Simulation of a MERIS FLH Product

This is an abridged version of a report prepared as part of a 1996-98 project for the Canadian Space Agency, as part of the Canadian Expert Support Laboratory (CESL) for MERIS: MacDonald Dettweiller Associates Ltd. (www.mda.com) prime contractors.

Introduction

As part of earlier work in 1974 to 1988, Gower and Borstad developed a completely new method for remote sensing of phytoplankton concentration which as been shown to have advantages for monitoring coastal areas where there is influence from dissolved organic material and suspended solids. This method, based on solar-stimulated fluorescence from chlorophyll pigments has been taken up especially by the Europeans, but also by other nationals as an important method for use in the coastal zone. The MERIS (the Medium Resolution Imaging Spectrometer) is the best evidence of this. MERIS has several channels that allow it to be used to image solar stimulated in vivo fluorescence from phytoplankton chlorophyll at 685 nm. The European Space Agency (ESA) will MERIS launch in 1999.

This project was intended to evaluate the capability of MERIS to measure the solar-stimulated chlorophyll fluorescence signal (Fluorescence Line Height, or FLH) to provide estimates of phytoplankton concentration, and information on primary productivity. Algorithms based on chlorophyll fluorescence originally developed in Canada by Gower and Borstad (1990) are the prime focus. The objective of this project was to demonstrate a potential Fluorescence Line Height (FLH) product using CASI airborne image data acquired with MERIS-like spectral channels. In this project, CASI data acquired by Borstad Associates in September 1992 over the Bay of Quinte in Lake Ontario (Figure 1) was used to simulate an FLH product from MERIS. The data used were at-sensor radiances (9,000'); no atmospheric corrections were applied. In situ ground truth obtained at the time of imaging was used to calibrate the FLH image in units of Chlorophyll concentration (mg/m³).

Figure 1. Location of the Bay of Quinte, in Lake Ontario

Methods

The ‘FLH’ algorithm represents a simple mathematical combination of spectral bands, which provides a measure of an apparent peak in a radiance or reflectance spectrum. In the case of solar stimulated fluorescence, this is the ‘excess’ upwelling radiance at 685 nm caused by fluorescence from phytoplankton chlorophyll a. In fact in the relatively turbid environment of the Bay of Quinte, and with the bands available for this study (the data was acquired for other purposes and the band at 681 is wider and displaced to shorter wavelengths) true fluorescence was not measurable. Instead, a ‘remnant red-edge’ is strongly visible – where the sharply increasing red edge reflectance above 700 nm, of concentrated phytoplankton chlorophyll is not completely absorbed by the strongly increasing water absorption above 700 nm. The shift in center wavelength of the red reflectance peak at high chlorophyll concentrations has been documented before. Gower et al (1996) have modelled the effect and have confirmed the suggestion that the shifted peak is not fluorescence but an interaction between water absorption and chlorophyll reflectance.

Figure 2. Upwelling Radiance signal at the aircraft altitude (without atmospheric correction) for 25 locations within the Bay of Quinte, showing a range of chlorophyll and suspended sediment concentrations.

From an operational point of view, it does not matter whether the phenomenon being measured is solar stimulated fluorescence or the ‘remnant red edge’. In either case, the FLH algorithm can be used to measure the apparent radiance peak – in the first case at 681 nm, in the second case at 692 or 705 nm.

Following the algorithm of Gower and Borstad (1990), we calculate the apparent radiance of the peak above a straight baseline between two bands on either side of the feature. Centered at 681 nm (the band is narrower and shifted slightly to shorter wavelengths to avoid the oxygen absorption at 687 nm), and using a baseline between 665 and 753 nm the calculation is:

Similarly, the calculation for the ‘remnant red edge’ is

In the 1992 project, CASI image data from forty individual flight lines were mapped and mosaiced to form several large tile mosaics of the entire Bay of Quinte. The original image data was acquired at 4 m ground pixel size, and was radiometrically calibrated at the sensor altitude (9,000') – it was not corrected for atmospheric effects. As the focus of the 1992 work was not water colour, the spectral bands are not identical to the MERIS band passes. For this work, the original tile mosaics on archive at Borstad Associates were mosaiced to form a very large mosaic and then block averaged at 250 m ground pixels to simulate MERIS data. Land areas and other extraneous features were zeroed and not included in the block averaging. Table 1 summarizes the channels of the 1992 data set.

For this work, channels 7, 8, 10 and 11 (665 nm, 678 nm, 705 nm and 753 nm) were used to calculate two separate FLH images. Channels 2 and 3 were used to simulate a Green/Blue ratio of the 490 and 520 nm MERIS channels. Note that that the 678 nm band is wider then the MERIS 681.5 nm channel.

Table 1. Spectral band configuration of the Bay of Quinte CASI data from which the FLH image mosaic was constructed, as compared to MERIS bands. Shaded bands were used in this study.

Results

Figure 3 illustrates that, as we expected, the 678 nm band center was not suitable for measurement of fluorescence in the Bay of Quinte. However, the 692 and 705 nm bands were strongly correlated with surface in situ Chlorophyll concentration. A direct comparison of the two FLH calculations for the 37 stations in the study area, show that they are very similar (FLH692 = 0.73*FLH705-62; r2 = 0.98), and compare equally well with in situ chlorophyll, when all stations are considered. FLH 692 and 705 signals both exhibit a better correlation with in situ chlorophyll than a Green/Blue ratio (Table 2). All three indices are negatively affected by increased inorganic sediment, and in fact direct comparison of the FLH705 and Ratio 550/520 shows good agreement except for stations 10, 11, GL, GL1, GL2 and GL3 (Figures 4 and 5). These stations are all contiguous, at the eastern-most sharp bend in the Bay where the low chlorophyll waters of Lake Ontario meet the much more eutrophic waters of the Bay of Quinte. While we do not have subsurface data, it seems possible that the Ratio/FLH plots are pointing out a difference in vertical distribution in this area.

Figure 3. Comparison of Fluorescence Line Height calculated using three different center wavelengths and in situ Chlorophyll concentration, and the predictive equations. The 678 nm band was not correctly placed to calculate FLH.

Figure 4. Comparison of a Green/Blue ratio with bands approximating MERIS bands 2 and 3 (490 and 520 nm).

Figure 5. Relationship between FLH705 and the Ratio of 550 nm/r520nm upwelling Radiances.

Figure 6. A simulated MERIS FLH product – surface Chlorophyll concentration in the Bay of Quinte, September 21-22, 1992.

Table 2. Predictive equations for Chlorophyll concentration and their correlation coefficients.

Conclusions

In the Bay of Quinte, a measure of the ‘shifted peak’ resulting from the interaction between the red edge and water absorption provided an excellent index of surface chlorophyll concentration. The FLH algorithm, using bands centered on the shifted peak provided an alternative index of chlorophyll with a higher correlation with pigment content than did a Ratio of 550 nm/ 520 nm. The differences between the two methods may be a result of differences in vertical distribution of phytoplankton and the different depth of penetration of the two indices.

More work is required to separate true solar-stimulated fluorescence from the remnant red edge phenomenon. Although it greatly increases the complexity and cost of the study, the in situ ground truth for future studies should include information concerning subsurface pigment distribution.

References

Brown, R. M. and G. A. Borstad. 1982. Chlrophyll a determinations from solar stimulated in vivo fluorescence. Unpublished report by Seakem Oceanography Ltd. for Department of Fisheries and Oceans. DSS Contract 04SB.FP833-1-0905

Borstad, G. A., R. Kerr, D. Truax and D. Pan. 1987. Using an imaging spectrometer to map phytoplankton chlorophyll. Unpublished report by G. A. Borstad Ltd. for the Department of Fisheries and Oceans, DSS contract 06Sb.FP941-6-0259. 95pp.

Gower, J. F. R. and G. A. Borstad. 1990. Mapping of phytoplankton by solar-stimulated fluorescence using an imaging spectrometer. International Journal of Remote Sensing, 11:313-320.

Gower, J. F. R., R. Doerffer and G. A. Borstad. 1998. Interpretation of the 685 nm peak in water leaving radiance spectra in terms of fluorescence, absorption and scattering, and its observations by MERIS. submitted to International Journal of Remote Sensing.

   
Last update 19 March 2007
The contents of this site are © copyright 1993 - 2009, ASL Borstad Remote Sensing Inc..