The range of fault slip behaviors near the trench at subduction plate boundaries is critical to know, as this is where the world's largest, most damaging tsunamis are generated. Our knowledge of these behaviors has remained largely incomplete, partially due to the challenging nature of crustal deformation measurements at offshore plate boundaries. Here we present detailed seafloor deformation observations made during an offshore slow-slip event (SSE) in September and October 2014, using a network of absolute pressure gauges deployed at the Hikurangi subduction margin offshore New Zealand. These data show the distribution of vertical seafloor deformation during the SSE and reveal direct evidence for SSEs occurring close to the trench (within 2 kilometers of the seafloor), where very low temperatures and pressures exist.

No abstract

After a large subduction earthquake, crustal deformation continues to occur, with a complex pattern of evolution. This postseismic deformation is due primarily to viscoelastic relaxation of stresses induced by the earthquake rupture and continuing slip (afterslip) or relocking of different parts of the fault. When postseismic geodetic observations are used to study Earth's rheology and fault behaviour, it is commonly assumed that short-term (a few years) deformation near the rupture zone is caused mainly by afterslip, and that viscoelasticity is important only for longer-term deformation. However, it is difficult to test the validity of this assumption against conventional geodetic data. Here we show that new seafloor GPS (Global Positioning System) observations immediately after the great Tohoku-oki earthquake provide unambiguous evidence for the dominant role of viscoelastic relaxation in short-term postseismic deformation. These data reveal fast landward motion of the trench area, opposing the seaward motion of GPS sites on land. Using numerical models of transient viscoelastic mantle rheology, we demonstrate that the landward motion is a consequence of relaxation of stresses induced by the asymmetric rupture of the thrust earthquake, a process previously unknown because of the lack of near-field observations. Our findings indicate that previous models assuming an elastic Earth will have substantially overestimated afterslip downdip of the rupture zone, and underestimated afterslip updip of the rupture zone; our knowledge of fault friction based on these estimates therefore needs to be revised.

[1] On 11 March 2011, the devastating M9.0 Tohoku Earthquake occurred on the interface of the subducting Pacific plate, and was followed by a huge tsunami that killed about 20,000 people. Several geophysical studies have already suggested that the very shallow portion of the plate interface might have played an important role in producing such a large earthquake and tsunami. However, the sparsity of seafloor observations leads to insufficient spatial resolution of the fault slip on such a shallow plate interface. For this reason, the location and degree of the slip has not yet been estimated accurately enough to assess future seismic risks. Thus, we estimated the coseismic slip distribution based on terrestrial GPS observations and all available seafloor geodetic data that significantly improve the spatial resolution at the shallow portion of the plate interface. The results reveal that an extremely large (greater than 50 m) slip occurred in a small (about 40 km in width and 120 km in length) area near the Japan Trench and generated the huge tsunami. The estimated slip distribution and a comparison of it with the coupling coefficient distribution deduced from the analysis of the small repeating earthquakes suggest that the 2011 Tohoku Earthquake released strain energy that had accumulated over the past 1000 years, probably since the Jogan Earthquake in 869. The accurate assessments of seismic risks on very shallow plate interfaces in subduction zones throughout the world can be obtained by improving the quality and quantity of seafloor geodetic observations.

Fig. 4. An example of the localized structures for u = 0.6 x ohm-' m-' and E = 0.05. The image covers an area of 0.1 7 cm by 0.1 7 cm.ulated, yielding an amplitude that varied only slightly over the image. The spatial average An(t) was then studied separately for the four modes. A 30-min segment of A,(t) for the right-traveling zig and zag rolls for E = 0.01 is shown in Fie. 3. A t times one u