Using altimetry to help explain patchy changes in hydrographic carbon measurements

Abstract

[1] Here we use observations and ocean models to identify mechanisms driving large seasonal to interannual variations in dissolved inorganic carbon (DIC) and dissolved oxygen (O 2 ) in the upper ocean. We begin with observations linking variations in upper ocean DIC and O 2 inventories with changes in the physical state of the ocean. Models are subsequently used to address the extent to which the relationships derived from short-timescale (6 months to 2 years) repeat measurements are representative of variations over larger spatial and temporal scales. The main new result is that convergence and divergence (column stretching) attributed to baroclinic Rossby waves can make a first-order contribution to DIC and O 2 variability in the upper ocean. This results in a close correspondence between natural variations in DIC and O 2 column inventory variations and sea surface height (SSH) variations over much of the ocean. Oceanic Rossby wave activity is an intrinsic part of the natural variability in the climate system and is elevated even in the absence of significant interannual variability in climate mode indices. The close correspondence between SSH and both DIC and O 2 column inventories for many regions suggests that SSH changes (inferred from satellite altimetry) may prove useful in reducing uncertainty in separating natural and anthropogenic DIC signals (using measurements from Climate Variability and Predictability's CO 2 /Repeat Hydrography program).

Trends of anthropogenic CO<sub>2</sub> storage in North Atlantic water masses

“…It must be noticed that F b,l,c values are not always identical at every station on a given cruise, layer and basin: there is a high short-scale spatial variability linked with the variability of the mesoscale field (Rodgers et al, 2009). In addition, the WOA05 gridded fields have been largely smoothed and have less spatial resolution (1 • × 1 • , i.e., each of WOA05's pixels may include more than one station from the same cruise) than the observations from the cruises.…”

“…The LSW in the ENA basin fades progressively towards the Iberian Peninsula, where the BD3 shows the LSW θ and S maximum values compared to what is observed in the rest of the basin, i.e., a homogeneous distribution of such tracers (Table 2c). The low AOU signal in 1991 and 1993 stems from the fast and far-reaching spreading of LSW after the strong convection period that occurred during that time (Read and Gould, 1992;Sy et al, 1997).…”

Inorganic carbon and pH levels in the Rockall Trough 1991–2010

“…The two main assumptions of the eMLR method are (1) that the linear multiple-parameter model is able to describe a regions spatially varying hydrographic C T distribution and (2) that the underlying natural correlations of C T with the selected independent parameters do not change over the period of interest. A number of studies have highlighted that natural variability in oceanic C T can be comparable to or larger than the anthropogenic signal sought by the decadal repeat occupations due to the influence of ocean circulation on annual variations in C T Levine et al, 2008;Rodgers et al, 2009). This eMLR technique accounts for the natural variations of carbon due to both hydrographic (temperature and salinity) and biogeochemical (nutrients and oxygen) parameters incorporated into the model.…”

Role of regression model selection and station distribution on the estimation of oceanic anthropogenic carbon change by eMLR

“…In these restricted areas, mapping errors can be as much as half the size of the anthropogenic signal (about ±10 mol m −2 in absolute terms). These unsampled regions are also zones experiencing the highest magnitude of temporal carbon variability in the North Atlantic in the simulations (Rodgers et al, 2009). When integrated over the basin, mapping errors are smaller than other sources of uncertainties, however.…”

“…The model was initialized with World Ocean Atlas (2001) temperature, salinity and nutrients, GLODAP carbon and forced with the NCEP-derived CORE representation of atmospheric fields and fluxes Yeager, 2004, 2009;Griffies et al, 2009) The strategy used to isolate the anthropogenic carbon concentration from the model is described by Rodgers et al (2009). Briefly, the model was spun up for two repeating CORE cycles with fixed pre-industrial atmospheric CO 2 concentration after initialization.…”