MODIS: Fluorescence, Primary productivity, and some applications V. Banzon

1
MODIS Fluorescence, Primary productivity, and
some applicationsV. Banzon

• Climate change oceans
• Ocean net primary productivity
• Passive natural fluorescence
• Particulate inorganic carbon
• Future Applications/Algorithms
• red tides
• dust

2
Global carbon cycle increase in atmospheric CO2
contributes to global warming climate change
SLOW processes occur in ocean, which provides
long-term CO2 REMOVAL mechanism
3
Primary production is a main mechanism for carbon
fixation
Biomass? Rates? Feedback?
4
Need to characterize ocean Variability and
Response to climate change

• will the oceans role in carbon cycling change in
  terms of
• Changes in circulation and temperature
• Shifts in ecosystem structure and carbon export
• (e.g., in analogy to vegetation shifts on land in
  response to precipitation changes)

5
MODIS OCEAN NET PRIMARY PRODUCTION (ONPP)
W.E. Esaias, NASA GSFC K.R.TURPIE, D. THOMAS,
R. VOGEL, A. BHATTI July 23, 2002 MODIS Science
Team Meeting wayne.esaias_at_gsfc.nasa.gov http//mo
dis-ocean.gsfc.nasa.gov
6
MODIS Ocean Net Primary Productionprovides
measure of C fixation by phytoplankton2 Models
P1 and P2
P1 Behrenfeld Falkowski VGCM NPP f(Chl a,
PAR, Pb opt) Integrated over the Euphotic zone
(1) Pb opt f (SST) 7th order
polynomial P2 Howard, Yoder, Ryan NPP f
(Chl a, PAR, Pmax) Integrated over the upper
Mixed Layer Depth (MLD) Pmax (Platt) f (SST)
Eppley Peterson exponential Produced for 8-day
intervals at 4.6 km resolution. Binned files -
4.6 km equal area, Map files - 4.6 km , 36 km ,
1 degree in linear lat-lon fields - P1, P2, MLD,
PAR, SST, Chl parameters - mean, sd, n, N,
quality flags
7
ONPP INPUT FIELDS
Chlor_a_3 MODIS Chlorophyll concentration
MLD (FNMOC) Mixed Layer Depth (for P2)
January 25, 2001
SSTD MODIS SST day
PAR (GSFC DAO) Photosynthetically Available
Radiaiton
January 25, 2001
8
P1 Behrenfeld-Falkowski
Daylength (hrs)
Optimal photosynthetic yield
Depth of euphotic zone
9
P2 Howard-Yoder-Ryan
Depth of mixed layer
Carbon fixation/volume over depth of mixed layer
Avg radiative energy over mixed layer depth
a .1124
10

• P1
• 31 Oct- 7Nov 2000
• (Austral Spring)
• P2
• lt-higher

11
Difference between P1 P2 primarily due to MLD
effect in temperate zones Southern Ocean
12
Oceanic regions used to assess trends in primary
production
13
Chl_a3 uses nutrient depletion temperature which
increases chlorophyll relative to SeaWiFS
like chl_a2 when surface temperature lt
climatological value
South Pacific
14
ONPP_P1 shows seasonality in different basins
-e.g., difference between N S
hemispheres -noisy if number of data pixels is
small
15
ONPP estimates affected by data quality
0 best data, only oligotrophic, open ocean
(case 1) waters 1 good data includes high
productivity coastal (case 2) waters 2 fair
Examples of quality levels
QL0
QL1
QL2
16
We recommend that modellers use QL 0,1 for
ONPP includes shallow regions greater coverage
17
Annual ONPP is based on empirical relation from
SeaWiFS
1999
-to investigate interannual and long-term changes
2000
18
FUTURE new P1 formulation will include nutrient
and light limitation (compare mixed vs. euphotic
depth)
new Pbopt f(Z1 , MLD, Nut)
M.J.Behrenfeld
lt MLD
gt MLD
19
Pbopt from PhotAcc (NDT)
PhotoAcc model provides more realistic estimate
of photosynthetic yield
Polynomial
Pbopt (mgC / mgChl / h)
20
15
10
5
25
0
20
ONPP with PhotoAcc Model, Nut. Depletion T
Use of PhotoAcc results in better representation
of seasonality
January July
Behrenfeld et al.
21
MODIS Estimates of Passive Fluorescence

• Ricardo Letelier, Mark Abbott,
• Jasmine Nahorniak
• College of Oceanic and Atmospheric Sciences
• Oregon State University

Acknowledgment Robert Evans et al. University
of Miami
22
Outline

• What is fluorescence (F) and what is its
  relation to photosynthesis?
• Measurement of chlorophyll natural fluorescence
  from space and validation
• Using fluorescence to estimate sea surface
  chlorophyll concentration
• Using fluorescence to improve Ocean Primary
  Productivity algorithms

23
Light Harvesting, Fluorescence and Photosynthesis
Light energy not used for photosynthesis is lost
as heat and fluorescence
Fp Ff Fh 1
24
Blue light induced chlorophyll fluorescence in
Tobacco leaf. (From Krause and Weis, 1988)
photosynthesis has been blocked by the herbicide
duiron (DCMU). No DCMU
A. photographed in white light.
B. taken in the low steady state of fluorescence,
5 min after the onset of illumination.
u
e-
LHC
PSI
(ATP NADPH2)
L683
heat
Fp Ff an be estimated using DCMU
Assume Fp Ff Fh 1 Fh constant
u
DCMU
LHC
PSI
L683
heat
25
Why do we want to use fluorescence as another
measure of chlorophyll?

• ADVANTAGE Absorption-based algorithms fail in
  waters where there are other materials that
  absorb and scatter and are not correlated with
  chlorophyll
• Sediment
• Dissolved organic matter
• Chlorophyll fluorescence is specific to
  chlorophyll
• LIMITATION it also depends on physiology

26
Absorption signal gtgt Fluorescence signal
Fluorescence can be expected to be useful only in
high chl conditions.
MOCE-5 TSRB reflectance at the time of CTD cast
Light absorption by algal pigments
Light emitted by chlorophyll
Oligotrophic -----gthighly productive
27
Regular method to calculate Chl fluorescence
use Fluoresence Line Height
Sea surface Upwelling irradiance (calculated
using 10 mg Chl m-3 )
FLH Lu683 Baseline Baseline Lu1 -
(Lu1-Lu2)/(lLu2-lLu1)(683-lLu1)
28
MODIS FLH bands avoid oxygen absorbance at 687
nm
Weighting factor used to compensate for
off-center FLH
29
MODIS successfully estimates FLH from space even
in low chlorophyll case 1 waters
MODIS FLH
Goddard DACC weekly declouded 36 km starting
08/05/2001 (Quality0 L2 V 4.2.2)
30
Validation using airborne laser
A
B
C
(From Frank Hoge)
31
If MODIS can measure natural Fluorescence, what
information can be derived?

• F Chl x (PAR x a) x fF
• where F fluorescence
• chl chlorophyll concentration
• PAR photosynthetically available radiation
• a chlorophyll specific absorption
• ?F fluorescence quantum yield
• ARP a x chl x PAR gt a ARP / (chl x
  PAR)
• where ARP absorbed radiation by phytoplankton
• Chl Fluorescence Efficiency (CFE) F/ARP
• ? ?F
  

32
Historical Use of Natural Fluorescence

• Estimation of chlorophyll concentration
• assumes ?F ? constant
• Estimation of primary production
• assumes a predictable relationship
• between ?F and ?p

33
MODIS Chlorophyll and In situ fluorescence
mg m-3
Latitude
Longitude
34
Field measurements and Remote Sensing Chl data
are well-correlated (Mesoscale Survey August
2000And MODIS Image from August 2)
MODIS chl_2 (mg m-3)
MODIS FLH, W m-2 um-1 sr-1
In situ chl (mg m-3)
In situ chl (mg m-3)
(In situ chl derived from the calibration of the
flow through fluorometer with HPLC chlorophyll
determinations ) -Blue all mesoscale survey
data (July 31st August 7th) -Red Within 0.5
days of the MODIS Image Time stamp
35
OSU Direct Broadcast October 04, 2001
MODIS_Chl MODIS_FLH MODIS_CFE
MODIS_ARP
MODIS data shows chl not always in spatial
correspondence to Fluorescence
Physiological parameters also vary spatially
36

• Brief summary of the use of FLH to
• estimate chlorophyll concentrations
• FLH appears to vary consistently with sea surface
  chl concentration,
• BUT some significant differences in spatial
  distribution of chlorophyll and fluorescence
  present in MODIS images.
• These differences may be due to
• Differences in the optical depth for the
  derivation of chlorophyll
• concentration and that of chlorophyll
  fluorescence
• 2) Variability in the specific absorption of
  chlorophyll (a) and the fluorescence yield
  (ff).
• 3) Error propagation due to the sensitivity of
  the algorithms involved.

37
SeaWiFS-estimated primary productivity
Satellites provide good spatial-temporal coverage
but
Present chlorophyll-based productivity models
-focus on light absorption processes -attempt
to use physical data and/or climatology to
estimate light utilization potential (SST,
MLD, etc.)
38
Antarctic satellite-based Estimates of Primary
Productivity vary by factor of 2
Similar situation for global estimates
Most of the variability in estimates is due to
the uncertainty in the physiological parameters
embedded in the models.
39
East Coast Image 2001095.1605
a ARP / (chl x PAR)
CFE F/ARP ? ?F
Variability in a
Variability in CFE (or Ff)
Fluorescence-related MODIS products have
physiological information and show variability of
photosynthetic process
40
Can we use MODIS CFE to improve the Primary
Productivity algorithm?
PP chl x (PAR x a) x Fp (1) If
Fp Ff Fh 1 Fh constant then Fp
constant Ff (2) Replacing Fp with (2)
in (1) PP chl x (PAR x a) x (constant
Ff) or PP ? ARP x (constant - FLH/ARP)
? (constant x ARP) - FLH
CFE
41
In Situ Observations of F/chl suggest it can be
a proxy for ff
Initial slope proportional to ?F
42
East Coast Image 2001095.1605
Oregon Coast DB Image 2001150
offshore
inshore
Shows 2 principal groups with different CFE May
indicate source of new production (inshore group)
43
FLH/chl vs. Fv/Fm as Function of SST
FLH / chl, W m-2 µm-1 sr-1 (mg m-3)-1 (proxy
for Ff)
Cold SST shows upwelling group, warm SST indicate
offshore group
44
Conclusions

• Chlorophyll fluorescence can be measured
  accurately from space.
• Variability in phytoplankton physiological
  parameters such as a (chl specific absorption)
  and Ff (fluorescence yield) can be assessed.
• Variability in FLH signal and ARP (absorbed
  radiation by phytoplankton) is dominated by
  chlorophyll but, the ratio of ARPFLH is not
  constant and reflects changes in Ff.
• Variability in Ff provides information regarding
  the variability in Fp. However, more research is
  needed to develop regional algorithms that will
  improve our estimates of primary production from
  space

45
Coccolith-Calcite Concentrations Validation and
Status

• William Balch, Bruce Bowler, Dave Drapeau, Emily
  Booth, Joaquim Goes
• Bigelow Laboratory for Ocean Sciences
• W. Boothbay Harbor, ME
• Howard Gordon and Katherine Kilpatrick
• University of Miami,Miami, FL

46
Road Map

• What are coccoliths and coccolithophores?
• MODIS Products
• Detached coccolith concentration 21
• Particulate Inorganic Carbon (PIC) 22
• Pigment concentration in coccolithophore blooms
  20
• Basis of the Gordon two-band PIC algorithm
• Ship measurements of PIC and backscattering
• Chalk-experiment
• Natural bloom
• Satellite validation of the PIC algorithm

47
Sources of scattering- various species of
coccolithophores
SEMs courtesy of Dr. Delors Blasco, Institute
de Ciencias del Mar, Barcelona, Spain Other
in-water scattering source sediments
48
Particulate Inorganic Carbon increases nLw by
increasing in-water scattering
49
Among all PIC, coccoliths likely play the major
role in light scattering, especially at 412-550nm
and when gt10,000 ml-1.
50
The 2-band PIC algorithm look-up table
nLw(550) (mW cm-2 um Sr)
nLw(440) (mW cm-2 um Sr)
51
PIC vs bb relationship the basis of the 2-band
algorithm-1991 Iceland coccolithophore bloom
52
Data limits suggested better accuracy at high
PIC
Is data spread due to effect of different of size
shape particles?
53
Also need some higher PIC concentrations Chalk-ex

• Do it yourself coccolithophore bloom
• Need only 13T of coccolith chalk to make a patch
  visible from space
• Most of marine sediments are chalk so
  environmental impact is minimal
• Can deploy on clear-sky days
• no problem of scheduling ships around rare bloom
  events!

54
Chalk concentration is highly correlated to its
backscattering
55
Loading Chalk In Portland, ME
56
Spreading 0500h-0930h, steaming in an expanding
ellipse, 1.5 x 0.5 km
57
MODIS view of Chalk-Ex Patch 2 551nm, 1Km
data, 15 November 2001
Two patch pixels 39.81oN x 67.78oW (9.73 W
m-2 um-1 sr -1 ) 39.80oN x 67.76oW (10.24 W
m-2 um -1 sr -1 )
58
Ship-measured/contoured surface bb showing four
most intense MODIS pixels
59
How do satellite-derived estimates compare to
ship measurements?
Ship chalk-ex
SeaWIFS Gulf of Maine
MODIS chalk-ex
MODIS Gulf of Maine
60
Mother natures chalk-ex
61
Ship and satellite 7/02 bloom observations
MODIS
62
Putting all the data together
Algorithm RMS error still about 28 ugC/L due to
natural variability in calcite particles and
their bb
63
Summary

• PIC algorithm works, as defined using original
  Iceland coccolithophore bloom.
• Is PIC product validated?
• The absolute RMS accuracy is /- 30mg C/L for
  mixtures of unknown PIC particles.
• The relative precision is much better in features
  with one type of particle. For example, Gulf of
  Maine bloom where algorithm detected relative
  variability of PIC at levels of 5 mg PIC/L, or
  Chalk-Ex with uniform PIC particles.

64
Satellite identification of red tidesK.Carder,
J. Cannizzaro, F.R.Chen
If Rrs(?) ? bb(?) / a(?) where bb(?) bbp (?)
bbw(?) and a (?) ap(?) ag(?) aw(?)
Note bbw(?) and aw(?) are knowns Then,
how do ap(?), ag(?), and bbp(?) influence Rrs(?)
in K. brevis blooms???
Red tides pose a human health risk and can harm
marine resources
65
Particulate absorption at 443nm is highly
correlated with Chl a (Note Major K. brevis
pigments are similar to other chl-c containing
phytoplankton) Therefore, pigment absorption in
K. brevis blooms is not unique and cannot be
used to identify red tide blooms from space. K.
Brevis blooms exhibit low chlorophyll -specific
detrital absorption at 443nm perhaps indicative
of low grazing.
66
N470
83
K. Brevis blooms (?)exhibit low pheo/chl ratios
another indicator of low grazing
67
EcoHAB 1999 - 2001
Estuarine high chl, high backscatter
Oligotrophic low chl, low backscatter
Red tide high chl, low backscatter
68
Surface Underway Data No Red Tide until after
EH0801
69
MODIS semi-analytical algorithm
Outputs Chl a, aph(443), ag(400), and
bbp(550) Allows us to examine the bbp(550) vs
chl relationships without interference by
gelbstoff absorption.
70
3 October 2001 443, 488, 551nm SeaWiFS
False color does not show where red tide bloom is
71
3-band image R chl a/bbp(550) G chl a B
bbp(550) Whitebottom effects Pinkish huesred
tide-rich
October 3, 2001 SeaWiFS image
72
bbp vs chl a (MODIS algorithm products)
10/3/01 SeaWiFS image
Flagged pixels (red) indicate red- tide-rich
regions w/o false flagging
73
K. brevis dominated regions on 3 October flagged
red using bb vs. chl mask
74
(No Transcript)
75
Summary K. brevis dominated regions have low
Pheo/Chl, low detritus, and low
backscattering per unit chlorophyll due to
reduced grazing in toxin-rich waters.
gelbstoff plumes can also exhibit low
reflectivity leading to false flagging if chl and
gelbstoff are not separated. MODIS algorithms
separate gelbstoff from chlorophyll and provide
particle backscattering coefficients.
Field-validated imagery of low chlorophyll-specifi
c backscattering regions with ag(440) / aph(440)
lt 2.0 were dominated by K. brevis patches with
concentrations from 1 13 mg Chl m-3
76
Application of the Spectral Matching Algorithm
to Recover Chlorophyll Time Series During the
Arabian Sea SW Monsoon
by V. Banzon, R. Evans, H. Gordon and R. Chomko
25 N
25 N
15 N
15 N
50 E
60 E
70 E
50 E
60 E
70 E
77
Year 2000 Monthly Chl_a
mg/m3
Standard Processing -gt Data Gaps due to Cloud
Masking
78
2000 Aerosol Index (TOMS)
Jan
Feb
Mar
Apr
May
Jul
Jun
Aug
Sep
Oct
Nov
Dec
ftp//jwocky.gsfc.nasa.gov/pub/eptoms/images/month
ly_averages
79
How to increase data retrieval?

• Disable masking CLOUD and HIGHLT
• Apply spectral matching algorithm (SMA)
• Offers a range of absorbing aerosol (dust) models
• Uses full spectral information to select model
• Simultaneously computes for best solution for
  water and atmospheric components

80
DUST has more absorption in blue relative to
green band
More constant at bands gt 670 nm Thus, cannot
distinguish from scattering models using near-IR
only
81
Lt(l) Lr(l) La(l) Lra(l) tLw(l,C)

• Spectral Matching
• Bottom-of the-atmosphere approach
• Atmosphere and in-water solved simultaneously
• 18 absorbing models (or 16 non-absorbing)
• First-guess chl needed for bio-optical model (a,
  bb),
• Model spectrum (generated using t, aerosol model,
  integrated to TOA)
• compared with nIR and visible

• Standard
• Top-of-the-atmosphere (TOA) approach
• Atmospheric term solved first
• 15 non-absorbing models 1 weakly absorbing
• nIR bands used to select aerosol model

82
1 Apr 2000 No dust
mg/m3
Std Chl
S M A
SMA Chl
Std
40
20
60
80
gt100
0
Data Density
83
28 Jul 2000 Some dust
mg/m3
Std Chl
S M A
SMA Chl
Std
0
40
20
60
80
gt100
Data Density
84
Weekly Chl Standard SeaWiFS processing
Masks LAND, HIGLINT, ATMFAIL, HIGHLT, CLDICE
85
Weekly Mean Chl SMA processing
Masks LAND, HIGLINT, EPSILON (ATMFAIL)
86

• Problem Cannot use scattering and absorbing
  aerosol models simultaneously
• Test Solution for SeaWiFS
• use 510670 ratio limited to open water
• Possibilities for MODIS
• use 2.1 mm band or other aerosol product

87
(No Transcript)
88
Dust Event 18 July 2000
89
Dust Event 17 July 2000