A Dye-Tracer Forward-Modeling Framework for Deglacial Meltwater Reconstruction

Freshwater input from melting polar ice sheets can profoundly alter ocean circulation, in particular the Atlantic Meridional Overturning Circulation (AMOC), with far-reaching climatic consequences. Yet the sensitivity of the AMOC to freshwater forcing remains highly uncertain: models exhibit divergent responses depending on source location, background climate state, and circulation regime, while the instrumental record is too short to unambiguously detect and characterise a melt-driven weakening.

Palaeoclimate archives, especially from the last deglaciation, provide ample evidence of melt events through indicators such as surface-ocean δ¹⁸O and biomarkers (e.g. BIX) in sediment cores and speleothems. However, the spatial and temporal characteristics of the underlying meltwater forcing remain poorly constrained. While meltwater discharge into the North Atlantic may be local, rapid, and event-like, its redistribution and impact on the AMOC unfold over centuries, complicating direct inference from surface-ocean proxies. Consequently, in deglacial general circulation model simulations, meltwater forcing is typically inferred indirectly from ice-sheet reconstructions or expected climate responses, resulting in a wide spread of applied forcings that propagates into substantial uncertainty.

Here we introduce a new forward-modelling approach aimed at strengthening the estimation and detection of regionally distinct and temporally evolving surface-ocean meltwater signals in proxy archives. We develop an empirical Green’s-function (impulse-response) framework based on a new suite of HadCM3 simulations, in which conservative tracers track meltwater originating from different source regions under distinct AMOC modes representative of deglacial conditions. Signals at terrestrial proxy sites are inferred using atmospheric back-trajectory analysis. The resulting kernels encode the system’s response for different source regions across multiple time lags, allowing any transient meltwater history to be reconstructed through discrete convolution with a derived 500-year response function. Applied to the last deglaciation, the framework demonstrates how differences between ice-sheet reconstructions (e.g. GLAC-1D versus ICE-6G) translate into distinct surface-ocean meltwater anomalies in the North Atlantic. The model highlights the critical role of meltwater amount, timing, and injection location, as well as the underlying AMOC circulation mode, in shaping surface-ocean proxy signals. It further provides quantitative estimates of how meltwater-related surface anomalies propagate to proxy sites distributed across the North Atlantic. Notably, transitions between AMOC modes can effectively mask even massive meltwater pulses, such as Meltwater Pulse 1A, at certain proxy locations. This forward-modelling approach thus offers an alternative perspective on deglacial freshwater forcing in the proxy realm and represents a step towards data-constrained reconstructions of past surface-ocean freshening and AMOC resilience.