Assessing the Impact of Ocean Observing

Systems in Support of U.S. IOOS

Andy Moore1, Hernan Arango2, Chris Edwards1,Julia Levin2, Brian Powell3 & John Wilkin2

1: Dept. of Ocean Sciences, UC Santa Cruz2: Dept. of Marine and Coastal Sciences, Rutgers University

3: Dept. of Oceanography, University of Hawaii

Assessing the Impact of Ocean Observing

Systems in Support of U.S. IOOS

Andy Moore1, Hernan Arango2, Chris Edwards1,Julia Levin2, Brian Powell3 & John Wilkin2

1: Dept. of Ocean Sciences, UC Santa Cruz2: Dept. of Marine and Coastal Sciences, Rutgers University

3: Dept. of Oceanography, University of Hawaii

U.S. Integrated Ocean Observing System (IOOS)

AOOS

NANOOS

CeNCOOS

SCCOOS

GCOOS

CariCOOS

GLOS

MARACOOS

NERACOOS

SECOORA

PacIOOS

11 Regional Associations

The charge:• Observe• Analyze• Forecast• Products

IOOS Stakeholders

Search & rescue

Fisheries

Water quality

AOOS

CeNCOOS

SCCOOS

GCOOS

CariCOOS

GLOS

MARACOOS

NERACOOS

SECOORA

Ocean Observing Systems

Remote Sensing

AOOS

NANOOS

CeNCOOS

SCCOOS

GCOOS

CariCOOS

GLOS

MARACOOS

NERACOOS

SECOORA

PacIOOS

11 Regional Associations

MARACOOS : Mid-Atlantic Regional Association Coastal Ocean Observing SystemCeNCOOS: Central and Northern California Ocean Observing SystemPacIOOS: Pacific Islands Ocean Observing System

U.S. Integrated Ocean Observing System (IOOS)

Outline• Methodology• A MARACOOS

example• Summary

A Typical Sequential Analysis-Forecast Procedure

4D-Var 4D-Var

Obs impact onforecast skill?

Data Assimilation & Observation Impacts

xa

= xb+BGT GBGT +R( )

-1

y- H(xb)( )

analysis

background

backgrounderror cov

TL modelat obs pts

obserror cov

obs

obsoperator

( )I xScalar function: (e.g. transport, forecast skill,…)

Change due to 4D-Var: ( ) ( )I I I = −a bx x

Analysis equation:

GBGT +R( )-1

GBMT ¶I ¶x( )xb

Observation Impact

Impact of observations on is given by:DI

Let’s look at what this really means…Langland and Baker (2004)

Observation Impacts

Zonal shear flow

GBGT +R( )-1

GBMT ¶I ¶x( )xb

“Target” line of

delta functions

Observation Impacts

Zonal shear flow

GBGT +R( )-1

GBMT ¶I ¶x( )xb

Adjoint Model

Adjoint Model

Zonal shear flow

Observation Impacts

GBGT +R( )-1

GBMT ¶I ¶x( )xb

A weighted sum of Green’s

functions

Zonal shear flow

Observation Impacts

GBGT +R( )-1

GBMT ¶I ¶x( )xb

A weighted sum of Green’s

functions

Covariance

Covariance

Zonal shear flow

Observation Impacts

GBGT +R( )-1

GBMT ¶I ¶x( )xb

Zonal shear flow

Observation Impacts

GBGT +R( )-1

GBMT ¶I ¶x( )xb

Tangent Linear Model

sampled at obs points

Tangent Linear Model

sampled at obs points

Zonal shear flow

Observation Impacts

GBGT +R( )-1

GBMT ¶I ¶x( )xb

Tangent Linear Model

sampled at obs points

Zonal shear flow

× Observations

Observation Impacts

GBGT +R( )-1

GBMT ¶I ¶x( )xb

××

×

×

altimeter track

mooring

Zonal shear flow

Observation Impacts

GBGT +R( )-1

GBMT ¶I ¶x( )xb

Tangent Linear Model

sampled at obs points

Zonal shear flow

Observation Impacts

GBGT +R( )-1

GBMT ¶I ¶x( )xb

Remove covariance between obs locations

Zonal shear flow

Observation Impacts

GBGT +R( )-1

GBMT ¶I ¶x( )xb

Remove covariance between obs locations

Observation

impact

(CGsolver)TGBMT ¶I ¶x( )xb

Zonal shear flow

Adjoint of CG solver

Observation Impact: Take 2

Zonal shear flow

Observation Impact: Take 2

(CGsolver)TGBMT ¶I ¶x( )xb

Adjoint of CG solver

Observation

sensitivity

Model surface and boundary forcing:Surface forcing derived from NAM [NCEP NOMADS]

USGS daily average flow [waterdata.USGS.gov]

Mercator open boundary conditions

Assimilation data sets: [real-time source]

Regional CODAR hourly [RU TDS]

IOOS glider T,S (1-hr delay) [RU ERDDAP]

AVHRR IR passes 6/day [MARACOOS TDS]

AMRS2+OceanSat mu-wave SST [NASA PODAAC]

Jason-2 & 3, CryoSat, AltiKa [RADS.tudelft.nl]

GTS XBT/CTD, Argo floats [OSMC NOAA ERDDAP]

Pioneer glider+mooring [RU ERDDAP]

Data assimilation system: ROMS ~7km, 40 levels4-dimensional variational (4D-var) data assimilationDual formulation (augmented RPCG)2 outer-loops, 7 inner-loops3 day assimilation windows

Regional Ocean Modeling System (ROMS): MARACOOS

Per 3 day cycle: SST ~105

HF radar ~104

in situ ~5X103

Altimetry ~103

Wilkin, Levin, & Arango

A Typical Sequential Analysis-Forecast Procedure

4D-Var 4D-Var

I

A Forecast Example

Impact Sensitivity

degrade

improve

I = change in mean-squared 3-day forecast error in surface velocity due to assimilating obsand evaluated at all HF radar obs locations.

Metric=MSE velocity

in situ T

in situ S

u & v

SST

SSH

in situ T

in situ S

u & v

SST

SSH

I

I = 1 N ui

f - ui

o( )2

+i=1

N

å vi

f - vi

o( )2

• Some platforms appear to “borrow strength” from other platforms – “corroborating evidence”

• In the case of surface velocity obs from HF radar:- obs include Ekman and pressure driven flow, but only the

latter is “seen” by satellite remote sensing (but it’s presence is corroborated by u&v obs)

- most of the energy is in potential form -> T&S best (but circulation features corroborated by u&v obs)

• Observation impact and observation sensitivity provide important complementary quantitative information about the synergy between observations.

Summary

I1= 1 t u

ndz ds

-h

0

òS

ò dt0

t

ò

Cross-shelf volume transport:

Cross-Shelf Exchange Circulation Metrics

200 m isobathtarget

Historical context:Garvine et al (1989)Linder and Gawarkiewicz (1998)Chen & He (2014)OOI Pioneer endurance array

Obs Impact Obs Sensitivity

in situ T

in situ S

u & v

SST

SSH

RMS impact of SST observations on cross-shelf volume transport during 2017.

RMS impact on cross-shelf volume transport of excludingSST during 2017. in situ T

in situ S

u & v

SST

SSH

log10

Sv

log10

Sv

rms contributionof SST to transport

rms change in transport is SST excluded

RMS Obs Impact RMS Obs Sensitivity