Detecting Blooms of the Dinoflagellate
Karenia brevis From Space
Introduction
Blooms of the toxic dinoflagellate Karenia brevis (formerly
Gymnodinium breve ) occur periodically off the Gulf coast of
Florida in the fall and early
winter. These blooms, often referred to as "red tides" because
their presence discolors the water a reddish hue, pose a
significant risk to human health and detrimentally affect regional
economies and marine resources. The adverse impacts of these
blooms may be mitigated if their presence could be detected early.
Satellite ocean color imagery provides the synoptic and
repeated coverage appropriate to detect these biological events well
offshore where they are thought to initiate. The blooms are likely to
be distinguishable from most other water conditions in ocean color
imagery owing to their relatively unique coloration.
The objective of this project was to develop
a prototype algorithm to detect blooms of the toxic
dinoflagellate Karenia brevis blooms off southwestern
Florida in ocean color imagery.
Approach
A supervised, multispectral classification algorithm was
constructed to detect K. brevis blooms in satellite ocean color
imagery. The empirical algorithm is based upon the spectral
signatures of K. brevis blooms and other common oceanic
conditions. The spectral signature of K. brevis blooms was
ascertained from Coastal Zone Color Scanner images
of November 14, 1978 (Tester, et al., 1997; Fig. 1) and October 28, 1983
(Carder et al, 1985), dates with contemporaneous in sea-truth measurements.
Spectral signatures of common oceanographic conditions were obtained from the
literature (Brown and Yoder, 1994). Decision boundary values of the
algorithm were established to allow the blooms to be spectrally
distinguished from the other oceanic conditions.
Fig. 1. True-color composite of Coastal Zone Color Scanner image from
November 14, 1978. Contemporaneous in-situ sampling sites
are indicated by yellow crosses.
Results and Discussion
The classified counterpart of the image presented in Figure 1 is presented
below (Fig. 2). Though the algorithm is preliminary, the results are
encouraging. The K. brevis class is located in the region known
to be occupied by these blooms.
Fig. 2. Classified counterpart of CZCS image presented in Figure 1.
This algorithm was designed for use with CZCS imagery in waters off the
western coast of Florida.
The algorithm has been implemented to use Sea-viewing Wide
Field-of-view Sensor (SeaWiFS) data from the Gulf of Mexico. To see these
results, please click here.
Literature Cited
Brown, C.W. and J. Yoder. 1994. Coccolithophorid blooms in the global
ocean. Journal of Geophysical Research 99: 7467-7482.
Carder, K. L. and R. G. Steward. 1985. A remote-sensing reflectance model
of a red tide dinoflagellate off West Florida. Limnology and
Oceanography 30: 286-298.
Tester, P.A., R.P. Stumpf, and K.A. Steidinger. 1997. Ocean color
imagery: What is the minimum detection level for Gymnodinium breve
blooms? In:B. Reguera, J. Blanco, M. Fernandez, and T. Wyatt (eds),
Harmful Microalgae. Proceedings of the VIII International Conference on
Harmful Algae, Vigo, Spain, 25-29 June 1997. Xunta de Galicia and IOC
of UNESCO Publishers.
Last Revised: March 2, 1998