Effect of sampling bias on global estimates of ocean carbon export

Abstract

Shipboard sampling of ocean biogeochemical properties is necessarily limited by logistical and practical constraints. As a result, the majority of observations are obtained for the spring/summer period and in regions relatively accessible from a major port. This limitation may bias the conceptual understanding we have of the spatial and seasonal variability in important components of the Earth system. Here we examine the influence of sampling bias on global estimates of carbon export flux by sub-sampling a biogeochemical model to simulate real, realistic and random sampling. We find that both the sparseness and the ‘clumpy’ character of shipboard flux observations generate errors in estimates of globally extrapolated export flux of up to ~ ± 20%. The use of autonomous technologies, such as the Biogeochemical-Argo network, will reduce the uncertainty in global flux estimates to ~ ± 3% by both increasing the sample size and reducing clumpiness in the spatial distribution of observations. Nevertheless, determining the climate change-driven trend in global export flux may be hampered due to the uncertainty introduced by interannual variability in sampling patterns.

Observational and Numerical Modeling Constraints on the Global Ocean Biological Carbon Pump

“…A variety of in situ methods have been used to estimate surface ocean export flux estimates (∼F 100 ) including drifting sediment traps and 234 Th deficits. Global Biogeochemical Cycles 10.1029/2024GB008156 situ data is required, which generates uncertainties in the derived estimates due to the underlying data sparsity (Henson et al, 2024). Typically, satellite data are used to build an empirical relationship between flux and readily derived variables, such as sea surface temperature or chlorophyll concentration.…”