Back to homepage

Broadly speaking, my research interests can be classified into two subcategories: (1) Earthquake source and (2) the Structure. Both of these go hand-in-hand, where understanding of one is impossible without the knowledge of the other.



3D numerical simulations

Simulation for a M 5.7 earthquake in Nepal with reverberations in Indo-Gangetic basin.	3D velocity model of southern Alaska (modified from Eberhart-Phillips et. al. 2006) with velocity model of Cook Inlet basin (Shellenbaum et. al. 2010) embedded in it.
Wavefield movie for an earthquake in southern Alaska showing effects (amplification) due to (low velocity) sedimentary basin.	Synthetic waveforms filtered between 6-50s computed with (blue) and without (red) embedding the Cook Inlet basin.

      
References:
Silwal, V., C. Tape, and E. Casarotti, Wavefield Simulation of earthquakes in southern Alaska for tomographic inversion, AGU Fall Meeting 11-15 Dec. 2017, New Orleans
Silwal, V., et al., The 2023 MW 5.7 Nepal earthquake: Source mechanism and structural response of Indo-Gangetic basin (in Review)
Silwal, V., 2018, Earthquake source mechanisms and three-dimensional wavefield simulations in Alaska, PhD Thesis 

We prepare a moment tensor catalog of >250 events (2007-2017) that are distributed in Alaska and 40 earthquakes in Himalayas.
Earthquakes in general occur on a fault by shear dislocation, which can be modeled as a double couple moment tensor. A double couple moment tensor is a 3 × 3 symmetric matrix whose eigenvalues are (λ, 0, −λ). We are concerned with estimating the magnitude and orientation (strike, dip, rake) of the moment tensor. Alternative terms for double couple moment tensors are ‘fault-plane solution’ or ‘focal mechanism.’ Our approach to moment tensor estimation can also be applied to ‘full’ moment tensors, which contain an additional two parameters (Alvizuri & Tape (2016)).

We perform moment tensor inversion using the ‘cut-and-paste’ (CAP) method of Zhu & Helmberger (1996). The best moment tensor is obtained by minimizing the difference between the data with the synthetics.  The synthetics are computed for all possible moment tensors using a 1D reference structural model. Different bandpass filters are applied to the body waves and surface waves when comparing the data with the synthetics. 


Cryoseismology
We acquired seismic dataset from Satopanth Glacier (collaborator: Dr. B. Shekar, IITBombay, and Dr. Arjun Dutta, IISER Pune) and deployed equipment in Antarctica with support from NCPOR, Goa. We are developing automated detection and classification approaches for understanding icequakes and ice dynamics.


Event detection using attentive deep learning

We have used an attention-based deep learning model for automatically detecting events, picking seismic phases and estimating magnitude. In this study, we also developed automated selection criteria to filter out the less reliable detection based on the probability of detection and picked phases; this saved lots of computational time when working with sizeable seismic data sets. Our deep learning based model was applied from January 2013 to October 2013, and we were able locate approximately 4.5 times more earthquakes than those in the ISC catalogue within a very short time. Magnitude estimation clearly shows that the attention deep learning model is quite efficient in detecting low magnitude and recurring events that were not previously detected by manual picking or other automated algorithms. Our model enhanced the region’s seismicity and could map several earthquakes along the neotectonic likely active faults of western Pamir, which further supports the western extrusion of Pamir rocks due to the collision of the Pamir Plateau in the Tajik Depression.

Seismotectonics and Fault delineation
Distribution of similar mechanism faulting for earthquakes in a  seismic zone can be used as a strong evidence for the presence of a fault(s). This is specially useful when faults are sub-surface or overlain by a layer of sediment as in a basin environment and there is no surface feature.
These studies are generally supplemented with earthquake relocation, gravity and magnetic measurements, and revisit historical catalogs to get a better perspective on tectonics at regional scale.

(left) Delineated are the proposed faults in Minto Flats fault zone (interior Alaska). Note the similar moment tensor solutions for major earthquakes near the fault zone.
(right) Consistent thrust faulting mechanism for small-to-intermediate magnitude earthquakes around Cook Inlet and Susitna basin, Alaska.

References:
Tape, C., V. Silwal, C. Ji, L. Keyson, M. E. West, and N. Ruppert, 2015, Transtensional Tectonics of the Minto Flats Fault Zone and Nenana Basin, Central Alaska, Bulletin of the Seismological Society of America, Vol. 105, No. 4, pp. 2081–2100, August 2015, doi: 10.1785/0120150055
Silwal, V., C. Tape, and A. Lomax, Crustal earthquakes in the Cook Inlet and Susitna region of southern Alaska (submitted to Tectonophysics).