How data set characteristics influence ocean carbon export models

Author Posting. © American Geophysical Union, 2018. This article is posted here by permission of American Geophysical Union for personal use, not for redistribution. The definitive version was published in Global Biogeochemical Cycles 32 (2018): 1312-1328, doi:10.1029/2018GB005934.Ocean biological processes mediate the transport of roughly 10 petagrams of carbon from the surface to the deep ocean each year and thus play an important role in the global carbon cycle. Even so, the globally integrated rate of carbon export out of the surface ocean remains highly uncertain. Quantifying the processes underlying this biological carbon export requires a synthesis between model predictions and available observations of particulate organic carbon (POC) flux; yet the scale dissimilarities between models and observations make this synthesis difficult. Here we compare carbon export predictions from a mechanistic model with observations of POC fluxes from several data sets compiled from the literature spanning different space, time, and depth scales as well as using different observational methodologies. We optimize model parameters to provide the best match between model‐predicted and observed POC fluxes, explicitly accounting for sources of error associated with each data set. Model‐predicted globally integrated values of POC flux at the base of the euphotic layer range from 3.8 to 5.5 Pg C/year, depending on the data set used to optimize the model. Modeled carbon export pathways also vary depending on the data set used to optimize the model, as well as the satellite net primary production data product used to drive the model. These findings highlight the importance of collecting field data that average over the substantial natural temporal and spatial variability in carbon export fluxes, and advancing satellite algorithms for ocean net primary production, in order to improve predictions of biological carbon export.NASA Ocean Biology and Biogeochemistry Program Grant Numbers: NNX16AR49G, NNXA122G, NNX16AR47G, OBB16_2‐0031; National Science Foundation2019-03-1