Importance of understanding the evolution of crustal permeability for the apparent friction coefficients in Japanese subduction zones

The Nankai subduction zone in Southwest Japan is vulnerable to megathrust earthquakes posing a significant risk to the infrastructure and population it accommodates. This region has gained recent interest after the Hyuga-nada earthquake on 8th August 2024, because a megathrust earthquake, which has not occurred for the last 80 years despite its cycle known as 100 – 150 years, can be triggered by the event. Understanding earthquake mechanisms can mitigate the potential damage. The frictional condition at the plate interface is one of the key factors in estimating the location and magnitude of the potential megathrust earthquake. A previous study used numerical modelling that includes frictional heat to find the best apparent friction coefficient (μ') to explain the observed seafloor heat flow. However, hydrothermal circulation (HC) was not considered in this previous model although it significantly affects the thermal structure and the seafloor heat flow by redistributing heat energy. Therefore, we conducted numerical modelling that includes HC to find μ' values for the two subduction zones known for high risks of potential megathrust earthquakes – the Nankai and Tohoku (Northeast Japan) subduction zones. The results show that a wide range of μ' (0.00 – 0.30 and 0.00 – 0.12 for the Nankai and Tohoku subduction zones, respectively) can explain the observed seafloor heat flow depending on the vigour and extent of HC. This indicates that μ' cannot be constrained using heat flow observations before the evolution of the aquifer permeability is understood. Here, we suggest that the age of the oceanic crust and bending-induced faulting play a crucial role in the evolution of the aquifer permeability, resulting in a slowly decreasing permeability. Therefore, to better understand the frictional condition within a subduction zone, various fields of research – magnetic and seismological surveys, field and laboratory measurements, etc. – should work together as well as computational modelling.