Auxiliary Material Submission for Paper 2005GB002530

Inverse Estimates of Anthropogenic Carbon Uptake, Transport, and Storage by the OceanS. 

 \authors{S. E. Mikaloff Fletcher, \altaffilmark{1}
 N. Gruber, \altaffilmark{1}
 A. R. Jacobson, \altaffilmark{2}
 S. C. Doney, \altaffilmark{3}
 S. Dutkiewicz, \altaffilmark{4}
 M. Gerber, \altaffilmark{5}
 M. Follows, \altaffilmark{4}
 F. Joos, \altaffilmark{5}
 K. Lindsay, \altaffilmark{6}
 D. Menemenlis, \altaffilmark{7}
 A. Mouchet \altaffilmark{8}
 S. A. M\"uller \altaffilmark{5}
 and J. L. Sarmiento \altaffilmark{2}}

\altaffiltext{1}{Department of Atmospheric and Oceanic Sciences and
the Institute for Geophysics and Planetary Physics, University of
California, Los Angeles, California, USA.}

\altaffiltext{2}{Program in Atmospheric and Oceanic Sciences
Department, Princeton, New Jersey, USA.}

\altaffiltext{3}{Marine Chemistry and Geochemistry, Woods Hole
Oceanographic Institution, Woods Hole, Massachusetts, USA.}

\altaffiltext{4}{Department of Earth, Atmosphere, and Planetary
Sciences, Massachusetts Institute of Technology, Cambridge,
Massachusetts, USA.}

\altaffiltext{5}{Climate and Environmental Physics, Physics
Institute, University of Bern, Bern, Switzerland.}

\altaffiltext{6}{Climate and Global Dynamics Division, National
Center for Atmospheric Research, Boulder, Colorado, USA.}

\altaffiltext{7}{Estimating the Circulation and Climate of the
Ocean, Jet Propulsion Laboratory, Pasadena, California, USA.}

\altaffiltext{8}{Astrophysics and Geophysics Institute, University
of Liege, Liege, Belgium.}
Mikaloff Fletcher, S. E., et al. (2006), Inverse estimates of anthropogenic CO2 uptake, transport, and storage by the ocean, Global Biogeochem. Cycles, 20, GB2002, doi:10.1029/2005GB002530.
Introduction
This folder contains the following materials: 

Supplemental Text----------------------------

2005GB002530-text01.tex-- This latex file contains additional
information about the methods.  This includes:1. Further details about the model simulations used to generate the basis functions, discussed in Section 2.2 of the manuscript  2. Additional discussion of the region aggregation and the covariance matrix  3. The specific formulation of the CFC-11 skill score and the model 
weighting scheme  4. A complete description of how the scenarios for possible biases in the 
data-based anthropogenic carbon estimates were constructed. 

2005GB002530-text02.tex-- This file contains  brief  descriptions of each of the OGCMs used in this study. 

2005GB002530-text03.tex-- This file contains discussion of the
differences between the UL model an the other participating models in the North Atlantic.  

Supporting files needed to compile the latex files:  

agu2001.cls

agu2001.sty

agu.bst

Figures----------------------------

2005gb002530-fs01.eps-- The locations of the observations used in this
study.  The colors indicate the different time periods of the observations:
prior to 1985 (purple), 1985 to 1990 (blue) 1990 to 1995 (green), since 1995 
(red).

2005gb002530-fs02a.eps-- Column inventory of the time-dependent basis function (mol dye/m3) for regions 1-12 of the 24 model regions shown in 
Figure 2 of the manuscript in 1995, calculated using the PRINCE-2 OGCM. 

2005gb002530-fs02b.eps-- Column inventory of the time-dependent basis function (mol dye/m3) for regions 13-24 of the 24 model regions shown in 
Figure 2 of the manuscript in 1995, calculated using the PRINCE-2 OGCM. 

2005gb002530-fs03.eps -- Comparison between the atmospheric CO2
perturbation and the global, annual mean air-sea CO2 flux using four
different OGCMs: MIT, NCAR, PRINCE, and UL.  The top panel shows the
temporal evolution of the atmospheric CO2 perturbation (left axis, solid line) and the modeled annual mean air-sea CO2 flux normalized to the modeled annual mean flux in 1990 (right axis, dashed lines).The bottom panel shows the correlation between the atmospheric perturbation
and the normalized modeled annual mean air-sea CO2 flux (dashed lines). The solid line represents the idealized case where the atmospheric
perturbation is exactly proportional to the surface flux.

2005gb002530-fs04.eps -- Covariance between regional flux estimates for
the 24 region aggregation (10^5 Pg C yr-1), scaled to 1995).

2005gb002530-fs05.eps-- The weighted mean inverse
estimates of anthropogenic carbon (Pg C/yr) from this study (white bars) compared with the inverse anthropogenic carbon estimates of
Gloor et al., [2003] (gray bars).  The flux estimates from this study were
interpolated to the 13 model regions used in Gloor et. al., 2003 based on the spatial pattern of the surface fluxes used for the basis functions and
scaled to 1990 in order to be directly comparable to the Gloor study.  
The thick, gray error bars show the weighted standard deviation of the
models, and the thin, black error bars represent the total range of the models.

2005gb002530-fs06.eps-- Column inventory of the time-dependent basis function (mol dye/m2) for region 9 in 1995, calculated using the nine OGCMs.  A map of
region 9 is shown bottom, center.

2005gb002530-fs07.eps--  Zonal mean of the time-dependent basis function (mol dye/m3) for region 9 in 1995, calculated using the nine
OGCMs.  A map of region 9 is shown bottom, center.

2005gb002530-fs08.eps-- Comparison between the atmospheric CO2
perturbation and the anthropogenic carbon storage from a forward simulation
of the PRINCE-2 model integrated over several regions in the Atlantic Ocean.
The correlations between the atmospheric perturbation and the modeled
anthropogenic carbon storage for regions 1-8, normalized to the storage
at the end of the simulation, are shown as dashed lines.  The solid line
represents the idealized case where the atmospheric perturbation is exactly
proportional to the anthropogenic carbon storage.

2005gb002530-fs09.eps-- Zonally, vertically integrated anthropogenic carbon 
transport by the global, (top), Atlantic (center), and Indo-Pacific (bottom) 
Oceans from 1765-1995.  Positive (negative) values indicate northward 
(southward) transport. 

2005gb002530-fs10.eps-- The zonally averaged hypothetical bias added to the
data-based anthropogenic carbon estimates to construct the 'Matsumoto
corrected' scenario.  The color scale indicates the total change to the 
anthropogenic carbon estimates (micro mol kg-1), and the contours
indicate the change relative to the GLODAP anthropogenic carbon
inventory.  

2005gb002530-fs11.eps-- Zonally, vertically integrated difference
between the forward and inverse anthropogenic carbon storage estimates
over the global, (top), Atlantic (center),and Indo-Pacific (bottom) Oceans
from 1765-1995. Forward simulations are from OCMIP-2 (Watson and Orr, 2003). Positive (negative) values indicate that the forward simulation finds 
more (less) anthropogenic carbon storage than the inversion.

Tables----------------------------

2005gb002530-Tablel.text-- Summarizes differences between the five different
configurations of the PRINCE model. 

2005gb002530-Table2.text-- Inverse anthropogenic carbon flux estimates
based on basis functions from ten different OCGMS and their weighted
mean, weighted standard deviation, and range (Pg C yr-1, scaled to 1995). 
Positive values indicate flux into the ocean.

2005gb002530-Table3.text-- Optimal coefficients and standard deviations of a 
multiple linear regression fit of anthropogenic carbon concentrations estimated
using lower and upper limits of the C:O ratio to observed AOU and 
anthropogenic carbon as discussed in Section 4 of the Extended Methods in the 
online supplement.  

2005gb002530-Table4.text.text-- Inverse anthropogenic carbon flux 
estimates from the PRINCE-2 model based on the GLODAP anthropogenic carbon 
data set and three scenarios designed to test the sensitivity of the inverse 
estimates to biases associated with estimating the anthropogenic carbon from 
the observations (Pg C yr-1, scaled to 1995).  Positive values indicate flux into the ocean.

2005gb002530-Table5.text--The difference between the anthropogenic carbon uptake in the forward model simulations in 1995 and the inverse flux
estimates using the same models scaled to 1995 (Pg C yr-1). Positive (negative)
values indicate that the forward model simulations find more (less) anthropogenic carbon uptake than the inverse model. Forward simulations for the model version used here are not available for the Bern-3D, ECCO, and MIT models.