Neural network-based estimates of Southern Ocean net community production from in-situ O<sub>2</sub> / Ar and satellite observation: a methodological study

Abstract

Southern Ocean organic carbon export plays an important role in the global carbon cycle, yet its basin-scale climatology and variability are uncertain due to limited coverage of in situ observations. In this study, a neural network approach based on the self-organizing map (SOM) is adopted to construct weekly gridded (1° × 1°) maps of organic carbon export for the Southern Ocean from 1998 to 2009. The SOM is trained with in situ measurements of O₂ / Ar-derived net community production (NCP) that are tightly linked to the carbon export in the mixed layer on timescales of 1–2 weeks, and six potential NCP predictors: photosynthetically available radiation (PAR), particulate organic carbon (POC), chlorophyll (Chl), sea surface temperature (SST), sea surface height (SSH), and mixed layer depth (MLD). This non-parametric approach is based entirely on the observed statistical relationships between NCP and the predictors, and therefore is strongly constrained by observations. <br><br> A thorough cross-validation yields three retained NCP predictors, Chl, PAR, and MLD. Our constructed NCP is further validated by good agreement with previously published independent in situ derived NCP of weekly or longer temporal resolution through real-time and climatological comparisons at various sampling sites. The resulting November–March NCP climatology reveals a pronounced zonal band of high NCP roughly following the subtropical front in the Atlantic, Indian and western Pacific sectors, and turns southeastward shortly after the dateline. Other regions of elevated NCP include the upwelling zones off Chile and Namibia, Patagonian Shelf, Antarctic coast, and areas surrounding the Islands of Kerguelen, South Georgia, and Crozet. This basin-scale NCP climatology closely resembles that of the satellite POC field and observed air-sea CO₂ flux. The long-term mean area-integrated NCP south of 50° S from our dataset, 14 mmol C m^&ndash;2 d^&ndash;1, falls within the range of 8.3–24 mmol C m^&ndash;2 d^&ndash;1 from other model estimates. A broad agreement is found in the basin-wide NCP climatology among various models but with significant spatial variations, particularly in the Patagonian Shelf. Our approach provides a comprehensive view of the Southern Ocean NCP climatology and a potential opportunity to further investigate interannual and intraseasonal variability

Satellite estimates of net community production indicate predominance of net autotrophy in the Atlantic Ocean

“…Similarly, our NCP C algorithm yielded estimates of 23 mmol O 2 m −2 d −1 over the SeaWiFS time series for these months, though values from NCP D were lower (11 mmol O 2 m −2 d −1 ). Chang et al (2014), using a neural network approach, reported mean area-integrated NCP south of 50°S in the Southern Ocean of 18 mmol C m − 2 d −1 . For the SSTC region, they reported N30 mmol C m −2 d −1 , which is similar to our estimate of 33 mmol O 2 m − 2 d − 1 using NCP C (assuming PQ = 1.1), though this was double using NCP D , which may suggest that higher PQ values are more appropriate for this region.…”

“…Satellite models of NCP, based on net primary production and export, have also been proposed for the Southern Ocean (Nevison et al, 2012). Chang et al (2014) developed a neural network approach based on self-organizing maps to construct weekly gridded maps of organic carbon export for the Southern Ocean. The maps were trained with in situ measurements of O 2 /Ar to estimate NCP linked to potential predictors of NCP through statistical relationships with photosynthetically available radiation (PAR), Chla and mixed layer depth (MLD).…”

“…All other data were used for validation only. production to estimate NCP in specific regions of the Atlantic Ocean (Duarte, Agusti, Aristegui, Gonzalez, & Anadon, 2001;Serret, Robinson, Fernandez, Teira, & Tilstone, 2001;Serret, Fernandez, & Robinson, 2002), in the Southern Ocean (Chang, Johnson, & Cassar, 2014;Nevison et al 2012) and at global scales (Westberry, Williams, & Behrenfeld, 2012). Serret et al (2009) derived empirical relationships between measurements of in vitro O 2 based NCP and 14 C particulate organic carbon production ( 14 C POC) to predict NCP in either eutrophic or oligotrophic environments.…”

Characteristics of the surface water DMS and pCO2 distributions and their relationships in the Southern Ocean, southeast Indian Ocean, and northwest Pacific Ocean

“…Recently, Kameyama et al [2013] found evidence for a strong positive relationship between DMS and net community production (NCP) in the North Pacific Ocean. Thus, it is possible that the DMS could be reconstructed by remotely derived NCP [Chang et al, 2014]. Thus, it is possible that the DMS could be reconstructed by remotely derived NCP [Chang et al, 2014].…”

Interannual variability in net community production at the Western Antarctic Peninsula region (1997–2014)

“…For example, Chang et al . [] and Li and Cassar [] recently estimated NCP for the Southern Ocean and the global oceans using in situ O 2 /Ar‐NCP estimates and various statistical methods including Artificial Neural Networks (ANNs), Support Vector Regression (SVR), and Genetic Programming (GP). Although these approaches estimate NCP relatively well, uncertainties remain significant, in part due to factors controlling NCP varying from one area to another.…”