Sagar Masuti

Imaging the distribution of transient viscosity after the 2016 Mw 7.1 Kumamoto earthquake

Crustal rock strength from outer space The response of crustal rock to stresses is challenging to estimate yet vital for determining risks from events such as earthquakes. Moore et al. take advantage of the recent Mw 7.1 Kumamoto earthquake in Japan to determine the rheology of crustal rocks in the region. The observed inversion of the crustal strain rates demonstrates that certain areas have stiff rock and others (e.g., under the Aso volcanic complex) have much weaker rock. The results match up with expectations, which means that the method can successfully measure rock properties over a wide range of strength and large spatial and temporal scales. Science, this issue p. 163 The combination of GPS and InSAR data after the Kumamoto earthquake in Japan allows for an estimate of regional rock rheology. The deformation of mantle and crustal rocks in response to stress plays a crucial role in the distribution of seismic and volcanic hazards, controlling tectonic processes ranging from continental drift to earthquake triggering. However, the spatial variation of these dynamic properties is poorly understood as they are difficult to measure. We exploited the large stress perturbation incurred by the 2016 earthquake sequence in Kumamoto, Japan, to directly image localized and distributed deformation. The earthquakes illuminated distinct regions of low effective viscosity in the lower crust, notably beneath the Mount Aso and Mount Kuju volcanoes, surrounded by larger-scale variations of viscosity across the back-arc. This study demonstrates a new potential for geodesy to directly probe rock rheology in situ across many spatial and temporal scales.

Upper-mantle water stratification inferred from observations of the 2012 Indian Ocean earthquake

Water, the most abundant volatile in Earth’s interior, preserves the young surface of our planet by catalysing mantle convection, lubricating plate tectonics and feeding arc volcanism. Since planetary accretion, water has been exchanged between the hydrosphere and the geosphere, but its depth distribution in the mantle remains elusive. Water drastically reduces the strength of olivine and this effect can be exploited to estimate the water content of olivine from the mechanical response of the asthenosphere to stress perturbations such as the ones following large earthquakes. Here, we exploit the sensitivity to water of the strength of olivine, the weakest and most abundant mineral in the upper mantle, and observations of the exceptionally large (moment magnitude 8.6) 2012 Indian Ocean earthquake to constrain the stratification of water content in the upper mantle. Taking into account a wide range of temperature conditions and the transient creep of olivine, we explain the transient deformation in the aftermath of the earthquake that was recorded by continuous geodetic stations along Sumatra as the result of water- and stress-activated creep of olivine. This implies a minimum water content of about 0.01 per cent by weight—or 1,600 H atoms per million Si atoms—in the asthenosphere (the part of the upper mantle below the lithosphere). The earthquake ruptured conjugate faults down to great depths, compatible with dry olivine in the oceanic lithosphere. We attribute the steep rheological contrast to dehydration across the lithosphere–asthenosphere boundary, presumably by buoyant melt migration to form the oceanic crust.

Relax-Miracle: GPU parallelization of semi-analytic fourier-domain solvers for earthquake modeling

Effective utilization of GPU processing capacity for scientific workloads is often limited by memory throughput and PCIe communication transfer times. This is particularly true for semi-analytic Fourier-domain computations in earthquake modeling (Relax) where operations on large-scale 3D data structures can require moving large volumes of data from storage to the compute in predictable but orthogonal access patterns. We show how to transform the computation to avoid PCIe transfers entirely by reconstructing the 3D data structures directly within the GPU global memory. We also consider arithmetic transformations that replace some communication-intensive 1D FFTs with simpler, data-parallel analytical solutions. Using our approach we are able to reduce computation times for a geophysical model of the 2012 Mw8.7 Wharton Basin earthquake from 2 hours down to 15 minutes (speedup of ≈8x) for grid sizes of 512-512-256 when comparing NVIDIA K20 with a 16-threaded Intel Xeon E5-2670 CPU (supported by Intel-MKL libraries). Our GPU-accelerated solution (called Relax-Miracle) also makes it possible to conduct Markov-Chain Monte-Carlo simulations using more than 1000 time-dependent models on 12 GPUs per single day of calculation, enhancing our ability to use such techniques for time-consuming data inversion and Bayesian inversion experiments.

Limits of FPGA acceleration of 3D Green's Function computation for geophysical applications

FPGA-based accelerators can outperform multi-core, GPU and Xeon Phi based platforms by at as much as 2.8× for 3D Greens Function processing in geophysics while delivering superior energy efficiency. FPGAs can efficiently implement a complex mixture of compute patterns that include data-parallelism, reductions, dataflow and streaming computations using spatial parallelism to deliver these speedups and power benefits. Since 3D Greens Function is highly-parallel but communication bound, we optimize the FPGA implementation by considering loop restructuring and tiling optimizations to minimize and regularize off-chip accesses. Furthermore, we configure the FPGA to implement the key compute intensive kernels at double-precision as well as single-precision to exploit the uncertainty in measurements of earthquake monitoring sensors. For 512×512×512 problem size, the Xilinx SX475T (Maxeler MAX3) outperforms the fastest architecture by 1.1-1.4× (double-precision), 2.2-2.8× (single-precision) with 1.2× better energy efficiency.