garçon v 20150707_1700_upmc_jussieu_-_amphi_herpin (conflit lié au codage unicode)
TRANSCRIPT
Reconstruction of super resolution oceanic pCO2 from remotely sensed data and multi-resolution
analysis: an application in the South Eastern Atlantic
www.oceanflux-upwelling.eu
Véronique Garçon
Ismaël Hernandez-Carrasco, Joël Sudre, Hussein Yahia, Christoph Garbe
Aurélien Paulmier, Boris Dewitte, Séréna Illig
LEGOS, Toulouse, France
University of Heidelberg, Germany
Géostat-INRIA, Bordeaux, France
Karlsruhe Institute for Technology, Germany
(Bakker et al., 2013)
Surface water fCO2 (atm)
SOCAT (Surface Ocean CO2 Atlas,
Version 2)
3
Constraining surface fluxes by Earth Observation data
Temporal evolution of a tracer c in the atmosphere :
Solubility α and transfer velocity k, from EO data, pair
GHG assumed constant over the upwelling region under study, So pocean
GHG obtained from inverse modeling and satellite data at VCD (GOSAT : Vertical Column Density of CO2) resolution or from CARBON TRACKER (CT2013)
F = k (pairGHG – pocean
GHG)
Properties diagrams pCO2 vs SST and pCO2 vs Chl-a
10 years of IPSL present : 1990-2000 with downscaled winds over the Benguela upwelling
(Machu et al., 2015)
SAfE 1/4°
Child domain 1/12°
pC
O2
pC
O2
SST
Chl-a
Multiscale properties of seawater pCO2
SST simulated from
ROMS-BIOEBUS coupled model
pCO2 simulated from
ROMS-BIOEBUS coupled model
Microcanonical Multiscale Formalism (MMF)
The multiscale functional:
- α (x): coefficient dependent of metrics and scaling unit - r: radius of a ball centred on x - h(x): exponent at each point x - o(rh(x)): a negligible term
The exponents give information on the degree of regularity at each point
Hierarchical organization in the image: (singular varieties)
Remote sensing context: exponent values represent the transition fronts
Properties diagrams of SE: pCO2 vs SST and pCO2 vs Chl-a
SAfE 1/4°
Child domain 1/12°
10 years of IPSL present : 1990-2000 with downscaled winds over the Benguela upwelling
(Machu et al., 2015) Si
ngu
lari
ty E
xpo
ne
nt
of
pC
O2
Singularity Exponent of SST
Singularity Exponent of Chl-a
Sin
gula
rity
Exp
on
en
t o
f p
CO
2
Reconstruction of super resolution sea water pCO2 signal using dual ROMS- BIOEBUS simulations at various resolutions
Propagation of the sea water pCO2 signal
across the scales of the multiresolution
analysis determined from the Singularity
Exponents
SST
Chla
pCO2 SE(pCO2)
SE(Chla)
SE(SST)
16 21°C
0.5 2 mg/m3
300 370 µatm
SE(pCO2HR) (x,t) = a(x)SE(SST)(x,t)+b(x)SE(Chla)(x,t)+c(x)SE(pCO2LR)(x,t)+d(x)
Multi-linear regression from ROMS-BIOEBUS model outputs
1) Obtention of coefficients a(x), b(x), c(x), and d(x) (360 images being considered, 10 years of simulation, 1 image every 10 days) 2) For any signal acquisition (SST, Chlorophyll a, pCO2) construction of a SE proxy of pCO2 super resolution 3) Wavelets decomposition through pyramidal algorithm (microcanonical cascade) of the proxy obtained in step 2 and injection of pCO2 low resolution satellite or Carbon Tracker into approximation of image 4) Ascent of microcanonical cascade to finally derive the Super Resolution pCO2
Oceanic reconstructed pCO2
Super resolution (1/12°)
Multi-linear regression from ROMS-BIOEBUS model outputs
Oceanic pCO2
Low resolution (1/4°)
Oceanic original pCO2 Super resolution (1/12°)
(Sudre et al., 2015, submitted)
SST OSTIA 21 September 2006 Chlorophyll a Globcolour 21 September 2006
Benguela Upwelling Inference for 2006 and 2008
Three product combinations
Spatial distribution of the time average over both 2006 and 2008 years of the inferred pCO2 values using: a) High resolution OSTIA SST - MERIS
Chl-a b) High resolution OSTIA SST – GSM-
GLOBCOLOUR Chl-a c) High resolution MODIS SST – GSM-
GLOBCOLOUR Chl-a d) Map with the spatial distribution
of the standard deviation for the inferred pCO2 among the different combination of datasets
Histograms of pCO2 values for 2006/2008
(Hernandez-Carrasco et al., 2015)
Statistical Errors (2006/2008)
Longitudinal comparison of the daily and monthly CT2013 and inferred
pCO2 with in situ observations
Monthly CT2013 produces absolute errors 2
atm higher than daily CT2013 on the inferred
pCO2
To wrap up:
Novel method to reconstruct maps of ocean pCO2 at super resolution (~4km) from CarbonTracker CO2 fluxes data at low resolution (~110 km). Inferred representation of pCO2 improves the description provided by CarbonTracker. Merged products (Globcolour/OSTIA) as the best product combinations since more coverage and best results in the validation exercise Good inference of super-resolution pCO2 using monthly LR pCO2 Very promising result, opens up new work for DMS inference Use satellite-derived CO2 vertical column densities (Carbonsat) as our LR pCO2 image…and obtain for the first time SR-pCO2 from space
Thank you for your attention
(Law et al., 2013)
Complex coupled biogeochemistry/dynamics
Many interactions with the climate system
What is net impact on Earth’s radiation budget?
How are these regions changing under the multiple stressors of warming, stratification, acidification, deoxygenation, etc.?
Eastern Boundary Upwelling Zones and Oxygen Minimum Zones
CarbonTracker 1° x 1°
CO2 fluxes
Inference of DMS air-sea fluxes
Inference of DMS (dimethylsulphide) using an appropriate planktonic functional dependence (haptophytes and dinophytes (prymesiophytes and dinoflagellates), see Brewin et al., 2011, PFTs versus PSCs, nanoplankton (5 – 20 m) Low resolution DMS : 1° by 1° monthly DMS concentration climatology (SOLAS-BODC) (no daily product available)
Monthly CT LR pCO2
Inferred HR pCO2
Daily CT LR pCO2
Inferred HR pCO2
DMS - GO Atlas
http://saga.pmel.noaa.gov/dms/