Remote Assessment of Remote Assessment of Phytoplankton Functional TypesPhytoplankton Functional Types

Using RetrievalsUsing Retrievalsof the of the Particle Size DistributionParticle Size Distribution

from Ocean Color Datafrom Ocean Color Data

Tihomir Kostadinov, David Siegel, Stéphane MaritorenaTihomir Kostadinov, David Siegel, Stéphane Maritorena

ICESS, University of California Santa BarbaraICESS, University of California Santa Barbara

NASA Ocean Color Research Team Meeting,NASA Ocean Color Research Team Meeting,

New Orleans, LA, May 12, 2010New Orleans, LA, May 12, 2010

OutlineOutline

• Introduction & Motivation• Kostadinov et al. (2009) PSD Algorithm

– Algorithm theoretical basis & operation– Uncertainty analysis– Validation

• Phytoplankton Functional Types Retrieval (Biogeosci. Disc., submitted)

– Definition of PFT’s– Validation– Global climatology, seasonal succession

Why PFT’s are ImportantWhy PFT’s are Important• PFT’s are groups of

phytoplankton with simil2耀 biology & biogeochemical roles, e.g.:– physiology– sinking – CO2 sequestration– DMS production– silicate drawdown

• Cell SIZE– is a characteristic feature of

PFT’s– determines structure and

function of pelagic ecosystems

• Global RS retrieval of the PFT’s is needed

Chisholm, 2000

PSD’s PSD’s PFT LinkPFT Link• PSD’s # & V in any size class• Case I assumption – particle load dominated by

Chl and covariates• Size-defined PFT in terms of % volume = f(PSD

parameters):– 3 classes – pico, nano, micro– definition does not explicitly take into account

taxonomy/biology

• Existing methods for PFT retrieval are based on HPLC pigments & Chl (e.g. Uitz, Alvain); phytoplankton absorption (e.g. Mouw, Devred)

Describing the PSDDescribing the PSDPower-law Junge-type

Size Distribution

= PSD slope

Do = 2 m

No = N(Do), [m-4]

oo D

DNDN )(

Example PSD measured by LISST-100X

July 21, 2008Santa Barbara Channel

California

: 3.91No: 16.7 m-4

log1

0 o

f

34o12.26’N 119o55.69’ W

Link to Optics - Mie Scattering TheoryLink to Optics - Mie Scattering Theory

• Single particle optical properties depend on:– Complex index of refraction mr() = nr – i*nr’()

– Size relative to the incident wavelength– Shape & internal composition

• Mie modeling solves the Maxwell equations for the IOP’s of homogeneous spherical particles

dDD

DNmDQDb

oobb

D

D

bp

),,(

4)(

max

min

2

Retrievable spectrallyGoal of retrievalbbp() efficiency solved by Mie theory

PSD Algorithm SchemePSD Algorithm Scheme

Use the LUT’s and bbp(440) & maps to calculate algorithm base products:

PSD slope = N(2 m) = No

Calculate derived products: Particle # & V in different size classes

PFT’s

Input Mie model parameters: = 2.5 to 6

m() = n – m’()i Dmin; Dmax

Run Monte Carlo simulation of Mie model with various input

combinations & create two mean LUT’s:

= f-1() log10(bbp(440)/No) = g-1()

Operational Satellite Processing

Retrieve spectral bbp() and its slope from Rrs() via Loisel et al. (2006)

Theoretical LUT Development

oobpbp bb )()(

Global bGlobal bbpbp(440) and (440) and ClimatologyClimatology

oobpbp bb )()(

Algorithm LUT’sAlgorithm LUT’s

log10(particles*m-4)

Mission mean of (Sept. 1997 – Dec. 2007)

Mission mean of (Sept. 1997 – Dec. 2007)

Global Global & & NNoo Climatology Climatology

Endogenous UncertaintiesEndogenous Uncertainties

•Due to Dmax and m

• () is small compared to its variability

• (log10(No)) higher, due to n

PSD Validation w/ Coulter CounterPSD Validation w/ Coulter Counter

Regional validation uses GAC monthly data instead (N = 363):• OK for , great for No!

In-situ

Sea

WiF

S

In-situ S

eaW

iFS

N =22Slope = 1.34 R2 = 0.24

N =22Slope = 2.05 R2 = 0.26

Partitioning Number ConcentrationPartitioning Number Concentration

Picoplankton, # m-3 (0.5 m to 2 m)

Microplankton, # m-3 (20 m to 50 m)

Nanoplankton, # m-3 (2 m to 20 m)

Pico’s vary ~100 times

Nano’s vary ~ 10,000 times

Micro’s vary ~ 106 times

log10(particles/m3)

PFT’s Definition by % VolumePFT’s Definition by % Volume

• Partitioning by volume makes more sense– related to biomass, POC, living C

• Three PFT’s quantitatively defined as %

volume concentration contribution = f():– Picoplankton (0.5 – 2 m equiv. sphere cell

diameter)– Nanoplankton (2 – 20 m)– Microplankton (20 – 50 m)

max

min

3

6

D

D oo dDD

DNDV

(_ 4min

4max

4min

4max f

DD

DDPFTPBv PFTPFT

PFT’s = f(PSD slope)PFT’s = f(PSD slope)

Partitioning Biovolume – the PFT’sPartitioning Biovolume – the PFT’sPicoplankton % (0.5 m to 2 m)

Microplankton % (20 m to 50 m)

Nanoplankton % (2 m to 20 m)

Pico’s dominate oligotrophic ocean (>90%)

Nano’s in transition regions (~50%)

Micro’s only found in upwelling zones & high latitudes (<60%)

PFT Validation w/ HPLC DataPFT Validation w/ HPLC Data

• Uses in-situ HPLC diagnostic pigments (Vidussi et al., 2001)• Matched with daily SeaWiFS 9 km data.• Satisfactory for pico & micro, poor for nano.

N =48Slope = 1.58 R2 = 0.34

N =48Slope = 1.01 R2 = 0.41

Sea

WiF

S %

pic

o

Sea

WiF

S %

mic

ro

In-situ % pico In-situ % micro

BATS Seasonal SuccessionBATS Seasonal Succession

BATS Seasonal SuccessionBATS Seasonal Succession

ConclusionsConclusions• First global assessment of PFT’s via the

PSD from space• Spatial patterns are consistent with current

understanding– Oligotrophic oceans have high PSD slopes,

low abundances & are dominated by pico’s– Bloom regions have lower PSD slope & are

dominated by nano’s & micro’s– Pico’s vary over few orders of magnitude,

micro’s – over many.

• Seasonal succession and relationships to Chl-a are consistent with expectations

AcknowledgementsAcknowledgements

• David Siegel, Stéphane Maritorena

• Funding from the NASA Ocean Biology & Biogeochemistry Program

• Mike Behrenfeld, Hubert Loisel, Emmanuel Boss, Curtis Mobley, Mary Jane Perry, Collin Roesler, Wayne Slade, Giorgio Dall’Olmo, Toby Westberry

The EndThe End