Satellite Retrieval of Phytoplankton Community

Size Structure in the Global OceanColleen Mouw

University of Wisconsin-Madison

In collaboration with Jim Yoder

Woods Hole Oceanographic Institution

Photo David Doubilet

Ecological Importance of Cell Size

Chisholm, 2000

• Small cells:• recycled within euphotic

zone•utilizing regenerated

nutrients•Prefer stratified high

light conditions• Large cells:

• sink out of the euphotic zone

•utilize new nutrients efficiently

•Prefer turbulent, low light conditions

Many biogeochemical processes are directly related to the distribution of phytoplankton size class (Longhurst 1998), and is a major biological factor that governs the functioning

of pelagic food webs (Legendre and Lefevre 1991).

Optical Importance of Cell Size

• Despite the physiological and taxonomic variability, variation in spectral shape can be defined by changes in the dominant size class.

(Ciotti et al. 2002)

a*ph()= [(1-Sf) a*

pico()] + [Sf a*micro()]

Package effect

Motivation

• Rrs() imagery also contains information about cell size in addition to chlorophyll and CDM concentration.

O’Reilly et al. 1998

Chl (mg m-3)

R=log{(Rrs443 > Rrs 490 > Rrs510)/Rrs555}

•SeaWiFS standard chlorophyll algorithm (OC4v4).

Effect of Phytoplankton Concentration on Rrs()

O’Reilly et al. 1998

Maximum band shifts from 443 to 490 to 510 nm with increasing chlorophyll concentration

Effect of [Chl] on water-leaving radiance

Spectral shift

Effect of Cell Size on Rrs()

Magnitude shift!

Sf varyingConstant [Chl] = 0.5 mg m-3

Constant aCDM/NAP(443) = 0.002 m-1

Wavelength (nm)

Rrs (sr-

1)

Hydrolight simulations

Effect of CDM/NAP on Rrs()

In addition to the magnitude shift of cell size, effects of CDM/NAP must

be considered.

aCDM/NAP(443) varyingConstant Chl = 0.5 mg m-3

Constant Sf = 50%

Wavelength (nm)

Rrs (sr-

1)

How can phytoplankton cell size be retrieved

from satellite imagery?

Mouw & Yoder (2009) Remote Sensing of Environment, submitted

HPLC in situ observations

n=4,564

Perc

en

t m

icro

pla

nkto

n

Log10 in situ [Chl] (mg m-3)

The relative biomass proportions of pico-, nano-, and microplankton can be estimated from the concentrations of pigments which have a taxonomic significance and associated to a size class (Bricaud et al. 2004; Vidussi et al. 1996).

Look-up-table Construction

• Full factorial design

• Independently varied [Chl], Sf, & aCDM/NAP over expected ranges for the global ocean

• For a given combination of IOPs, AOPs are calculated via radiative transfer

Look-up-table Construction

Full Factorial Design:Chl, Sf,

aCDM/

NAP(443)

Optical model

Hydrolight

Rrs()

Log10 in situ [Chl] (mg m-3)

Log10 GSM01 [Chl] (mg m-3)

Perc

en

t M

icro

pla

nkto

nG

SM

01

aC

DM

/NA

P(44

3)

m-1

n = 44,343

Wavelength (nm)

Detectable Ranges

If LUT ∆nRrs(443) > SeaWiFS NE∆nRrs(443) –Beyond detection

Rrs (sr-

1)

Chlorophyll (mg m-3)

aC

DM

/NA

P (443) (m

-

1)

SeaWiFS has the sensitivity to retrieve Sf...

• chlorophyll 0.05 - 1.75 mg m-3

• aCDM/NAP(443) < 0.17 m-1

Of decadal mean imagery,84% of [Chl]99.7% of aCDM/NAP(443)

fall within thresholds

LUT Retrieval

SeaWiFS Rrs() imagery

If aCDM(443) > threshold MaskIf aCDM(443) < threshold Continue

HydrolightNormalized Rrs (443)

(Sf range)

GSM01 Chl

GSM01 aCDM/NAP(443)

SfSeaWiFS

Normalized & Corrected Rrs(443)

If [Chl] above/below threshold < MaskIf [Chl] within threshold Continue0.05 - 1.75 mg m-3

< 0.17 m-1

(443/555)

Guide search space

in LUT

Land/Cloud

Size Retrieval

Masked regions that are outside of thresholds for Sf retrieval.

Estimated Sf for May 2006

High CDM/Chl

Low Chl

No flag

Validation

• 85% within 1 standard deviation

• 11%, 2 std. dev.

• 4%, 3 std. dev.

Sf in situ

Sf re

trie

val

Comparison with other functional type retrievals

Uitz et al. 2006

June 2000

Sf - SeaWiFS first 10 years

How do the Sf temporal and spatial patterns compare with [Chl]?

Sf and [Chl] Decadal Climatology

Individual EOF – Mode 1

• [Chl] - adjustments to seasonal cycle

• Sf - ENSO relations– Smaller Sf deviations

until until 2002 (Equatorial Pacific) when deviations become negative

Joint EOF – Mode 1

• Amplitude time series – mirror over zero of

individual Sf mode 1

- Variance driven by Sf

Summary

• Satellite Sf estimates agree well with previous observations

• Regions of the ocean where Sf and [Chl] are decoupled

• ENSO variability more apparent in Sf than [Chl]

• Non-linear response between Sf & [Chl] points to the importance of additional ecological information in the interpretation of [Chl] distributions

Moving Forward

• Much more to investigate with Sf time series…– Further investigation of Sf changes over the

decadal record– Flux estimates with assistance from

numerical models– Production estimates considering cell size

(Mouw & Yoder 2005)– Other suggestions/ideas…

Acknowledgements

• Jim Yoder (WHOI)• Jay O’Reilly (NOAA, NMFS)• Tatiana Rynearson (URI, GSO)• Benjamin Beckmann (MSU)• Maureen Kennelly (URI, GSO)• Kim Hyde (NOAA, NMFS)• Primary Funding

– RI Space Grant/Vetlesen Climate Change Fellowship

– NASA Earth and Space Science Fellowship– URI GSO Alumni Fellowship