Gayana 68(2) supl. t.I. Proc. : 252-258, 2004 ISSN 0717-652X

SATELLITE FLUORESCENCE AS A MEASURE OF OCEAN SURFACE CHLOROPHYLL

Jim Gower & Stephanie King

Fisheries and Ocean Canada, Institute of Ocean Sciences, PO Box 6000, Sidney BC, Canada V8L 4B2, Email: gowerj@pac.dfo-mpo.gc.ca

ABSTRACT

Both the MODIS optical sensors on the NASA Aqua and Terra spacecraft and the MERIS instrument on the ESA Envisat measure spectral radiance in bands designed to allow detection of the fluorescence signal from surface chlorophyll in sea water, stimulated by ambient sun and sky light. We present images of fluorescence as derived from the level 1 MERIS data and compare the signal levels of the fluorescence with measurements of extracted chlorophyll from research cruises during 2002 to 2004, with satellite estimates of chlorophyll from the blue to green ratio observed by MERIS, and with fluorescence measurements made by MODIS. We show that the fluorescence signal is consistent between MERIS and MODIS, and that relation between fluorescence and chlorophyll concentration in surface waters is consistent with a simple model accounting for absorption of stimulating and emitted radiation by chlorophyll pigments. Considerable scatter is observed, suggesting variable fluorescence efficiency and effects of coloured dissolved organic matter.

INTRODUCTION

The waters of the west coast of Canada provide a variety of conditions suitable for testing the capability of satellite optical imagers for measuring the fluorescence of chlorophyll a. These include the frequent occurrence of high surface chlorophyll concentrations over a range of seasons on the continental shelf, out to 50 to 100 km from shore and low surface sediment concentrations over most of the area. There are several sources of surface data including research cruises in these waters by the Institute of Ocean Sciences and the University of British Columbia in Canada, and university and government groups in the US, weather buoys with added biological instruments, and an instrumented ferry in the inshore waters.

Satellite data on fluorescence are available from both MODIS and MERIS. MODIS data are available over the web from archives maintained by NASA, such as the Distributed Active Archive Center at the Goddard Space Flight Center, http://daac.gsfc.nasa.gov/data/. We have reported on fluorescence observations with MODIS [1]. MERIS data are more restricted, but are becoming more widely available. We have received data as part of a research agreement with the European Space Agency, and this report concentrates on the MERIS data. In section 7 we show that MODIS and MERIS measurements of fluorescence agree closely.

MERIS DATA RECEIVED

As part of the MERIS validation process, ESA has provided images for 52 days in 2002 (starting on May 17), 75 in 2003 and 39 in 2004 (up to August). MERIS gives optimal coverage of an area once every 3 days, so this represents 156 out of about 270 expected total scenes in this period. Each image covers about a 600 km length (north-south) of the 1150 km wide MERIS swath. Images are selected to cover the south-east part of Vancouver Island, showing exposed waters on the south-west coast and the sheltered waters of the Strait of Georgia and other straits and inlets.

We have received level 1 data (calibrated radiances in each of the 15 spectral bands of the baseline band-set) for all the above 156 scenes, and level 2 data (atmospherically-corrected radiances, and derived products) for 111 scenes, a total of 42 in 2002, 33 in 2003 and 35 in 2004. All the above data are at the reduced resolution (RR) of 1200 m, comparable to SeaWiFS and MODIS. One important capability of MERIS is providing full resolution (FR) data at 300 m resolution. However, we have received FR data for only 30 days up to August 2004. This is the preferred data type, giving more spatial detail and demonstrating MERIS' capability for imaging in narrow inlets and close to the shore, areas often of highest interest and priority.

MERIS FLUORESCENCE OBSERVATIONS

In Figure 1, a difference spectrum is computed by taking the 15 MERIS above-atmosphere radiances for an area of relatively high chlorophyll, and subtracting the radiances for a nearby area of apparently "clear" water. The level 1 radiances are very similar for the two areas. Scattering of sunlight by the atmosphere increases the radiance at short wavelengths near 460 nm to about 4 times the value at 650 nm. The dip at 760 nm due to atmospheric oxygen is clearly shown by the three bands placed close to this feature. The difference spectrum is characterized by a peak in radiance at 681 nm (band 8). This band is positioned to include the peak of fluorescence at 685 nm [1,2], while still avoiding the strong oxygen absorption feature at 687 nm and longer wavelengths. The fluorescence signal (Fluorescence Line Height, FLH) is extracted from level 1 or level 2 radiance data by computing the radiance excess in band 8 above a linear baseline computed from the radiances in bands 7 and 9 (665 and 709 nm). Values computed using the level 1 data in this way give a negative "offset" of about 0.15 mW/(m².nm.sr), in clear offshore water. We show below that the fluorescence signal causes a radiance increase at 681 nm of about 0.5 mW/(m².nm.sr) for a chlorophyll concentration of 10 to 20 mg.m^-3. In Fig. 1 the difference spectrum also shows the effect of pigment absorption near 500 nm.

In Fig. 2, we show a comparison of the FLH fluorescence signal (top) and the Algal 1 chlorophyll product (bottom). Patterns in the two images are similar. Scatter plots of FLH versus Algal 1 chlorophyll were presented and discussed in for MODIS in [1] and for MERIS in [3].

4. FLUORESCENCE AS A FUNCTION OF CHLOROPHYLL CONCENTRATION

The expected level of the fluorescence signal depends on chlorophyll concentration, the ambient light level, the fluorescence efficiency of the chlorophyll and its distribution with depth. In [1] we that FLH should vary with chlorophyll concentration C according to the relation:

FLH = a . C / (1 + 0.2 C) - k................ (1)

where the factor a depends on the sun elevation angle, fluorescence efficiency and depth distribution of the phytoplankton, and instrument band placing. A negative offset represented by k is found for MERIS level 1 data. The factor 0.2 is an average value for the ratio of absorption of fluorescence by chlorophyll pigments at C = 1 mg.m^-3 to the absorption of fluorescence by pure water. In equation 1 the absorption reduces the fluorescence signal by a factor 2 at C = 5 mg.m^-3.

We have validated (1) by comparing MERIS data on FLH with C measured from a variety of sources, including research cruises and instrumented buoys. With lower confidence, we have also used Algal 1 chlorophyll data from MERIS. In general, we find that Algal 2 data are subject to larger errors.

In 2003 we were able to make use of surface chlorophyll measurements collected for the US/Canada ECOHAB study, which aims to understand the timing and movement of harmful algal blooms in the area shown in Fig. 2. Fig. 3 shows the positions of stations occupied on research cruises for ECOHAB in June and September of 2003. Background images are MERIS Level 1 RR FLH for June 6 and September 16. Ship measurements provide a large amount of data covering a wide range of FLH and chlorophyll concentration.

Figure 1: MERIS Level 1 radiance values in 15 bands for pixels showing high and low fluorescence signals (Sept 5 2002, coastal inlet). The radiance difference is plotted against the expanded scale on the right hand axis. These spectra are plotted as linear interpolations between the 15 MERIS radiance values, giving relatively coarse approximations to the true spectra.

Figure 2. FLH radiance computed from MERIS Level 1 data (top) compared with the Algal 1 level 2 data product for the coasts of southern Vancouver Island and northern Washington State on the west coast of North America.

Figure 3. Stations at which surface chlorophyll and other surface and sub-surface water and phytoplankton properties were measured for ECOHAB in June and September 2003. Background images are rectified MERIS Level 1 FLH images with added digital coastlines, rivers and international boundary.

Figure 4. shows a comparison between the fluorescence measured from a MERIS image and surface extracted chlorophyll concentration for the June stations. The curved line indicates the model (equation 1, with values of a and k chosen to fit these data points). Points are plotted for MERIS/ship comparisons up to 1 day apart in time. Points of different colours show the dates of MERIS images used.

MERIS FLUORESCENCE SIGNAL LEVEL AS A FUNCTION OF CHLOROPHYLL

Fitting the expected relation (1) to data, as shown in Fig 3, provides estimates of the parameters a and k for different areas and dates. Combining data from different areas, we derive a relation for a range of sun zenith angles and test the hypothesis that a and k are the same for all areas, but vary as cos (Z) where Z is the solar zenith angle.

Figs 5 and 6 tend to confirm the hypothesis, in that the expected variation with sun zenith angle is apparent, but the cause of the scatter about the solid red lines needs further investigation. The best fit model for all data in our study then gives a = 0.180 cos (Z) mW/(m².nm.sr) per mg.m^-3 of chlorophyll, and k = 0.265 cos (Z) mW/(m².nm.sr), , that is:

FLH = (0.18 C / (1 + 0.2 C) ­ 0.265) cos (Z) mW/(m².nm.sr) (2)

Where C is the concentration of chlorophyll in mg.m³.

Figure 5: Values of the scaling factor a (equation 1) derived by comparing MERIS fluorescence data with chlorophyll measurements from research cruises off the west coast of Vancouver Island, from an instrumented meteorological buoy and from MERIS level 2 Algal 1.

Figure 6: Values of the scaling factor k (equation 1) derived by comparing MERIS fluorescence data with chlorophyll measurements from research cruises off the west coast of Vancouver Island, from an instrumented meteorological buoy and from MERIS level 2 Algal 1.

EXTENSION TO HIGHER CHLOROPHYLL CONCENTRATIONS

Although the observed MERIS fluorescence signal computed from level 1 radiances using a linear baseline algorithm gives a consistent relation with surface data for the range of chlorophyll concentrations considered here (up to about 20 mg.m^-3) model simulations suggest that it will be impossible to extend measurements to much higher concentrations using only the bands 7, 8 and 9 (665, 681 and 709 nm). A simple model of water colour for a range of values of chlorophyll concentrations can be derived based on the work of Morel and others [2,4]. The model demonstrates the contribution of scattering and absorption by chlorophyll a and pure water in controlling the reflectance spectrum of ocean, coastal and inland waters.

In Fig 7, the chlorophyll a fluorescence is modeled by adding a "Gaussian" shaped peak centered at 685 nm, with width to half height of 25 nm and amplitude determined by the observed variation with chlorophyll concentration (C): 0.19 C / (1 + 0.2 C) mW/(m².nm.sr) per mg.m^-3 of chlorophyll for sun at the zenith [1,3]. For the extra-terrestrial solar irradiance at 685 nm of 1500 mW/(m².nm), this implies a reflectance due to fluorescence of 0.0003 C / (1 + 0.2 C), allowing for a factor 0.75 to account for two-way atmospheric absorption and scattering.

Fig 7 shows that for C=3 mg.m^-3 bands 7, 8 and 9 can be used to give a good estimate of FLH using a linear baseline algorithm. For C=300, however, absorption by water and chlorophyll pigments have combined to distort the shape of the spectrum such that a linear baseline is no longer appropriate. In fact the model suggests that such an algorithm would produce a negative value. The peak near 705 nm can be measured using the MERIS band at 709 nm, giving a potentially useful method of detecting the high chlorophyll concentrations indicating bloom conditions [2,4]. A combination of bands 7, 8, 9 and 10 should be able to provide estimates of both fluorescence and chlorophyll concentration using an inverse modeling approach similar to that now used for deriving the MERIS Algal 2 product.

Figure 7: Modelled water reflectance spectra (see [2,4]) for chlorophyll concentrations of 3, 30 and 300 mg.m³ with fluorescence (solid lines) and without fluorescence (dotted).

MERIS FLUORESCENCE COMPARED WITH MODIS FLUORESCENCE

Fig. 8 shows a scatter plot of MERIS versus MODIS (Terra) fluorescence, for images on 21 Sept 2002, navigated to the same geographic projection for an area off the coast of Vancouver Island. Images from the two sensors appear very similar. They are expected to differ in amounts of fluorescence radiance detected, because of differences in the locations and widths of the spectral bands of the two sensors. Fig. 8 shows that on average on this day, the best unbiased regression fit between MERIS and MODIS fluorescence has a slope of 1.39. Correcting this for the small difference in solar zenith angles between the passes of the two instruments gives a value of 1.34. The correction is small since the two satellites pass overhead less than 30 minutes apart. Some of the scatter about the mean relation will be due to the striping that is evident in MODIS FLH data.

Comparison of ten pairs of MODIS (Terra) and MERIS images for dates between July 2002 and February 2003, shows an average ratio of 1.34 + 0.03 after correction for solar elevation differences. The increased signal for MERIS is due to spectral band placement made possible by the improved band shape provided by the imaging spectrometer design. This allows the 681 nm fluorescence band to be placed closer to the optimum position at the peak of the fluorescence emission, in spite of the presence of a strong oxygen absorption feature at 687 nm. The expected ratio is about 1.35, assuming chlorophyll fluorescence is centered at 685 nm with a 25 nm width to half radiance, and that there are no errors in the relative calibration between MODIS and MERIS. This is in close agreement with the above observed ratio.

Figure 8: Scatter plot of MERIS versus MODIS (Terra) FLH for an area about 60 km square off Vancouver Island on Sept 21 2002. MERIS measures about 30% more fluorescence signal than MODIS (Terra), due to its spectral band placement.

CONCLUSIONS

A series of comparisons of the MERIS fluorescence signal with a variety of sources of surface chlorophyll estimates appears to give a consistent relation for a variety of seasons and solar zenith angles in the area studied, off the coast of Vancouver Island on the west coast of Canada. The relation needs to be confirmed or modified using observations from other areas. Also, techniques should be investigated to provide improved estimates of fluorescence radiance and high chlorophyll concentrations using MERIS spectral bands at 665, 681, 709 and 753 nm.

ACKNOWLEDGEMENTS

Funding for this project came from the Canadian Government (Fisheries and Oceans Canada and the Canadian Space Agency's GRIP program). Image data were provided by ESA under the MERIS AO program.

REFERENCES

Gower, J.F.R., L. Brown, & G.A. Borstad, 2004, Observation of chlorophyll fluorescence in west coast waters of Canada using the MODIS satellite sensor Canadian Journal of Remote Sensing, 30, 17-25. [         [ Links ]1]

Gower, J.F.R., R. Doerffer, & G.A. Borstad, 1999, "Interpretation of the 685 nm peak in water-leaving radiance spectra in terms of fluorescence, absorption and scattering, and its observation by MERIS," Int. J. Remote Sensing, 9, 1771-1786. [         [ Links ]2]

Gower, J.F.R., & King, S., 2003, Validation of chlorophyll fluorescence derived from MERIS on the west coast of Canada, Proceedings of Envisat Symposium, ESRIN, Italy, November 2003,http://envisat.esa.int/workshops/meris03/participants /133/paper_27_gower.pdf [         [ Links ]3]

Gower, J.F.R., King, S., Wei Yan, Borstad, G. & Brown, L., 2003, Use of the 709 nm band of MERIS to detect intense plankton blooms and other conditions in coastal waters, Proceedings of Envisat Symposium, ESRIN, Italy, November 2003, http://envisat.esa.int/workshops/ meris03/participants/135/paper_47_gower.pdf [         [ Links ]4]