arXiv subjects

Qingkai Kong

Publications and source records attributed to Qingkai Kong.

17 recordsLinked to original sources

arXiv2025

Evaluating Multi-station Phase Picking Algorithm Phase Neural Operator (PhaseNO) on Local Seismic Networks

Reliable automatic phase picking is important for many seismic applications. With the development of machine learning approaches, many algorithms are proposed, evaluated and applied to different areas. Many of these algorithms are single station based, while recent proposed methods start to combine surrounding stations into consideration in the problem of phase picking. Among these algorithms, the Phase Neural Operator (PhaseNO) shows promising results on regional datasets comparing to existing algorithms. But there are many use cases for the local seismic networks in our community, therefore in this paper we evaluate the performance of PhaseNO on 4 different local datasets and compared the results to PhaseNet and EQTransformer. We used both individual phase picking metrics as well as association metrics to illustrate the performance of PhaseNO. With manually reviewing the newly detected events, we find the PhaseNO model outperformed the single station-based approaches in the local-scale use cases due to its consideration of coherent signals from multiple stations. We also explored PhaseNO's behaviors when only using one station, as well as gradually increase the number of stations in the seismic network to understand it better. Overall, using the off-the-shelf machine learning based phase pickers, PhaseNO demonstrated its good performance on local-scale seismic networks.

physics.geo-ph↗

arXiv2025

WellPINN: Accurate Well Representation for Transient Fluid Pressure Diffusion in Subsurface Reservoirs with Physics-Informed Neural Networks

Accurate representation of wells is essential for reliable reservoir characterization and simulation of operational scenarios in subsurface flow models. Physics-informed neural networks (PINNs) have recently emerged as a promising method for reservoir modeling, offering seamless integration of monitoring data and governing physical equations. However, existing PINN-based studies face major challenges in capturing fluid pressure near wells, particularly during the early stage after injection begins. To address this, we propose WellPINN, a modeling workflow that combines the outputs of multiple sequentially trained PINN models to accurately represent wells. This workflow iteratively approximates the radius of the equivalent well to match the actual well dimensions by decomposing the domain into stepwise shrinking subdomains with a simultaneously reducing equivalent well radius. Our results demonstrate that sequential training of superimposing networks around the pumping well is the first workflow that focuses on accurate inference of fluid pressure from pumping rates throughout the entire injection period, significantly advancing the potential of PINNs for inverse modeling and operational scenario simulations. All data and code for this paper will be made openly available at https://github.com/linuswalter/WellPINN.

physics.flu-dyn↗

arXiv2025

Real-Time Reconstruction of Ground Motion During Small Magnitude Earthquakes: A Pilot Study

This study presents a pilot investigation into a novel method for reconstructing real-time ground motion during small magnitude earthquakes (M < 4.5), removing the need for computationally expensive source characterization and simulation processes to assess ground shaking. Small magnitude earthquakes, which occur frequently and can be modeled as point sources, provide ideal conditions for evaluating real-time reconstruction methods. Utilizing sparse observation data, the method applies the Gappy Auto-Encoder (Gappy AE) algorithm for efficient field data reconstruction. This is the first study to apply the Gappy AE algorithm to earthquake ground motion reconstruction. Numerical experiments conducted with SW4 simulations demonstrate the method's accuracy and speed across varying seismic scenarios. The reconstruction performance is further validated using real seismic data from the Berkeley area in California, USA, demonstrating the potential for practical application of real-time earthquake data reconstruction using Gappy AE. As a pilot investigation, it lays the groundwork for future applications to larger and more complex seismic events.

physics.geo-ph↗

arXiv2025

Reducing Frequency Bias of Fourier Neural Operators in 3D Seismic Wavefield Simulations Through Multi-Stage Training

The recent development of Neural Operator (NeurOp) learning for solutions to the elastic wave equation shows promising results and provides the basis for fast large-scale simulations for different seismological applications. In this paper, we use the Fourier Neural Operator (FNO) model to directly solve the 3D Helmholtz wave equation for fast seismic ground motion simulations on different frequencies, and show the frequency bias of the FNO model, i.e. it learns the lower frequencies better comparing to the higher frequencies. To reduce the frequency bias, we adopt the multi-stage FNO training, i.e., after training a 1st stage FNO model for estimating the ground motion, we use a second FNO model as the 2nd stage to learn from the residual, which greatly reduced the errors on the higher frequencies. By adopting this multi-stage training, the FNO models show reduced biases on higher frequencies, which enhanced the overall results of the ground motion simulations. Thus the multi-stage training FNO improves the accuracy and realism of the ground motion simulations.

physics.geo-ph↗

arXiv2024

A Staged Deep Learning Approach to Spatial Refinement in 3D Temporal Atmospheric Transport

High-resolution spatiotemporal simulations effectively capture the complexities of atmospheric plume dispersion in complex terrain. However, their high computational cost makes them impractical for applications requiring rapid responses or iterative processes, such as optimization, uncertainty quantification, or inverse modeling. To address this challenge, this work introduces the Dual-Stage Temporal Three-dimensional UNet Super-resolution (DST3D-UNet-SR) model, a highly efficient deep learning model for plume dispersion prediction. DST3D-UNet-SR is composed of two sequential modules: the temporal module (TM), which predicts the transient evolution of a plume in complex terrain from low-resolution temporal data, and the spatial refinement module (SRM), which subsequently enhances the spatial resolution of the TM predictions. We train DST3DUNet- SR using a comprehensive dataset derived from high-resolution large eddy simulations (LES) of plume transport. We propose the DST3D-UNet-SR model to significantly accelerate LES simulations of three-dimensional plume dispersion by three orders of magnitude. Additionally, the model demonstrates the ability to dynamically adapt to evolving conditions through the incorporation of new observational data, substantially improving prediction accuracy in high-concentration regions near the source. Keywords: Atmospheric sciences, Geosciences, Plume transport,3D temporal sequences, Artificial intelligence, CNN, LSTM, Autoencoder, Autoregressive model, U-Net, Super-resolution, Spatial Refinement.

cs.LG↗

arXiv2024

A Comprehensive Analysis of Real-World Accelerometer Data Quality in a Global Smartphone-based Seismic Network

The proliferation of low-cost sensors in smartphones has facilitated numerous applications; however, large-scale deployments often encounter performance issues. Sensing heterogeneity, which refers to varying data quality due to factors such as device differences and user behaviors, presents a significant challenge. In this research, we perform an extensive analysis of 3-axis accelerometer data from the MyShake system, a global seismic network utilizing smartphones. We systematically evaluate the quality of approximately 22 million 3-axis acceleration waveforms from over 81 thousand smartphone devices worldwide, using metrics that represent sampling rate and noise level. We explore a broad range of factors influencing accelerometer data quality, including smartphone and accelerometer manufacturers, phone specifications (release year, RAM, battery), geolocation, and time. Our findings indicate that multiple factors affect data quality, with accelerometer model and smartphone specifications being the most critical. Additionally, we examine the influence of data quality on earthquake parameter estimation and show that removing low-quality accelerometer data enhances the accuracy of earthquake magnitude estimation.

physics.geo-ph↗

arXiv2024

Evaluating Physics Informed Neural Network Performance for Seismic Discrimination Between Earthquakes and Explosions

Combining physics with machine learning models has advanced the performance of machine learning models in many different applications. In this paper, we evaluate adding a weak physics constraint, i.e., a physics-based empirical relationship, to the loss function (the Physics Informed manner) in local distance explosion discrimination in the hope of improving the generalization capability of the machine learning model. We compare the proposed model to the two-branch model we previously developed, as well as to a pure data-driven model. Unexpectedly, the proposed model did not consistently outperform the pure data-driven model. By varying the level of inconsistency in the training data, we find this approach is modulated by the strength of the physics relationship. This result has important implications for how to best incorporate physical constraints in Machine Learning models.

physics.geo-ph↗

arXiv2024

Multi-fidelity Fourier Neural Operator for Fast Modeling of Large-Scale Geological Carbon Storage

Deep learning-based surrogate models have been widely applied in geological carbon storage (GCS) problems to accelerate the prediction of reservoir pressure and CO2 plume migration. Large amounts of data from physics-based numerical simulators are required to train a model to accurately predict the complex physical behaviors associated with this process. In practice, the available training data are always limited in large-scale 3D problems due to the high computational cost. Therefore, we propose to use a multi-fidelity Fourier neural operator (FNO) to solve large-scale GCS problems with more affordable multi-fidelity training datasets. FNO has a desirable grid-invariant property, which simplifies the transfer learning procedure between datasets with different discretization. We first test the model efficacy on a GCS reservoir model being discretized into 110k grid cells. The multi-fidelity model can predict with accuracy comparable to a high-fidelity model trained with the same amount of high-fidelity data with 81% less data generation costs. We further test the generalizability of the multi-fidelity model on a same reservoir model with a finer discretization of 1 million grid cells. This case was made more challenging by employing high-fidelity and low-fidelity datasets generated by different geostatistical models and reservoir simulators. We observe that the multi-fidelity FNO model can predict pressure fields with reasonable accuracy even when the high-fidelity data are extremely limited. The findings of this study can help for better understanding of the transferability of multi-fidelity deep learning surrogate models.

stat.ML↗

arXiv2023

Predicting Wind-Driven Spatial Deposition through Simulated Color Images using Deep Autoencoders

For centuries, scientists have observed nature to understand the laws that govern the physical world. The traditional process of turning observations into physical understanding is slow. Imperfect models are constructed and tested to explain relationships in data. Powerful new algorithms can enable computers to learn physics by observing images and videos. Inspired by this idea, instead of training machine learning models using physical quantities, we used images, that is, pixel information. For this work, and as a proof of concept, the physics of interest are wind-driven spatial patterns. These phenomena include features in Aeolian dunes and volcanic ash deposition, wildfire smoke, and air pollution plumes. We use computer model simulations of spatial deposition patterns to approximate images from a hypothetical imaging device whose outputs are red, green, and blue (RGB) color images with channel values ranging from 0 to 255. In this paper, we explore deep convolutional neural network-based autoencoders to exploit relationships in wind-driven spatial patterns, which commonly occur in geosciences, and reduce their dimensionality. Reducing the data dimension size with an encoder enables training deep, fully connected neural network models linking geographic and meteorological scalar input quantities to the encoded space. Once this is achieved, full spatial patterns are reconstructed using the decoder. We demonstrate this approach on images of spatial deposition from a pollution source, where the encoder compresses the dimensionality to 0.02% of the original size, and the full predictive model performance on test data achieves a normalized root mean squared error of 8%, a figure of merit in space of 94% and a precision-recall area under the curve of 0.93.

cs.LG↗

arXiv2023

Crowdsourcing Felt Reports using the MyShake smartphone app

MyShake is a free citizen science smartphone app that provides a range of features related to earthquakes. Features available globally include rapid post-earthquake notifications, live maps of earthquake damage as reported by MyShake users, safety tips and various educational features. The app also uses the accelerometer to detect earthquake shaking and to record and submit waveforms to a central archive. In addition, MyShake delivers earthquake early warning alerts in California, Oregon and Washington. In this study we compare the felt shaking reports provided by MyShake users in California with the US Geological Survey's "Did You Feel It?" intensity reports. The MyShake app simply asks "What strength of shaking did you feel" and users report on a five-level scale. When the reports are averaged in spatial bins, we find strong correlations with the Modified Mercalli Intensity scale values reported by the USGS based on the much more complex DYFI surveys. The MyShake felt reports can therefore also be used to generate shaking intensity maps.

physics.geo-ph↗

arXiv2022

Combining Deep Learning with Physics Based Features in Explosion-Earthquake Discrimination

This paper combines the power of deep-learning with the generalizability of physics-based features, to present an advanced method for seismic discrimination between earthquakes and explosions. The proposed method contains two branches: a deep learning branch operating directly on seismic waveforms or spectrograms, and a second branch operating on physics-based parametric features. These features are high-frequency P/S amplitude ratios and the difference between local magnitude (ML) and coda duration magnitude (MC). The combination achieves better generalization performance when applied to new regions than models that are developed solely with deep learning. We also examined which parts of the waveform data dominate deep learning decisions (i.e., via Grad-CAM). Such visualization provides a window into the black-box nature of the machine-learning models and offers new insight into how the deep learning derived models use data to make the decisions.

physics.geo-ph↗

arXiv2021

Deep Convolutional Autoencoders as Generic Feature Extractors in Seismological Applications

The idea of using a deep autoencoder to encode seismic waveform features and then use them in different seismological applications is appealing. In this paper, we designed tests to evaluate this idea of using autoencoders as feature extractors for different seismological applications, such as event discrimination (i.e., earthquake vs. noise waveforms, earthquake vs. explosion waveforms, and phase picking). These tests involve training an autoencoder, either undercomplete or overcomplete, on a large amount of earthquake waveforms, and then using the trained encoder as a feature extractor with subsequent application layers (either a fully connected layer, or a convolutional layer plus a fully connected layer) to make the decision. By comparing the performance of these newly designed models against the baseline models trained from scratch, we conclude that the autoencoder feature extractor approach may only perform well under certain conditions such as when the target problems require features to be similar to the autoencoder encoded features, when a relatively small amount of training data is available, and when certain model structures and training strategies are utilized. The model structure that works best in all these tests is an overcomplete autoencoder with a convolutional layer and a fully connected layer to make the estimation.

physics.geo-ph↗

arXiv2021

Detecting Damage Building Using Real-time Crowdsourced Images and Transfer Learning

After significant earthquakes, we can see images posted on social media platforms by individuals and media agencies owing to the mass usage of smartphones these days. These images can be utilized to provide information about the shaking damage in the earthquake region both to the public and research community, and potentially to guide rescue work. This paper presents an automated way to extract the damaged building images after earthquakes from social media platforms such as Twitter and thus identify the particular user posts containing such images. Using transfer learning and ~6500 manually labelled images, we trained a deep learning model to recognize images with damaged buildings in the scene. The trained model achieved good performance when tested on newly acquired images of earthquakes at different locations and ran in near real-time on Twitter feed after the 2020 M7.0 earthquake in Turkey. Furthermore, to better understand how the model makes decisions, we also implemented the Grad-CAM method to visualize the important locations on the images that facilitate the decision.

cs.CV↗

arXiv2019

Towards Global Earthquake Early Warning with the MyShake Smartphone Seismic Network Part 1 -- Detection algorithm and simulation platform

The MyShake project aims to build a global smartphone seismic network to facilitate large-scale earthquake early warning and other applications by leveraging the power of crowdsourcing. The MyShake mobile application first detects earthquake shaking on a single phone. The earthquake is then confirmed on the MyShake servers using a "network detection" algorithm that is activated by multiple single-phone detections. In this part one of the two paper series, we present a network detection algorithm and a simulation platform to test earthquake scenarios at various locations around the world. The proposed network detection algorithm is built on the DBSCAN classic spatial clustering algorithm, with modifications to take temporal characteristics into account and the association of new triggers. We test our network detection algorithm using real data recorded by MyShake users during the M4.4 January 4th, 2018, Berkeley and the M5.2 June 10th, 2016, Borrego Springs earthquakes to demonstrate the system's utility. In order to test the entire detection procedure and to understand the first order performance of MyShake in various locations around the world representing different population and tectonic characteristics, we then present a software platform which can simulate earthquake triggers in hypothetical MyShake networks. Part two of this paper series explores our MyShake early warning simulation performance in selected regions around the world.

physics.geo-ph↗

arXiv2019

Towards Global Earthquake Early Warning with the MyShake Smartphone Seismic Network Part 2 -- Understanding MyShake performance around the world

The MyShake project aims to build a global smartphone seismic network to facilitate large-scale earthquake early warning and other applications by leveraging the power of crowdsourcing. The MyShake mobile application first detects earthquake shaking on a single phone. The earthquake is then confirmed on the MyShake servers using a "network detection" algorithm that is activated by multiple single-phone detections. In part two of this two paper series, we report the first order performance of MyShake's Earthquake Early Warning (EEW) capability in various selected locations around the world. Due to the present sparseness of the MyShake network in most parts of the world, we use our simulation platform to understand and evaluate the system's performance in various tectonic settings. We assume that 0.1% of the population has the MyShake mobile application installed on their smartphone, and use historical earthquakes from the last 20 years to simulate triggering scenarios with different network configurations in various regions. Then, we run the detection algorithm with these simulated triggers to understand the performance of the system. The system performs best in regions featuring high population densities and onshore, upper crustal earthquakes M<7.0. In these cases, alerts can be generated ~4-6 sec after the origin time, magnitude errors are within ~0.5 magnitude units, and epicenters are typically within 10 km of true locations. When the events are offshore or in sparsely populated regions, the alerts are slower and the uncertainties in magnitude and location increase. Furthermore, even with 0.01% of the population as the MyShake users, in regions of high population density, the system still performs well for earthquakes larger than M5.5. For details of the simulation platform and the network detection algorithm, please see part one of this two paper series.

physics.geo-ph↗

arXiv2019

MyShake: Detecting and characterizing earthquakes with a global smartphone seismic network

MyShake harnesses private/personal smartphones to build a global seismic network. It uses the accelerometers embedded in all smartphones to record ground motions induced by earthquakes, returning recorded waveforms to a central repository for analysis and research. A demonstration of the power of citizen science, MyShake expanded to 6 continents within days of being launched, and has recorded 757 earthquakes in the first 2 years of operation. The data recorded by MyShake phones has the potential to be used in scientific applications, thereby complementing current seismic networks. In this paper: (1) we report the capabilities of smartphone sensors to detect earthquakes by analyzing the earthquake waveforms collected by MyShake. (2) We determine the maximum epicentral distance at which MyShake phones can detect earthquakes as a function of magnitude. (3) We then determine the capabilities of the MyShake network to estimate the location, origin time, depth and magnitude of earthquakes. In the case of earthquakes for which MyShake has provided 4 or more phases (21 events), either P- or S-wave signals, and has an azimuthal gap less than 180 degrees, the median location, origin time and depth errors are 2.7 km, 0.2 s, and 0.1 km respectively relative to USGS global catalog locations. Magnitudes are also estimated and have a mean error of 0.0 and standard deviation 0.2. These preliminary results suggest that MyShake could provide basic earthquake catalog information in regions that currently have no traditional networks. With an expanding MyShake network, we expect the event detection capabilities to improve and provide useful data on seismicity and hazards.

physics.geo-ph↗

arXiv2019

Earthquake Early Warning and Beyond: Systems Challenges in Smartphone-based Seismic Network

Earthquake Early Warning (EEW) systems can effectively reduce fatalities, injuries, and damages caused by earthquakes. Current EEW systems are mostly based on traditional seismic and geodetic networks, and exist only in a few countries due to the high cost of installing and maintaining such systems. The MyShake system takes a different approach and turns people's smartphones into portable seismic sensors to detect earthquake-like motions. However, to issue EEW messages with high accuracy and low latency in the real world, we need to address a number of challenges related to mobile computing. In this paper, we first summarize our experience building and deploying the MyShake system, then focus on two key challenges for smartphone-based EEW (sensing heterogeneity and user/system dynamics) and some preliminary exploration. We also discuss other challenges and new research directions associated with smartphone-based seismic network.

eess.SY↗