Pattern recognition of marine provinces

Transcription

1 International Journal of Remote Sensing Vol. 26, No. 7, 10 April 2005, Pattern recognition of marine provinces K.-H. SZEKIELDA The Graduate School and University Center, Earth and Environmental Center, Geography Department, Hunter College, City University of New York, 695 Park Avenue, New York, NY, 10021, USA; (Received 27 August 2003; in final form 5 October 2004 ) The objective of this Letter is to introduce a concept for identifying marine provinces by applying two band ratios that are indicative of chlorophyll distribution patterns. The ratios of water-leaving radiances 443 nm/551 nm and 678 nm/ 667 nm are close to the two absorption bands of chlorophyll; however, as photon penetration depth at the applied wavelengths varies, each ratio responds to different depths. This allows a qualitative interpretation of separated clusters in scatter diagrams. Pattern classification separates the major biogeochemical provinces as documented with the Peruvian upwelling system and the convergence zone in the Brazil Falkland/Malvinas current system. 1. Introduction Phytoplankton pigment analysis and measurements of productivity on a global scale provide insight into the marine food chain and are inputs into models for understanding the global carbon cycle. Suspended matter, in general, is spatially inhomogeneous and the biogeochemical provinces in the productive ocean surface layer are determined by physical and chemical changes in the marine environment (Longhurst et al. 1995). The success of the global monitoring of chlorophyll by satellite sensors revealed the heterogeneity in chlorophyll distribution at temporal and spatial scales; see for instance, Kahru and Mitchell (2001) and Gregg et al. (2002). The most common approach to measure chlorophyll concentration remotely is the use of the spectral region where chlorophyll and related accessory pigments have strong absorption bands. The radiance ratio 443nm/551 nm is directly related to chlorophyll concentrations and has been measured with the Coastal Zone Colour Scanner (CZCS), the Sea-viewing Wide Field-of-view Sensor (SeaWiFS) and more recently with the Moderate Resolution Imaging Spectrometer (MODIS). In addition, primary productivity (Field et al. 1998, Behrenfeld et al. 2001), particulate carbon (Stramski et al. 1999), and fluorescence (Gower and Borstad 1981, Gower et al. 1999) have been derived from satellite sensors. The objective of this study is to introduce a concept for identifying biogeochemical provinces with satellite remote sensing data. As MODIS radiometric specifications have the appropriate spectral bands that are close to the absorption and fluorescence bands of chlorophyll, it is useful to apply the data in the spectral region of the second absorption band of chlorophyll near 683 nm. This spectral region is not only chlorophyll specific but is also less influenced by the presence of other optically relevant substances such as suspended inorganic matter and chromophoric dissolved organic matter. The concept of using a ratio has been evaluated in coastal water with high concentrations of chlorophyll (Szekielda et al. International Journal of Remote Sensing ISSN print/issn online # 2005 Taylor & Francis Group Ltd DOI: /

2 1500 K.-H. Szekielda 2003). In the following, an approach will be introduced to apply the ratio technique using MODIS bands 13 and 14, which have their full width at half maximum (FWHM) at 667 nm and 678 nm, respectively, and to relate them to chlorophyll values based on the common retrieval technique using the ratio 443 nm/551 nm. Normalized water-leaving radiance from MODIS was used as described in detail by Gordon and Voss (1999). The spectral region considered in the following includes that portion of the spectrum where the natural fluorescence of phytoplankton is close to the second chlorophyll absorption band. Significant correlation was found between chlorophyll concentrations and water-leaving radiance (Szekielda et al. 2003) in this very same spectral region where natural fluorescence of chlorophyll is present (Gower and Borstad 1981, Bricaud et al. 1995, Gower et al. 1999). Natural chlorophyll fluorescence was used by Hoge and Swift (1987) to study the ocean colour spectral variability. The distinctive second chlorophyll absorption band is located at 670 nm but its absorption intensity at concentrations.20 mg m 23 is offset by the scattering of cell walls (Gitelson 1992). Natural fluorescence is diagnostic of chlorophyll at 685 nm but can be significantly altered by re-absorption at chlorophyll concentrations mg m 23 (Gitelson 1992). It follows that the spectral region between 670 nm and 685 nm should be useful for detecting elevated concentrations of chlorophyll. 2. Approach The division of multi-spectral measurements into distinct regions through interactive analysis of two-dimensional scatter plots permits the identification of clusters that relate to specific oceanic regimes. The principle of identifying a pattern classifier to compare different depth horizons, according to photon penetration depth, was tested over several oceanic regions with MODIS data, but only two test sites will be introduced to demonstrate the potential to detect various biogeochemical provinces. Yearly averaged MODIS data from the year 2002 were selected, because at present, they are the only complete dataset available over most parts of the oceans; however, it should be realized that climatological datasets smooth the dynamic oceanic field as well as the concentration of chlorophyll. 3. Results Test sites in the present study include the upwelling regime along the Peruvian coast and the Brazil Falkland/Malvinas current system. For the upwelling study, the Peruvian coast was selected. The western coast of South America is exposed to the longest eastern boundary current (EBC) and covers the distance from 42u S to the equator and represents one of the major upwelling ecosystems. The Humboldt Current flows from the Antarctic to the equator with Ekman-induced upwelling that is typically about 150 km wide, and fresh water input and sediment transport into the coastal regions are negligible. The other test site presented here is the Brazil Falkland/Malvinas convergence. The convergence is the result of warm and salty water carried by the Brazil Current in a pole-ward direction along the continental shelf up to approximately 39u S where it collides with the northward branch of the Antarctic Circumpolar current and the Falkland/Malvinas current (Bianchi et al. 1993). Highly organized circulation cells develop in both pole-ward and equator-ward western boundary currents with a narrow zone of nutrient rich water.

3 Remote Sensing Letters 1501 Figure 1 gives the two-dimensional scatter diagrams for both test sites with ratios 443 nm/531 nm and 667 nm/678 nm as well as the corresponding clusters (regions of interest, ROI) identified by interactive classification. The maps shown in figure 2(a) and (c) were generated to show the geographical location of the identified clusters. As the hydrography of the Brazil Falkland/Malvinas current system is more complicated than that observed in the upwelling regime along the Peruvian coast, the scatter diagram shown in figure 2 for this current system also displays a more distinct pattern. By exporting the ROIs from the cluster interpretation, the resulting image in figure 2(d ) demonstrates that various provinces can be isolated. Results presented here are preliminary findings and a more thorough interpretation would be required. Nevertheless, it can be stated that the differentiation of various biogeochemical provinces can be achieved with the use of the spectral region where chlorophyll has its two absorption bands. It is further based on the fact that plankton growth depends on the vertical stratification of the upper water column. Nutrient support is limited under highly stratified conditions, whereas in upwelling regimes nutrients support high productivity close to the surface. Consequently, the depth location of chlorophyll is a factor that must be considered when working at longer wavelengths especially over the second absorption band of chlorophyll. Based on the photon penetration depth as a function of wavelength, it can be (a) (b) (c) (d) Figure 1. Scatter diagrams for the Peruvian upwelling regime ((a) and (b)) shown in figure 2(a). (c) and (d ) show scatter diagrams for the Brazil Falkland/Malvinas current system as also shown in figure 2(c). The colour presentation in (a) and (c) represents the data density in rainbow colours, whereby red indicates the highest density of the data to blue the lowest. The clusters in (b) and (d ) were separated and the same corresponding colour coding given in (b) and (d ) was applied in the presentations of their regional distribution in figures 2(b) and (d ).

4 1502 K.-H. Szekielda (a) (b) (c) (d) Figure 2. (a) and (c) give the chlorophyll distribution in mg m 23.(b) and (d ) show the clusters (regions of interest) identified in figures 1(b) and 1(d ). (a) and (b) represent the Peruvian upwelling regime and (c) and (d ) show the Brazil Falkland/Malvinas current system. postulated that the chlorophyll estimation based on the ratio 443 nm/551 nm is not necessarily correlated with the ratio 678 nm/667 nm, the latter being responsive to high chlorophyll concentration in the upper part of the water column. The 443 nm/ 551 nm ratio is affected greatly by the absorption of detritus, chromophoric dissolved organic matter (CDOM) as well as scattering by all suspended matter while the red range is much less affected by CDOM absorption. In addition, variability in the biogeneous compartment and decoupling between chlorophyll, chromophoric dissolved organic matter, inorganic suspended matter, organic detritus and bacteria may be the cause for recognition of biogeochemical provinces in cluster diagrams. Therefore, distinctive clustering appears in the spectral space allowing the use of two-dimensional scatter plots to design a pattern classifier. It can be concluded that the above approach can be used for separating biogeochemical provinces in connection with chlorophyll measurements based on the use of the first and second absorption band of chlorophyll. Acknowledgments The satellite sensor data used in this study were acquired as part of NASA s Earth Science Enterprise. The algorithms for sea surface temperature and chlorophyll were developed by the MODIS Science Teams. The data were processed by the MODIS Adaptive Processing System (MODAPS) and Goddard Distributed Active Archive Center (DAAC) and are archived and distributed by the Goddard DAAC,

5 Remote Sensing Letters 1503 Greenbelt, MD, USA. I also acknowledge the comments on this manuscript by three anonymous reviewers. References BEHRENFELD, M.J., RANDERSON, J.T., MCCLAIN, C.R., FELDMAN, G., LOS, S.O., TUCKER, C.J., FALKOWSKI, P.G., FIELD, C.B., FROUIN, R., ESAIAS, W.E., KOLBER, D.D. and POLLACK, N.H., 2001, Biosphere primary production during an ENSO transition. Science, 291, pp BIANCHI, A.A., GIULIVI, C.F. and PIOLA, A.R., 1993, Mixing in the Brazil Malvinas confluence. Deep-Sea Research, 40, pp BRICAUD, A., ROESLER, C. and ZANEVELD, J.R.V., 1995, In situ methods for measuring the inherent optical properties of ocean waters. Limnology and Oceanography, 40, pp FIELD, C.B., BEHRENFELD, M.J., RANDERSON, J.T. and FALKOWSKI, P., 1998, Primary production of the biosphere: integrating terrestrial and oceanic components. Science, 281, pp GITELSON, A., 1992, The peak near 700 nm on radiance spectra of algae and water: relationship of its magnitude and position with chlorophyll concentration. International Journal of Remote Sensing, 13, pp GORDON, H.R. and VOSS, K.J., 1999, MODIS normalized water-leaving radiance algorithm theoretical basis document (MOD 18) Version 4, NASA Contract Number NAS503163, 92 pp. GOWER, J.F. and BORSTAD, G., 1981, Use of the in vivo fluorescence line at 685 nm for remote sensing surveys of surface chlorophyll a. In Oceanography from Space, J.F. Gower, (Ed.), pp (Plenum Press, New York). GOWER, J.F.R., DOERFER, R. and BORSTAD, G.A., 1999, Interpretation of the 685 nm peak in water-leaving radiance spectra in terms of fluorescence, absorption and cattering, and its observation by MERIS. International Journal of Remote Sensing, 20, pp GREGG, W.W., CONKRIGHT, M.E., O REILLY, J.E., PATT, F.S., WANG, M.H., JODER, J.A. and CASEY, N.W., 2002, NOAA-NASA Coastal Zone Color Scanner reanalysis effort. Applied Optics, 41, pp HOGE, F. and SWIFT, R., 1987, Ocean color spectral variability studies using solar induced chlorophyll fluorescence. Applied Optics, 26, pp KAHRU, M. and MITCHELL, B.G., 2001, Seasonal and nonseasonal variability of satellitederived chlorophyll and colored dissolved organic matter concentration in the California Current. Journal of Geophysical Research, 106, pp LONGHURST, A., SATHYENDRANATH, S., PLATT, T. and CAVERHILL, C., 1995, An estimate of global primary production in the ocean from satellite radiometer data. Journal of Plankton Research, 17, pp STRAMSKI, D., REYNOLDS, R.A., KAHRU, M. and MITCHELL, B.G., 1999, Estimation of particulate organic carbon in ocean from satellite remote sensing. Science, 285, pp SZEKIELDA, K.-H., GOBLER, C., MOSHARY, F., GROSS, B. and AHMED, S., 2003, Spectral reflectance measurements of estuarine waters. Ocean Dynamics, 53, pp