Remote sensing of assimilation number for marine phytoplankton
Stéphane Saux Picart, Shubha Sathyendranath, Mark Dowell, Trevor Platt
3rd Coastcolour user consultation meeting - October 2011
Aim of the study
Challenge: Estimate phytoplankton primary production from space
Primary production can be computed
using a photosynthesis-light model:
B)
P B (z) = P B (I (z); αB , Pm
Superscript B indicates
normalisation to chlorophyll
biomass B.
P B Normalised production;
z: depth; I : irradiance;
αB : initial slope;
B : assimilation number
Pm
Aim of the study
Challenge: Estimate phytoplankton primary production from space
Primary production can be computed
using a photosynthesis-light model:
B)
P B (z) = P B (I (z); αB , Pm
Superscript B indicates
normalisation to chlorophyll
biomass B.
P B Normalised production;
z: depth; I : irradiance;
αB : initial slope;
B : assimilation number
Pm
Specific objective of the study:
Estimate the assimilation number globally from remote sensing data
General approach
I
B
If no photo-inhibition, maximum primary production Pmax = BPm
I
Maximum primary production is also given by Pmax =
where Cp is the phytoplankton carbon concentration
I
Maximum growth rate: µmax =
dCp
|
dt max
1 dCp
|
Cp dt max
Cp
B =µ
Assimilation number: Pm
max B = χµmax
where χ is the carbon-to-chlorophyll ratio
General approach
I
B
If no photo-inhibition, maximum primary production Pmax = BPm
I
Maximum primary production is also given by Pmax =
where Cp is the phytoplankton carbon concentration
I
Maximum growth rate: µmax =
dCp
|
dt max
1 dCp
|
Cp dt max
Cp
B =µ
Assimilation number: Pm
max B = χµmax
where χ is the carbon-to-chlorophyll ratio
Eppley (1972) defines the maximum
growth rate as a function of
temperature:
µmax = 0.851(1.066T )
ln 2
24
Three models
Two ways of estimating carbon-to-chlorophyll ratio:
(1.81+0.63∗log10 (Chl))
1. Sathyendranath et al. 2004: χs = 10
Chl
(gives an upper limit of χ)
i−1
h
[N]
2. Cloern et al. 1995: χc = 0.003 + 0.0154 exp(0.05T ) exp(−0.059I ) K +[N]
N
where N is the nutrient concentration
Three models
Two ways of estimating carbon-to-chlorophyll ratio:
(1.81+0.63∗log10 (Chl))
1. Sathyendranath et al. 2004: χs = 10
Chl
(gives an upper limit of χ)
i−1
h
[N]
2. Cloern et al. 1995: χc = 0.003 + 0.0154 exp(0.05T ) exp(−0.059I ) K +[N]
N
where N is the nutrient concentration
Combining these two model and the maximum growth rate formulation from Eppley (1979), we
can approach computation of assimilation number as follows:
I
B = f (Chl, T ) = χs µ
Model 1: Pm
max
I
B = f (I , N, T ) = χc µ
Model 2: Pm
max
I
Model 3:
h
B = f (I , N, Chl, T ) = (χs )−1 + 0.0154 exp(0.05T ) exp(−0.059I )
Pm
K
[N]
N +[N]
i−1
µmax
Results: in-situ
B , Chl, T , mixed layer depth,
Model applied to a large database (> 700 measurements): Pm
surface PAR, nitrates
=⇒ Gaps filled with climatological data (for surface PAR and nitrates)
=⇒ North West Atlantic, subtropics (Gulf of Mexico and Arabian Sea), Middle Atlantic
Results: in-situ
B , Chl, T , mixed layer depth,
Model applied to a large database (> 700 measurements): Pm
surface PAR, nitrates
=⇒ Gaps filled with climatological data (for surface PAR and nitrates)
=⇒ North West Atlantic, subtropics (Gulf of Mexico and Arabian Sea), Middle Atlantic
North West Atlantic
Model 1
f (Chl, T )
Model 2
f (I , N, T )
15
15
10
15
10
bias = 0.59
rmse = 1.59
r
= 0.71
5
00
20
PmB estimated [mgC(mgChl)−1 h−1 ]
20
PmB estimated [mgC(mgChl)−1 h−1 ]
PmB estimated [mgC(mgChl)−1 h−1 ]
20
Model 3
f (I , N, Chl, T )
North West Atlantic
5
10
15
PmB measured [mgC(mgChl)−1 h−1 ]
20
10
bias = -0.25
rmse = 1.85
r
= 0.51
5
00
5
10
15
PmB measured [mgC(mgChl)−1 h−1 ]
20
bias = 1.67
rmse = 2.34
r
= 0.73
5
00
5
10
15
PmB measured [mgC(mgChl)−1 h−1 ]
20
Results: in-situ
B , Chl, T , mixed layer depth,
Model applied to a large database (> 700 measurements): Pm
surface PAR, nitrates
=⇒ Gaps filled with climatological data (for surface PAR and nitrates)
=⇒ North West Atlantic, subtropics (Gulf of Mexico and Arabian Sea), Middle Atlantic
North West Atlantic, subtropics and Middle Atlantic
Model 1
f (Chl, T )
Model 2
f (I , N, T )
15
15
10
bias = 0.59
rmse = 1.59
r
= 0.71
Arabian S/G. Mexico
off European shelf
North West Atlantic
5
10
15
PmB measured [mgC(mgChl)−1 h−1 ]
15
10
bias = -1.49
rmse = 4.13
r
= 0.51
5
00
20
PmB estimated [mgC(mgChl)−1 h−1 ]
20
PmB estimated [mgC(mgChl)−1 h−1 ]
PmB estimated [mgC(mgChl)−1 h−1 ]
20
Model 3
f (I , N, Chl, T )
20
bias = -0.25
rmse = 1.85
r
= 0.51
5
00
10
bias = -0.77
rmse = 2.30
r
= 0.52
5
10
15
PmB measured [mgC(mgChl)−1 h−1 ]
20
bias = 1.24
rmse = 2.22
r
= 0.57
bias = 1.67
rmse = 2.34
r
= 0.73
5
00
5
10
15
PmB measured [mgC(mgChl)−1 h−1 ]
20
Results: remote sensing application
Data used (all monthly composites), 2004:
I
SeaWiFS cholophyll-a concentration
I
MODIS sea surface temperature
I
SeaWiFS surface PAR
I
World ocean atlas surface nitrate concentration
Results: remote sensing application
Data used (all monthly composites), 2004:
I
SeaWiFS cholophyll-a concentration
I
MODIS sea surface temperature
I
SeaWiFS surface PAR
I
World ocean atlas surface nitrate concentration
Month: 04
20
15
10
5
0
0
5
10 15 20
Temperature [ ◦ C]
Month: 07
25
in-situ data
Model 1: PmB = f(Chl,T)
Model 2: PmB = f(I,N,T)
Model 3: PmB = f(I,N,Chl,T)
PmB [mgC(mgChl)−1 h−1 ]
PmB [mgC(mgChl)−1 h−1 ]
25
25
30
20
15
10
5
0
0
5
10 15 20
Temperature [ ◦ C]
25
30
Results: remote sensing application
Data used (all monthly composites), 2004:
I
SeaWiFS cholophyll-a concentration
I
MODIS sea surface temperature
I
SeaWiFS surface PAR
I
World ocean atlas surface nitrate concentration
Month: 10
20
15
10
5
0
0
5
10 15 20
Temperature [ ◦ C]
Month: 11
25
in-situ data
Model 1: PmB = f(Chl,T)
Model 2: PmB = f(I,N,T)
Model 3: PmB = f(I,N,Chl,T)
PmB [mgC(mgChl)−1 h−1 ]
PmB [mgC(mgChl)−1 h−1 ]
25
25
30
20
15
10
5
0
0
5
10 15 20
Temperature [ ◦ C]
25
30
Results: remote sensing application
I
Model 1 (f (Chl, T )) overestimates
B in oligothrophic waters
Pm
I
Model 2 and 3 have stronger
latitudinal seasonal variations
because of the light
I
B are too
Model 3: estimates of Pm
low
Model 1
f (Chl, T )
Model 2
f (I , N, T )
Model 3
f (I , N, Chl, T )
Results: remote sensing application
I
Model 1 (f (Chl, T )) overestimates
B in oligothrophic waters
Pm
I
Model 2 and 3 have stronger
latitudinal seasonal variations
because of the light
I
B are too
Model 3: estimates of Pm
low
Model 1
f (Chl, T )
Model 2
f (I , N, T )
Model 3
f (I , N, Chl, T )
Implementation in coastal waters
B = f (B, T , N, I )
Assimilation number: Pm
B
I
I (0), K
T
N
=⇒
=⇒
=⇒
=⇒
=⇒
Coastcolour product (case 2 water)
I (z) = I (0) exp(−Kz)
Coastcolour products (case 2 water)
Satellite sea surface temperature
Climatological data (World Ocean Atlas)
Conclusion
I
B
We have tested three different models for estimating Pm
I
Comparison with in-situ data has shown three contrasting results, and each model has skill in
a different region of the globe
I
Overall Model 2 (f (I , N, T )) seems to be able to estimate the assimilation number
reasonably well at global scale
Future work:
I
Develop a model to estimate the initial slope of the photosynthesis-light curve (αB )
I
Estimate primary production
Aknowledgments
CoastColour - ESA project
NERC Earth Observation Data Acquisition and Analysis Service (NEODAAS)
Thank you
Contact
Stéphane SAUX PICART: [email protected]