Estimation of unbiased Empirical Green’s
Function using Deep Neural Networks from
Seismic Ambient Noise
Implementing Organization
Indian Institute of Science
Principal Investigator
Dr. Priyanshu Gupta
Indian Institute Of Science
priyanshug2007@gmail.com
Project Overview
Imaging the Earth’s interior structure is crucial for solving many practical problems such as carbon
sequestration, metal mining, ground water studies, exploration of mineral oils, and natural gases.
The primary objective of this research is to extract useful information about the Earth’s interior
structure in seismically stable regions where earthquakes do not occur frequently. In such regions,
seismic ambient noise is processed to perform tomography (i.e., imaging of Earth’s interior structure).
Ambient noise is generated mainly by the interaction of ocean swells with the seafloor. Ambient
noise generated in this manner is comprised of primary and secondary microseisms.
Ambient noise tomography (ANT) relies on cross-correlating long duration ambient noise data
recorded at pairs of seismic stations, estimating the empirical Green’s function (EGF). Theoretically,
ambient noise has the potential to obtain complete Green’s functions (i.e. the path effect due to
a point source in ideal case), provided that noise sources are distributed uniformly in both space
and time. Real-world complexities such as anisotropic and non-stationary noise sources undermine
the effectiveness of conventional methods of extracting EGF, therefore in practical seismic studies,
retrieval of seismic waves in EGF is often restricted to surface waves.
Unequal energy distribution of ambient noise can amplify incoherent noise relative to coherent
signal components, causing bias in the estimated EGF. These biases impact the subsequent analyses
like dispersion curve extraction and their inversion, eventually leading to limited resolution and
inaccuracies in imaging the Earth’s subsurface structure. Additionally, retrieving surface waves
across all frequencies is challenging due to highly directional dominant noise sources, and the recovery
of higher-order modes of surface wave propagation is even harder due to multipath fading. High
frequency surface waves are useful for sampling the shallow subsurface, while low-frequency surface
waves are more effective for probing deeper layers. Relying on incomplete observed bandwidth of
surface waves is insufficient for effectively constraining velocity interfaces. Therefore, extraction of
the body wave signals, along with the complete bandwidth of surface waves, is essential for accurately
imaging subsurface structure.
Disclaimer:
Information available on this portal is sourced from various organizations and is provided for informational purposes only. Users are advised to verify details from the respective official sources.
Please enter your details
Please provide your name and email to continue. Your details are saved in this browser for future use.
Latest Updates
Loading…
⚠️
You are leaving this website
You are about to be redirected to an external website that is not operated by
India Science, Technology & Innovation (ISTI) Portal.