Hunting red tides from sea and space

1
Hunting red tides from sea and space Jennifer
P. Cannizzaro, David C. English and K.L.
Carder College of Marine Science, University of
South Florida, St. Petersburg, FL 33701
1. Introduction
4. Application to shipboard data
5. Application to satellite data
Independent shipboard data were collected during
four WFS and two RT research cruises
(2005-2006) on the west Florida shelf to test our
K. brevis bloom detection technique.
Classification criteria for detecting K. brevis
blooms were developed. The spatial and temporal
variability of optical properties for the WFS1006
cruise are examined.
Harmful algal blooms (HAB) of the toxic
dinoflagellate Karenia brevis occur regularly in
the Gulf of Mexico, typically in late summer and
fall along the west Florida shelf (WFS). K.
brevis blooms can cause bird, fish, and marine
mammal mortalities along with human eye and
respiration irritation. As a result, tourism and
commercial fishing industries are negatively
impacted when blooms occur, with Florida losses
estimated at 25M/year.
Moderate Resolution Imaging Spectroradiometer
(MODIS) data collected aboard the Aqua satellite
(1km resolution) and FWRI K. brevis cell
concentration data for 2005 were used to validate
our HAB detection technique. The 2005 bloom event
with manatee deaths in March, dead marine life
washing ashore in July, and a severe benthic
mortality event in August was the worst in Tampa
Bay since 1971.
Data
Data

• MEASURMENTS
• Above-water Rrs(?)
• Surface discrete data
• Chl
• a(?) (phytoplankton, detritus, and CDOM)
• K. brevis cell concentrations (Florida Fish and
  Wildlife Research Institute (FWRI))
• Surface underway data
• temperature and salinity
• beam attenuation (480,660nm)
• Chl and CDOM fluorescence
• bbp(488,676nm)

• MEASUREMENTS
• K. brevis cell concentration data
• filtered based on sample depth (surface only),
  location (estuarine and lt25oN omitted) and bottom
  depth (gt5m only)
• Kareniaceae (K.brevis, K. mikimotoi, K.
  papilionacea, K. selliformis, K.spp.) combined
• MODIS data products (SeaDAS ver. 5.0)
• Chl (OC3M OReilly et al.,00)
• bbp(551) (Carder et al., 99) (NOTE Usage of
  alternative backscattering algorithms that
  perform
• better in absorption-rich waters
  or that correct for bottom reflectance leads to
  decreased
• separation between bloom and
  non-bloom data in Chl vs bbp(550) plots and
  therefore are
• not recommended.)
• FLH (fluorescence line height) (Letelier and
  Abott, 96) (NOTE useful for discriminating
  between
• true phytoplankton bloom and
  false high-chlorophyll CDOM-rich regions Hu et
  al., 2005)

To mitigate the human health risks, ecosystem
damage, and negative economic impacts associated
with K. brevis blooms, an accurate system for
detecting blooms is desired. Here, we present
the development and application of an optical
technique for detecting K. brevis blooms that can
provide near-real time data across a wide range
of spatial and temporal scales from a variety of
ocean-observing platforms (e.g. ships, moorings,
AUVs, gliders, aircrafts, satellites, etc.).

• Station locations for 2005-2006 cruises

• 2005 FWRI K. brevis cell concentration data
  locations

2. Background
Classification criteria
Classification criteria
K. brevis blooms can appear various shades of
color depending on how light is absorbed and
scattered by the particulate and dissolved
material in the ocean. The relationship between
remote sensing reflectance, Rrs(?), (a measure of
ocean color) and the absorption, a(?), and
backscattering, bb(?), properties of seawater can
be described as follows

• Classification criteria (shipboard data)
• Chl gt 1.5 mg m-3 and
• bbp(550) / Chl lt 0.0065 m2 mg-1

• Classification criteria (satellite data)
• Chl gt 1.0 mg m-3,
• FLH gt 0.01 mW cm-2 ?m-1 sr-1 and
• bbp(551) lt Morel (1988) relationship
• 81 of bloom data (gt104 cells l-1) flagged
  successfully

85 of bloom data (gt104 cells l-1) successfully
flagged
where the terms ph, d, CDOM, w, and p
represent phytoplankton, detritus, colored
dissolved organic matter, water, and
particulates, respectively.

• Spectral variability
• aw(?), bbw(?) constant and known
• aph(?) 440 and 675nm peaks due to chl a and
  accessory pigments
• aCDOM(?) and ad(?) ? exponentially with ??
• bbp(?) ? or remains constant with ??

Spatial and temporal variability
Challenge Can we differentiate between toxic and
non-toxic phytoplankton blooms from optical
measurements of oceanic waters?
Classification process
STEP MODIS DATA OMIT
FLAGGED IMAGE
3. Development of detection technique
(1) (2) (3)
Latitude (deg. N)
7 Oct. 2006

• Ecology and Oceanography of Harmful Algal Blooms
  (ECOHAB) station locations (1998-2001)
• Optics measured on eighteen cruises
• Data collected Rrs(?), aph(?), ad(?), aCDOM(?),
  bbp(?), chlorophyll a concentrations (Chl), and
  K. brevis cell counts

Chl lt 1 mg m-3 (oligotrophic regions) FLH
lt 0.01 mW cm-2 ?m-1 sr-1 (regions with high
apparent Chl, but low actual Chl
(e.g. Big Bend region)) bbp(551) gt Morel 88
function (regions containing
other algal blooms, ? bottom reflectance, ?
suspended sediments)
Chlor_a

• Spatial variability of WFS1006 data collected 2-6
  October 2006 with FWRI K. brevis cell
  concentrations (cells l-1) overlaid on upper
  right plot
• Chl and salinity poor bloom indicator
• bbp(550) good bloom indicator (i.e. low values
  generally associated with high concentrations)

• K. brevis blooms on the WFS are 4xs less
  reflective than waters with similar chlorophyll
  concentrations and lower K. brevis cell
  concentrations

7 Oct. 2006
FLH
GOAL to determine which factor (aph, ad, aCDOM,
/or bbp) is responsible for this decreased
reflectance
7 Oct. 2006
Oct. 4-7 FWRI data ? lt 104 cells l-1 ? gt 104
cells l-1
bbp(550)
6. Conclusions

• Best-fit linear regression relationships for K.
  brevis bloom (gt104 cells l-1 ------) and non-K.
  brevis bloom (lt104 cells l-1 ?? ) data
• are significantly different for Chl vs
  aph(520-570), ad(443), and bbp(550)
• are not significantly different for Chl vs
  aph(400-520, gt570) and aCDOM(400)
• Model results (not shown) indicate that relative
  to non-K. brevis bloom Rrs(?)
• lower aph and ad coefficients in K. brevis blooms
  will increase Rrs(?) slightly and
• lower bbp coefficients in K. brevis blooms will
  decrease Rrs(?) significantly
• THEREFORE,
• Differences in BACKSCATTERING (and not
  absorption) are responsible for the 4-fold lower
  Rrs(?) values observed in K. brevis blooms and
• Chl and bbp measurements (e.g. in situ) or
  estimates from Rrs(?) (e.g. satellites) can be
  used to distinguish between toxic and non-toxic
  blooms

• K. brevis blooms can be detected accurately from
  shipboard (85 accuracy) and satellite (81
  accuracy) chlorophyll and backscattering data
• Classification criteria differ depending on the
  data source (e.g. shipboard or satellite)
• Flagging for 1) high CDOM using river discharge,
  salinity, CDOM fluorescence, /or FLH data, 2)
  storm history using wind data, and 3) bottom
  reflectance (Cannizzaro and Carder, 2006) can
  help reduce misclassification events

• Temporal variability of WFS1006 data (2-6 October
  2006). FWRI cell counts(cells l-1) shown in top
  plot. Chl and aCDOM/aph determined from underway
  Chl and CDOM fluorescence data.
• gt104 K. brevis cells l-1 typically associated
  with Chls gt 1.5 mg m-3 and bbp(550)s lt 0.0065
  m2 mg-1
• Alternative detection techniques (e.g.
  microscopy, genetic probes, etc.) required in
  Charlotte Harbor where aCDOM/aph values exceeded
  5 and salinity values dropped below 30 psu

REFERENCES Cannizzaro, J.P., Carder, K.L., Rem.
Sens. Env. 101 13-24 (2006). Letelier, R.M.,
Abbott, M.R., Rem. Sens. Env. 58 215-223
(1996). Cannizzaro, J.P., et al. Cont. Shelf Res.
(in press).
Morel, A., J. Geophys. Res. 93 10,749-10,768
(1988). Carder, K.L., et al., J. Geophys. Res.
104 5403-5422 (1999). O'Reilly,
J.E., et al., NASA Tech. Memo. 2000-206892,
Vol.11, pp.9-23, (2000). Hu, C. et al., Rem.
Sens. Env. 97 311-321 (2005)
Wynne et al., Harm. Algae
4992-1003 (2005)
(details in Cannizzaro et al., in press)