Ariane Verdy, Matt Mazloff

Data assimilation of carbon and other biogeochemical constraints

Lynne Talley, Sharon Escher, Bruce Cornuelle, Isa Rosso, Natalie Freeman, Joellen Russell, Jorge Sarmiento, Ken Johnson, Emmanuel Boss, Matt Long, John Dunne, Eric Galbraith

OCB CMIP

6 Work

shop

2018

Southern Ocean Climate and Carbon Observations and Modelling

bgc-Argo: oxygen, nitrate, pH, optics

OCB CMIP

6 Work

shop

2018

State estimationmodels are used to hindcast the ocean state (T, S, V, SSH)

using inputs: initial conditions & atmospheric forcing

adjust those inputs to bring the model closer to observations of the actual ocean state

minimize the “cost function” :

Σ (weighted model-observations misfit)2 + Σ (weighted adjustment to inputs)2

B-SOSE: biogeochemical + physical state optimized together

* *

**

***

o

e.g. Southern Ocean 2013-2017e.g. T & S from Argo maps, winds & air temp, etc. from ECMWF e.g. from Argo profiles, satellites, …

4D-Var, “adjoint” method

new estimate1st guess

?

?

?

ecco.jpl.nasa.gov

Biogeochemical model

pCO2

pH

Fe

ALK

DIC phytoplankton community production

carbon system

chlBlg Bsm DOP

O2

NO3PO4

light temperature

scavenging

sediments

dust

remineralization

air-sea flux air-sea flux

DON

Bdia

all and variables are estimated; can be compared / constrained to observationsprognostic diagnostic

“N-bling”

B-SOSE product2008-2012, 1/3 degree

Verdy and Mazloff (2017), A data assimilating model for estimating Southern Ocean biogeochemistry, JGR-Oceans

2013-2017 in production (with SOCCOM floats constraints)

pCO2

DIC ALK pH NO3

PO4

O2

O2

pH Fe

No

rma

lize

d c

ost

0

5

10

15

20

25

30

35

40

SOCATv4 GLODAPv2 Argo GoShip GEOTRACES

prior

B-SOSE

B-SOSE indep

B-SOSE vs climatologywith SOCCOM float observations

B-SOSE vs climatologywith SOCCOM float observations

Comparisons with in situ observations* Argo profiles (T,S)* calibrated bgc-Argo (O2)* SOCCOM floats

* SOCAT (pCO2)* GLODAPv2 (carbon, nutrients)* CTD (T, S, O2, chl)* XBT, MEOP, PIESGEOTRACES

Comparisons with gridded products* ocean color (chl, POC)* altimetry* microwave SST* sea ice Argo monthly mapped productGLODAPv2, WOA13, SOCAT climatologiesLandschützer monthly mapped product

Validation* = assimilated

mean difference

standard dev. difference

Comparisons with in situ observations7 m O2 in B-SOSE 2013-2017 is compared to bgc-Argo

2008 2009 2010 2011 2012 2013

pC

O2 (µ

atm

)

320

340

360

380

400

420 SOCATv4B-SOSEB-SOSE subsampled

Monthly-averaged pCO2 in Drake Passage (75°W to 55°W, south of 50°S) from SOCATv4 observations (black) [Bakker et al., 2016; Munro et al., 2015a, 2015b], and from B-SOSE (area average in pink; subsampled at the location of observations in red). Summer months are shaded gray.

pCO2 in Drake Passage

Assimilating ocean color observationscost function = surface chlorophyll from VIIRS satellite, 2013

red = where adding iron would reduce the misfit with observations

Higher resolution, multi-grid assimilation

Budgets

Rosso et al. (2017), Space and time variability of the Southern Ocean carbon budget. JGR-Oceans.

DIC in the top 650 m, 2008-2012

mol

/m2 /y

r

DIC

NO

330-40°S

40-50°S

50-60°S

60-70°S

70-80°S

as a tool for explaining observed C:N ratios

surface top 150m

Budgets

plots show seasonal variability of DIC and NO3 in B-SOSE (black) and SOCCOM floats (red) in latitude bands

arrows between seasonal averages = tendencies from DIC and NO3 budgets

budget analysis shows how different mechanisms are contributing to the tendency

30-40°S 40-50°S 50-60°S

60-70°S 70-80°S

(1) dilution is dominant at high latitudes (sea ice)(2) gas exchange is important in mid-latitudes(3) ocean dynamics (advection + mixing) is non-negligible

What explains deviations from Redfield?

Getting the products

B-SOSE output: sose.ucsd.edu+ validation+ documentation

MITgcm BLING model and adjoint: github.com/MITgcm/MITgcm

use model + observations to estimate carbon system over the past ~10 years

multi-year estimate = a continuous model run, which has closed budgets for mass / heat / salt / BGC tracers (DIC, O2, NO3, …)

the output is available online (sose.ucsd.edu) for analysis and comparisons with modelsSnapshot of the simulated dissolved inorganic

carbon (DIC) concentration at 100 m on 2/1/09

summary