The detection of deformation in GNSS time series associated with (a)seismic events down to a low magnitude is still a challenging issue. The presence of a considerable amount of noise in the data makes it difficult to reveal patterns of small ground deformation. Traditional analyses and methodologies are able to effectively retrieve the deformation associated to medium to large magnitude events. However, the automatic detection and characterization of such events is still a complex task, because traditionally-employed methods often separate the time series analysis from the source characterization. Here we propose a first end-to-end framework to characterize seismic sources using geodetic data by means of deep learning, which can be an efficient alternative to the traditional workflow, possibly overcoming its performance. We exploit three different geodetic data representations in order to leverage the intrinsic spatio-temporal structure of the GNSS noise and the target signal associated with (slow) earthquake deformation. We employ time series, images and image time series to account for the temporal, spatial and spatio-temporal domain, respectively. Thereafter, we design and develop a specific deep learning model for each data set. We analyze the performance of the tested models both on synthetic and real data from North Japan, showing that image time series of geodetic deformation can be an effective data representation to embed the spatio-temporal evolution, with the associated deep learning method outperforming the other two. Therefore, jointly accounting for the spatial and temporal evolution may be the key to effectively detect and characterize fast or slow earthquakes.

The early stage of the postseismic phase is characterized by a large deformation rate. Its analysis is thus key to decipher the role played by different mechanisms (afterslip and viscoelasticity) at various time scales. Here, we process GPS data to obtain 30-seconds kinematic position time series recording the surface deformation following the Mw 9.0 Tohoku-Oki megathrust earthquake (2011), and combine them with static solutions over 9 years. We analyze the temporal evolution of the time series and use these observations to image the postseismic slip. We find that the first month of deformation following Tohoku-Oki can be explained by an afterslip mechanism, that exhibits an “Omori-like” decay, with a p-value around 0.75 almost everywhere with the exception of a small region around Ibaraki prefecture where p~1 is observed. This p<1 indicates that the postseismic displacements do not increase logarithmically with time as predicted by rate-and-state rheology. Instead, we argue that early afterslip is associated to a transient brittle creep mechanism. We use numerical simulations to show that an exponent of p<1 can be explained by a combination of thermal activation of local slips and elastic interactions. Over longer time scales, an additional mechanism is required to explain the observed deformation signal, and the transient brittle creep mechanism is combined with viscoelastic relaxation modeled by a Newtonian flow. The spatial analysis reveals two distinct afterslip regions, a major one on the North, associated with a p-value around 0.75, and a smaller one close to the Ibaraki aftershock, associated to p~1.

An acceleration of the background seismicity and a shortening of the slow slip events on the Boso peninsula (Japan) recurrence intervals suggest a slow decoupling of the Philippine Sea-North America (PHS-NAM) subduction interface from 1990 to 2011. Motivated by these observations, we used GPS (Global Positioning System) time series to study the 14-year evolution of interface coupling offshore Honshu with a specific focus on the Kanto region. We processed the GPS data in double difference and analyze them with a trajectory model that accounts for seismic and aseismic variations, and that includes an inter-seismic acceleration term. We inverted the surface acceleration obtained, on both the Pacific-North America (PAC-NAM) and the PHS-NAM interfaces. The inverted slip rate changes over time compares well with previous studies: we observe slip deceleration between 39$^o$-41$^o$ N and slip acceleration between 37$^o$-39$^o$ N, with a maximum amplitude of 3.45 mm/yr$^2$ corresponding to an equivalent geodetic coupling change of 0.64. Our analysis reveals a novel and robust slip acceleration South of 36.5$^o$ N that we interpret as a decoupling of the PAC-NAM interface. It is located noticeably far from the 2011 Tohoku earthquake rupture and is therefore unlikely connected to it. We link the slip rate changes to the background seismicity changes and retrieve the slip acceleration from either the seismicity rate or the surface displacement. Our results further demonstrate that inter-seismic slip rate can significantly evolve over years to decades, and suggest a simple relationship between the background seismicity and the slip on the subduction interface.