SCOPE 42 - Biogeochemistry of Major World Rivers

2.1 INTRODUCTION

Recent progress made in remote multispectral radiance measurements and methods of interpretation has opened new ways for monitoring rivers, estuarine and coastal waters (NASA 1986). The major water constituents, whose concentration and distribution can be determined by optical remote sensing are suspended matter, phytoplankton and Gelbstoff, which is a fraction of dissolved organic matter. The distribution of these substances is important for various questions of the marine and estuarine ecology and the biogeochemical budget of an area. Furthermore, they can be used as a natural tracer to investigate the transport of dissolved or suspended substances. Phytoplankton also plays a role as a sink for carbon dioxide, which is an important factor in climate.

Up to now we have only a poor knowledge of the horizontal distribution and temporal variation of these water constituents. Only remote sensing measurements from aircraft or satellite provide us with a synoptic view of larger water areas, and thus may improve the estimation of biomass, dissolved organic matter and the concentrations of inorganic particles. Furthermore, the problem of finding representative locations for measurements from ships and moored buoys can hardly be solved without synoptic maps.

The principle of remote sensing is based on the interaction between electromagnetic radiation and constituents present in the water column. Because of the absorption spectrum of water molecules, only the visible part of the radiation between 400 and 700 nm penetrates deep enough into water for the detection of substances.

For a detection of several water substances, commonly used techniques such as simple colour ratios, although successfully applied for open oceans (Gordon 1980), are not sufficient. The similarity of the spectral scattering and absorption coefficients for all optically active water substances poses problems for finding an adequate procedure for their detection. Consequently, for their detection and separation a more advanced interpretation method is needed. Only the sun-stimulated chlorophyll fluorescence has been found to be a specific spectral feature for phytoplankton within the ocean-leaving spectral range.

A good approach for the development of an appropriate evaluation technique is the determination of the information content of multispectral radiance measurements with either an eigenvalue or factor analysis (Doerffer 1981; Fischer 1985). Following the results of these statistical approaches, an inverse modelling technique for the detection and separation of the most important water substance classes was developed. Such a procedure was successfully applied to multispectral aircraft (Fischer 1983) and satellite radiance measurements (Fischer and Doerffer 1987).

Besides the interpretation of multispectral radiance measurements, taken with radiometers on aircraft or satellite, we also look towards simpler techniques, e.g. aerial photography (Doerffer 1985). For the monitoring of the River Elbe such a method has successfully been applied. From a time series of photographs, taken from a low flying single engine Cessna 172, the horizontal distribution and its variation due to currents were studied.

Furthermore, an accurate correction of the atmospheric path radiance and specular reflected sky radiance is a prerequisite for remote monitoring of water substances, because often more than 80% of the signal received at satellites was derived from atmospheric scattering. Although the determination of the atmospheric path radiances is one of the most critical parts in all water substance detection procedures, we concentrate here on the techniques for the detection and separation of water constituents. An overview of atmospheric correction procedures used for Coastal Zone Color Scanner (CZCS) measurements is given elsewhere (Gordon and Morel 1983).

2.2 SUBSTANCES IN RIVERS, ESTUARIES AND COASTAL WATERS

Suspended matter has numerous components, such as phytoplankton, detritus, inorganic particles, Gelbstoff and bacteria, which vary in type and concentration as well as with location and time. To give an idea of the water substance variability in rivers, estuaries and coastal waters we selected in situ measurements from the River Elbe and the German Bight (Doerffer 1979);

The concentrations of suspended matter in the Elbe reach up to 100 mg/l, while in the turbidity zone amounts of more than 200 mg/1 have been recorded. In the Elbe river plume the concentrations are reduced due to mixing processes with sea water and the reduced turbulence as indicated in Figure 2.1a. In most parts of the German Bight the concentrations are below 10 mg/l.

Chlorophyll concentrations are a function of season and have a minimum in winter months. The horizontal distribution is very patchy, with concentrations varying from below 1 mg/m³ to 30 mg/m³. The relation between chlorophyll and salinity shows no significant correlation, because different populations exist in different zones of the estuary and coastal area, while the concentration of phaeophytin reaches a maximum within the turbidity zone where most of the fresh water phytoplankton dies (Figure 2.1b).

Gelbstoff is mainly transported by rivers into coastal waters. Due to the dilution with sea water it is well correlated with salinity. Since Gelbstoff is chemically more stable it can be used as a measure of diffusion of river water into coastal areas. The absorption coefficients at l = 380 nm vary between 3 and 5 m for the Elbe, with a steady decline to 0.5 to 1.0 m in the German Bight (Figure 2.1c).

The classification after Morel (1980) into ocean water, which contains mainly phytoplankton ('case 1 water'), and waters in which phytoplankton, Gelbstoff and suspended matter are present in not covarying concentrations ('case 2 water'), is useful for the development of remote sensing procedures. The compression of all water substances into only three groups, namely phytoplankton, suspended matter and Gelbstoff, although the specific absorption of estuarine phytoplankton is rather variable, is based on measurements made by Doerffer (1979) for the Elbe Estuary and German Bight and by Prieur and  Sathyendranath (1981) for the Mediterranean and the Atlantic Ocean. These substances are generally considered to be responsible for significant modifications observed in the absorption and scattering properties of estuarine and coastal waters.

The determination of the absorption coefficient and the scattering function of these three groups is very difficult. Thus an adequate choice of optical parameters for each radiative transfer model is a serious problem. The optical properties used here stem from real measurements as far as possible, while missing parameters, such as the scattering function, are estimated from Mie theory, for which the particle size distribution and the complex refractive index are needed.

The size of algal cells often surmounts the wavelengths of visible light. Most of the total suspended matter particles have sizes (r) less than 1 µm and are in the range 0.1 < r < 1.0 µm (Paltridge and Platt 1976). The scattering parameters of both particle groups have therefore to be determined using Mie theory, which is exact only for homogeneous spheres, but may be applied to irregular particles in the ocean at least for deriving integrated parameters like extinction coefficients.

The absorption may be expressed by the imaginary part of a complex refractive index: m = n_r - in_i. According to Gordon and Morel (1983), phytoplankton scatters similarly to suspended matter, but with a stronger absorption. We used a mean absorption coefficient relating to a mean imaginary part n_i with 0.007 < n_i < 0.01. The real part nr lies in the range 1.02 < n_r < 1.08 (Bricaud et al. 1979; Reuter 1980) and depends on the outer shell composition and the size of a cell.

The scattering or phase functions depend strongly on the complex refractive index and the size distribution of the particles. Since only a few measurements of particle size distributions in the ocean exist and since no generally accepted measured size distribution types have evolved, the distribution is approximated here by a junge size-distribution (Fischer and Grassl 1984). We will now discuss problems which are specific for each of the three water substance groups.

2.2.1 PHYTOPLANKTON

The optical properties of plankton vary with species and their physiological state, as for example the chloroplast distribution within a cell. A mean spectral absorption spectrum for chlorophyll a is taken from Prieur and Sathyendranath (1981) (Figure 2.2), who analysed a set of 80 measured spectral curves.

A serious problem is the relationship between the absorption coefficient and the concentration of chlorophyll a. There is a strong disagreement between authors with respect to the specific absorption coefficients a* at l = 440 nm in m^-l (mg/m³)^-1 (per unit 'pigment concentration'). The published values range from 0.168 for chlorophyll concentrations ( £ 1 mg/m³ (Smith and Baker 1978) to 0.018 for ( ³ 1 mg/m³ (Prieur and Sathyendranath 1981).

It is difficult to estimate whether these differences are due to actual variations in the absorption characteristics of phytoplankton or to the measuring technique itself. We use measurements from Doerffer (1979), taken in the North Sea and Elbe estuary, which result in the desired relation between pigment concentration , specific absorption coefficient a* and absorption coefficient a^P :

(2.1)

Figure 2.1 Concentration or attenuation versus chloride concentration (per thousand) (Doerffer 1979) for: (a) suspended matter, (b) chlorophyll and phaeophytin , (c) Gelbstoff

Figure 2.2 Normalized absorption coefficients for chlorophyll (¾) and yellow substance ( ...) as well as normalized extinction for suspended matter (- -)depending on wavelength l; the reference wavelength is l = 440 nm; the wavelengths used in the radiative transfer model are also indicated (Fischer et al. 1986)

Before this or any other algorithm can be used for other environments additional in situ measurements must be performed.

2.2.2 SUSPENDED MATTER 

All the non-chlorophyllous particles of biological, benthic and/or fluvial origin are summarized as 'suspended matter'. The presence of suspended matter is highly dependent on river outflow, biological by-products, the ocean bottom and turbulence, and even the circulation of the ocean itself. Considering the variables involved, no general set of optical properties for suspended matter can be expected. The number of experiments designed especially for measuring suspended matter concentration and its optical properties are small. A general relationship between absorption or extinction coefficients and suspended matter concentrations, such as outlined in literature, cannot be derived. Prieur and Sathyendranath (1981) use the scattering coefficient and Doerffer (1979) the diffuse extinction coefficient (measured in a special cuvette) to define a relationship between extinction coefficient C^S and suspended matter concentration (S). Measurements of the latter author are taken to find the following relation, where b* is the specific extinction coefficient which has been proven as valid for a wide range of different waters in the North Sea and Elbe estuary.

(S) = b* CS = 8 mg/m2 CS(l = 550 nm); (S) in mg/l

(2.2)

The wavelength dependence of the extinction coefficient of suspended matter also originates from measurements taken in the German Bight (Figure 2.2) (Doerffer 1979). The absorption of suspended matter as used here considers a weak absorption of inorganic material and a certain absorption of non-chlorophyllous particles of biological origin. This leads to a single scattering albedo, w_o = b/c, defined as the ratio of the scattering and extinction coefficient, of 0.92.

2.2.3 GELBSTOFF 

The term 'Gelbstoff' or yellow substance (Kalle 1961) is used for characterizing a mixture of highly polymerized organic substances showing a yellow colour when dissolved in water. It is mainly composed of degradation products of organisms. A further characteristic is its chemical stability. From the point of view of optical oceanography, Gelbstoff is that water constituent, which passes a filter with a pore size of < 450 nm and strongly absorbs radiation in the UV and blue range while it scatters light like water molecules.

In marine environments the Gelbstoff absorption spectrum is observed to be nearly constant. The exponential decrease of the absorption coefficient from the blue to the red part of the visible spectrum (Figure 2.2) is given by

This expression is valid for various ocean areas, in which a^(YS)(lo) is the absorption coefficient at any fixed wavelength lo between l = 280 and l = 450 nm (Hojerslev 1979). Experimental results from Hojerslev (1979), Bricaud et al. (1979) and Prieur and Sathyendranath (1981) indicate that the exponent in Equation (2.3) varies up to ±20% .For estuarine water (Elbe), Doerffer (1979) found an exponent of -0.011.

2.3 MULTISPECTRAL RADIANCES: MEASUREMENTS AND SIMULATIONS 

The upward directed radiance spectrum¾as recorded by a remote sensor at high altitudes over water areas¾is determined by a large number of scattering and absorption processes which occur in the atmosphere, at the sea/air interface, in the water body and at the sea bottom. The optical processes within the water column are the link between the concentration parameters, which we want to derive, and the radiance spectrum measured. All other processes, particularly those which happen in the atmosphere, perturb the signal. Thus, the problem of any procedure in evaluating a radiance spectrum with respect to sea water constituents comprises two parts: (1) we have to separate the signal of the water column from all perturbations and (2) we have to retrieve the concentration of water constituents from the corrected radiance spectrum. Since the overall number of independent variables which determines the radiance spectrum theoretically exceeds the number of independent parameters which can be derived from the spectrum, it is necessary to investigate, for realistic conditions, if some of the parameters can be assumed as constant.

For this task we have analysed a set of data comprising measured radiance spectra, concentrations of sea water constituents as well as parameters of the atmosphere and sea surface. Because of the difficulties involved in measuring all the necessary optical parameters from a ship and because of the limited number of in situ measurements available, such a data set has to be incomplete. Thus, we have also used radiative transfer calculations to simulate the required complete set of radiance spectra. These calculations are based on realistic substance concentrations and their variations, as well as realistic boundary conditions of sun elevation, atmospheric turbidity and sea surface roughness.

The use of simulated spectra for investigating the information content and for deriving evaluation procedures requires a validation with measured spectra. The main data set used for such a validation was collected during the Marine Remote Sensing Experiment in the North Sea (MARSEN) (Doerffer et al. 1982), an area with a large variety of different water bodies in 1979.

Figure 2.3 Measured (--)and calculated (-) multispectral radiances just above the ocean surface for (Fischer et al. 1986): (I) chlorophyll concentration (P = 16 mg/m³, suspended matter concentration (S = 11.8 mg/l, yellow substance absorption a^(YS) = 2.0 m^-l and solar zenith angle Q_o = 60° ; (II) (P = 8 mg/m³, (S = 2.5 mg/l, a^(YS) = 1.5 m^-1, Q_o = 45° ; (III)  (p = 7 mg/m³, (!5 = 1.9 mg/l, a^(YS) = 1.9 m^-l, Q_o = 45°

We found three types of spectra (Figure 2.3): chlorophyll a amount varied between 7 and 16 mg/l, suspended matter between 2 and 12 mg/l and Gelbstoff absorption coefficients between 1.0 and 2.5 m^-1. These values and actual sun elevation have been used as input parameters for the radiative transfer model. The influence of the atmosphere and the sea surface was approximated by observations of the horizontal visibility and the sea state. Figure 2.3 also depicts a comparison between three typical radiance spectra, actually measured and those calculated for 16 wavelengths according to the spectrometer used. The good agreement between both measured and calculated radiances encouraged us to use also calculated spectra for the discussions of the following paragraphs.

2.3.1 INFORMATION CONTENT OF RADIANCE SPECTRA

We saw in Section 2.2 that the scattering and absorption spectra of the water constituents are similar and that only chlorophyll and chlorophyll-like pigments have a specific signature between 650 and 700 nm due to absorption and fluorescence. Thus we have to ask, if it is possible to identify and to determine the concentration of all three substances from the radiance spectrum. To answer this question we analysed a set of calculated radiance spectra by eigenvalue and factor analysis. The advantage of using simulated spectra is that the independent variables¾in our case, chlorophyll, suspended matter and Gelbstoff concentration¾can be kept uncorrelated. Covariances between the concentrations¾as is often the case in natural waters¾would prevent us from calculating the maximum possible number of independent variables which are retrievable from multispectral radiance measurements by eigenvalue and factor analysis.

However, since the specific optical coefficients of all three substances have to be kept constant in a model because the natural variations are unknown, the variability may be underestimated compared to measured spectra where the coefficients per unit concentration may vary, particularly those of different phytoplankton populations.

We will now try to show the ideas, underlying the use of eigenvalue and factor analysis for our task, without going into details. Both techniques are given in Fischer (1985) and Fischer et al. (1986).

Both techniques are based on an analysis of the covariances of a set of parameters, which in our case are the radiances at n different wavelengths. Eigenvectors are calculated as a series of n vectors, orthogonal to each other , within the n-parameter space achieving a maximum variance with a minimum number of eigenvalues. The eigenvalues indicate how much each parameter contributes to the total variance of the analysed data set. Thus, the number of independent parameters is equal to the number of eigenvalues exceeding the error variance of the radiance measurements. If the error variance is not well known, the ranked series of eigenvalues can be used as a guide to retrieve the number of independent significant variables and also to separate these ranked eigenvalues from those which are determined by data noise. In the ideal case, eigenvectors with high eigenvalues would indicate¾as in our case¾independent factors, which determine the variance of a set of radiance spectra; they should be clearly separated from those eigenvectors with low eigenvalues which represent variance due to noise or random errors of the spectra.

Figure 2.4 Eigenvalues of radiance spectra measured along a 90 km flight profile over the North Sea in order of their variance (Doerffer 1981)

An example given in Figure 2.4 may help to clarify the idea of this technique. It is calculated from a set of about 2000 spectra sampled along a flight track through the North Sea with a 17-channel multispectral radiometer (Amann and Doerffer 1983). The eigenvalue series reveals that two eigenvectors are clearly separated from the rest. After suppressing data noise by a moving average filter this separation is even enhanced. However, the eigenvectors found are only a formal representation of the data.

Factor analysis goes one step further. With this technique one tries to search for factors which are really present in the physical system and tries to analyse the relationship between these factors and the variables measured, the latter in our case radiances at different wavelengths. The result is a limited number of factors, which can be interpreted as factors which determine the variance of the radiance spectra as for example substances in water or atmospheric aerosol. The relationship between these factors and the radiances at a single wavelength is expressed as factor loading. Using these loadings one can calculate furthermore, a quantity, the factor score, which expresses how each factor determines a single sample (spectrum). A correlation between factor scores and concentration should finally prove whether the factors are really identical with chlorophyll or suspended matter concentration.

Table 2.1 Correlation coefficients between factor scores P1-P3 and phytoplankton, suspended matter and Gelbstoff

For our simulated set of MARSEN spectra (16 wavelengths, including a chlorophyll fluorescence channel at 685 nm) we calculated the factor scores for each simulated sample, assuming three significant factors, and correlated these scores with the concentrations of phytoplankton chlorophyll, suspended matter and Gelbstoff. The result is summarized in Table 2.1 (Fischer et al. 1986). All three factors are significantly correlated with the concentrations. We may conclude from this analysis that the spectrum contains information on three substances. The next section addresses the problem of how to retrieve concentrations from multispectral measurements.

2.3.2 DETECTION PROCEDURES

The first method used to estimate remotely the chlorophyll concentration in the sea evaluates the green/blue colour ratio, e.g. using 550 nm and 440 nm radiances (maximum of chlorophyll backscattering and absorption). It was used by Arvesen et al. (1973) and later also successfully applied to CZCS data of open ocean areas (see overview in Gordon et al. 1983). However, this simple technique applies only to areas where only phytoplankton is present or where other substances, such as organic detritus, are correlated with phytoplankton. In coastal areas with high amounts of inorganic suspended matter and Gelbstoff the green/blue ratio is also modified by these substances; in this instance this technique should no longer be applied for chlorophyll estimates. Ratios using more than two channels may produce significant correlations with in situ concentrations of all three substances, but they are in most cases only valid for a specific investigation. Thus, we have to look for other procedures. One possibility is to extract the sunlight-simulated chlorophyll fluorescence from the radiance at 685 nm. This technique was first applied by Neville and Gower (1977) when measuring.the radiance spectrum under the Brewster angle with a polarizing filter in order to suppress specularly reflected skylight. Doerffer (1981) found a high linear relationship between the fluorescence parameter and the chlorophyll concentration even with an airborne spectrometer pointing to the nadir and without a polarizing filter (Section 2.4.2).

  Here the fluorescence line height at  l_F = 685 nm above a linear base line fixed at  l₁ = 645 nm and l₂ = 711 nm, is used to analyse multispectral radiances

with a modification (F2) if the base line is formed by l₁ = 645 nm and l₂ = 670 nm, thus accounting for the absorption of chlorophyll at l = 670 nm and subsequently extrapolating to l_F .

The advantage of the fluorescence line height for the retrieval of chlorophyll in coastal waters is demonstrated by the different use of radiance spectra, as given in Figure 2.5. Colour ratio and fluorescence line height (F2) are compared in Figure 2.6a and 2.6b. While the scatter of colour ratios shows no clear dependence on chlorophyll concentration, the fluorescence line height predicts well the chlorophyll.

The detection procedures as given above are restricted to the retrieval of only one parameter, here the chlorophyll concentration. In some cases the colour ratio technique is applied to the detection of suspended matter, but this is only useful where suspended matter is the dominating water substance.

Following the results of the eigenvalue analysis (see Section 2.3.1), three independent quantities may be derived from the four visible multi spectral  CZCS measurements (Fischer 1985). However, this can only be achieved if the atmospheric turbidity is low and errors due to the atmospheric correction process are below 2%.

Figure 2.5 Calculated multi spectral radiances just below the ocean surface for coastal water according to substance variations as found in the North Sea (Fischer et al. 1986)

For the detection and separation of water substances from multispectral radiances we choose an inverse modelling technique, based on a radiative transfer model for ocean plus atmosphere. The basic idea is to minimize an error function, here the c²-function. Input for this function consists of measured and simulated multispectral radiances. The optical properties, such as absorption and extinction coefficients of suspended matter, chlorophyll and Gelbstoff, are varied until an optimal fit of the radiative transfer model radiances with measured radiances is reached. Of course, any result of such a procedure is, at best, as good as the description of the physical processes in the radiative transfer model. Furthermore, if relying on simulations only, a sufficiently accurate set of optical properties of water substances is obligatory and this set has to be adapted to different water areas. The principles of this inverse technique are described in detail elsewhere (Fischer 1983).

Figure 2.6 (a) Blue-green ratio depending on chlorophyll concentration; (b) Fluorescence line height (F2) using the chlorophyll absorption at l = 670 nm depending on chlorophyll concentration; input data are radiances from Figure 2.5

2.4 AIRCRAFT DATA APPLICATIONS

Satellite remote sensing has unveiled spectacular insight into the horizontal structure of the ocean and its dynamics. However, for some questions and applications the operation of an aircraft can provide measurements, otherwise not obtainable, for surveying the sea:

• whenever the temporal and spatial scales of mainly tidal driven processes, as present in estuaries, require a resolution of minutes or hours or an extremely high geometric resolution;
• in cases where a remote survey is necessary during a multiplatform experiment at a specific time in areas where cloud coverage is a problem and the research strategy allows for aircraft operations in low altitudes below clouds;
• in areas where cloud-free scenes are restricted to specific seasons or weather situations, and where the analysis of only these scenes would lead to conclusions, which are not valid in general;
• when the interference of the atmosphere has to be kept small in order to yield the required radiometric accuracy;
• when the survey has to address an event such as a pollution accident.

  In order to demonstrate the potential of airborne remote sensing two examples will be discussed here:

1. A study of the suspended matter dynamics in the Elbe estuary with aerial photography where a high temporal and spatial resolution is required.
2. A study of the development of a spring plankton bloom with airborne radiometers for one month under cloudy conditions.

2.4.1 AERIAL PHOTOGRAPHY AS A SIMPLE REMOTE SENSING TECHNIQUE TO STUDY SUSPENDED MATTER DYNAMICS

The horizontal suspended matter distribution of rivers and estuaries usually shows very complex patterns in the form of large 'fields' and 'clouds', which are growing, moving and declining during the tidal cycle as a function of the tidal current velocity and the turbulence. Because of the small-scale variations in time and space it is impossible to get synoptic maps of a momentary distribution from ships. However, such maps are critical, e.g. to calculate transport vectors and the suspended matter budget of a river section. Even the problem to find representative locations for measurements from ships and moored buoys can hardly be solved without a series of synoptic maps.

One simple method for mapping the suspended matter distribution in estuaries is aerial photography (Doerffer 1985). The example we discuss here is from a field experiment in the upper Elbe estuary (BILEX 82), which had the task of estimating with model calculations the transport of suspended material through a 1 km section during a tidal cycle (Michaelis 1983). The specific goal of the aerial survey was to study the short-term, tide-dependent variability of suspended matter distribution within a 1 km² section of the estuary. These data were also used to determine how representative ship- borne data are for estimating suspended matter distributions.

The method is simply based on the fact, that suspended particles backscatter solar radiation (see Section 2.3) and thus increase the density on a photographic film. Investigations of the optical conditions of the Elbe estuary by Doerffer (1979) have shown that the radiance at wavelengths in the red part of the spectrum (around 650 nm) is a non-linear function of mainly suspended matter concentration, while Gelbstoff and phytoplankton pigments are of minor importance in this spectral range.

Thus one has to consider as the three main light sources responsible for the density of the red sensitive layer of a true colour film: (1) the light backscattered by suspended matter in the river; (2) the sun or skylight specularly reflected from the water surface; and (3) the light which is scattered by the atmosphere.

The main reason why we selected a camera with photographic film as the remote sensor is the high geometric resolution obtainable, which is necessary for resolving single suspended matter clouds. Furthermore, this method requires only a light aircraft and thus reduces the costs as compared to those invoked in the operation of an optoelectronic scanner.

The camera used is a remotely controlled Hasselblad with a Zeiss Distagon  f = 30 mm wide-angle lens using 70 mm Kodak Ektachrome true colour reversal film. The camera was mounted vertically outside the luggage door of a single engine Cessna 172. The wide-angle lens was chosen such that sufficient landmarks could be included from altitudes of 800-1000 m for the geometric correction.

Figure 2.7 Geometrical correction of aerial photographs: (a) Original raw image; (b) after reprojection from an angular to a distance proportional (normal) image and after correction of aircraft altitude; (c) after scale adjustment and projection into a map of scale 1 : 25 000; the film density is calibrated in terms of suspended matter concentration

The same area could be photographed with an interval of 5- 7 minutes due to the turn around time of the aircraft. Sequences with this interval were taken over 30-90 minutes during different phases of the tidal cycle.

The prerequisite for a quantitative analysis is a digitization of all three colour bands (colour film layers) of the images. We used an Optronics C 4500 film-scanner and a scan resolution of 0.1 mm x 0.1 mm.

One particular problem is the geometric correction of the images since the Distagon lens has an angular proportional projection (fish eye type). The correction was performed digitally on a computer and includes three parts: the conversion of the projection; the correction of camera attitudes not  parallel to the ground; and an adjustment of the required scale. Figure 2.7 shows the results of the individual steps: Figure 2.7(a) is the original raw image; Figure 2.7(b) shows the result after reprojection from an angular to a distance proportional (normal) image and after correction of aircraft altitude; and Figure 2.7(c) after scale adjustment and projection into a map of scale 1:25 000. In this latter map each pixel represents an area of 5 m x 5 m.

For deriving concentrations from film densities a procedure was developed and successfully tested in the form of a proof-of-concept experiment with one image- Therefore, the film density has to be calibrated by photographing a set of grey plates with diffuse reflectivities of 2,4,6 and 11% before each flight.

However, for studying the dynamics of the suspended matter distribution it turned out that raw images are sufficient in many cases; thus, in the following paragraph, we will discuss how a series of just raw images can help to understand the dynamics of the suspended matter transport.

In order to get an overview of the horizontal variability within a one-hour period, we have selected a series of photographs which were taken during an increasing ebb current with intervals of 6 minutes. The area comprises a section of the Elbe River with a depth of 10-15 m and the mouth of a shallow side arm with a depth of about 2-3 m. The images (Figure 2.8) shown here are without geometric correction in order to retain the fine structure in suspended matter clouds. In the first image (Bilex #25) of the series the main stream (fairway) has low concentrations while in the shallow areas near the banks and within the side arm the surface concentration is high and separated by sharp fronts. In the following images one can observe that the fronts of both sides are moving into the middle of the stream; finally, after only 50 minutes, nearly the whole river shows high concentrations at the surface. Observing this rapid development, one can imagine that it is hardly possible to reproduce the true horizontal distribution by observations from either an anchored or profiling ship.

Figure 2.8 Aerial photographs showing the development of suspended matter distribution in the Lühesander Süderelbe during the ebb phase; photographs are without geometric correction

 Figure 2.9 Schematic description of the development of a suspended matter front due to the upward directed current component induced by an underwater slope

Furthermore, the following phenomena detected in the images are of interest: the suspended matter front is moving from the side arm into the Elbe against the ebb current. This phenomenon can be explained in the following way (Figure 2.9): the suspended matter field observed in the side arm is not static and is not moving by advection but is formed by a steady flow of material which is transported to the surface by the upward directed current component. With increasing ebb current and descending water level the front where the high concentrations appear at the surface is moving towards deeper areas, which in this case is against the current.

Suspended matter clouds also appear behind steaming ships and are forming long bands of clouds with a scale similar to that of naturally formed clouds (Figure 2.8, Bilex #44). We must therefore assume that turbulences induced by the ship and the ship propellers transport suspended matter from deeper layers up to the surface.

Furthermore, we can observe that¾in accordance with turbulence theory¾suspended matter clouds are smaller over shallow water and larger over deeper water.

Summarizing these observations we have to assume that suspended material is transported in small clouds. The main factor controlling the horizontal distribution is the current and turbulence regime which is controlled by the bottom topography.

2.4.2 AERIAL SURVEY OF THE DEVELOPMENT OF A SPRING PHYTOPLANKTON BLOOM

Another example demonstrating the potential of aerial survey is the Fladenground Experiment FLEX 76 (Amann and Doerffer 1983). The objective of this international research programme was to investigate the hydrographical, chemical and biological interactions which determine the development of the spring plankton bloom, which is the most important event in the biogeochemical cycle in the northern oceans. The task of the aerial survey was to provide a time series of maps of chlorophyll and sea surface temperature distribution within a 100 km x 100 km area in the northern North Sea. During the one-month period of the aircraft operation, only one day was cloud free. This further illustrates the advantage of using plane-borne techniques for measuring from below clouds (i.e. 200 to 500 m altitude) and where satellite observations are impossible. 

Figure 2.10 Horizontal profiles of chlorophyll concentration along the flight pattern in the FLEX box estimated from the radiance measurements (Amann and Doerffer 1983) (note that scale on right column is valid for all columns)

Figure 2.11 Development of the spring plankton bloom; the line represents the in situ data at the central station, the dots show the corresponding aircraft values and the bars indicate the variability in the FLEX box as detected from the aircraft (Amann and Doerffer 1983)

For the determination of the chlorophyll concentration, multichannel radiometers were used, These recorded the radiance spectra in the visible along a flight track, which was organized in a pattern of six parallel lines, each 15 km apart (Figure 2.10).

Figure 2.11 shows the development of the phytoplankton bloom in terms of the chlorophyll concentration as measured from the ship which was moored in the centre of the research area. The dots indicate the concentration derived from the radiance reflectance ratio R(560)/R(445), measured when passing the ship. The bars represents the variability of all six profiles during a flight.

During this experiment another method (already discussed in Section 2.3) was also tested to derive chlorophyll concentration, It uses the sunlight- stimulated chlorophyll fluorescence as a measure of the chlorophyll concentration. The fluorescence is remotely measured as the radiance at 685 nm- the fluorescence radiance peak-subtracted from the baseline radiance, which is the radiance at 685 nm without the fluorescence calculated from neighbouring channels at 645 and 725 nm, Figure 2.12 shows the relation between the fluorescence derived with this method from an altitude of 600 m and the chlorophyll concentration derived from water samples at 2 m depth,

Figure 2.12 Calibration of the fluorescence measurements with sea truth chlorophyll values (Doerffer 1981)

Figure 2.13 The horizontal chlorophyll distribution in the FLEX box derived from airborne fluorescence measurements (Amann and Doerffer 1983)

Both data sets were collected along a 90 km-long profile on the same day. Using this regression the horizontal chlorophyll distribution was calculated and is shown in Figure 2.13.

Although the sampling strategy is similar to ship profiles and does not produce the high resolution in both horizontal dimensions as present in satellite data, it has the advantage of a quasi-synoptic view since the whole area, i.e. the track system of 650 km length, could be covered within 3-3.5 hours and the survey could be repeated every day.

2.5 SATELLITE DATA APPLICATIONS

The advantage of satellite measurements is the observation of large areas at nearly the same time. From Coastal Zone Color Scanner (CZCS) measurements spectacular observations of the horizontal distribution of ocean colour and also of ocean dynamics are obtained. CZCS on board the experimental Nimbus 7 satellite was the only instrument in space, which had been especially designed for the monitoring of coastal waters. Images were recorded between August 1978 and mid-1986.

For the monitoring of rivers and estuaries the spatial resolution of 1 km² of the CZCS is usually not sufficient. Furthermore, the low frequency of overpasses (every three days) limits the applicability of these satellite measurements for the observation of processes which are strongly time dependent.

The present lack of ocean colour measurements can partly be replaced by Thematic Mapper (TM, on board the Landsat 5 satellite) measurements. The high spatial resolution of 30 m favours this instrument for the monitoring of rivers, estuaries and even coastal waters. However, the limitations imposed by the spectral and radiometric resolution of TM reduce the detection of water substances, even though these waters are often contaminated with high suspended matter concentrations, which lead in turn to high values of backscattered radiation. From in situ measurements and additional radiative transfer calculations we expect a resolution of roughly 5 mg/l. Only an averaging over several pixels, which thus reduces spatial resolution, enhances the signal-to-noise ratio, for example for 300 m x 300 m by a factor of 10.

To demonstrate the potential of ocean colour measurements from space we chose a CZCS image of the North Sea from orbit 4384 (6 September 1979). During the NIMBUS 7 overflight the international Marine Remote Sensing Experiment, MARSEN, took place. Although the measurements taken cannot be used for a direct comparison with the derived water substances from CZCS, we have information on possible variations of phytoplankton, suspended matter and Gelbstoff in the German Bight at this time. For the atmospheric correction, additional aerosol measurements from the North Sea Forschungsplattform were used.

As already discussed in Section 2.3.2 only an inverse modelling technique enables us to estimate and separate suspended matter, phytoplankton and Gelbstoff by multi spectral measurements. Each pixel of a CZCS image is represented by four spectral radiances, to which this technique is applied.

First, we discuss results applying an atmospheric correction supported by ground-based aerosol measurements. A true colour composite of the raw data of this CZCS imagery is shown in Figure 2.14a. Obviously bright layers cover more than half of the scene. Since they cross the coastlines, we have to interpret these features as atmospheric haze. Consequently, for a successful correction of the atmosphere this phenomenon must be removed. As Figure 2.14b shows the successful correction, the larger scale phenomena no longer exist, but the small-scale features in the ocean are still present.

This true colour composite (Figure 2.14b) with contrast enhancement for the remaining spectral water-leaving radiances mainly displays radiances in the green and red aligned along the 30 m depth contour. The band composed of small-scale eddies with bright features is approximately 24 km wide. The high turbidity in the inner German Bight-in the lower right corner of Figure 2.14¾depicts the plume of the Elbe River.

Due to computer time constraints the inverse modelling technique is only applied to three lines of the shown CZCS image, as indicated in Figure 2.14b.

High suspended matter concentrations are found along the coast, especially within the Elbe and Weser River plumes (Figure 2.15a). Values between 15 and 25 mg/1 agree well with in situ measurements in this area. The high suspended matter concentrations, detected northwest of the Netherlands, also correspond well with in situ measurements (Eisma 1981). For the darker areas in the image suspended matter concentrations between 3 and 5 mg/1 are retrieved. They also agree with the in situ measurements of the MARSEN experiment. The processes, which cause the enhanced suspended matter concentrations of line 242, are not fully understood. Probably the circulation within the southern North Sea and the inflow through the English Channel are responsible for the suspensions.

The simultaneously retrieved chlorophyll concentrations are distributed in small-scale patches, varying horizontally between 2 and 15 km (Figure 2.15b). Along the coast (line 188) in comparably dark areas we find the highest amounts of chlorophyll, ranging up to 10 µg/l.

The horizontal distribution of Gelbstoff absorption (Figure 2.15c) agrees with in situ measurements: an increase in the concentration of Gelbstoff absorption is correlated with the outflow of the Elbe River. However, the small-scale patches, as found on line 188, are probably due to the difficulties in separating chlorophyll and Gelbstoff using CZCS measurements. The

Figure 2.14 CZCS measurements from Orbit 4384 (6 Sept. 1979); southeastern part of the North Sea: (a) true colour composite of raw data; (b) true colour composite of water leaving radiances after atmospheric correction. The inverse modelling technique is applied to the three indicated scan lines (Fischer and Doerffer 1987)

retrieved Gelbstoff absorption coefficient, outside the influence of the River Elbe plume, is about a^YS = 0.3 m^-l at l = 380 nm and this agrees with the later measurements by Reuter (1980).

This CZCS interpretation demonstrates the potential of ocean colour measurements from space especially for coastal waters. However, for the detection and separation of water substances from CZCS measurements, here the inverse modelling technique, based on a complete radiative transfer model, has to be simplified, in order to process full images with an acceptable effort. Furthermore, for the atmospheric correction additional aerosol measurements are needed, if unclear ocean areas are present in the satellite image. Looking forward, new instruments measuring also the chlorophyll fluorescence and aerosol properties will surely improve the detection of water substances, especially over coastal and estuarine waters (Doerffer and Fischer 1987).

2.6 SUMMARY AND CONCLUSIONS

Remote sensing of water substances has been shown to be a very useful tool for the monitoring of rivers, estuaries and coastal waters. The spatial and temporal variations of water constituents cannot be evaluated with in situ ship measurements alone. The interpretation method of remote sensing measurements strongly depends not only on the geographical area but also on properties under survey.

Figure 2.15 Retrieved water constituents from subsurface radiances of CZCS, for the lines indicated in Figure 2.14(b), pixels are counted from left to right (Fischer and Doerffer 1987) : (a) suspended matter concentrations in mg/l; (b) phytoplankton in µg/l; (c) Gelbstoff absorption coefficient in m^-l

Aircraft measurements are especially suitable to observe small-scale processes which may require a temporal resolution of minutes or hours. Even with simple colour photographs, suspended matter concentrations are detectable within acceptable errors when high suspended matter and Gelbstoff concentrations are present. With these aerial photographs suspended matter distribution and its temporal variations can be observed and used for dynamical studies. For the separation of different water constituents, however, multispectral radiances are needed.

Satellite measurements provide us with a synoptic overview of horizontal distribution of water colour over large areas. The interpretation of CZCS measurements using an inverse modelling technique is shown to be successful for evaluating suspended matter, phytoplankton and Gelbstoff concentrations of coastal waters. However, the errors for correcting the intervening atmosphere and the calibration errors have to be below 5%. For the monitoring of rivers and estuaries TM (Landsat 5) measurements are the only source of data existing today which provide the necessary spatial resolution. However, the radiometric and spectral resolution of this instrument only allow a detection of suspended matter in high concentrations, since only 5 mg/l variations can be resolved.

Future satellite missions promise new opportunities for the monitoring of rivers, estuaries and coastal waters. The present state of radiometer technology offers the possibility of higher spectral and spatial resolution. However, such instruments, as proposed for the Earth Observation Platform, will not be in space before the mid-1990s.

REFERENCES

Amann, V. and Doerffer, R. (1983) Aerial survey of the temporal distribution of phytoplankton during FLEX '76. In: Sündermann, J. and Lenz, W. (Eds) North Sea Dynamics, Springer Verlag, Berlin, Heidelberg, pp. 517-29.

Arvesen, J. C. , Millard, J. P. and Weaver, E. C. (1973) Remote sensing of chlorophyll and temperature in marine and fresh waters. Astronautica Acta 18, 229-39.

Bricaud, A., Morel, A. and Prieur, L. (1979) Spectral analysis of absorption by yellow substance. In: Symposium on Radiant Energy in the Sea. IUGG, IAPSO programme, pp. 71-2.

Doerffer, R. (1979) Untersuchungen über die Verteilung oberflächennaher Substanzen im Elbe-Ästuar mit Hilfe yon Fernmeßyerfahren. Arch. Hydrobiol. 43, 119-224.

Doerffer, R. (1981) Factor analysis in ocean color interpretation. In: Gower, J. F. R. (Ed.) Oceanography from Space, Plenum, New York, pp. 339-45.

Doerffer, R. (1985) Observations of the suspended matter distribution dynamics in the Elbe Estuary from time series aerial photographs. Int. Rev. Ges. Hydrobiol. 70, 127-50.

Doerffer, R. and Fischer, J. (1987) Sun stimulated chlorophyll fluorescence within a daily cycle. In: The Use of Chlorophyll Fluorescence Measurements from Space for Separating Constituents of Sea Water. ESA Contract No. RFQ 3-5059/84/NL/MD Vol.2.

Doerffer, R., van der Piepen, H. and Amann, V. (1982) The marine remote sensing experiment in the North Sea (MARSEN): a survey of phase 1 (Aug.-Sept. 1979). In: First Int. Conf. Meteor. and Air/Sea Interaction of the Coastal Zone, 10-14 May 1982. Amer. Meteor. Soc., Boston, Mass., USA.

Eisma, D. (1981) Supply and deposition of suspended matter in the North Sea. Spec. Publs. Int. Ass. Sediment. 5, 415-28.

Fischer, J. (1983) Remote sensing of suspended matter, phytoplankton and yellow substances over coastal waters, part 1: aircraft measurements. In: Degens, E. T., Kempe, S. and Soliman, H. (Eds) Transport of Carbon and Minerals in Major World Rivers, Pt.2. Mitt. Geol.-Paläont. Inst. Univ. Hamburg, SCOPE/UNEP Sonderbd. 55, pp. 85-95.

Fischer, J. (1985) On the information content of multi spectral radiance measurements over an ocean. Int. J. Remote Sensing 6, 773-86.

Fischer, J. and Doerffer, R. (1987) An inverse technique for remote detection of suspended matter, phytoplankton and yellow substance from CZCS measurements. Adv. Space Res. 7(2), (2)21-(2)26.

Fischer, J. and Grassl, H. (1984) Radiative transfer in an atmosphere-ocean system: an azimuthally dependent matrix-operator approach. Appl. Opt. 23, 1032-9.

Fischer, J., Doerffer, R. and Grassl, H. (1986) Factor analysis of multispectral radiances over coastal and open ocean water based on radiative transfer calculations. Appl. Opt. 25, 448-56.

Gordon, H. R. (1980) Phytoplankton pigments from the Nimbus 7 Coastal Zone Colour Scanner: comparisons with surface measurements. Science 210, 63-6.

Gordon, H. R. and Morel, A. (1983) Remote Assessment of Ocean Color for Interpretation of Satellite Visible Imaginery. A Review. Springer Verlag, New York, Berlin, Heidelberg, Tokyo, 114pp.

Gordon, H. R., Clark, D. K., Brown, J. W., Brown, O. B., Evans, R. H. and Bromkov, W. A. (1983) Phytoplankton pigment concentrations in the Middle Atlantic Bight: comparison of ship determinations and CZCS estimates. Appl. Opt. 22, 20-36.

Hojerslev, N. K. (1979) On the origin of yellow substance in marine environment. In: Workshop on Eurasep OCS Experiment, 13-28.

Kalle, K. (1961) What do we know about gelbstoff? In: Jerlov, N. G. (Ed.) Symposium on Radiant Energy in the Sea. IUGG Monograph 10.

Michaelis, W. (1983) Untersuchungen zum Schwebstoff- und Schwermetalltransport in der Elbe. GKSS Forschungsberichte GKSS 83/E/39, GKSS Forschungszentrum Geesthacht.

Morel, A. (1980) In-water and remote sensing of ocean color. Bound. Layer Meteor. 18, 177-201.

NASA (1986) MODIS¾Instrumental Panel Report. Earth Observation System, Vol.2b.

Neville, R. A. and Gower, J. F. R. (1977) Passive remote sensing of phytoplankton via chlorophyll a fluorescence. J. Geophys. Res. 82, 3487-93.

Paltridge, G. W. and Platt, C. M. R. (1976) Radiative Processes in Meteorology and Climatology, Elsevier. Amsterdam, Oxford, New York.

Prieur, L. and Sathyendranath, S. (1981) An optical classification of coastal and oceanic waters based on the specific absorption curves of phytoplankton pigments, dissolved organic matter, and other particulate materials. Limnol. Oceanogr. 26, 671-86.

Reuter, R. (1980) Characterization of marine particle suspensions by light scattering, I. Numerical predictions from Mietheory, II. Experimental results. Oceanologica Acta 3(3), 317-24.

Smith, R. C. and Baker, K. S. (1978) Optical classification of natural waters. Limnol. Oceanogr. 23, 260-7.