Lauriane Cayet-Boisrobert – PgDip in Remote Sensing, UCL- PRN: 59048211
Remote Sensing of Oceans coursework: Red tides Detection using satellite ocean imagery Submitted 23 March 2006 Introduction The colour of the ocean depends not only on the weather and light conditions, but also on what the water contains. Occasionally, the surface of the water can appear green or even dark red in patches. Such events happen when the environmental conditions become suitable for phytoplankton to bloom. Approximately 5000 species of microalgae exist world-wide, of which about 100 are toxic (Fisher et al. 2003). These produce potent neurotoxins which that can be transferred through the food web where they affect and even cause mortalities of zooplankton, shellfish, wild and farmed fish, birds, marine mammals such as sea lions (Sholin et al. 2000). They can even affect humans by intoxications or irritation of the eyes and respiratory systems (Anderson 1995, Hu et al. 2005, NOAA/CSCOR/COP 2006). Therefore, harmful algae blooms (HABs) can affect dramatically the coastal economies and livelihoods (Fisher et al. 2003, Hu et al. 2005). Moreover, HAB events worldwide appear to be increasing in frequency and severity (Smayda 1989 cited Fisher et al. 2003, EC and NSF 2003). It is debated that this expansion may be attributed to ship ballast waters or global warming (EC and NSF 2003). Several countries have warning systems in place such as the Gulf of Mexico HAB bulletin, issued beginning of 1999 by the National Oceanic and Atmospheric Administration (http://www.csc.noaa.gov/crs/habf/). These systems are reliable to some extent, but much research still need to be undertaken to improve their reliability. Hence, collaborative regional programs, such as EUROHAB for Europe, ECOHAB for the United States, and GEOHAB as an international program have been launched to focus on co-ordinated research, technologies, and shared resources. Among a variety of research focuses, timely detection and observation of HABs is critical for the prediction and mitigation of red tides impacts. Sample collection and analysis are limited in both time and space frequency. However, Earth Observation sensors can provide a wide synoptic coverage, frequent revisit capability, at relatively low coast (Hu et al. 2005). The present paper focuses on red tide detection using satellite ocean colour sensors. High chlorophyll concentration is the surrogate used to infer a potential threat red tide. The paper will show that ancillary data and expertise in the region are required for reliable red tides detection. The first part of the report gives the eco-physiological background and pre-requisite knowledge of the colour imagery sensors. The second part explains the physical principles and basis behind red tides detection from colour imagery. The third part, gives a non-exhaustive but most frequent methods encountered in the literature. The final part will reframe the issue in the operational context, explaining that reliable red tide detection should rely on an approach integrating various sources of data. 1. Background Red tides temporal and spatial distribution has been well documented, especially in the developed countries. Much is known about fate, duration, frequency, and distribution according to Hu et al. (2005). However, to date the mechanisms causing most HABs remain unclear. It seems that their initiation is different from one species to another and differs geographically. For instance, Karenia Brevis (West Florida Shelf) would initiate offshore in nutrient-poor environment waters, and then move towards shore under favourable winds and currents, where growth may be stimulated by additional nutrients form coastal run-off (Tester and Steidinger 1997 cited Hu et al. 2005). Other studies suggest that toxic Pseudo nitzschia species off the Pacific Northwest coast development are coincident with the seasonal Juan de Fuca eddy, a nutrient rich retentive feature off the Washington state coast which might serve as an initiation site for the growth of phytoplankton (ECOHAB PNW 2006). The red tide issues is complex because some species are commonly reported as toxic in one region but not in another. For example, Pseudo Nitzschia is the only species off Spain that is toxic whereas the Page 1 of 8
Lauriane Cayet-Boisrobert – PgDip in Remote Sensing, UCL- PRN: 59048211
same species can be non-toxic off the US Pacific Northwest and toxic in waters further south, off the coast of California as stated above (EC and NSF 2003). Additionally, waters conditions are different from one region to an other and the water leaving radiance (Lw) detected by the optical sensors would be influenced by other matters than phytoplankton bloom’s backscatter. Phytoplankton organisms like plants are capable to produce their carbonic constituents from solar energy in addition to H2O and CO2 by photosynthesis trough pigments called chlorophyll a and b. Since, empirical relationships exist between chlorophyll concentration and specific phytoplankton cells number, thus chlorophyll can be used as a surrogate to detect phytoplankton blooms. Chlorophyll pigments absorb blue and red light, but deflect green light (for photosynthesis). Thus, it appears green to the eyes and on true colour composite imagery (Hu et al. 2005). These particular biooptical properties are the basis of the red tides detection by ocean colour imagery. Satellite ocean colour research began in the late 1970s with the Coastal Zone Colour Scanner (CZCS) which acquired data from October 1978 to June 1986. Then, SeaWiFS was designed to have much higher radiometric sensitivity and additional spectral bands to aid in atmospheric correction and designed to avoid some of the limitations of the CZCS (Hooker et al. 1993 cited Islam and Tat 2001). More recently, several new ocean colour sensors have been launched by various countries: OCM (India), OSMI (Korea), OCI (Taiwan), NASA’s MODIS Aqua/Terra (USA), MERIS on ENVISAT (ESA) 300-resolution mode (Europe), NASDA’s ADEOS on ADEOS-II (Japan). Over the last 30 years, one of the successes of optical oceanography is the capability to measure chlorophyll concentration from measurements of water colour. Scientists use the balance between blue and green (the blue to green ratio) to calculate how much chlorophyll the water contains. This allows them to create world-wide maps of chlorophyll from satellite images of ocean colour. In the open ocean (Case I conditions), substances or particles that scatter light and colour the water green, are almost certainly phytoplankton. Thus, chlorophyll concentration retrieved can be retrieved with confidence (Robinson 2004). New studies (Lee and Hu 2006) warn that not all open waters fall under case I, and case I distribution varies with season (figure 1). Hence, it is important to keep in mind that they are large deviations in the chlorophyll concentration estimation even in open waters (Lee and Hu 2006). When looking at the coastal (often turbid) areas (case II conditions), where most of the red tides events occur and becomes the subject of many socio-economic impact considerations, chlorophyll mapping is unreliable. In case II waters, chlorophyll concentration tend to be overestimated (Robinson 2004), because of the presence coloured dissolved organic matter (CDOM) or suspended material (SSM) released from rivers estuaries.
Figure 1. Global distribution of case I (in blue) and Case II (in green) waters (Lee and Hu 2006).
Page 2 of 8
Lauriane Cayet-Boisrobert – PgDip in Remote Sensing, UCL- PRN: 59048211
2. Physical basis and principles behind red tides detection The detection of red phytoplankton blooms uses chlorophyll as a surrogate. Chlorophyll concentration can be measured by remote sensors because a real dependence exists between wavelength and chlorophyll concentration (figure 2). As chlorophyll concentration increases, reflectance decreases for λ < 550nm; whereas reflectance increases for λ < 550nm. However, reflectance around 550nm does not vary much, and thus, is independent of chlorophyll concentration. Consequently, there are at least two ways to retrieve phytoplankton properties. 2.1. Reflectance Many algorithms make use of the reflectance behaviour at λ
