publications

2024

1. arXiv

   Domain decomposition of the modified Born series approach for large-scale wave propagation simulations

   Swapnil Mache, and Ivo M Vellekoop

   2024

   Abs arXiv Bib HTML PDF

   The modified Born series method is a fast and accurate method for simulating wave propagation in complex structures. Currently, its main limitation is that the size of the simulation is limited by the working memory of a single computer or graphics processing unit (GPU). Here, we present a domain decomposition method that removes this limitation. We show how to decompose large problems over subdomains while maintaining the accuracy, memory efficiency, and guaranteed monotonic convergence of the method. We demonstrate our approach by solving the Helmholtz problem for a complex structure of size 315×315×315 wavelengths in just 379 seconds on a dual-GPU system.

   @article{mache2024domain,
     title = {Domain decomposition of the modified Born series approach for large-scale wave propagation simulations},
     author = {Mache, Swapnil and Vellekoop, Ivo M},
     year = {2024},
     eprint = {2410.02395},
     archiveprefix = {arXiv},
     primaryclass = {physics.comp-ph},
   }

2023

1. IEEE TCI

   Introducing Nonuniform Sparse Proximal Averaging Network for Seismic Reflectivity Inversion

   Swapnil Mache, Praveen Kumar Pokala, Kusala Rajendran, and 1 more author

   IEEE Transactions on Computational Imaging, 2023

   Abs Bib HTML

   We consider the problem of seismic reflectivity inversion, which pertains to the high-resolution recovery of interface locations and reflection coefficients from seismic measurements, which are vital for estimating the subsurface structure. We develop two model-based neural networks within the framework of deep-unfolding. First, we propose a nonuniform minimax concave penalty regularized formulation for reflectivity inversion and unfold the resulting iterative algorithm into a network. Second, we propose a nonuniform sparse model that relies on a combination of regularizers (composite regularization) and develop the nonuniform sparse proximal averaging network (NuSPAN). We demonstrate the efficacy of the proposed approaches over the benchmark techniques through numerical experiments on synthetic 1-D seismic traces and 2-D wedge models. We also report validations on the 2-D Marmousi2 simulated model and 3-D real field measurements from the Penobscot 3D survey off the coast of Nova Scotia, Canada. The accuracy of the proposed approaches is higher than the state-of-the-art techniques in terms of amplitude and support recovery. Further, for Marmousi2, the proposed deep-unfolding networks result in 600× faster inference than the fast iterative shrinkage-thresholding algorithm (FISTA). In terms of combined training and inference times, the learned iterative shrinkage-thresholding algorithm (LISTA) is the fastest. The inference speed-up is significant given that the volume of data is typically large in seismic signal processing.

   @article{mache2023introducing,
     author = {Mache, Swapnil and Pokala, Praveen Kumar and Rajendran, Kusala and Seelamantula, Chandra Sekhar},
     journal = {IEEE Transactions on Computational Imaging},
     title = {Introducing Nonuniform Sparse Proximal Averaging Network for Seismic Reflectivity Inversion},
     year = {2023},
     volume = {9},
     number = {},
     pages = {475-489},
     keywords = {Reflectivity;Imaging;Optimization;Iterative methods;Convolution;Seismic measurements;Signal processing algorithms;Algorithm unrolling;geophysics;non-convex optimization;reflectivity inversion;seismic imaging;sparsity},
     doi = {10.1109/TCI.2023.3277629},
     issn = {2333-9403},
     month = {},
   }

2. Marine Geol.

   Temporally variable recurrence regimes of mega-tsunamis in the 6500 years prior to the 2004 Indian Ocean event

   Jaishri Sanwal, C.P. Rajendran, Mohammad Heidarzadeh, and 3 more authors

   Marine Geology, 2023

   Abs Bib HTML

   The analyses of cores retrieved from three sites near Port Blair (South Andaman) revealed out-of-sequence deposits at various depths. They are identified as previous episodes of tsunami by their sediment characteristics and microfossil content, using the 2004 event deposition as a template. These deposits have median ages of 601 cal. yr BP, 837 cal. yr BP, 1440 cal. yr BP, 3018 cal. yr BP, 3591 cal. yr BP, 4712 cal. yr BP, 5607 cal. yr BP, and 6357 cal. yr BP and are chronologically equivalent of those from the far-field locations in the Indian Ocean region. The distant deposits that are correlated with the South Andaman sites most likely owe their origin to the 2004-type events, as indicated by tsunami simulations in the study region. The long-term record presented here is characterized by an early phase of a quasiperiodic recurrence regime that transitions into a distinct interval of temporally clustered events. The quasiperiodic regime that appears around the mid-Holocene with an inter-event interval of 980 ± 385 years is followed by a sizable quiescent period of 1605 ± 245 years, before being succeeded by a regime of temporally clustered events. The chronology of nine tsunami events in the last 6500 years from the Indian Ocean region, thus implies a nonlinear pattern for the causative earthquakes. As demonstrated in the subduction zones elsewhere, the temporal variability of tsunamigenic great earthquakes is supported by the theoretical models espousing the characteristics of long-term stress recycling processes active within the subduction zones and transfer processes between the lower viscoelastic layer and the upper seismogenic crust.

   @article{sanwal2023temporally,
     title = {Temporally variable recurrence regimes of mega-tsunamis in the 6500 years prior to the 2004 Indian Ocean event},
     journal = {Marine Geology},
     volume = {460},
     pages = {107051},
     year = {2023},
     issn = {0025-3227},
     doi = {https://doi.org/10.1016/j.margeo.2023.107051},
     url = {https://www.sciencedirect.com/science/article/pii/S0025322723000634},
     author = {Sanwal, Jaishri and Rajendran, C.P. and Heidarzadeh, Mohammad and Mache, Swapnil and Anandasabari, K. and Rajendran, Kusala},
     keywords = {Indian Ocean, The 2004 earthquake, Subduction zone, Tsunami recurrence, Supercycles},
   }

2022

1. BSSA

   Hilbert–Huang Transform and Energy Rate Functions for Earthquake Source Characterization—A Study from the Japan Trench

   Swapnil Mache, Avigyan Chatterjee, Kusala Rajendran, and 1 more author

   Bulletin of the Seismological Society of America, Sep 2022

   Abs Bib HTML

   The Hilbert–Huang Transform (HHT) has been sparsely applied to problems in seismology, although previous studies have pointed to its broad scope. In this maiden attempt, we use the HHT to represent earthquake energy release duration and frequency content and compare the results with two conventional inversion methods. By selecting examples from interplate, intraplate, and intraslab settings, we demonstrate that the HHT has the power to discriminate energy release of earthquakes with different tectonic affiliations. We observe that the dominant frequencies for energy release are higher for intraslab earthquakes than for interplate and intraplate events. We use the empirical mode decomposition‐based HHT and introduce a new parameter, which we name the energy rate function (ERF), to quantify the energy release. By employing empirical Green’s functions to remove the path and site effects and using a linear combination of a select set of intrinsic mode functions, we generate the station‐specific relative measure of energy that we refer to as relative ERFs (RERFs). Averaged over RERFs from multiple stations, the ERF represents a measure of the total relative energy release, comparable to the moment rate functions (MRFs) and SCARDEC source time functions (STFs). Results for six of the seven earthquakes we analyzed show high cross correlation with the STFs (0.84 ± 0.03) and MRFs (0.79 ± 0.06), but there are mismatches between ERFs and MRFs or STFs when the energy release is complex and involves multisegment or bilateral ruptures. The proposed method is computationally efficient, requiring only 3.46 ± 2.62 s on average, compared to  20 min ( 1200 s) for the teleseismic inversion method we employ. With its ability to represent the seismic source in terms of energy release, the ERF method has the potential to evolve not as an alternative to waveform inversion but as a rapid time–frequency analysis tool, useful for earthquake hazard assessment.

   @article{mache2022hilbert,
     author = {Mache, Swapnil and Chatterjee, Avigyan and Rajendran, Kusala and Seelamantula, Chandra Sekhar},
     title = {{H}ilbert–{H}uang Transform and Energy Rate Functions for Earthquake Source Characterization—A Study from the {J}apan Trench},
     journal = {Bulletin of the Seismological Society of America},
     volume = {112},
     number = {6},
     pages = {2847-2858},
     year = {2022},
     month = sep,
     issn = {0037-1106},
     doi = {10.1785/0120220099},
     url = {https://doi.org/10.1785/0120220099},
     eprint = {https://pubs.geoscienceworld.org/ssa/bssa/article-pdf/112/6/2847/5740109/bssa-2022099.1.pdf},
   }

2. ICIP

   An Ensemble of Proximal Networks for Sparse Coding

   Kartheek K Reddy Nareddy, Swapnil Mache, Praveen Kumar Pokala, and 1 more author

   In 2022 IEEE International Conference on Image Processing (ICIP), Oct 2022

   Abs Bib HTML

   Sparse coding methods are iterative and typically rely on proximal gradient methods. While the commonly used sparsity promoting penalty is the ℓ1 norm, alternatives such as the minimax concave penalty (MCP) and smoothly clipped absolute deviation (SCAD) penalty have also been employed to obtain superior results. Combining various penalties to achieve robust sparse recovery is possible, but the challenge lies in parameter tuning. Given the connection between deep networks and unrolling of iterative algorithms, it is possible to unify the unfolded networks arising from different formulations. We propose an ensemble of proximal networks for sparse recovery, where the ensemble weights are learnt in a data-driven fashion. We found that the proposed network performs superior to or on par with the individual networks in the ensemble for synthetic data under various noise levels and sparsity conditions. We demonstrate an application to image denoising based on the convolutional sparse coding formulation.

   @inproceedings{reddey2022ensemble,
     author = {Reddy Nareddy, Kartheek K and Mache, Swapnil and Pokala, Praveen Kumar and Sekhar Seelamantula, Chandra},
     booktitle = {2022 IEEE International Conference on Image Processing (ICIP)},
     title = {An Ensemble of Proximal Networks for Sparse Coding},
     year = {2022},
     volume = {},
     number = {},
     pages = {1251-1255},
     keywords = {Training; Gradient methods; Image coding; Noise reduction; Image restoration; Iterative methods; Task analysis; Ensemble networks; deep-unfolding; non-convex optimization; sparse coding; image denoising},
     doi = {10.1109/ICIP46576.2022.9897607},
     issn = {2381-8549},
     month = oct,
   }

2021

1. arXiv

   NuSPAN: A Proximal Average Network for Nonuniform Sparse Model – Application to Seismic Reflectivity Inversion

   Swapnil Mache, Praveen Kumar Pokala, Kusala Rajendran, and 1 more author

   Oct 2021

   Abs arXiv Bib HTML PDF

   We solve the problem of sparse signal deconvolution in the context of seismic reflectivity inversion, which pertains to high-resolution recovery of the subsurface reflection coefficients. Our formulation employs a nonuniform, non-convex synthesis sparse model comprising a combination of convex and non-convex regularizers, which results in accurate approximations of the l0 pseudo-norm. The resulting iterative algorithm requires the proximal average strategy. When unfolded, the iterations give rise to a learnable proximal average network architecture that can be optimized in a data-driven fashion. We demonstrate the efficacy of the proposed approach through numerical experiments on synthetic 1-D seismic traces and 2-D wedge models in comparison with the benchmark techniques. We also present validations considering the simulated Marmousi2 model as well as real 3-D seismic volume data acquired from the Penobscot 3D survey off the coast of Nova Scotia, Canada.

   @misc{mache2021nuspan,
     title = {{NuSPAN}: A Proximal Average Network for Nonuniform Sparse Model -- Application to Seismic Reflectivity Inversion},
     author = {Mache, Swapnil and Pokala, Praveen Kumar and Rajendran, Kusala and Seelamantula, Chandra Sekhar},
     year = {2021},
     eprint = {2105.00003},
     archiveprefix = {arXiv},
     primaryclass = {physics.geo-ph},
   }

2. arXiv

   DuRIN: A Deep-unfolded Sparse Seismic Reflectivity Inversion Network

   Swapnil Mache, Praveen Kumar Pokala, Kusala Rajendran, and 1 more author

   Oct 2021

   Abs arXiv Bib HTML PDF

   We consider the reflection seismology problem of recovering the locations of interfaces and the amplitudes of reflection coefficients from seismic data, which are vital for estimating the subsurface structure. The reflectivity inversion problem is typically solved using greedy algorithms and iterative techniques. Sparse Bayesian learning framework, and more recently, deep learning techniques have shown the potential of data-driven approaches to solve the problem. In this paper, we propose a weighted minimax-concave penalty-regularized reflectivity inversion formulation and solve it through a model-based neural network. The network is referred to as deep-unfolded reflectivity inversion network (DuRIN). We demonstrate the efficacy of the proposed approach over the benchmark techniques by testing on synthetic 1-D seismic traces and 2-D wedge models and validation with the simulated 2-D Marmousi2 model and real data from the Penobscot 3D survey off the coast of Nova Scotia, Canada.

   @misc{mache2021durin,
     title = {{DuRIN}: A Deep-unfolded Sparse Seismic Reflectivity Inversion Network},
     author = {Mache, Swapnil and Pokala, Praveen Kumar and Rajendran, Kusala and Seelamantula, Chandra Sekhar},
     year = {2021},
     eprint = {2104.04704},
     archiveprefix = {arXiv},
     primaryclass = {physics.geo-ph},
   }

3. AGU

   Explainable Deep Neural Networks for Seismic Reflectivity Inversion

   Swapnil Mache, Praveen Kumar Pokala, Kusala Rajendran, and 1 more author

   Dec 2021

   AGU Fall Meeting 2021

   Abs Bib Poster

   Reflectivity inversion/model building is an important inverse problem in seismic imaging. Considering a convolutional model by assuming a piecewise-constant impedance profile, we present a novel deep neural network architecture, namely, Nonuniform Sparse Proximal Average Network (NuSPAN), to solve the problem within the framework of model-based prior learning (Mache et al., 2021a, 2021b). The architecture is inspired by sparse recovery algorithms. Greedy and iterative techniques formulate the problem as an l1-norm regularization problem. Although this is a widely used sparsity enforcing regularizer, it suffers from estimation bias (Candes et al., 2008). Recently, neural networks have been deployed for solving the reflectivity inversion problem. For example, the feedforward network by Kim and Nakata (2018) outperformed conventional techniques in support recovery but showed suboptimal amplitude recovery. We develop data-driven nonuniform sparse regularization based on a composite prior constructed from a convex combination of weighted convex and nonconvex penalties. Learning accurate priors from seismic data instead of using a fixed one allows one to estimate the sparse reflectivity accurately. We develop a model-based prior learning network called NuSPAN within the paradigm of deep-unfolding (Gregor and LeCun, 2010). NuSPAN combines the advantages of iterative and data-driven techniques. Deep unfolding gives rise to interpretable architectures, unlike ad hoc networks. We demonstrate the efficacy of NuSPAN for amplitude and support recovery considering both synthetic and simulated data (Marmousi2 model) and show that the accuracy is higher than the state-of-the-art techniques. For the Marmousi2 model, NuSPAN results in 600x faster inference than FISTA, which is the next best technique (Beck and Teboulle, 2009). Such a speedup is an attractive feature when handling large amounts of data.

   @conference{mache2021explainable,
     author = {Mache, Swapnil and Pokala, Praveen Kumar and Rajendran, Kusala and Seelamantula, Chandra Sekhar},
     booktitle = {AGU Fall Meeting 2021},
     title = {Explainable Deep Neural Networks for Seismic Reflectivity Inversion},
     year = {2021},
     month = dec,
     day = {13},
     language = {English},
     note = {AGU Fall Meeting 2021},
   }

2020

1. AGU

   Energy Rate Functions: An Overview of HHT-based Earthquake Source Characterization using Strong Motion Data

   Swapnil Mache, Nishant Chauhan, Avigyan Chatterjee, and 1 more author

   Dec 2020

   AGU Fall Meeting 2020

   Abs Bib HTML Video

   Subduction zone earthquakes show varying energy release patterns and frequency content, based on their tectonic settings and hypocentral depths. Resolving these features from the nonlinear and non-stationary seismograms is a challenge. Our work in the Japan Trench follows studies by Huang et al. (1998, 2001) and Zhang et al. (2003), who demonstrated the use of empirical mode decomposition to separate records into multiple timescales, or intrinsic mode functions (IMFs). Zhang et al. observed that IMFs 2-5 represented the source rupture process for the 1994 Northridge earthquake. Chauhan (master’s thesis, 2019) used time-frequency distributions, short-time Fourier and continuous wavelet transforms, of IMFs of strong-motion data for a pair of interplate-intraslab earthquakes to identify the dominant, short duration, low-frequency energy release for the intraslab event. He found a high correlation between the original signal and a linear combination of IMFs 3 and 4, possibly representing the source. Chatterjee et al. (AGU, 2018) observed an association between time-frequency-energy distributions of certain IMFs and moment rate functions (MRFs) from teleseismic waveform models, for five earthquakes. Chatterjee et al. (AGU, 2019) and Mache et al. (AGU, 2019) used Hilbert spectral analysis (Huang et al., 1998) of IMFs selected based on their frequency and energy and observed better match between the two. This new function, which they regard as the Energy Rate Function (ERF), can reproduce the MRF’s essential elements, i.e., its duration and shape, but Mache (master’s thesis, 2020) observed that results depended on the selection of stations. As the next step, Mache and Rajendran (JpGU-AGU, 2020) based the selection criteria on the slip distribution, strike, and JMA intensity distribution maps (JMA 1996) and applied the method to 7 earthquakes from various tectonic settings of the Japan Trench. Here we present an overview of the various methods for analyzing KiK-net strong-motion data for selected earthquakes to extract information on their time-frequency-energy distributions. The ERF generated through this analysis is a physically compatible expression of the MRF and, therefore, more useful in predicting the shaking effects of earthquakes.

   @conference{mache2020energyb,
     author = {Mache, Swapnil and Chauhan, Nishant and Chatterjee, Avigyan and Rajendran, Kusala},
     booktitle = {AGU Fall Meeting 2020},
     title = {Energy Rate Functions: An Overview of {HHT}-based Earthquake Source Characterization using Strong Motion Data},
     year = {2020},
     month = dec,
     day = {1},
     language = {English},
     note = {AGU Fall Meeting 2020},
   }

2. Energy release patterns and shaking effects of earthquakes in the Japan Trench: A Hilbert-Huang Transform approach

   Swapnil Mache, and Kusala Rajendran

   Jul 2020

   JpGU-AGU Joint Meeting 2020

   Abs Bib HTML Video

   Subduction zones showcase the multiplicity of earthquakes—interplate, intraplate and intraslab—with shallow, intermediate, or deep focus, associated with different energy release patterns and frequency contents. An understanding of the duration and frequencies associated with various pulses of energy is useful for damage assessment. Empirical Mode Decomposition (EMD) of strong-motion records and the application of Hilbert transform have been suggested to overcome the limitations of the Fourier spectral analysis in dealing with highly non-linear strong-motion records (Huang et al., 1998, Zhang et al., 2003). Following the same approach, we have been trying various methods of analysis using the KiK-net strong-motion records to explore the efficacy of these techniques in representing the source of the rupture, in terms of energy release and frequency distribution. Our previous studies used EMD and time-frequency analysis tools such as spectrogram, scalogram, and Hilbert spectrum, using Intrinsic Mode Functions (IMFs) of the original signals as inputs. Nishant (2019) made random picks of IMFs to represent sources by correlating the sum of the selected IMFs with the original signal but found that the results were station dependent. We selected IMFs based on their frequency content (0.1 to 3 Hz) and used their linear combinations to develop the Energy Release Functions (ERF) for individual earthquakes (Mache et al., 2019). They reported that the ability to capture the signature of the original signal using the IMFs varied between earthquakes and stations. Next, we selected stations based on the direction of rupture inferred from teleseismic waveform models. The use of appropriate combinations of individual IMFs, chosen based on the direction of slip, resulted in ERFs whose shapes compared better with the Moment Rate Functions (MRFs) obtained from the teleseismic models. To further explore the station dependence on the resolution of ERFs viz-a-viz the MRFs, we used the instrumental seismic intensity distribution maps (JMA 1996, Shabestari and Yamazaki 2001) to select the stations. We analyzed five earthquakes; two interplate (Mw 7.2 2005 Miyagi, and Mw 6.9 2008/07/19), two intraplate (Mw 7.0 2003 Sendai, and Mw 7.2 2012 Kamaishi) and one intraslab (Mw 7.1 2011 Miyagi), following the above methodologies. This abstract presents the initial results of our study, which to our knowledge, is the first of its kind and holds significant potential in understanding the spatial and temporal patterns of energy release and their associated frequencies. On the use of IMFs based on their frequencies, we find that a linear combination of appropriate signals can lead to ERFs that compare well with their respective MRFs. The selection of stations in the direction of rupture generates better-resolved spectra. While using the seismic intensities, we find that for values three and higher, stations located along the direction of rupture propagation produce ERFs that correlate better with their respective MRFs, as observed for the 2011 Miyagi earthquake. The use of stations located along the direction of the trench also shows a good correlation. For seismic intensities lower than 3, there is a decay in the energy release and hence a poor reproduction of the ERFs. For complex ruptures (2003 Sendai, 2005 Miyagi, and 2012 Kamaishi), the ERFs are not smooth, with their energy distributed in bands of varying frequencies. It could be due to changes in slip direction or generation of sub-events, but the fact that the shapes of both the MRF and ERF are comparable adds credence to our analysis. We find that the local geology also plays an essential role in limiting the energy distribution within a frequency range, an issue that needs further exploration using more examples.

   @conference{mache2020energya,
     title = {Energy release patterns and shaking effects of earthquakes in the Japan Trench: A Hilbert-Huang Transform approach},
     author = {Mache, Swapnil and Rajendran, Kusala},
     year = {2020},
     month = jul,
     day = {12},
     note = {JpGU-AGU Joint Meeting 2020},
     abbv = {JpGU-AGU}
   }

2019

1. AGU

   An Alternate Method for Earthquake Source Characterization through Empirical Mode Decomposition and Spectral Analysis of Strong-Motion Records

   Swapnil Mache, Avigyan Chatterjee, and Kusala Rajendran

   Dec 2019

   AGU Fall Meeting 2019

   Abs Bib HTML Poster

   The occurrence of a large number of earthquakes in the inter and intraplate settings of the Japan Trench leads to ruptures with varying frequencies. To capture the temporal distribution of energy and their ranges of frequencies, we have used the Intrinsic Mode Functions (IMFs) derived from the vertical components of the strong-motion records. Here we present an “energy release function”, which is yet another way of representing frequency-dependent energy release. Without the assumptions of the area of slip and elastic moduli, this provides a new representation of the energy released at the source. Choice of the appropriate IMF and thus the range of frequencies representing the source was based on the best fitting teleseismic model for the same earthquake. We analysed the strong-motion records for three earthquakes (all in the magnitude range of 7.1 to 7.3), representing interplate, intraplate, and intraslab settings and used borehole data from the KiK-net. These were the Miyagi 2005 (Interplate), Tohoku 2011 (Intraslab), and Honshu 2012 (intraplate). We used the Hilbert-Huang Transform, a combination of Empirical Mode Decomposition (EMD) and Hilbert Transform (HT) to develop the spectra for vertical components of each of these earthquakes. A combination of the IMFs within the frequency band (0.1 to 3 Hz) that mostly represent the frequency range used for teleseismic source inversion (0.01 to 2 Hz) was used to develop the spectra in each case. The shape of the spectra generally mimics that of the moment rate function. Where the moment-rate function follows a single pulse, the spectrum is able to generate its shape, and the sub-events are represented through independent pulses of energy. We believe that the representation of an earthquake source based on its frequency content and temporal pattern has important applications in predicting the shaking effects of an earthquake.

   @conference{mache2019alternate,
     author = {Mache, Swapnil and Chatterjee, Avigyan and Rajendran, Kusala},
     title = {An Alternate Method for Earthquake Source Characterization through Empirical Mode Decomposition and Spectral Analysis of Strong-Motion Records},
     year = {2019},
     month = dec,
     day = {9},
     language = {English},
     note = {AGU Fall Meeting 2019},
   }

2. AGU

   Insights on the source characteristics of Japanese earthquakes from source models developed from teleseismic and strong-motion records

   Avigyan Chatterjee, Swapnil Mache, and Kusala Rajendran

   Dec 2019

   AGU Fall Meeting 2019

   Abs Bib HTML

   Large earthquakes within the Japanese trench are known for their variable levels of regional high frequency ground shaking. Previous studies have demonstrated that inter and intraslab earthquakes exhibit different styles of rupture and radiation of energy. Intraslab earthquakes in particular are known to have high damage potential, disproportionate to their magnitudes. This study was motivated by observations of such variations in the source characteristics of earthquakes within the Japanese subduction zone. We used teleseismic data and strong-motion records in tandem, to understand the temporal pattern of energy release and frequency content of the maximum energy. We used 8 earthquakes (Mw> 6.5) representing different tectonic settings of the Japan Trench and developed solutions using two techniques. We used Kikuchi-Kanamori’s program that inverts teleseismic body waves to determine the mechanism, rupture pattern and slip of the earthquakes. In the next step we used the near-field strong-motion data to understand the nature of moment release and the centroid time. Empirical Mode Decomposition (EMD) of the strong ground motion recordings (0.1-20 Hz) from K-NET and KiK-net enabled the segregation of Intrinsic Mode Functions (IMFs) which closely represent the source and site effects of energy release. The IMFs that replicate similar signatures of frequency content as that of the teleseismic data were then combined and their Hilbert spectrum was treated as equivalent to the moment release at the source. Although there are notable differences between the solutions derived from the near and far field data, the attributes of the sources remain unaltered. While the teleseismic model gives fault parameters, the Hilbert-transform based solution provides a way of looking at the moment rate through its frequency dependent energy distribution, which is useful in assessing the damage potential of earthquakes.

   @conference{chatterjee2019insights,
     author = {Chatterjee, Avigyan and Mache, Swapnil and Rajendran, Kusala},
     title = {Insights on the source characteristics of {J}apanese earthquakes from source models developed from teleseismic and strong-motion records},
     year = {2019},
     month = dec,
     day = {9},
     language = {English},
     note = {AGU Fall Meeting 2019},
   }