Multiple linear regression analysis (MLRA) applied to sediment trap data has been highly influential in identifying a plausible ‘ballasting’ mechanism that directly links the settling fluxes of particulate organic carbon (POC) to those of denser, inorganic minerals. However, analysis to date has primarily been carried out at the global scale, missing spatial variability in the flux relationships that may be important. In this paper, Geographically Weighted Regression (GWR) is applied to an updated deep (>1500 m) sediment trap database (n = 156), using the MLRA approach of Klaas and Archer (2002) but now allowing the carrying coefficients to vary in space. While the global mean carrying coefficient values for CaCO3, opal, and lithogenics are broadly consistent with previous work, the GWR analysis reveals the existence of substantial and statistically significant spatial variability in all three carrying coefficients. In particular, the absence of a strong globally uniform relationship between CaCO3 and POC in our spatial analysis calls into question whether a simple ballasting mechanism exists. Instead, the existence of coherent spatial patterns in carrying coefficients, which are reminiscent of biogeochemical provinces, points toward differences in specific pelagic ecosystem characteristics as being the likely underlying cause of the flux relationships sampled by sediment traps. Our findings present a challenge to ocean carbon cycle modelers who to date have applied a single statistical global relationship in their carbon flux parameterizations when considering mineral ballasting, and provide a further clue as to how the efficiency of the biological pump in the modern ocean is regulated.
- POC flux
- ballast hypothesis
- geographically weighted regression
- organic carbon export