Auxiliary material for Paper 2007JC004153

Impact of ocean carbon system variability on the detection of temporal increases in anthropogenic CO2

Naomi Marcil Levine
Massachusetts Institute/Woods Hole Oceanographic Institution Joint Program, Woods Hole, Massachusetts, USA

Scott C. Doney
Marine Chemistry and Geochemistry Department, Woods Hole Oceanographic Institution, Woods Hole, Massachusetts, USA

Rik Wanninkhof
Atlantic Oceanographic and Meteorological Laboratory (NOAA), Miami, Florida, USA

Keith Lindsay
Climate and Global Dynamics, National Center for Atmospheric Research, Boulder, Colorado, USA

Inez Y. Fung
Berkeley Atmospheric Sciences Center, University of California, Berkeley, Berkeley, California, USA

Levine, N. M., S. C. Doney, R. Wanninkhof, K. Lindsay, and I. Y. Fung (2008), Impact of ocean carbon system variability 
on the detection of temporal increases in anthropogenic CO2, J. Geophys. Res., 113, C03019, doi:10.1029/2007JC004153.

Introduction

To estimate the short-term variability in anthropogenic CO2 in the ocean (Canthro), 
we compare a pair of historical (1958-2004) ocean simulations.  The two runs, a control 
run and a transient run, have identical surface forcing and nearly identical physical circulation.  
We estimate the error in delta Canthro(predicted) by comparing the true estimate of Canthro, 
calculated by differencing the control run from the transient run, to the low-pass filter 
estimate of Canthro, calculated using a spline fit. Figure S1 shows the difference between 
these two estimates for three representative latitudes (54.9N, 2.2S, and 29.1S) 
at 190m.  Also plotted are the dissolved inorganic carbon concentrations normalized to 
year zero so that they plot  on the same scale.  The mean RMS error between the "true" 
and "predicted" Canthro values  is 0.37 mol/kg per decade (200-2000m).  This is 
considerably smaller than either the natural  variability or the errors in the 
delta C* and multiple linear regression techniques.


1. 2007JC004153-fs01.eps
Time series of historical (1958-2004) ocean-only model output along the A16 transect 
at 190 m and (a) 54.9N, (b) 2.2S, and (c) 29.1S.  The transient run's dissolved inorganic carbon (DIC) concentrations, 
normalized to year zero, are shown in light gray.  The true Canthro estimates, calculated 
as the difference between a control run and a  transient run with identical surface 
forcing and nearly identical physical circulation, are shown in black.  The low-pass 
filter estimates of Canthro, calculated using a spline  fit to the transient run, are 
shown in dark gray.