Lecture: HABs Part 2 - Transcript

Harmful Algal Blooms

 

Slide 1 

Welcome to the online course on oceanographic satellite data products, produced by NOAA’s CoastWatch Program. In this module we will discuss how satellite data can be used for harmful algal bloom applications. The information presented here builds on the concepts discussed in several of the course’s earlier modules, including Satellite 101, Ocean Color, Water Quality, and Intro to Harmful Algal Blooms. You might want to review those modules first before viewing this module. My name is Betty Staugler, and I am Sea Grant’s liaison to NOAA CoastWatch. The materials I will present in this video were produced from a collaboration with members of NOAA’s CoastWatch Program including Shelly Thomlinson, Joaquin Trinanes, V Wegman and myself. 

Slide 2 

HAB forecasts are based on understanding the causes of HABs and how they respond to changing weather and ocean conditions. The other critical component of a HAB forecast is the ability to routinely and remotely detect HABs, their toxins, and environmental conditions that foster blooms and enhance their toxicity. There are generally three components to forecasts. If remote sensing is suitable, it is used to initiate models or show where algal cells are located. A hydrodynamic model is then used to move the cells via currents and winds. The final component is validation, which relies on monitoring data collected by state agencies and other groups. Across the country there are several bloom forecasting methods employed with each method tailored to decision-making needs and species of interest. In the next several slides we will look at the remote sensing component of forecasting HABs.  

Slide 3 

Optical Remote Sensing deals with the parts of electromagnetic spectrum characterized by the wavelengths from the visible to the near infrared (NIR) up to thermal infrared, collecting radiation reflected and emitted from the observed surfaces. Compared to pure water, most HABs have distinct spectral characteristics due to the pigments they contain. HABs tend to occur in specific locations – close to the coasts, in estuaries and in freshwater systems. And the formation of a HAB results from an intricate interaction of biological, chemical, physical and geological processes with associated contributors. It is important to consider all of these factors when using remote sensing for HABs. 

Slide 4 

The electromagnetic spectrum used for ocean color algorithms include the visible wavelengths from 400 nm to 700 nm (blue to red), some near infrared wavelengths and higher wavelengths for atmospheric corrections.  So, the way it works is light comes down from the sun and hits the water. About 90-95% of that light is absorbed by particles in the air as it makes its way down, so you lose a lot of the signal just from the atmosphere. Once the light hits the water it can do one of two things: it can scatter, or absorb, and that’s based on both the particles and the dissolved pigments that are in the water. Phytoplankton will absorb light based on the pigments it contains. CDOM and other particles containing absorbing compounds such as detritus also absorb light. And then particles of sediment, and organic material such as detritus and phytoplankton can scatter light. Some phytoplankton cells scatter light more than others. For instance, diatoms have a cell wall comprised of rigid silica, which causes a lot of scattering. What ends up getting returned to the satellite is a result of the light left over after the pigments have absorbed the light and the particles have scattered it. Also keep in mind, as light goes back thru the atmosphere, there’s more scattering and absorption going on.  

Slide 5 

You might ask, what affects the variations in color within the ocean? First of all, I should mention that what the satellite is seeing is called the first optical depth, which we tend to associate with about a secchi depth. So, the satellite sensor cannot see all the way to the bottom in turbid or deeper waters. The color you see is related to these absorption and scattering properties. For instance, far offshore the water appears very blue. This is because the water has less absorption in the blue wavelengths (400 nm) as there is less dissolved organic matter from land, and less phytoplankton and chla, which absorbs blue and red light. There are also less scattering particles, such as sediment, in the red wavelengths (600-700 nm) which would cause red reflectance back to the satellite. As you move closer to the coast, the increase in absorbing and scattering particles causes the blue Rrs to go down, as the red Rrs increases. The relative combination of these various scattering and absorbing materials is what affects the color of the ocean in different regions. So, as we see more phytoplankton in the water, we’ll see a shift from blue to more green, or if it’s a dinoflagellate maybe more red or brown. Sediment is very reflective, so in areas of high sediment you get more backscattering which will makes things appear brighter. What this means is that in coastal waters, things get a lot more complicated in trying to tease out the phytoplankton from all the other stuff that is in the water.  

Slide 6 

Here we are looking at reflectance measured by the Ocean Land Colour Imager (OLCI) aboard Sentinel-3. We are most interested in what is happening in the visible to NIR. OLCI is a multispectral imager which measures reflectance at discrete spectral bands. The green bars represent the placement of the spectral bands for OLCI. Along the reflectance curve you can see where different things are happening optically. The algorithms used for algal detection look at the changes in these peaks. You can think of reflectance as the inverse of absorption, so when there is more absorption occurring at a specific wavelength, there is less reflectance back to the satellite of that wavelength of light.  

Slide 7  

When looking at just the effects of chlorophyll a, which can provide an estimate of the amount of phytoplankton in the water, you may often hear of blue-green ratios. Common ocean color algorithms look at the changes in the blue to green peaks in the reflectance spectra to determine the concentration of chla in the water. From this figure, you can see that as the chla increases in concentration, the reflectance in the blue wavelengths is decreasing. This is the result of an increase in absorption by chla. Another thing to note is that there is a peak occurring around 680 nm which becomes more predominant as chla increases. This is related to a peak in chlorophyll-a fluorescence. 

Slide 8 

In ocean color, you will often hear waters classified as case 1 or case 2. Case-1 waters are those waters whose inherent optical properties (IOPs) are dominated by phytoplankton. This entails most open ocean waters, whereas Case-2 waters are all other waters, and these waters may be significantly influenced by constituents such as mineral particles, sediments and CDOM. A lot of ocean color algorithms are based on offshore case 1 water where the only effect on ocean color is the amount of phytoplankton in the water, but when you get closer to shore (class 2 waters) there are a lot of other things going on. HABs tend to occur in coastal, nearshore and freshwaters, which are strongly associated with optically complex waters. As a result, regional algorithms must be tuned to their unique optical and biological properties. There’s no one size fits all approach. So, what you’re looking at in the right figure is a traditional chlorophyll algorithm based on a blue/green ratio, and you can see how it picks up phytoplankton offshore well, but it over detects in the estuaries. On the left, a red-edge algorithm is used, which does really well at reducing errors in the estuaries but is unable to detect the phytoplankton offshore.  

Slide 9 

Chla absorbs light most strongly in the blue and red but poorly in the green portions of the electromagnetic spectrum. Most of this red and blue light is absorbed by the water fairly close to the surface. At greater depths, algae and cyanobacteria need help from accessory pigments. Accessory pigments determine both an organism’s color and its ideal depth. Cyanobacteria are often referred to as blue-green algae because of the accessory pigment phycocyanin which gives many of them a blue-green appearance. Phycocyanin is also extremely useful for discriminating cyanobacteria in ocean color applications. 

Slide 10 

Measuring algal abundance using satellite data depends on the type of algal bloom and the region of interest. Let’s begin by looking at the spectral image on the left from Lake Erie. When it comes to cyanobacteria, we look for phycocyanin and chla very specifically. Phycocyanin absorption occurs around 620 nm and is helpful for identifying cyanobacteria. Right around 681 nm is chla absorption, and cyanobacteria do not fluoresce well. The cells themselves have gas vacuoles which cause scattering around 709 nm. Therefore, we get a dip in the spectra at 681 nm as a result of this absorption and scattering and use this to quantify cyanobacteria blooms. In eukaryotes such as dinoflagellates, chlorophyll fluoresces at 681 nm, leading to increased remote sensing reflectance that obscures chlorophyll absorption. We see this in the spectra on the right for the dinoflagellate Margelefidinium, in the Chesapeake Bay estuary. Here even though the line is still decreasing at 620 nm, the curve around both 620 and 681 are positive. The changes in the shape of these key spectra can give us insight into the type of phytoplankton on the surface of the water.  

Slide 11 

In this example we see from the spectra that Karenia brevis fluoresces strongly at 681 nm with very little scatter. On the right we are looking at the fluorescence of Karenia brevis after applying a fluorescence algorithm. Fluorescence works particularly well in the clearer waters of the Gulf of Mexico.  

Slide 12 

The Cyanobacteria Index (CI) measures a proxy of chla absorption and provides the cyanobacterial biomass. It is calculated by assessing the spectral shape around 681 nm. In cyanobacteria, Chl-a absorption dominates the radiance signal from the water at 681 nm, causing the reflectance at 681 nm to decrease relative to 665 and 709 nm. 

Slide 13 

The “new bloom” anomaly is one method for bloom detection. This method relies on current satellite imagery and imagery from the previous 60 days. The difference of those two images identifies a bloom while eliminating any persistent false positives. This is the conventional method for K. brevis bloom detection by satellite. It is best for the beginning of a bloom and becomes ineffective during long duration blooms when high chlorophyll is no longer anomalous.  

Slide 14 

So based on everything I’ve discussed about scattering, absorption, and changes in fluorescence, we have a suite of algorithms that we use to separate out different blooms. This is really coming in handy in estuaries and coastal areas like Chesapeake Bay because there are multiple bloom formers that occur throughout the year and we’re trying to tease out what’s there (in order to get closer to what’s under the magnifying glass or microscope). If you want more info about the various algorithms, the citations are listed on the slide. 

Slide 15 

Products derived using the most applicable algorithm for a particular location are available for a number of regions. These can be viewed online or downloaded as GeoTIFF files. Additionally, a portal containing numerous derived products are available thru Coastwatch for the Chesapeake Bay. This portal was created as a response to Covid-19 when state agencies couldn’t get out on the water, but still needed to know what was happening out there.  

Slide 16 

The next two slides show how different algorithms can be used to characterize blooms. On this slide, in May of 2022, the Patapsco River region of Chesapeake Bay experienced a Prorocentrum minimum bloom. A Fluorescence algorithm, also called Relative Band Difference (or RBD) shows a bloom is there. Applying a red edge algorithm (in this case RE10) provides an indication of how much of a bloom is there. We also see that a cyanobacteria specific algorithm does not detect any bloom. 

Slide 17 

Margelefidinium polyrikoides produces a rust tide that is common in the York, Elizabeth, and James Rivers during the summer months. M. polykrikoides cause an issue for shellfish and in this area both aquaculture and recreation shellfish occur. In this example we see a bloom of Margelefidinium polyrikoides in the York River particularly well using a red edge (RE10) algorithm. Recall from earlier examples, M. polykrikoides fluoresces, so in the clearer open bay waters, we can use fluorescence to track bloom. 

Slide 18 

We are working on new higher spatial resolution algorithms to pick up the streakiness and patchiness of the blooms, and it gives us the ability to get into narrower areas and a better ability to delineate the size of the bloom. So here you see on top two sentinel 3 products alongside a true-color image and on the bottom an enhanced false color image from Sentinel 2. In the Sentinel 2 image you can clearly see a M. polykrikoides (rust tide) bloom which was confirmed by VIMS sampling at the 2 stars.  

Slide 19 

Several different satellites can detect HABs in fresh and salt water but there are tradeoffs in spatial, temporal, and spectral resolutions among the satellites. Current efforts to monitor and detect HABs with ocean color sensors include use of sensors such as the Ocean and Land Colour Instrument (OLCI) aboard Sentinel-3 that has a marginal spatial resolution of 300-meters, but very good temporal and spectral resolution. Sentinel-2 MultiSpectral Instrument (MSI) is also being explored for shorelines and narrow tributaries. Sentinel-2 has a lower temporal and spectral resolution but a significantly higher spatial resolution of 20-meters. Additionally, higher spatial resolution imagery of 3-meters can be obtained from satellites operated by Planet, but it is not available for government distribution.  

Slide 20 

Land is very bright compared to water due to scattering. So, we generally throw out pixels that are mixed water and land (for instance the footprint has land and water in them). In order to accurately resolve a bloom, it is best to have 3 water pixels across, which in the case of OLCI imagery is about 1 km wide. In this example, even though there are four pixels on the lake to the left that are solely water, the best we can do is 2 across without hitting a mixed pixel, so we would need to exclude that lake. The lake on the right has at least 3 pixels in a row that are solely water, so the lake can be identified. 

Slide 21 

The CyAN project is a multiagency cyanobacteria assessment project that was developed as an early warning indicator system to detect algal blooms in U.S. freshwater systems. The primary satellite sensors provided through CyAN are the current Sentinel-3 Ocean and Land Colour Instruments (OLCI), and their predecessor Medium Resolution Imaging Spectrometer (MERIS), run by EUMETSAT and the European Space Agency’s Copernicus program. These satellites provide daily images of the continental U.S. going back to 2002. Note, that MERIS stopped operating in April of 2012, so there is a gap in data coverage until the launch of Sentinel 3 in 2016. Due to their 300 m spatial resolution, we are able to pick up about 2000 (< 1%) of U.S. National Lake Assessment lakes and reservoirs and resolves 33% of public water surface intakes.  

Slide 22 

The CyAN app is available as two versions: CyANWeb app and the CyAN Android™ app. Both are free apps that require an internet connection and provide the same information using different platforms.  

Users can view cyanobacteria concentrations on a national-scale or can zoom in to see data for a specific lake or reservoir. Because states and localities may address HABs differently, the CyAN app and web-based interface allow users to set their own thresholds for cyanobacteria concentrations. Multiple water bodies can also be compared at once, allowing for better-informed decisions based on recent changes at specific locations. The web-based interface also provides the functionality to automatically cycle through available imagery for a user’s location, which can help provide some insight into how that area has changed over time.  

Slide 23 

New satellites equipped with ocean color imagers are coming online or scheduled to be launched within the next decade. The Plankton, Aerosol, Cloud, ocean Ecosystem (PACE) instrument, is a 1 km hyperspectral research instrument in Polar orbit, launched by NASA in February of 2024. PACE's primary sensor, the Ocean Color Instrument (OCI), is capable of monitoring global phytoplankton distribution and abundance with unparalleled spectral specificity. The first geostationary ocean color imager in the US, the Geostationary Littoral Imaging and Monitoring Radiometer (GLIMR), is anticipated to launch in 2027 to monitor a wide area, centered on the Gulf of Mexico. GLIMR will be able to capture multiple views per day in narrow wavebands to provide improved capacity in bloom monitoring and tracking as well as in bloom dynamics and forecasting. The next generation of US Geostationary satellites (GeoXO) launched by NOAA is expected to host an ocean color instrument (OCX). Launch of the first satellite is expected to come in the early 2030s and will watch over the US Economic Exclusive Zone (EEZ). This geostationary hyperspectral imager will view ocean, lake and coastal conditions more frequently and greatly improve the chance of cloud-free observations. Cloud cover is an issue over the nation’s most valuable fisheries, protected species populations and for coastal communities.  

Slide 24 

Here is a quick look at the different resolutions for the next generation of ocean color imagers. Recall OLCI aboard Sentinel-3, which is now widely used for HAB detection, has a spatial resolution of 300 m and a temporal resolution is 1-2 days. Unlike the hyperspectral imagers here, OLCI is a multi-spectral imager. In the next slide I’ll explain the difference.   

Slide 24 

Multispectral satellites have been the gold standard for ocean color sensing. Multispectral satellites can only capture data along 5-10 bands of the spectrum, most often all three primary colors and a few blocks in the infrared. On the other hand, a hyperspectral satellite can detect data along thousands of different bands within the light spectrum. Our hope is with hyperspectral data we can start to really separate out some of the phytoplankton species based on their pigment suite and their scattering properties. However, how much will still greatly depend on spatial resolution which we’ve already discussed. 

Slide 25 

In the remaining slides of this module, we will take a look at some regional models for HABs that have been developed to produce short-term forecasts (<1 week). Seasonal or other long-term forecasts are not satellite driven, and as such are not a focus of this module. Additionally, I will only briefly touch on validation in the final example. 

Slide 26 

Our first example looks at California, which has a mix of problems. Domoic acid is a significant one that not only impacts shellfish (mussels, farmed oysters), but also sea lions, pelicans, and sea otters. New detection methods (which are also used in the Pacific Northwest, especially by the Quileute and Quinault tribes), reduce the quantity of expensive analysis that must be conducted.  Current research indicates that conditions causing the blooms can be observed and forecasted. The Southern California Coastal Ocean Observing System (SCCOOS) in collaboration with the NOAA CoastWatch West Coast Node, developed a model that tries to predict domoic acid concentrations/levels along the coast. The bloom is forecasted using a series of satellite-derived products (reflectances, chl-a, salinity, and temperature), relationships established through logistic regression approaches, combined with a hydrodynamic model.  

Slide 27 

The forecast is run by CoastWatch, and the maps go out daily onto a website hosted by CeNCOOS, the Central and Northern California Coastal Ocean Observing System, who has been a major partner in this work since 2014. The online forecast shows the oceanography, cell density, and predicted toxin levels. In addition to the latest conditions and forecasted conditions out to 3 days, a portal allows users to look at all the previous predictions since it began. A survey helps us to address concerns or improve the way we display the maps and communicate with the public and with managers. 

Slide 28 

In our next example, western Lake Erie has been plagued by an increase of HAB intensity over the past decade. These blooms consist of cyanobacteria or blue-green algae, which are capable of producing toxins that can contaminate drinking water and pose a risk to human and animal health, foul coastlines, and impact communities and businesses that depend on the lake.  

Slide 29 

NOAA produces a weekly nowcast/forecast and a seasonal forecast for Lake Erie, typically from July to October when warmer water creates favorable bloom conditions. The weekly forecast relies on a hydrodynamic model developed by the Great Lakes Environment Research Lab, which is used to move the cells around and predict short-term movement over a 4-day period once a bloom is detected with satellite. The seasonal forecast predicts how bad the bloom is going to be based on the previous year using phosphorus data provided by Heidelberg University.  

Slide 30 

And for our final example, in recent years due to changes in ocean circulation and convergence zones, more Sargassum is entering the Caribbean and Gulf of Mexico where it strands along beaches causing considerable concern for beach economies, ecosystem health and even human health as bacteria accumulates in decaying sargassum and noxious gases are released during the decomposition. Because of the negative impacts that Sargassum may sometimes have on coastal communities, the Caribbean, Gulf of Mexico, and Atlantic OceanWatch Node of CoastWatch - housed at NOAA’s Atlantic Oceanic and Meteorological Laboratory (OAR/AOML) - produces a Sargassum Inundation Report (SIR) Links to an external site. in collaboration with the University of South Florida (USF). The Sargassum Inundation Report combines satellite observations, models, and in-situ data.  

Slide 31 

Several indices based on satellite imagery such as the Maximum Chlorophyll Index (MCI), Alternative Floating Algae Index (AFAI), or the multi-index (MSI) have been developed to detect Sargassum aggregations. These indices quantify the magnitude of red-edge reflectance of vegetation on the ocean surface using data derived from different satellites. 

Slide 32 

Data integration across satellites fills spatial, temporal, and spectral gaps for improved resolution.  

Slide 33 

The Sargassum Inundation Report (SIR) Links to an external site. is An experimental weekly risk assessment Links to an external site. for beaches at risk of inundation that combines satellite-based detections of the seaweed with ocean current models to identify coastal areas that may be at risk of being inundated by Sargassum. The Sargassum Inundation Report is available as a PDF, or a KMZ file can also be loaded into Google Earth Engine.  

Slide 34 

Validation for the product is conducted via in situ observations, much of which come from Community Science programs. The Caribbean, Gulf of Mexico, and Atlantic OceanWatch Node developed and maintains an OceanViewer Links to an external site. tool that maps satellite data, including Sargassum concentration. Additionally, NOAA maintains a database that collects digital photos and written descriptions of Sargassum from several repositories, including community science projects organized by NOAA and partners. These Sargassum observations are also a layer in the OceanViewer.  

Slide 35 

In conclusion, rules for processing and analyzing remotely sensed satellite data for HABs require consideration of optical, spatial, and ecological factors. Based on these factors a suite of algorithms has been developed. However, there are tradeoffs in temporal, spatial and key spectral resolution between satellites that must be considered. As a result, there is no one size fits all approach for detection and monitoring of HABs using satellites. HAB forecasts typically rely on three components. We briefly looked at three HAB forecasts that use satellite data to initiate the forecast models. Complimentary CoastWatch modules for additional information on satellite detection of HABs include Ocean Color and Water Quality.