Acquisition and analysis of remote sensing imagery of harmful algal blooms.

Abstract

Remote sensing was long considered an obvious tool for studying the distribution of harmful algal bloom (HAB) organisms over larger spatial and shorter time scales than is possible with ship-based sampling (Tester et al. 1991; Keafer and Anderson 1993). Legacy and nextgeneration instrumentation and sensors, including SeaWiFS, MODIS, MERIS, and the OLCI sensor on Sentinel-3, are dramatically improving the ability to determine constituents in the coastal ocean. Satellite altimeters and scatterometers also provide geophysical fields such as dynamic height (current patterns) and local winds (e.g. upwelling indices). Currently, MODIS Aqua and VIIRS are still operational, while the replacement for MERIS, OLCI, is now operational. In some regions, remote sensing has already become a valuable tool for helping to predict the onset, location, and transport of HABs. For example, in the Florida Shelf and Gulf of Mexico, SeaWiFS and MODIS imagery has been incorporated into the U.S. NOAA HAB Bulletin reports to identify potential red tide events, while feature-tracking has been used to follow the spatial transport of these events (e.g. Tester et al. 1991; Tester and Steidinger 1997). Progress has also been made on the use of inherent optical properties, derived from ocean color inversion algorithms, to identify functional phytoplankton groups based on fundamental biophysical properties (e.g. Lohrenz et al. 2003; Schofield et al. 1999). Although multi-spectral scanners (e.g. MODIS) can be used to detect the reflectance of chlorophyll a and other pigments with some accuracy, these efforts have been constrained by the inability of the sensors to discriminate phytoplankton populations at the species level. This is, of course, a fundamental requirement of HAB programs. Instead, progress has been made by first linking specific water masses to HAB organisms and then identifying and tracking that water mass with an appropriate remote sensing technique. In particular, remotely-sensed sea surface temperatures (SST) have been used to follow the movement of fronts, water masses, or other physical features where HAB species accumulate. A fundamental problem for identifying HAB events, however, is that the imagery is still limited to identification of chlorophyll or other biomass proxies rather than individual organisms (at the genus or even functional group level). Satellite imagery by itself will simply not provide the specificity needed to identify particular organisms. Recent advances have begun to extend our ability to use remote sensing beyond simple bulk chlorophyll measurements, however. For example, considerable work has gone into identifying phytoplankton functional groups, or groupings of optically similar organisms such as diatoms, dinoflagellates, and coccolithophorids. In some specific cases, optical estimates (either from in-water measurements or remote sensing) can be used to identify particular organisms, as some have unique optical properties. This includes Karenia brevis, Trichodesmium spp., and cyanobacterial (blue-green) algal blooms (Alvain et al. 2008; Stumpf et al. 2003; Westberry et al. 2005; Wynne et al. 2008). While diatoms and dinoflagellates are very similar optically, and both can cause high biomass events, there appear to be enough differences to discriminate between dinoflagellate- and diatomdominated surface waters as well (Dierssen et al. 2006; Palacios 2012). In addition to the limitations of optical methods (including remote sensing) for the identification of specific HAB organisms, another problem arises when imaging high biomass blooms. When the biomass exceeds ~50 mg/m3 total chlorophyll, standard satellite algorithms (e.g. MODIS OC3 or MERIS Algal-2) often fail because the water-leaving radiances are high enough to trigger atmospheric correction failures. This results in consistent underestimates of high biomass events in coastal waters. This can be remedied relatively easily by the use of non-standard ocean color products. For example, Kahru and Mitchell (2008) showed that the 250 m resolution bands on the MODIS satellite can be used to develop a “particle index” that closely tracks red tides, while also providing the highest possible spatial resolution. Hu et al. (2005) advocated the use of fluorescence bands for the same reason; a second advantage is that only chlorophyll-containing particles strongly fluoresce, solving the issue of working in optically complex coastal waters. Chen et al. (2009) extended this by using multiple bands (fluorescence line height (FLH), backscatter, etc) to develop a “machine learning” algorithm that can detect red tides. Given enough data it is also possible to develop region-specific algorithms that work better than the global methods (Kahru et al. 2012). To summarize, using modern methods and data freely available from several ocean color sensors, it is currently possible to identify high biomass HAB events (e.g., red tides), although this requires application of non-standard products. The biomass estimates can be further categorized into phytoplankton functional types, potentially useful for identifying subclasses of blooms such as high biomass dinoflagellate events. These methods require more effort and access to some laboratory or field optical measurements to parameterize the models. It is not currently possible (and is unlikely to become possible) to identify species of algae from space. When combined with other data streams such as currents, field measurements, and in-water monitoring programs, unusual events can be identified, tracked, and the subsequent impacts predicted if there are independent means of identifying the organisms. This is most effective when remote sensing is combined with in-water observations as part of an ocean observing program (see Chapter 3; Frolov et al. 2013; Kudela et al. 2013).