11We characterise the aftershock sequence following the 2016 Mw=7.8 Pedernales earthquake. 12More than 10,000 events were detected and located, with magnitudes up to 6.9. Most of the 13 aftershock seismicity results from interplate thrust faulting, but we also observe a few normal 14 and strike-slip mechanisms. Seismicity extends for more than 300 km along strike, and is 15 constrained between the trench and the maximum depth of the coseismic rupture. The most 16 striking feature is the presence of three seismicity bands, perpendicular to the trench, which 17 are also observed during the interseismic period. Additionally, we observe a linear 18 dependency between the temporal evolution of afterslip and aftershocks. We also find a 19 temporal semi-logarithmic expansion of aftershock seismicity along strike and dip directions, 20 further indicating that their occurrence is modulated by afterslip. Lastly, we observe that the 21 spatial distribution of seismic and aseismic slip processes is correlated to the distribution of 22 bathymetric anomalies associated with the northern flank of the Carnegie Ridge, suggesting 23 that slip in the area could be influenced by the relief of the subducting seafloor. To explain 24 our observations, we propose a conceptual model in which the Ecuadorian margin is subject 25 to a bimodal slip mode, with distributed seismic and aseismic slip mechanically controlled by 26 the subduction of a rough oceanic relief. Our study sheds new light on the mechanics of 27 subduction, relevant for convergent margins with a complex and heterogeneous structure 28 such as the Ecuadorian margin. 29

High-Rate (HR) GPS time series following the 2016 M w 7.8 Pedernales earthquake suggest significant postseismic deformation occurring in the early postseismic period (i.e. first few hours after the earthquake) that is not captured in daily GPS time series. To understand the characteristics of early postseismic deformation, and its relationship with the mainshock rupture area, aftershocks and longer-term postseismic deformation, we estimate the spatiotemporal distribution of early afterslip with HR-GPS time series that span ~ 2.5 minutes to 72 hours after the earthquake, and compare with afterslip models estimated with daily GPS time series spanning a similar postseismic time period and up to 30 days after the earthquake. Our inversion technique enables us to image the nucleation of afterslip in the initial hours after the earthquake, bringing us closer to the transition between the coseismic and postseismic phases. The spatial signature of early afterslip in the region updip of the mainshock rupture area is consistent with longer-term afterslip that occurs in the 30-day postseismic period, indicating that afterslip nucleated updip of and adjacent to peak coseismic slip asperities, in two localized areas, and subsequently continued to grow in amplitude with time in these specific areas. A striking difference, however, is that inversion of the 72-hour HR-GPS time series suggests early afterslip within the mainshock rupture area, but which may have been short-lived. More interestingly, we find that postseismic slip starts immediately after the earthquake at a rapid rate. Indeed, we find that early afterslip represents a significant contribution to the postseismic geodetic moment-afterslip in the first 72 hours is ~60 % greater than that estimated with daily GPS time series that span the first three post-earthquake daily GPS positions (i.e. covering the same time window). The results of our study demonstrate that imaging the spatio-temporal evolution of afterslip using subdaily GPS time series is important for evaluating postseismic slip 3 | P a g e budgets, and provides additional insights into the postseismic slip behaviour of faults.