Search for: All records

Mantle viscosity exerts important controls on the long-term (i.e., >106 years) dynamics of the mantle and lithosphere and the short-term (i.e., 10 to 104 years) crustal motion induced by loading forces including ice melting, sea-level changes, and earthquakes. However, mantle viscosity structures inferred from modeling observations associated with mantle dynamic and loading processes may differ significantly and remain a hotly debated topic over recent decades. In this study, we investigate the effects of mantle viscosity structures on observations of the geoid, mantle structures, and present-day crustal motions and time-varying gravity by considering five representative mantle viscosity structures in models of mantle convection and glacial isostatic adjustment (GIA). These five viscosity models fall into two categories: 1) two viscosity models derived from modeling the geoid in mantle convection models with ~100 times more viscous lower mantle than the upper mantle, and 2) the other three with less viscosity increase from the upper to lower mantles that are derived from modeling the late Pleistocene and Holocene relative sea level changes and other observations in GIA models. Our convection models use the plate motion history for the last 130 Myrs as the surface boundary conditions and depth- and temperature-dependent viscosity to predict the present-day convective mantle structure of subducted slabs and the intermediate wavelength (degrees 4–12) geoid. Our GIA models using different ice history models (e.g., ICE-6 G and ANU) compute the GIA-induced present-day crustal motions and time-varying gravity. Our calculations demonstrate that while the viscosity models with a higher viscosity in the lower mantle (~2 × 1022 Pa.s) reproduce the degrees 4–12 geoid and seismic slab structures, they significantly over-predict the geodetic (i. e., GPS and GRACE) observations of crustal motions and time varying gravity. Our calculations also show that while two viscosity models derived from fitting the RSL data with averaged mantle viscosity of ~1021 Pa.s for the top 1200 km of the mantle reproduce well the geodetic observations independent of ice models, they fail to explain the geoid and seismic slab structures. Therefore, our study highlights the persisting conundrum of mantle viscosity structures derived from different observations. We also discuss a number of possible ways including transient, stress-dependent and 3-D viscosity to resolve this important issue in Geodynamics. 

Abstract Deep earthquakes at depths below 500 km are under prohibitive pressure and temperature conditions for brittle failure. Individual events show diverse rupture behaviors and a coherent mechanism to explain their rupture nucleation, propagation, and characteristics has yet to be established. We systematically resolve the rupture processes of 40 large deep earthquakes from 1990 to 2023 and compare the rupture details to their local metastable olivine wedge (MOW) structures informed from thermo‐mechanical simulations in seven subduction zones. Our results suggest that these events likely initiate from metastable olivine transformations within the cold slab core and rupture beyond the MOW due to sustained weakening from molten rock at the rupture tip. Over half of the earthquakes likely rupture beyond the MOW boundary and are controlled by both mechanisms. Rupturing outside the MOW boundary leads to greater moment release, increased geometric complexity, and a reduction in rupture length, causing greater stress drops. 

Acoustic ranging is a technique for estimating the distance between two objects using acoustic signals, which plays a critical role in many applications, such as motion tracking, gesture/activity recognition, and indoor localization. Although many ranging algorithms have been developed, their performance still degrades significantly under strong noise, interference and hardware limitations. To improve the robustness of the ranging system, in this paper we develop a Deep learning based Ranging system, called DeepRange. We first develop an effective mechanism to generate synthetic training data that captures noise, speaker/mic distortion, and interference in the signals and remove the need of collecting a large volume of training data. We then design a deep range neural network (DRNet) to estimate distance. Our design is inspired by signal processing that ultra-long convolution kernel sizes help to combat the noise and interference. We further apply an ensemble method to enhance the performance. Moreover, we analyze and visualize the network neurons and filters, and identify a few important findings that can be useful for improving the design of signal processing algorithms. Finally, we implement and evaluate DeepRangeusing 11 smartphones with different brands and models, 4 environments (i.e., a lab, a conference room, a corridor, and a cubic area), and 10 users. Our results show that DRNet significantly outperforms existing ranging algorithms.