MEsoSCale dynamical Analysis through combined model, satellite and in situ data

MEsoSCale dynamical Analysis through combined model, satellite and in situ data

PI: Bruno Buongiorno Nardelli1

Co-PI: Ananda Pascual2 & Marie-Hélène Rio3

Participants: F.Bignami1, S. Guinehut3, G. Larnicol3, S. Mulet3, S.Ruiz2, J. Tintorè2

External expert: Y. Drillet4

2009 MyOcean CALL for R&D proposals

1 2 3 4

MESCLA concept• Main objective and approach• Scientific motivation• Main activities and expected results

Project organization • Workpackage Breakdown• Time schedule (GANTT) and required effort• Scientific meetings

First results

Out

line

MESCLA approach includes:

• the diagnosis and comparison of vertical velocity fields from MyOcean model output and synthetic 3D fields obtained by combining in situ and satellite data (e.g. MyOcean ARMOR-3D products)

• the adaptation of ARMOR-3D processing chain to ingest higher resolution MyOcean SST L4 products and the development of new extrapolation methodologies to obtain experimental higher resolution 3D re-analyses (requiring the development and test of a high resolution SSS L4 product).

• the analysis of the interannual variability of the vertical velocity fields

MESCLA project is focused on the estimation and analysis of the vertical exchanges associated to mesoscale dynamics and of their interannual variability, concentrating on selected areas of the global oceans.

Mai

n ob

jecti

ve a

nd a

ppro

ach

MESCLA approach includes:

• the diagnosis and comparison of vertical velocity fields from MyOcean model output and synthetic 3D fields obtained by combining in situ and satellite data (e.g. MyOcean ARMOR-3D products)

• the adaptation of ARMOR-3D processing chain to ingest higher resolution MyOcean SST L4 products and the development of new extrapolation methodologies to obtain experimental higher resolution 3D re-analyses (requiring the development and test of a high resolution SSS L4 product).

• the analysis of the interannual variability of the vertical velocity fields

MESCLA will thus also contribute to the scientific improvement of MyOcean products, developing and testing experimental techniques that could be used in the long-term to build the next generation of operational products.

MESCLA project is focused on the estimation and analysis of the vertical exchanges associated to mesoscale dynamics and of their interannual variability, concentrating on selected areas of the global oceans.

Mai

n ob

jecti

ve a

nd a

ppro

ach

FOCUS ON OBSERVATIONAL APPROACH

NEW PRODUCTS DEVELOPMENT/TEST

MODEL COMPARISON/VALIDATION

Why mesoscale dynamics?Sc

ienti

fic m

otiva

tion


ienti

fic m

otiva

tion

Mesoscale dynamics modulates biological responses:


ienti

fic m

otiva

tion


Vertical displacements and mixing influence biology:

upward transport of nutrients can lead to enhanced algal biomass in the photic zone… downward motions transport the biomass produced at the surface to deeper layers where it is remineralized, releasing once again nutrients into the dissolved phase...

Mesoscale variability may have indirect consequences on ecosystem functioning and carbon cycle/climate

Figures extracted from: Lévy, M, Klein, P. and A.-M. Treguier (2001). Impacts of sub-mesoscale physics on phytoplankton production and subduction, J. Mar. Res., 59,535-565 doi: 10.1357/002224001762842181


ienti

fic m

otiva

tion



ienti

fic m

otiva

tion

The dynamical balance in key areas of the global oceans is mantained by mesoscale


ienti

fic m

otiva

tion


EXAMPLE:• ACC frontal instabilities at mesoscale carry momentum downward and both heat and mass

poleward, counterbalancing the wind-driven upwelling along Antarctica

A typical SSH gradient ( α surface geostrophic velocity) field south of Australia and New Zealand (3 Jul 2002) From Sokolov and Rintoul, 2007


ienti

fic m

otiva

tion


EXAMPLE:• ACC frontal instabilities at mesoscale carry momentum downward and both heat and mass

poleward, counterbalancing the wind-driven upwelling along Antarctica

• Mesoscale has a primary role in the global heat, freshwater, carbon, and nutrient budgets impact on global climate system

net effect under debate

A typical SSH gradient (surface geostrophic velocity) field south of Australia and New Zealand (3 Jul 2002) From Sokolov and Rintoul, 2007

Scie

ntific

moti

vatio

nEffects of mesoscale on Climate and BiologyClimate models have low resolution (computational limitations) Effects of mesoscale on property diffusion/transport need to be parameterized (e.g. through Eddy diffusivity)data assimilation in operational models allows more accurate analyses of vertical exchanges, but results can be significantly influenced by specific model configurations (e.g. forcing, parameterization of smaller scale processes and spatial resolution)

Low resolution Medium resolution High resolution

What is missing? …in terms of understanding

Vertical velocities and space –time resolution are key factors to understand mesoscale processes and their impact at global scale

Scie

ntific

moti

vatio

nEffects of mesoscale on Climate and BiologyClimate models have low resolution (computational limitations) Effects of mesoscale on property diffusion/transport need to be parameterized (e.g. through Eddy diffusivity)data assimilation in operational models allows more accurate analyses of vertical exchanges, but results can be significantly influenced by specific model configurations (e.g. forcing, parameterization of smaller scale processes and spatial resolution)

What is missing? …in terms of understanding

Low resolution Medium resolution High resolution

New observational approaches can help to better understand the vertical exchanges associated with mesoscale processes, analyse their impact on ocean circulation and ecosystem and eventually help to define new parameterizations of mixing for climate models (far beyond MESCLA...)

What is missing? …in terms of parametersSc

ienti

fic m

otiva

tion

• Vertical velocity cannot be directly measured indirect methods require 3D density and velocity fields at high resolution (≤10 km, at least daily)

Scie

ntific

moti

vatio

n• Vertical velocity cannot be directly measured indirect methods require 3D density

and velocity fields at high resolution (≤10 km, at least daily, down to the bottom)• Vertical structure of the ocean is directly measured only by in situ observations sparse in space and time

Vertical profiles of temperature collected from ships in 1998

Daily coverage by autonomous density profilers

What is missing? …in terms of parameters

Scie

ntific

moti

vatio

n• Vertical velocity cannot be directly measured indirect methods require 3D density

and velocity fields at high resolution (≤10 km, at least daily)• Vertical structure of the ocean is directly measured only by in situ observations: sparse in space and time

• Satellite sensors only provide measurements of surface parameters different resolution/coverage depending on sensor

ALTIMETERsurface level, surface (geostrophic) currents only along track, low spatial coverage/repetitiveness

PASSIVE INFRARED/MICROWAVE SENSORSSea surface temperatureUp to <10 km, dailyonly provide surface measurements

What is missing? …in terms of parameters

Which improvements within MESCLA?

• Improve existing observational 3D fields (MyOcean ARMOR-3D) testing other multivariate techniques, merging in situ and satellite data and improving resolution

Taking advantage of:

higher resolution of satellite Sea Surface Temperature measurements

historical in situ dataset cover a high number of the possible states of the system

different parameters covariance

Spec

ific

task

s/ex

pect

ed re

sults

HIGHER RESOLUTION DATA MyO SST L4 New SSS L4

NEW VERTICAL EXTRAPOLATION TECHNIQUEMultivariate EOF Reconstruction

High resolution 3D fieldsTemperatureSalinityDensity

Which improvements within MESCLA?

• High resolution 3D velocity fields will be estimated from synthetic fields and model output through a diagnostic numerical model

These 3D velocity fields will be compared with estimates from MyOcean numerical model output

Spec

ific

task

s/ex

pect

ed re

sults

Q-vector formulation of the OMEGA equation

High resolution 3D fieldsTemperatureSalinityDensity

High resolution 3D velocity fields

z

V

y

U

z

U

x

Uf

z

V

y

V

z

U

x

VfQ ,

Qz

wfwN

22

2222

w Vertical velocity

U,V Horizontal geostrophic velocities

WP 1. Vertical velocity estimation from MyOcean productsTask 1.1 Adaptation of the Omega equation model to MyOcean productsTask 1.2 Application of the Omega equation model and comparison between different vertical velocity estimates

WP2: Development of 3D-fields experimental products from observationsTask 2.1 Ingestion of MyOcean SST L4 products in ARMORTask 2.2 Development of a high resolution SSS L4 productTask 2.3 Test of a multivariate EOF reconstruction for the extrapolation of vertical profiles from surface measurements

WP3: Analysis of the interannual variability of the vertical exchanges associated to mesoscale dynamicsTask 3.1 Re-analysis of vertical velocitiesTask 3.2 Analysis of the vertical velocities variability

Wor

kpac

kage

bre

akdo

wn

Project schedule (GANTT) and required effort

Task 1.1 Adaptation of the Omega equation model to MyOcean productsTask 1.2 Application of the Omega equation model and comparison between different vertical velocity estimates

Task 2.1 Ingestion of MyOcean SST L4 products in ARMORTask 2.2 Development of a high resolution SSS L4 productTask 2.3 Test of a multivariate EOF reconstruction for the extrapolation of vertical profiles from surface measurements

Task 3.1 Re-analysis of vertical velocitiesTask 3.2 Analysis of the vertical velocities variability

duration

Man

/months

First year Second year

01 02 03 04 05 06 07 08 09 10 11 12 01 02 03 04 05 06 07 08 09 10 11 12

WP 1 12

Task1.1 6 1.2

Task1.2 6 2.2

WP 2 16

Task2.1 2 1.2

Task2.2 6 2

Task2.3 10 3.7

WP 3 12

Task3.1 4 1.5

Task3.2 8 2

total 13.8

Kick Off Meeting First Scientific Meeting Final Scientific Meeting

Project schedule (GANTT) and required effort

Task 1.1 Adaptation of the Omega equation model to MyOcean productsTask 1.2 Application of the Omega equation model and comparison between different vertical velocity estimates

Task 2.1 Ingestion of MyOcean SST L4 products in ARMORTask 2.2 Development of a high resolution SSS L4 productTask 2.3 Test of a multivariate EOF reconstruction for the extrapolation of vertical profiles from surface measurements

Task 3.1 Re-analysis of vertical velocitiesTask 3.2 Analysis of the vertical velocities variability

duration

Man

/months

2010 2011 2012

04 05 06 07 08 09 10 11 12 01 02 03 04 05 06 07 08 09 10 11 12 01 02 03

WP 1 12

Task1.1 6 1.2

Task1.2 6 2.2

WP 2 16

Task2.1 2 1.2

Task2.2 6 2

Task2.3 10 3.7

WP 3 12

Task3.1 4 1.5

Task3.2 8 2

total 13.8

Kick Off Meeting First Scientific Meeting Final Scientific Meeting

MyO SDM

WP 1. Vertical velocity estimation from MyOcean productsTask 1.1 Adaptation of the Omega equation model to MyOcean productsTask 1.2 Application of the Omega equation model and comparison between different vertical velocity estimates

WP2: Development of 3D-fields experimental products from observationsTask 2.1 Ingestion of MyOcean SST L4 products in ARMORTask 2.2 Development of a high resolution SSS L4 product Task 2.3 Test of a multivariate EOF reconstruction for the extrapolation of vertical profiles from surface measurements

WP3: Analysis of the interannual variability of the vertical exchanges associated to mesoscale dynamicsTask 3.1 Re-analysis of vertical velocitiesTask 3.2 Analysis of the vertical velocities variability

Wor

kpac

kage

bre

akdo

wn

completed

completed

started

started

started

Task

1.1

Adaptation of the Omega equation to MyOcean products (resp. IMEDEA)

• Selection of input data:– PSY3V2R2: 1/4º resolution on mercator grid.

50 vertical levels– PSY2V3R1: 1/12º resolution on mercator grid.

50 vertical levels– ARMOR3D – V1 (v5e1 and v5e2)– Domain: NW Atlantic. Gulf Stream– Date test: 10/10/2007

• Computation of geostrophic velocities (Dynamic Heights):

– Sensitivity to the reference level: • 1000 m• surface SSH

• Vertical interpolation:– 10:10:1000 m

• Horizontal interpolation:– regular grid (1/4º, 1/12º)

• Conversion of data files to be ingested into the fortran code for the estimation of vector Q and vertical velocity

• Estimation of vector Q and vertical velocity

Original profileInterpolated profile

Example of a density profile

Task

1.2

PSY2V3R1 1/12º – ref level 1000 m

Applicaton of the Omega equation model and comparison between different vertical velocity estimates (resp. IMEDEA)

w-Omega (1/12°) w-model (1/12°)

SST DH

Task

1.2

PSY2V3R1 1/12º – ref surf SSH


w-Omega (1/12°) w-model (1/12°)

SST DH

Task

1.2

PSY2V3R1 1/12º – ref level 1000 m - zoom


w-Omega w-model

Task

1.2

PSY2V3R1 1/12º – ref level surf SSH - zoom


w-Omega w-model

Task

1.2

PSY3V2R2 ¼º – ref level 1000 m


w-Omega (1/4°) w-model (1/4°)

SST DH

Task

1.2

PSY3V2R2 ¼º – ref surf SSH



SST DH

Task

1.2

ARMOR-3D V1 projection of SLA and SST onto 24 Levitus vertical levels


w-Omega (1/4°)

SST DH

Task

1.2

ARMOR-3D V1 projection of SLA and SST onto 24 Levitus vertical levels + combination with in situ profiles


w-Omega (1/4°)

SST DH

Task

1.2




SST DH

Ingestion of MyOcean SST L4 product in Armor3D (Resp. CLS)Ta

sk 2

.1

Armor3D: 2-step method based on observations and statistical method

Step1: Vertical projection of satellite SLA+SST using a multiple linear regression methodStep2: Combination of synthetic fields and in-situ T/S profiles using an optimal interpolation method

• Historically, Armor3D-v0 uses Reynolds OI 1° SST

• Armor3D-v1 uses Reynolds AVHRR-AMSR 1/4° SST (real-time and reanalysis)

• Several global L4 SST products exists (MyOcean, other data centres):How do they compare?What is the impact on Armor3D?What are there performances compared to independent Argo data set?


sk 2

.1

Comparison of different SST products


sk 2

.1

Comparison of different SST product increments

weekly daily

daily daily


sk 2

.1

Synthetic T at 30m

SST

Synthetic T at 50m

Synthetic T at 100m

Reynolds 1° Reynolds 1/4° 27/12/2006

Impact of higher resolution on Armor3D

Development of a high resolution Sea Surface Salinity (SSS) L4 product (Resp. CNR)

Task

2.2

SSS possibly needed by new 3D reconstruction methodsnew product potentially useful in combination with SMOS data

Hypothesis:high correlation between sea surface temperature (SST) and sea surface salinity (SSS) variations can be expected (in the open ocean) at scales significantly smaller than the ones dominating atmospheric variability

similar correlation found by Jones et al. (1998) between SSH and SSTanalogous results recently obtained looking at SST and SSS variability by Aretxabaleta et al. (2009, communication at EGU) and Ballabrera-Poy et al. (2009).

Proposed technique:optimal interpolation (Bretherton-like) algorithm that includes satellite SST differences in the covariance estimation

)()( . backgroundobsbackgroundanalysis xyCRCxx 1

222

TSST

Lrt

eeeSSTtr ),,(C

Test strategy and input datasets

First tests: calculations performed on a single day (17th October 2007), limiting to a portion of the North Atlantic Ocean

Covariance function parameters (i.e. spatial (L), temporal (τ) and thermal (T) decorrelation scales)determined empirically minimizing errors vs independent surface observations

Salinity Observations (input data):MyOcean INSITU-TAC obs. from profiling floats, CTD and XCTD, referenced to 5 m depth

SST observations:ODYSSEA L4 SST different filtering applied to remove the large-scale variability associated to air-sea interactions, trying to keep only the signals associated with the different water masses and related to mesoscale structuresoriginal SST values high-pass filtered data (keeping the scales smaller than 500 and 1000 km respectively).

Background field:MyOcean CORIOLIS SSS objectively analyzed maps (covering the ½° MERCATOR grid)

Validation data:thermosalinograph obs. provided by the Global Ocean Surface Underway Data project (GOSUD) and by the French Sea Surface Salinity Observation Service (at LEGOS)

Task

2.2

Test datasets

in situ SSSRed dots (input) 30 days window, centered on interpolation dayBlue dots (validation) (only for interpolation day)

Task

2.2


Background SSS

Test DatasetsTa

sk 2

.2


Background

ODYSSEA L4 SST

Test datasetsTa

sk 2

.2

Empirical determination of OI parameters:

RMSE (black contour) and MBE (red) vs validation data (expressed as % with respect to CORIOLIS –SSS L4 errors), as a function of the spatial and thermal decorrelation scales. High-pass filtered SST (<1000 km)

diffe

rent

val

ues

of n

oise

-to-

sign

al ra

tio

different values of temporal decorrelation scale (τ)

(L)

(T)

Task

2.2

“Best” configuration and quantitative validation

lowest errors RMSE ~0.11 psu ~70% of CORIOLIS errorMBE ~0.01 psu ~30% of CORIOLIS errorL= 400 km τ=6 days T=2.75 °Csignal-to-noise=0.01High-pass filter: signals <1000 Km

RMSE (black contour) and MBE (red) vs validation data (expressed as % with respect to CORIOLIS–SSS L4 errors)

Task

2.2

Qualitative results:

Back

grou

nd

MES

CLA

M

ERCA

TOR

MESCLA high resolution SSS field and derived SSS gradient reveal more realistic and smaller scale structures than those visible in CORIOLIS-SSS product.MERCATOR 1/12° displays even smaller scales…

Task

2.2

Conc

lusi

ons

and

futu

re w

ork

First results promising:

• Omega gives accurate results at 1/12° (quantitative validation/analysis to be completed)

• ARMOR3D being improved in terms of resolution: soon getting the first 1/12° estimates (with standard 3D extrapolation)

• Test of multivariate 3D extrapolation will start soon (using combinations of SSH, SST, SSS as input)

• Method for SSS L4 estimate promising

THAN

KS!

Thanks for listening and for funding MESCLA!

OS3.2From physical to biogeochemical processes: ocean mesoscale and sub-mesoscale impact on marine ecosystem and climate variability

Convener: Ananda Pascual Co-Conveners: John Allen, Bruno Buongiorno Nardelli, Marina Lévy

…This session will provide a forum to properly address the new scientific challenges associated with mesoscale and sub-mesoscale variability (between 1 km and 300 km), based on observations (both in situ and satellite and multi-sensor approaches), theory, and numerical simulations. ….

IMPORTANT:Deadline for receipt of abstracts: 10 January 2011

MEsoSCale dynamical Analysis through combined model, satellite and in situ data

Documents

Transcript of MEsoSCale dynamical Analysis through combined model, satellite and in situ data