Isolating the Biological Carbon Pump: 3D Numerical Decoupling of Estuarine Carbon Reversals Under Extreme Flood Forcing

Authors:
Preprint
Publisher website
Google Scholar
Abstract
Extreme flood events deliver large terrestrial carbon pulses into coastal waters, where physical, thermodynamic, and biological signals become entangled and are difficult to isolate using conventional methods. We present a 3D numerical decoupling framework that integrates two inseparable components—a CR-NBR dual-run strategy isolating biological from physical signals, and explicit vertical resolution resolving their layer-specific expression across the halocline—applied via Delft3D-WAQ across a complete annual cycle (2018) in the Yatsushiro Sea, Japan. The extreme July 2018 flood triggered a two-stage biological response: riverine POM suppressed phytoplankton growth through light attenuation while fueling heterotrophic mineralization, driving inner-estuary surface pCO2 above 800 µatm—well above the NBR conservative mixing ceiling of ~485 µatm. As turbidity cleared, a delayed autotrophic bloom at the mid-bay generated a temporally bounded biological carbon sink (Ibio = -13.80 mmol C m⁻2; 20 July–30 September). The 3D structure revealed pronounced benthic-pelagic decoupling: haline stratification trapped respiration-derived CO2 within the bottom boundary layer while the surface maintained biological drawdown—a divergence that surface-only observations and 2D depth-averaged models cannot resolve. On the annual scale, the biological pump acted as a net carbon source amplifier at all stations (Ibio: +23.64 to +54.73 mmol C m⁻2), yet all stations remained net CO2 sinks, indicating that biological processes eroded rather than reversed the thermodynamic carbon uptake capacity. Neither the decoupling strategy alone nor 3D resolution alone could achieve these results; their integration provides a transferable framework for quantifying biological pump responses to extreme flood forcing in stratified river-dominated coastal systems.

This publication has 34 references indexed in Scilit: