Document Type


Date of Original Version





The shunt of photosynthetically derived particulate organic carbon (POC) from the euphotic zone and deep remineralization comprises the basic mechanism of the “biological carbon pump.” POC raining through the “twilight zone” (euphotic depth to 1 km) and “midnight zone” (1 km to 4 km) is remineralized back to inorganic form through respiration. Accurately modeling POC flux is critical for understanding the “biological pump” and its impacts on air‐sea CO2 exchange and, ultimately, long‐term ocean carbon sequestration. Yet commonly used parameterizations have not been tested quantitatively against global data sets using identical modeling frameworks. Here we use a single one‐dimensional physical‐biogeochemical modeling framework to assess three common POC flux parameterizations in capturing POC flux observations from moored sediment traps and thorium‐234 depletion. The exponential decay, Martin curve, and ballast model are compared to data from 11 biogeochemical provinces distributed across the globe. In each province, the model captures satellite‐based estimates of surface primary production within uncertainties. Goodness of fit is measured by how well the simulation captures the observations, quantified by bias and the root‐mean‐square error and displayed using “target diagrams.” Comparisons are presented separately for the twilight zone and midnight zone. We find that the ballast hypothesis shows no improvement over a globally or regionally parameterized Martin curve. For all provinces taken together, Martin's b that best fits the data is [0.70, 0.98]; this finding reduces by at least a factor of 3 previous estimates of potential impacts on atmospheric pCO2 of uncertainty in POC export to a more modest range [−16 ppm, +12 ppm].