eMLR Method Research Articles

Temporal changes in ventilation and the carbonate system in the Atlantic sector of the Southern Ocean

The Southern Ocean is the most important area of anthropogenic carbon (Cant) uptake in the world ocean, only rivalled in importance by the North Atlantic Ocean. Significant variability on decadal time-scales in the uptake of Cant in the Southern Ocean has been observed and modelled, likely with consequences for the interior ocean storage of Cant in the region, and implications for the global carbon budget. Here we use eight cruises between 1973 and 2012 to assess decadal variability in Cant storage rates in the southeast Atlantic sector of the Southern Ocean. For this we employed the extended multiple linear regression (eMLR) method. We relate variability in DIC (dissolved inorganic carbon) storage, which is assumed to equal anthropogenic carbon storage, to changes in ventilation as observed from repeat measurements of transient tracers. Within the Antarctic Intermediate Water (AAIW) layer, which is the dominant transport conduit for Cant into the interior ocean, moderate Cant storage rates were found without any clear temporal trend. In Subantarctic Mode Water (SAMW), a less dense water mass found north of the Subantarctic Front and above AAIW, high storage rates of Cant were observed up to about 2005 but lower rates in more recent times. The transient tracer data suggest a significant speed-up of ventilation in the summer warmed upper part of AAIW between 1998 and 2012, which is consistent with the high storage rate of Cant. A shift of more northern Cant storage to more southern storage in near surface waters was detected in the early 2000s. Beneath the AAIW the eMLR method as applied here did not detect significant storage of Cant. However, the presence of the transient tracer CFC-12 all through the water column suggests that some Cant should be present, but at concentrations not reliably quantifiable. The observed temporal variability in the interior ocean seems at a first glance to be out of phase with observed surface ocean Cant fluxes, but this can be explained by the time delay for the surface ocean signal to manifest itself in the interior of the ocean.

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

Abstract. Quantifying oceanic anthropogenic carbon uptake by monitoring interior dissolved inorganic carbon (DIC) concentrations is complicated by the influence of natural variability. The "eMLR method" aims to address this issue by using empirical regression fits of the data instead of the data themselves, inferring the change in anthropogenic carbon in time by difference between predictions generated by the regressions at each time. The advantages of the method are that it provides in principle a means to filter out natural variability, which theoretically becomes the regression residuals, and a way to deal with sparsely and unevenly distributed data. The degree to which these advantages are realized in practice is unclear, however. The ability of the eMLR method to recover the anthropogenic carbon signal is tested here using a global circulation and biogeochemistry model in which the true signal is known. Results show that regression model selection is particularly important when the observational network changes in time. When the observational network is fixed, the likelihood that co-located systematic misfits between the empirical model and the underlying, yet unknown, true model cancel is greater, improving eMLR results. Changing the observational network modifies how the spatio-temporal variance pattern is captured by the respective datasets, resulting in empirical models that are dynamically or regionally inconsistent, leading to systematic errors. In consequence, the use of regression formulae that change in time to represent systematically best-fit models at all times does not guarantee the best estimates of anthropogenic carbon change if the spatial distributions of the stations emphasize hydrographic features differently in time. Other factors, such as a balanced and representative station coverage, vertical continuity of the regression formulae consistent with the hydrographic context and resiliency of the spatial distribution of the residual field can be used to help guide model selection. The characteristic spatial scales of the modes of inter-annual to decadal variability in relation to the size of the North Atlantic, in concert with the station coverage available, place practical limits on the ability of eMLR to fully account for natural variability. Due to its statistical nature, eMLR only efficiently removes the natural variability whose spatial scales are smaller than the system analyzed.

Anthropogenic carbon changes in the Irminger Basin (1981–2006): Coupling δ13CDIC and DIC observations

The North Atlantic subpolar gyre is considered to be one of the strongest marine anthropogenic CO2 sinks, a consequence of extensive deep convection occurring during winter. Observations collected in this region since 1981 have shown large changes in Dissolved Inorganic Carbon (DIC) concentrations in intermediate and deep waters, which have been attributed to both anthropogenic CO2 penetration and natural variability in the ocean carbon cycle (Wanninkhof et al., 2010). In this context, we describe new δ13CDIC observations obtained in the Irminger Basin during two OVIDE cruises (2002 and 2006) which we compare to historical data (TTO-NAS 1981) in order to estimate the oceanic 13C Suess Effect over the more than twenty years that separates these surveys. The data reveal a significant decrease in δ13CDIC, of between −0.3‰ and −0.4‰ from 1981 to 2006. The anthropogenic change, extracted by using the extended Multi Linear Regression (eMLR) approach, explains 75% of this signal for oldest water mass and 90% for youngest. The reminding signal is due to the natural processes, such as remineralization and vertical mixing. The eMLR method was also applied to DIC measurements which i) reveal strong relationships between the increase of anthropogenic CO2 and the oceanic 13C Suess Effect over the whole water column during the 25-year period and ii) support the hypothesis of change in the Cant storage rate in the Irminger Basin between 1981 and 2006.

Impacts of temporal CO2and climate trends on the detection of ocean anthropogenic CO2accumulation

[1] A common approach for estimating the oceanic uptake of anthropogenic carbon dioxide (Canthro) depends on the linear approximation of oceanic dissolved inorganic carbon (DIC) from a suite of physical and biological ocean parameters. The extended multiple linear regression (eMLR) method assumes that baseline correlations and the resulting residual fields will remain constant with time even under the influence of secular climate changes. The validity of these assumptions over the 21st century is tested using a coupled carbon-climate model. Findings demonstrate that the influence of both changing climate and changing chemistry beyond 2–4 decades invalidates the assumption that the residual fields will remain constant resulting in significant errors in the eMLR estimate of Canthro. This study determines that the eMLR method is unable to describe Canthro uptake for a sampling interval of greater than 30 years if the error is to remain below 20% for many regions in the Southern Ocean, Atlantic Ocean, and western Pacific Ocean. These results suggest that, for many regions of the ocean basins, hydrographic field investigations have to be repeated at approximately decadal timescales in order to accurately predict the uptake of Canthro by the ocean if the eMLR method is used.