ArticlePDF Available

Airborne hyperspectral and satellite imaging of harmful algal blooms in the Great Lakes Region: Successes in sensing algal blooms

Authors:
  • Environment and Climate Change Canada

Abstract and Figures

Harmful algal blooms have become a more significant issue in recent years in many lakes and rivers, and it is a particularly significant issue in the western basin of Lake Erie. In response, several research organizations in the United States and Canada have increased their efforts to improve capabilities for the remote sensing of harmful algal blooms. Efforts are underway to improve the ability to monitor, assess and study harmful algal blooms using various remote sensing platforms. Research into improvements in remote sensing platforms and algorithms provide powerful new tools to study and understand spatial and temporal aspects of harmful algal blooms. These developing tools will also help stakeholders to assess the efficacy of harmful algal bloom mitigation efforts into the future. In this commentary we describe selected NASA, NOAA, and ECCC's ongoing research projects as well as brief descriptions on the overall goal and specific objectives of the programs. Specific results from these three agencies' investigations, including modeling efforts, are discussed in a number of papers found within a special section of this issue.
Content may be subject to copyright.
Commentary
Airborne hyperspectral and satellite imaging of harmful algal blooms in
the Great Lakes Region: Successes in sensing algal blooms
John Lekki
a,
, Steve Ruberg
b
,CarenBinding
c
, Robert Anderson
a
, Andrea Vander Woude
d
a
National Aeronautics and Space Administration Glenn Research Center, United States of America
b
National Oceanic and Atmospheric Administration, Great Lakes Environmental Research Laboratory, United States of America
c
Environment and Climate Change Canada, Canada
d
Cherokee Nation Businesses at National Oceanic and Atmospheric Administration, Great Lakes Environmental Research Laboratory, United States of America
abstractarticle info
Article history:
Received 31 March 2019
Accepted 31 March 2019
Available online 2 May 2019
Communicated by Robert Hecky
Harmful algal blooms have become a more signicant issue in recent years in many lakes and rivers, and it is a
particularly signicant issue in the western basin of Lake Erie. In response, several research organizations in
the United States and Canada haveincreased their effortsto improve capabilities for the remote sensing of harm-
ful algal blooms. Efforts are underway to improve the ability to monitor, assess and study harmful algal blooms
using various remote sensing platforms. Research into improvements in remote sensing platforms and algo-
rithms provide powerful new tools to study and understand spatial and temporal aspects of harmful algal
blooms. These developing tools will also help stakeholders to assess the efcacy of harmful algal bloom mitiga-
tion efforts into the future. In this commentary we describe selected NASA, NOAA, and ECCC's ongoing research
projects as well as brief descriptions on the overall goal and specic objectives of the programs. Specicresults
from these three agencies' investigations, including modeling efforts, are discussed in a number of papers
found within a special section of this issue.
© 2019 Published by Elsevier B.V. on behalf of International Association for Great Lakes Research.
Keywords:
Hyperspectral sensors
Remote sensing
HABs
Lake Erie
Introduction
The concern for harmful algal blooms (HABs) in Lake Erie has peaked
in recent years. The bloom in the western basin of Lake Erie in 2014
reached such a severe level that it triggered the State of Ohio to declare
a state of emergency. Because NASA Glenn Research Center (GRC) had
recently developed an airborne hyperspectral imaging system, the cen-
ter was requested by stakeholders to use this capability to help assess
the blooms in Lake Erie. Beginning in 2014, lakeshore communities
and stakeholders have requested NASA Glenn's participationin HAB ob-
servation. Also, since 2014, the State of Ohio, National Oceanic and At-
mospheric Administration (NOAA), U.S. Environmental Protection
Agency (EPA), United States Geological Survey (USGS) and Environ-
ment and Climate Change Canada (ECCC) have elevated their funding
and increased their eld activities for observing, monitoring, and ad-
dressing the root causes of HABs. These activities have provided an op-
portunity for NASA to work with these other organizations and leverage
these activities to advance the science of airborne hyperspectral remote
sensing of HABs in general and cyanoHABs in particular.
NASA Glenn has conducted airborne hyperspectral harmful algal
bloom observation campaigns every year since 2014. Focusing mostly
on Lake Erie, but also on a limited set of small inland lakes as well as
the Maumee and Ohio Rivers, the campaigns have proceeded in part-
nership with a large number of collaborators with strong interest in ma-
rine science, limnology, and remote sensing. Objectives of the NASA
research include distinguishing potential HABs from smaller algal
growths, determining bloom biomass and associated phytoplankton
pigment concentrations, as well as tracking bloom distribution with
augmented spatial and temporal resolution. Optimization of methods
for necessary atmospheric correction and sensor calibration have also
been addressed (Avouris et al., this issue;Moore et al., this issue;
Sawtell et al., this issue).
NOAA has been investigating HABs occurrence in the Western Basin
of Lake Erie since the early 2000s using a combination of in situ sam-
pling, remote sensing, and modeling approaches. NOAA produces a
weekly operational HABs bulletin for Lake Erie that combines satellite
detection of HABs witha hydrodynamic model to predict HABs trajecto-
ries. Further development by NOAA scientists is focused on the imple-
mentation of an operational web accessible version of the HAB
Bulletin. The NOAA Great Lakes Environmental Research Laboratory
(GLERL) HABs program includes the measurement of Inherent Optical
Properties (IOPs) of Western Basin waters to improve the remote sens-
ing algorithms used to detect and monitor HABs. Since August 2015,
NOAA/GLERL added an aircraft based hyperspectral imaging element
to its HABs Lake Erie investigations after the 2014 Toledo water intake
Journal of Great Lakes Research 45 (2019) 405412
Corresponding author.
E-mail address: john.d.lekki@nasa.gov (J. Lekki).
https://doi.org/10.1016/j.jglr.2019.03.016
0380-1330/© 2019 Published by Elsevier B.V. on behalf of International Association for Great Lakes Research.
Contents lists available at ScienceDirect
Journal of Great Lakes Research
journal homepage: www.elsevier.com/locate/jglr
crisis as part of the EPA Great Lakes Restoration Initiative. The aircraft
ies weekly throughout the HAB growing season and produces a
Cyanobacteria Index (CI) of HAB severity for the stakeholder commu-
nity and is also working towardsdeveloping a method to distinguish be-
tween phytoplankton types. The aircraft collects data under the clouds
and nearshore, providing needed information to the water resource
managers (Vander Woude et al., this issue).
ECCC is alsoactive in the investigation of the remote sensing of algal
blooms in the Great LakeBasin and other Canadian large lakes. Scientists
with the Canadian Center for Inland Waters utilize European Space
Agency's Ocean and Land Colour Instrument(OLCI), and its predecessor
MERIS, to monitor, report on, and further understand drivers of HABs.
Scientists have developed a suite of satellite-derived algal bloom indices
to characterize bloom intensity, spatial extent and duration, and carry
out algorithm development using a combination of extensive in situ
eld sampling of blooms from the Canadian coastguard ship CCGS
Limnos, measures of Inherent Optical Properties and radiative transfer
modeling (Binding et al., this issue;Soontiens et al., this issue).
In the following pages we describe NASA, NOAA, and ECCC's ongoing
research projects as well as brief descriptions on the overall goal and
specic objectives of the programs. Specic results from these three
agencies' major scientic investigations are discussed. Supporting
these investigations were modeling efforts utilizing satellite-based re-
mote sensing observations which are also discussed in this issue (Xu
et al., this issue;Manning et al., this issue;Sayers et al., this issue-a).
NASA-Glenn Research Center Investigation
The collaboration between NASA Glenn and NOAA/GLERL began in
2006 when the rst Glenn Hyperspectral Imager (HSI) was utilized for
ight operations along with NOAA/GLERL in situ measurements. There-
sults from this early work found the signal-to-noise ratio (SNR) of the
rst imager to be inadequate. This imager was redesigned; and a second
imager (HSI2), with adequate SNR, was built and deployed beginning in
2007. During the second eld study, in situ reectance data were ob-
tained from the EPA's RV Lake Guardian in conjunction with reectance
data obtained with the hyperspectral imager from overights of the
same locations. Comparison of these two data sets shows that the air-
borne hyperspectral imager closely matched measurements obtained
from instruments on the lake surface (Beck et al., this issue) and thus
supports its use for detecting and monitoring harmful algal blooms
(HABs).
Fig. 1. NASA Twin Otter Aircraft.
Photo NASA.
Fig. 2. HSI2 & HSI3 hypers pectral radiometer mounted in Twin Otter on a platfor m
including the INS system.
Photo NASA Glenn (Roger Tokars).
Fig. 3. HABs in Put-In-Bay, Lake Erie in 2015.
Photo NASA Glenn (Roger Tokars).
Fig. 4. HABs in Maumee Bay, Lake Erie, August 25, 2017.
Photo NASA Glenn (Roger Tokars).
406 J. Lekki et al. / Journal of Great Lakes Research 45 (2019) 405412
The primary objective of these early ights was to demonstrate a ca-
pability for detecting various concentrations of the cyanobacterial pig-
ment phycocyanin as an indicator of cyanoHAB presence. In general,
the airborne measurements showed very good agreement with the in
situ measurements. The largest area of variance is a higher reectance
of blue light measured by the HSI2 than in the in situ data, which has
been attributed to atmospheric optical effects and continues to be an
issue for research.
In 2009, ights wereconducted to acquire concurrent water samples
and overight of 75 specic locations. This data, while not published,
has allowed the evaluation of the capability of some of the different sug-
gested techniques for HAB identication and concentration estimation.
In 2014, the Great Lakes Workshop Series on Remote Sensing of
Water Quality laid the foundation for open collaboration in developing
a regional working strategy for remote sensing techniques, applications,
and data management methods. One project that stemmed from the
workshop was a collaborative effort including airborne remote sensing
and in situ measurements by the following partners: NASA Glenn,
NOAA GLERL, Michigan Tech Research Institute (MTRI), University of
Toledo (UT), Kent State University (KSU), and the U.S. Naval Research
Laboratory. The University of Cincinnati later joined this collaboration
as the project sought to include more inland water locations.
The rst of three ights was conducted before Toledo experienced
the drinking water ban due to microcystin detected in their treated
water. Once this occurred, the team changed plans to incorporate both
bloom observation and data acquisition for research purposes. In all,
fteen HSI2 ights were completed in 2014. Each of these ights cov-
ered 12 water intake locations ranging from Cleveland to Toledo. Also,
pre-arranged water sample and study locations were overown when
there were in situ measurements being taken.More than 60 concurrent
measurements were obtained. The hyperspectral data that were con-
currently obtained with water sample data have been used for the eval-
uation of various HAB algorithms.
A specic objective of the Glenn HABs program is tracking of algal
bloom community composition, mapping spatial extent, and determin-
ing concentration over time. This information is necessary to provide
the basis for effective measures to mitigate impacts and eventually to
curtail blooms in the future. Mapping the spatial extent and concentra-
tion of algal blooms is important for characterizing the larger macro-
scale features of blooms since algae do not always develop as regions
of homogenous concentrations, but rather as aggregations that form
strands and lament-like structures in the water.
The use of remote sensing and, specically hyperspectral remote
sensing, is key to adequately monitoring and researching these algal
blooms. In situ data collection is crucial for studying HABs, but airborne
Fig. 5. HABs in Ohio River 2015.
Photo NASA Glenn (Roger Tokars).
Fig. 6. HABs in western Lake Erie September 11, 2017.
Photo NASA Glenn (Roger Tokars).
Fig. 7. HABs in Maumee River, Ohio, September 25, 2017.
Photo NASA Glenn (Roger Tokars).
Fig. 8. HABs in Harsha Lake, Ohio, 2017.
Photo NASA Glenn (Roger Tokars).
407J. Lekki et al. / Journal of Great Lakes Research 45 (2019) 405412
remote sensing signicantly augments in situ measurements for
obtaininghigh resolution measurements over larger areas. Flying closer
to the surface than an orbiting satellite allows for higher spatial resolu-
tion and ying at different altitudes can allow for changing the balance
between spatial resolution and imaged swath width as desired (Lekki
et al., this issue). Additionally, the use of unmanned aerial vehicles
(UAVs) was evaluated with respect to monitoring HABs and is reported
in this issue (Becker et al., this issue).
In 2014, 2015, 2016, and 2017 the NASA Glenn Research Center used
the S-3 Viking and the DHC-6 Twin Otter as platforms for hyperspectral
sensors in order to study remote sensing algorithms for HAB detection
and discrimination (Beck et al., this issue). In the 2017 campaign, two
HSI imagers (HSI2 and HSI3) were mounted in the Twin Otter and oper-
ated simultaneously (Figs. 1 and 2).
One objective of the research team was to perform detailed studies
of HABs in multiple inland water locations. To expand thebreadth of ap-
plicability of the remote sensing techniques being developed, it was
necessary to show that they would work in small lakes, rivers as well
as in the Laurentian Great Lakes. Examples of HAB observations in
these various water types are shown in Figs. 38.
The airborne HSI2 and HSI3 imagers have been developed in-house
at NASA Glenn Research Center. These pushbroom imagers collect
three-dimensional (3D) hyperspectral data in the 400 to 900 nm wave-
length range, which is necessary for harmful algae identication. Addi-
tional airborne instrumentation includes a spectroradiometer
mounted inside the aircraft and directed above to collect at-aircraft
downwelling irradiance. A similar instrument is used at ground level
to capture the difference in the irradiance caused by atmospheric af-
fects. The nal system component is an inertial navigation system
(INS) which provides sensor attitude data and Global Positioning Sys-
tem (GPS) information. These data are combined with the image data
and used for georeferencing during processing (Sawtell et al.,this issue).
Two imagers were own in 2017. The HSI2 and HSI3 imagers were
operationally similar. The improvements in the HSI3 design included
lower noise, increased spatial resolution (1260 vs 496 pixels across
the swath) and a wider eld of view (74° vs 12°). Figs. 912 compare
imaging swaths from 2017 ights over the western basin of Lake Erie
and Harsha Lake in southern Ohio. The benets of the wider HSI3
swath are clear, providing near complete images of the own area.
Fig. 9. HSI2 Imaging Swaths acquired October3, 2017 over western Lake Erie.Background
images from Google Earth.
(Data products from NASA GLENN).
Fig. 10. HSI3 Imag ing Swaths acquired October 3, 2017 over western Lake E rie.
Background images from Google Earth.
(Data products from NASA GLENN).
Fig. 11. HSI2 Imaging Swaths acquired October 2, 2017 over Harsha Lake. Background
images from Google Earth.
(Data products from NASA GLENN).
Fig. 12. HSI3 Imaging Swaths acquired October 2, 2017 over Harsha Lake. Background
images from Google Earth.
(Data products from NASA GLENN).
408 J. Lekki et al. / Journal of Great Lakes Research 45 (2019) 405412
HSI data processing is handled in a way that is timely, automated,
and effective at transforming raw data into a product relevant to end
users, such as researchers, municipal water treatment plant operators,
or park departments with affected beaches. During data collection, the
HSI systems collect pixel intensity counts that are stored in a matrix.
The operator may observe this matrix in real time as a waterfall
image of the area being scanned in order to verify that the system is op-
erating as expected. The HSI2 or HSI3 data are converted post-ight to
georeferenced radiance data (Sawtell et al., this issue).
An important element of the Glenn investigations was obtaining
ground and water surface measurements and water samples in coor-
dination with the ights. The eld data is needed to aid in radiomet-
riccalibrationoftheHSIimagersaswellastoverifytheremote
sensing derived HABs products. This activity was coordinated by
NASA Glenn and performed by partners external to NASA. Many of
these partners have papers in this volume (Avouris et al., this
issue;Bosse et al., this issue;Sayers et al., this issue-b). The surface
in situ measurements allow Glenn to anchor its hyperspectral data
to yield information regarding the algal composition, and in ex-
change, the eld partners can extrapolate their single-point mea-
surements to larger areas of interest using the ight data. Fig. 13
shows various examples of the ground measurements undertaken
as part of this program.
This cooperative research effort in HAB hyperspectral sensing in
Lake Erie, which involves U.S. and Canadian Federal and State organi-
zationsaswellasacademicinstitutions,canboastofnumerous
Fig. 13. (Clockwise from upper left)Deploying mirrorsfor radiometric calibration (Kent State University), shore-based radiometry (Kent State University), boat-based radiometry (NASA,
Larry Liou), parking lot-based radiometry for aircraft calibration (Michigan Tech Research Institute).
Fig. 14. Resonon Pika IIcamera (lower leftcorner) positionedin the back of the contracted
single engine airplane that ies outof the Ann Arbor Airportclose to NOAA GLERL. The full
system in the isolation pod is shown in the right-hand side of the image (NOAA GLERL).
Fig. 15. Example true-color geo-referenced hyperspectralimages from the Resonon Pika II
camera system. Lower left image includes lighthouse #2 (white object) in the western
basin of Lake Erie, where the AERONET-OC SeaPRISM is located.
(Data products from NOAA GLERL).
409J. Lekki et al. / Journal of Great Lakes Research 45 (2019) 405412
accomplishments as detailed in this volume. The team has assembled
one of the mostif not the mostcomprehensive sets of airborne
hyperspectral measurements with coincident in situ data. The data
has been made available to all collaborators through the creation of
a centralized database hosted by the Ohio Supercomputing Center.
Weekly ights and in situ sampling of the same body of water over
a period of 3 to 4 months has allowed for unique insights into how
the greater research community can use hyperspectral data to better
characterize water quality. Findings from this exercise have very real
implications for the future of aquatic remote sensing, including the
development and implementation of NASA's PACE (Pre-Aerosols
Clouds and ocean Ecosystems) mission in 2023, and the Canadian
Space Agency's hyperspectral microsatellite mission WATERSAT.
An important aspect of the project was improving remote sensing
capability through research towards better atmospheric correction of
remote-sensed data from hyperspectral imagers HSI2 and HSI3, devel-
opment of advanced algorithms for extracting more information from
the data, and evolution of a number of systematic improvements in
data processing. Next-day Quicklook HAB processing activity in 2015
was successful in rapidly turning raw airborne imagery into informative
data products for use by water resource managers and other decision
makers. The streamlined process included geometric and radiometric
correction (NASA Glenn), followed by atmospheric correction algorithm
implementation, and product reporting (Michigan Tech Research Insti-
tute, Sawtell et al., this issue).
The project has also explored several strategies for atmospheric
correction of the HSI data, of which the most positive results were
obtained from several vicarious calibration methods, such as the em-
pirical line method and the empirical mirror-based calibration
method. Theoretical calculations provided a sound basis for the ef-
fectiveness of the empirical mirror-based calibration method
allowing varimax-rotated principal component analysis (VPCA)
spectral decomposition method (Ortiz et al., this issue)tobeapplied
successfully to NASA HSI data. The results show that it is relatively
insensitive to the choice of reectance calculation method. This
provides a means of comparison between sensors of variable spec-
tral, spatial, and temporal sampling that will prove helpful for evalu-
ation of future hyperspectral missions (Ortiz et al., this issue).
In the upcoming years the research group plans to build upon these
advances, further transitioning the task of monitoring Lake Erie algal
blooms to NOAA GLERL and providing data products for inclusion into
the NOAA HAB bulletin. The project seeks tofurther improve data deliv-
ery as well as the research into atmospheric correction and improved al-
gorithms; for example, operationally testing the VPCA algorithm and
Fig. 16. Images during the duration of the NOAA-GLERL hyperspectral ights showing diversity of bloom macroscale aggregations. August 14, 2017, September 22, 2017, September 25,
2017 and July 9, 2018.
(Images taken by the pilot, Zachary Haslick, Aerodata Associates Photography, LLC).
Fig. 17. HABTracker forecastfor the western basin of LakeErie on September 19, 2016and
corresponding true-color image from the Resonon Pik a II on the same day, near the
Monroe drinking water intake location.
(Data products from NOAA GLERL).
410 J. Lekki et al. / Journal of Great Lakes Research 45 (2019) 405412
also focus research on real-time atmospheric correction. Some potential
areas to improve data dissemination include development of a better
method to share comprehensive data products produced from the HSI
data and also creation of a centralized geospatial Web site where HSI-
derived products and in situ data can be accessed. Finally, plans are to
provide this comprehensive set of information and remote sensing
techniques to the larger research community in order to improve the
monitoring and understanding of algal blooms in complex inland and
coastal waters.
NOAA/GLERL HABs investigations
In August 2015, NOAA GLERL began weekly ights over Lake Erie to
monitor the presence of HABs from a Resonon Pika II HSI positioned in
the fuselage of a single engine airplane (Fig. 14) with an Ocean Optics
downwelling irradiance cosine detector out of the top of the airplane
(Vander Woude et al., this issue).
During the HAB season, the airplane ies out of the Ann Arbor
airport every Monday over the Cooperative Institute for Great
Lakes Research (CIGLR) monitoring sites and the Ohio and Michi-
gan drinking water intake locations and bi-weekly over Saginaw
Bay. Fig. 15 shows an example true color image from the Resonon
Pika II camera system in the western basin of Lake Erie. Fig. 16
shows examples of a variety of HAB events photographed by the
pilot.
NOAA GLERL logged 8 ights in 2015, 28 ights in 2016, 30 ights in
2017 and 22 ights in 2018. In 2015 there were many processing steps
and lessons to transition to the publicly accessible product that NOAA
GLERL provides today. The rst processing steps included learning
how to geo-reference each of the images based upon the inertial mea-
surement unit (IMU) header data on pitch, roll and yaw of the airplane
and to provide easily accessible Google format les to drinking water
managers. The rst success story for the HSI was a true-color image
that in combination with the 3-dimensional HAB Tracker forecasting
system was used to warn the drinking water manager at the Monroe fa-
cility in Michigan, that a HAB was near their drinking water intake
(Fig. 17).
The Google based les are now used to display the true-color imag-
ery post-ight through an online mapping service on the NOAA GLERL
website (https://www.glerl.noaa.gov/res/HABs_and_Hypoxia/
airSatelliteMon.html). The Google based kml and kmz les transitioned
to a product-driven hyperspectral bulletin of cyanobacteria levels in
2017 and 2018 (Fig. 18), allowing ready comparison with the
cyanobacteria index widely used for the NOAA HAB bulletin.
Fig. 18. Example of the NOAA GLERL hyperspectral bulletin from September 17, 2018 which featured cyanobacteria levels from 9 different drinkingwater intakes in Michigan and Ohio.
(Data products from NOAA GLERL).
411J. Lekki et al. / Journal of Great Lakes Research 45 (2019) 405412
The hyperspectral bulletin was provided to the Ohio Environmental
Protection Agency to disseminate to the drinking water managers and
the Ohio governor, the Monroe drinking water manager and the Mich-
igan Department of Natural Resources as a valuable resource for weekly
tracking of HABs near water intakes.
Starting in 2017, NOAA GLERL also began ying over the SeaPRISM
9-channel AERONET-OC station at lighthouse number 2 in Lake Erie
and that station was instrumented with a 12-channel SeaPRISM in
2018 (Moore et al., this issue). Both instruments and the Michigan
Tech Research Institute radiometer at the Maumee State Park are cur-
rently being used to test a suite of atmospheric correction routines (6S
and the empirical line correction method, respectively; Sawtell et al.,
this issue). The SeaPRISM data and the CIGLR monitoring data (includ-
ing IOP and AOP data) are also used to validate the hyperspectral imag-
ery and algorithm developmentduring eachof the eld seasons (Sayers
et al., this issue-b).
In 2019, NOAA GLERL will offer an experimental phytoplankton func-
tional type product to municipalities that will map phytoplankton types
beyond cyanobacteria. NOAA GLERL also purchased a new HSI in 2018
that will offer a wider eld of view than the Pika II, with a 23-degree
eld of view for the lens chosen for the Pika L compared to the 17-
degree eld of view of the Pika II. The new capability to offer weekly phy-
toplankton functional types mapped beneath clouds and nearshore will
aid in ecosystem dynamic modeling and contribute towards plans to inte-
grate the hyperspectral data into the HAB Tracker forecasting data. NOAA
GLERL will continue to offer a valuable product to municipalities in the
Great Lakes, from a valuable perspective of a hyperspectral sensor.
References
Avouris, D., Ortiz, J., 2019. Validation of 2015 Lake Erie MODIS image spectral decomposi-
tion using visible derivative spectroscopy. J. Great lakes Res. (this issue).
Beck, R., Xu, M., Zhan, S., Johansen, R., Liu, H., Tong, S., Yang, B., Shu, S., Wu, Q., Wang, S.,
Berling, K., Murray, A., Emery, E., Reif, M., Harwood, J., Young, J., Nietch, C., Macke, D.,
Martin, M., Stillings, G., Stumpf, R., Su, H., Ye, Z., Huang, Y., 2018. Comparison of sat-
ellite reectance algorithms for estimating turbidity and cyanobacterial concentra-
tions in productive freshwaters using hyperspectral aircraft imagery. J. Great lakes
Res. (this issue).
Becker, R., Sayers, M., Dehm, D., Shuchman, R., Quintero, K., Bosse, K., Sawtell, R., 2019.
UAS based spectromet er for monitoring HABs: A new paradigm in water quality
monitoring. J. Great lakes Res. (this issue).
Binding, C., Zastepa, A., Zeng, C., 2018. The impact of phytoplankton community compo-
sition on optical properties and satellite observations of the 2017 western Lake Erie
algal bloom. J. Great lakes Res. (this issue).
Bosse, K., Sayers, M., Shuchman, R., Fahnens tiel, G., Ruberg, S., Fanslow, D., Stuart, D.,
Johengen, T., Burtner, A., 2019. Spatial-temporal variability of in situ cyanobacteria
vertical structure in western lake erie: implications for remote sensing observations.
J. Great lakes Res. (this issue).
Lekki, J., Deutsch, E., Sayers, M., Bosse, K., Anderson, R., Tokars, R., Sawtell, R., 2019. Deter-
mining remote sensing spatial resolution requirements for the monitoring of harmful
algal blooms in the Great Lakes. J. Great lakes Res. (this issue).
Manning, N., Wang, Y., Long, C., Bertani, I., Sayers, M., Bosse, K., Shuchman, R., Scavia, D.,
2019. Extending the forecast model: predicting harmfulalgal blooms at multiple spa-
tial scales. J. Great lakes Res. (this issue).
Moore, T., Feng, H., Ruberg, S., Beadle, K., Constant, S., Miller, R., Muzzi, R., Johengen, T.,
DiGiacomo, P., Lance, V., Holben, B., Wang, M., 2018. SeaPRISM observations in the
western basin of Lake Erie in the summer of 2016. J. Great lakes Res. (this issue).
Ortiz, J., Avouris, D., Schiller, S., Luvall, J., Lekki, J., Tokars, R., Anderson, R., Shuchman, R.,
Sayers, M., Becker, R., 2019. Evaluating visible derivative spectroscopy by varimax-
rotated, principal component analysis of aerial hyperspectral images from the west-
ern basin of Lake Erie. J. Great lakes Res. (this issue).
Sawtell, R.,Anderson, R., Tokars, R., Lekki,J., Shuchman, R., Bosse,K., Sayers, M., 2019. Real
Time HABs Mapping Using NASA Glenn HyperspectralImager. J. Great lakes Res. (this
issue).
Sayers, M., Grimm, A., Shuchman, R., Bosse, K., Fahnenstiel, G., Ruberg, S., Leshkevich, G.,
2019a. Satellite monitoring of harmful algal blooms in the wester n basin of Lake
Erie: a 20-year time-series. J. Great lakes Res. (this issue).
Sayers, M., Bosse, K., Shuchman, R., Ruberg, S., Fahnenstiel, G., Leshkevich, G., Stuart, D.,
Johengen, T., Burtner,A., Palladino, D., 2019b. Trends in spatial and temporal variabil-
ity of inherent optical properties in western Lake Erie: Implications for water quality
remote sensing. J. Great lakes Res. (this issue).
Soontiens, N., Binding, C., Fortin, V., Mackay, M., Rao, Y., 2018. Algal bloom transport in
Lake Erieusing remote sensingand hydrodynamic modelling: sensitivityto buoyancy
velocity and initial vertical distribution. J. Great lakes Res. (this issue).
Vander Woude, A., Ruberg, S., Johengen, T., Miller, R., Stuart, D., 2019. Hyperspectral y-
overs for warning and detection of harmful algal blooms in the Great Lakes: spatial
and temporal scales of variability. J. Great lakes Res. (this issue).
Xu, M., Liu, H., Beck, R., Lekki, J., Yang, B., Sh u, S., Kang, E., Anderson, R., Johansen, R.,
Emery, E., Reif, M ., Benko, T., 2019. A spectral space partition guided ensemble
method for retrieving chlorophyll-a concentration in inland waters from Sentinel-
2A satellite imagery. J. Great lakes Res. (this issue).
412 J. Lekki et al. / Journal of Great Lakes Research 45 (2019) 405412
... In particular, drinking-water-supply systems require real-time monitoring to maintain the water quality standard and to prevent unexpected accidents, including the malfunctioning of water treatment processes and contamination of raw water. The natural water in lakes, rivers, and the sea can be effectively managed by an in situ real-time monitoring system where area-based monitoring with multi-or hyper-spectral imaging sensors are increasingly applied to collect data for wide-ranging areas [8][9][10][11]. The recent advances of ICT and sensing technologies in environmental engineering enable reliable measurements, and the transmission and management of massive environmental data at low costs [5,12,13], which can support decision-making processes ( Figure 1). The technologies are already widely used in various fields, such as early warning systems for meteorological issues and public health protection [5,14,15]. ...
... Chl-a concentration is determined from the empirical relationship between HSI and Chl-a concentration [10,50] Phycocyanin Phycocyanin concentration is determined from the empirical relationship between HSI and phycocyanin concentration [8,50,51] Cyanobacteria biomass Cyanobacteria biomass concentration is determined from the empirical relationship between HSI and cyanobacteria biomass [8,10] Physical status for water quantity monitoring ...
... Chl-a concentration is determined from the empirical relationship between HSI and Chl-a concentration [10,50] Phycocyanin Phycocyanin concentration is determined from the empirical relationship between HSI and phycocyanin concentration [8,50,51] Cyanobacteria biomass Cyanobacteria biomass concentration is determined from the empirical relationship between HSI and cyanobacteria biomass [8,10] Physical status for water quantity monitoring ...
Article
Full-text available
Water quality control and management in water resources are important for providing clean and safe water to the public. Due to their large area, collection, analysis, and management of a large amount of water quality data are essential. Water quality data are collected mainly by manual field sampling, and recently real-time sensor monitoring has been increasingly applied for efficient data collection. However, real-time sensor monitoring still relies on only a few parameters, such as water level, velocity, temperature, conductivity, dissolved oxygen (DO), and pH. Although advanced sensing technologies, such as hyperspectral images (HSI), have been used for the areal monitoring of algal bloom, other water quality sensors for organic compounds, phosphorus (P), and nitrogen (N) still need to be further developed and improved for field applications. The utilization of information and communications technology (ICT) with sensor technology shows great potential for the monitoring, transmission, and management of field water-quality data and thus for developing effective water quality management. This paper presents a review of the recent advances in ICT and field applicable sensor technology for monitoring water quality, mainly focusing on water resources, such as rivers and lakes, and discusses the challenges and future directions.
... Recent advances in hyperspectral technology have empowered the development of relatively small, scientifically useful instruments for the diverse range of applications for stakeholders with limited budgets. Applications of these sensors have included agricultural vegetation modelling [4,5]; algal bloom investigation [6] and even deep space applications [7] as well as the possibility of carrying out remote geological surveys [8]. Multiple technologies have been devised and built to support these applications, including liquid crystal tunable filters [9], tunable Fabry-Perot filters [10] and diffractive optics [11,12]. ...
... One sensor design shall not include a Bayer filter. 5 The sensor and supporting electronics shall consume at most 15 W. 6 The sensor shall occupy a volume not exceeding 2 U (100 × 100 × 200 mm). 7 ...
Article
Full-text available
Near infrared (NIR) remote sensing has applications in vegetation analysis as well as geological investigations. For extra-terrestrial applications, this is particularly relevant to Moon, Mars and asteroid exploration, where minerals exhibiting spectral phenomenology between 600 and 800 nm have been identified. Recent progress in the availability of processors and sensors has created the possibility of development of low-cost instruments able to return useful scientific results. In this work, two Raspberry Pi camera types and a panchromatic astronomy camera were trialed within a pushbroom sensor to determine their utility in measuring and processing the spectrum in reflectance. Algorithmic classification of all 15 test materials exhibiting spectral phenomenology between 600 and 800 nm was easily performed. Calibration against a spectrometer considers the effects of the sensor, inherent image processing pipeline and compression. It was found that even the color Raspberry Pi cameras that are popular with STEM applications were able to record and distinguish between most minerals and, contrary to expectations, exploited the infra-red secondary transmissions in the Bayer filter to gain a wider spectral range. Such a camera without a Bayer filter can markedly improve spectral sensitivity but may not be necessary.
... The portability of satellite-derived algorithms as applied to imagery with higher spectral (hyperspectral), spatial, and temporal resolutions is not well understood particularly with respect to assessing the life cycle characteristics of cyanobacteria bloom events. For example, hyperspectral sensors have greater utility for detecting and quantifying HAB indicators as they have hundreds of narrow spectral bands which allows for the identification of spectral features characteristic of HABs: green reflectance (550 nm), phycocyanin absorption (620 nm), chlorophyll a absorption (665 nm -680 nm), and cell backscattering or turbidity (709 nm) (Davis and Bissett, 2007;Lekki et al., 2019;Shen et al., 2012;Stumpf et al., 2016). ...
Article
Full-text available
Remote sensing technologies offer a consistent, spatiotemporal approach to assess water quality, which includes the detection, monitoring, and forecasting of cyanobacteria harmful algal blooms. In this study, a series of ex-situ mesoscale experiments were conducted to first develop and then monitor a Microcystis sp. bloom using a hyperspectral sensor mounted on an unmanned aircraft system (UAS) along with coincident ground sampling efforts including laboratory analyses and in-situ field probes. This approach allowed for the simultaneous evaluation of both bloom physiology (algal growth stages/life cycle) and data collection method on the performance of a suite of 41 spectrally-derived water quality algorithms across three water quality indicators (chlorophyll a, phycocyanin and turbidity) in a controlled environment. Results indicated a strong agreement between Lab and Field-based methods for all water quality indicators independent of growth phase, with regression R 2-values above 0.73 for mean absolute percentage error (MAPE) and 0.87 for algorithm R 2 values. Three of the 41 algorithms evaluated met predetermined performance criteria (MAPE and algorithm R 2 values); however, in general, algal growth phase had a substantial impact on algorithm performance, especially those with blue and violet wave bands. This study highlights the importance of co-validating sensor technologies with appropriate ground monitoring methods to gain foundational knowledge before deploying new technologies in large-scale field efforts.
... The vast majority of previous studies have been based on conventional point-based sporadic monitoring of HAB cell numbers with sampling time intervals mostly relying on direct in situ sampling [11][12][13][14]. The recent advent of remote sensing methods, such as aerial hyperspectral images, has enabled researchers to capture the spatial extent of algal blooms and their evolution at a significant degree of accuracy [15][16][17][18][19][20]. In particular, aerial survey equipment using unmanned aerial vehicles (UAVs) has been widely applied to monitor and investigate HABs in rivers and lakes [17,18,[21][22][23][24][25][26][27][28][29]. ...
Article
Full-text available
Harmful algal blooms (HABs) have been recognized as a serious problem for aquatic ecosystems and a threat to drinking water systems. The proposed method aimed to develop a practical and rapid countermeasure, enabling preemptive responses to massive algal blooms, through which prior to the algal bloom season we can identify HAB-prone regions based on estimations of where harmful algae initiates and develops significantly. The HAB-prone regions were derived from temperature, depth, flow velocity, and sediment concentration data based only on acoustic Doppler current profilers (ADCPs) without relying further on supplementary data collection, such as the water quality. For HAB-prone regions, we employed hot-spot analysis using K-means clustering and the Getis-Ord G*, in conjunction with the spatial autocorrelation of Moran’s I and the local index of spatial association (LISA). The validation of the derived HAB-prone regions was conducted for ADCP measurements located at the downstream of Nam and Nakdong River confluence, South Korea, which preceded three months of algal bloom season monitored by unmanned aerial vehicles (UAVs). The visual inspection demonstrated that the comparison resulted in an acceptable range of agreement and consistency between the predicted HAB-prone regions and actual UAV-based observations of actual algal blooms.
Article
Full-text available
The hyperspectral imaging system (HSI) developed by the NASA Glenn Research Center was used from 2015 to 2017 to collect high spatial resolution data over Lake Erie and the Ohio River. Paired with a vicarious correction approach implemented by the Michigan Tech Research Institute, radiance data collected by the HSI system can be converted to high quality reflectance data which can be used to generate near-real time (within 24 h) products for the monitoring of harmful algal blooms using existing algorithms. The vicarious correction method relies on imaging a spectrally constant target to normalize HSI data for atmospheric and instrument calibration signals. A large asphalt parking lot near the Western Basin of Lake Erie was spectrally characterized and was determined to be a suitable correction target. Due to the HSI deployment aboard an aircraft, it is able to provide unique insights into water quality conditions not offered by space-based solutions. Aircraft can operate under cloud cover and flight paths can be chosen and changed on-demand, allowing for far more flexibility than space-based platforms. The HSI is also able to collect data at a high spatial resolution (~1 m), allowing for the monitoring of small water bodies, the ability to detect small patches of surface scum, and the capability to monitor the proximity of blooms to targets of interest such as water intakes. With this new rapid turnaround time, airborne data can serve as a complementary monitoring tool to existing satellite platforms, targeting critical areas and responding to bloom events on-demand.
Article
Full-text available
Timely identification of color-producing agents (CPAs) in Lake Erie is a challenging, but vital aspect of monitoring harmful algal blooms (HABs). In particular, HABs that include large amounts of cyanobacteria (CyanoHABs) can be toxic to humans, posing a threat to drinking water, in addition to recreational and economic use of Lake Erie. The optical signal of Lake Erie is complex (Becker et al., 2009; Moore et al., 2017), typically comprised of phytoplankton, cyanobacteria, colored dissolved organic matter (CDOM), detritus, and terrigenous inorganic particles, varying in composition both spatially and temporally. The Kent State University (KSU) spectral decomposition method effectively partitions CPAs using a varimax-rotated, principal component analysis (VPCA) of visible reflectance spectra measured using lab, field or satellite instruments (Ali et al., 2013; Ortiz et al., 2017, 2013). We analyze 2015 imagery acquired by the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor and field samples collected during the early 2015 cyanoHAB season. We identified four primary CPA spectral signatures, and the spatial distribution of each identified CPA, in the reflectance spectra datasets of both the MODIS and lab-measured water samples. The KSU spectral decomposition method results in mixtures of specific pigments, pigment degradation products, and minerals that describe the optically complex water. We found very good agreement between the KSU VPCA spectral decomposition results and in situ measurements, indicating that this method may be a powerful tool for rapid CyanoHAB monitoring and assessment in large lakes using instruments that provide moderate resolution imagery (0.3 to 1 km ² ).
Article
Full-text available
NOAA GLERL has routinely flown a hyperspectral imager to detect cyanobacteria harmful algal blooms (cyanoHABs) over the Great Lakes since 2015. Three consecutive years of hyperspectral imagery over the Great Lakes warn drinking water intake managers of the presence of cyanoHABs. Western basin imagery of Lake Erie contributes to a weekly report to the Ohio Environmental Protection Agency using the cyanobacteria index (CI) as an indicator of the presence of cyanoHABs. The CI is also used for the weekly NOAA NCCOS cyanoHAB Lake Erie bulletin applied to satellite data. To date, there has not been a sensor comparison to look at the variability between the satellite and hyperspectral imagery on a pixel-by-pixel basis, as well as a time scale comparison between measurements from buoys and shipboard surveys. The spatial scale is a measure of size of a cyanobacteria bloom on a scale of meters to kilometers. The change in the spatial scale or spatial variability has been quantified from satellite and airborne imagery using a decorrelation scale analysis to find the point at which the values are not changing or are not correlated with each other. The decorrelation scales were also applied to the buoy and shipboard survey data to look at temporal scales or changes in time on hourly to daytime scales for blue-green algae, chlorophyll and temperature. These scales are valuable for ecosystem modelers and for those initiating sampling efforts to optimize sampling plans and to infer a potential mechanism in an observational study from a synoptic viewpoint.
Article
Full-text available
Remote sensing has provided expanded temporal and spatial range to the study of harmful algal blooms (cyanoHABs) in western Lake Erie, allowing for a greater understanding of bloom dynamics than is possible through in situ sampling. However, satellites are limited in their ability to specifically target cyanobacteria and can only observe the water within the first optical depth. This limits the ability of remote sensing to make conclusions about full water column cyanoHAB biomass if cyanobacteria are vertically stratified. FluoroProbe data were collected at nine stations across western Lake Erie in 2015 and 2016 and analyzed to characterize spatio-temporal variability in cyanobacteria vertical structure. Cyanobacteria were generally homogenously distributed during the growing season except under certain conditions. As water depth increased and high surface layer concentrations were observed, cyanobacteria were found to be more vertically stratified and the assumption of homogeneity was less supported. Cyanobacteria vertical distribution was related to wind speed and wave height, with increased stratification at low wind speeds (<4.9 m/s) and wave heights (<0.27 m). Once wind speed and wave height exceeded these thresholds the assumption of vertically uniform cyanobacteria populations was justified. These findings suggest that remote sensing can provide adequate estimates of water column cyanoHAB biomass in most conditions; however, the incorporation of bathymetry and environmental conditions could lead to improved biomass estimates. Additionally, cyanobacteria contributions to total chlorophyll-a were shown to change throughout the season and across depth, suggesting the need for remote sensing algorithms to specifically identify cyanobacteria.
Article
Full-text available
Blooms of harmful cyanobacteria (cyanoHABs) have occurred on an annual basis in western Lake Erie for more than a decade. Previously, we developed and validated an algorithm to map the extent of the submerged and surface scum components of cyanoHABs using MODIS ocean-color satellite data. The algorithm maps submerged cyanoHABs by identifying high chlorophyll concentrations (>18 mg/m³) combined with water temperature >20 °C, while cyanoHABs surface scums are mapped using near-infrared reflectance values. Here, we adapted this algorithm for the SeaWiFS sensor to map the annual areal extents of cyanoHABs in the Western Basin of Lake Erie for the 20-year period from 1998 to 2017. The resulting classified maps were validated by comparison with historical in situ measurements, exhibiting good agreement (81% accuracy). Trends in the annual mean and maximum total submerged and surface scum extents demonstrated significant positive increases from 1998 to 2017. There was also an apparent 76% increase in year-to-year variability of mean annual extent between the 1998–2010 and 2011–2017 periods. The 1998–2017 time-series was also compared with several different river discharge nutrient loading metrics to assess the ability to predict annual cyanoHAB extents. The prediction models displayed significant relationships between spring discharge and cyanoHAB area; however, substantial variance remained unexplained due in part to the presence of very large blooms occurring in 2013 and 2015. This new multi-sensor time-series and associated statistics extend the current understanding of the extent, location, duration, and temporal patterns of cyanoHABs in western Lake Erie.
Article
Full-text available
Since the early 2000s Lake Erie has seen a dramatic increase in phytoplankton biomass, manifested in particular by the rise in the severity of cyanobacteria blooms and the prevalence of potentially toxic taxa such as Microcystis. Satellite remote sensing has provided a unique capacity for the synoptic detection of these blooms, enabling spatial and temporal trends in their extent and severity to be documented. Algorithms for satellite detection of Lake Erie algal blooms often rely on a single consistent relationship between algal or cyanobacterial biomass and spectral indices such as the Maximum Chlorophyll Index (MCI) or Cyanobacteria Index (CI). Blooms, however, are known to vary significantly in community composition over space and time. A suite of phytoplankton and optical property measurements during the western Lake Erie algal bloom of 2017 showed highly diverse bloom composition with variable absorption and backscatter properties. Elevated backscattering coefficients were observed in the Maumee Bay, likely due to phytoplankton cell morphology and buoyancy regulating gas vacuoles, compared with typically Planktothrix dominated blooms in Sandusky Bay. MCI and CI calibrated to historical chlorophyll observations and applied to Sentinel 3's OLCI sensor accurately captured the 2017 bloom in Maumee Bay but underestimated the Sandusky Bay bloom by nearly 80%. The phycoerythrin-rich picocyanobacteria Aphanothece and Synechococcus were found in abundance throughout the western and central basins, resulting in substantial biomass underestimations using blue to green ratio-based algorithms. Potential misrepresentation of bloom severity resulting from phytoplankton optical properties should be considered in assessments of bloom conditions on Lake Erie.
Article
Harmful algal blooms (HABs) have become a major health and environmental concern in the Great Lakes. In 2014, severe HABs prompted the State of Ohio to request NASA Glenn Research Center (GRC) to assist with monitoring algal blooms in Lake Erie. The most notable species of HAB is Microcystis aeruginosa, a hepatotoxin producing cyanobacteria that is responsible for liver complications for humans and other fauna that come in contact with these blooms. NASA GRC conducts semiweekly flights in order to gather up-to-date imagery regarding the blooms' spatial extents and concentrations. Airborne hyperspectral imagery is collected using two hyperspectral imagers, HSI-2 and HSI-3. Hyperspectral imagery is necessary in order to conduct experiments on differentiation of algal bloom types based on their spectral reflectance. In this analysis, imagery from September 19, 2016 was utilized to study the subpixel variability within the footprint of arbitrary sized pixels using several analysis techniques. This particular data set is utilized because it represents a worst case scenario where there is significant potential for public health concern due to high concentrations of microcystin toxin found in the water on this day and the concurrent observational challenges to accurately measure the algal bloom concentration variability with a remote sensing system due to the blooms high spatial variability. It has been determined that the optimal spatial resolution to monitor algal blooms in the Great Lakes is at most 50 m, and for much lower error 25 m, thus allowing for greater ease in identifying high concentration blooms near the surface. This resolution provides the best sensitivity to high concentration areas that are of significant importance in regard to human health and ecological damage.
Article
Lake Erie has experienced dramatic changes in water quality over the past several decades requiring extensive monitoring to assess effectiveness of adaptive management strategies. Remote sensing offers a unique potential to provide synoptic monitoring at daily time scales complementing in-situ sampling activities occurring in Lake Erie. Bio-optical remote sensing algorithms require knowledge about the inherent optical properties (IOPs) of the water for parameterization to produce robust water quality products. This study reports new IOP and apparent optical property (AOP) datasets for western Lake Erie that encapsulate the May–October period for 2015 and 2016 at weekly sampling intervals. Previously reported IOP and AOP observations have been temporally limited and have not assessed statistical differences between IOPs over spatial and temporal gradients. The objective of this study is to assess trends in IOPs over variable spatial and temporal scales. Large spatio-temporal variability in IOPs was observed between 2015 and 2016 likely due to the difference in the extent and duration of mid-summer cyanobacteria blooms. Differences in the seasonal trends of the specific phytoplankton absorption coefficient between 2015 and 2016 suggest differing algal assemblages between the years. Other IOP variables, including chromophoric, dissolved organic matter (CDOM) and beam attenuation spectral slopes, suggest variability is influenced by river discharge and sediment re-suspension. The datasets presented in this study show how these IOPs and AOPs change over a season and between years, and are useful in advancing the applicability and robustness of remote sensing methods to retrieve water quality information in western Lake Erie.
Article
Lake Erie is a classic case of development, recovery from, and return to eutrophication, hypoxia, and harmful algal blooms. Forecast models are used annually to predict bloom intensity for the whole Western Lake Erie Basin, but do not necessarily reflect nearshore conditions or regional variations, which are important for local stakeholders. In this study we: 1) developed relationships between observed whole basin and nearshore bloom sizes, and 2) updated and extended a Bayesian seasonal bloom forecast model to provide new regional predictions. The western basin was subdivided into 5 km near-shore regions, and bloom start date, size, and intensity were quantified with MODIS-derived images of chlorophyll concentrations for July–October 2002–2016 for each subdivision and for the entire basin. While bloom severity within each subdivision is temporally and spatially unique, it increased over the study period in each subdivision. The models for the 5 km subdivisions explained between 83 and 95% of variability between regional sizes and whole bloom size for US subdivisions and 51% for the Canadian subdivision. By linking predictive basin-wide models to regional regression estimates, we are now able to better predict potential bloom impacts at scales and in specific areas that are vital to the economic well-being of the region and allow for better management responses.
Article
The Kent State University (KSU) spectral decomposition method provides information about the spectral signals present in multispectral and hyperspectral images. Pre-processing steps that enhance signal to noise ratio (SNR) by 7.37–19.04 times, enables extraction of the environmental signals captured by the National Aeronautics and Space Administration (NASA) Glenn Research Center's, second generation, Hyperspectral imager (HSI2) into multiple, independent components. We have accomplished this by pre-processing of Level 1 HSI2 data to remove stripes from the scene, followed by a combination of spectral and spatial smoothing to further increase the SNR and remove non-Lambertian features, such as waves. On average, the residual stochastic noise removed from the HSI2 images by this method is 5.43 ± 1.42%. The method also enables removal of a spectrally coherent residual atmospheric bias of 4.28 ± 0.48%, ascribed to incomplete atmospheric correction. The total noise isolated from signal by the method is thus <±7% based on analysis of multiple swaths. The method is essentially independent of the order of operations, extracting the same spectral components within error in all cases, spatial patterns that are very similar and explaining nearly identical amounts of information from each image. Based on comparison between multiple swaths the uncertainty in the variance associated with each component averages ±1.69% and is as low as ±0.08% and in all cases <±3.15%.