Upward mantle flow from novel seismic observations

ongoing

While downward mantle flow (subduction) is well constrained, a grand challenge in Earth sciences is to understand deep upward flow that cannot be explained by plate tectonics. The aim of UPFLOW is to develop an entirely new seismic imaging approach and to use it to constrain plume-like mantle upwellings at unprecendented resolution. Upward flow is critical for continental growth, for returning volatiles to the atmosphere and for producing Earth's largest melting events. These events coincide with major extinctions, supercontinent breakups and with changes in geodynamo behaviour. Seismology has come far by using travel-time and waveform (phase) information but has mostly ignored the amplitude information. Mantle plumes have just started to be robustly imaged, yet resolution is low and uncertainty quantification is lacking. I developed breakthrough seismic imaging methods based on big amplitude data, showing their strong resolution. With recent computing resources and explosion in datasets, it is now timely to use this unique expertise to lead a step change in observational seismology: rather than relying on pre-defined observables (e.g., travel-times), objective statistical techniques exploring the full richness of the data will be used to design observables that provide optimal constraints on Earth structure. I will lead a new off-shore experiment in the Azores-Madeira-Canary islands region, which is an under-studied natural laboratory comprising significant mantle upwellings that are poorly understood in general. The imaging technique to be developed will be applied to these new data and to existing data (Hawaii, Reunion) to obtain the sharpest ever images of mantle plumes along with their interdisciplinary interpretation. These results will have a major impact far beyond the study regions and across Earth Sciences (e.g., geology, geochemistry, geodynamics) since they will reveal ubiquitous, fundamental phenomena that control the global evolution of the planet.