shipboard measurements of spectral reflectance in the oligotrophic Keweenaw Bay, L. Superior (Gons & Auer, 2004). MERIS imagery might therefore be suitable to map Chla in the oligotrophic parts of the Great Lakes. In the present article an empirical relationship between fluorescence line height (FLH) and Chla concentration of the shipboard data is described. The actual MERIS application could be tested for a scene covering part of the Keweenaw Bay at the day where observations were made in situ.

The same MERIS bands used to derive FLH, together with the 779 nm spectral band, are expected to provide reasonably accurate estimates of Chla in eutrophic water (Gons et al., 2002). A previously derived semi-analytic algorithm mainly for European inland and coastal waters (Gons et al., 2002, Gons et al., 2005) was applied to shipboard spectral reflectance in Green Bay, L. Michigan, as well as oligotrophic to eutrophic waters of the Finger Lakes (New York). The results are reported in this paper. The above-mentioned MERIS scene fully included Green Bay, thus a MERIS application for eutrophic Great Lakes waters could be evaluated, even though in this case field observations were made a day earlier.