Research

I am a computational seismologist focused on building scalable data and software systems for geophysical discovery and structural health monitoring.
My work connects physical seismology, geodesy, and modern scientific computing, with emphasis on:

• smartphone-based structural monitoring (MyShake)
• real-time and near-real-time geophysical pipelines
• high-performance inversion and tomography workflows
• robust, reproducible research software for large-array data

Current Appointments

• Assistant Project Scientist, Berkeley Seismological Laboratory, UC Berkeley (September 2025 - Present), with Prof. Richard M. Allen
• Postdoctoral Researcher, UC Berkeley (September 2021 - September 2025), with Prof. Richard M. Allen
• Postdoctoral Researcher, UC Berkeley (September 2021 - August 2023), with Prof. Barbara Romanowicz
• Postdoctoral Researcher, Institute of Earth Sciences, Academia Sinica (December 2020 - September 2021), with Prof. Bor-Shouh Huang

Current Research Projects

MyShake Statewide Structural Health Monitoring (California)

Structural health monitoring (SHM) is essential for ensuring the safety, reliability, and lifecycle cost-effectiveness of buildings and infrastructure. Traditional SHM approaches often require custom sensor deployments and substantial investments in installation, maintenance, and operations.

This project develops a community-sourced SHM framework using the MyShake smartphone app, a free UC Berkeley application supported by Early Warning California and downloaded globally more than 4.5 million times. During significant shaking events, MyShake phones at rest can be triggered to record and transmit waveforms. Because phones are commonly located inside buildings, these records capture structural response that can be used to infer dynamic building behavior.

Using these recordings, we extract structural indicators such as natural frequencies and construct a continuously updated database of building modal properties. The approach is validated through controlled experiments and real-world observations to assess data quality and robustness.

By tracking frequency changes through time, this framework supports rapid detection of potential structural changes after earthquakes and other strong-loading events. The work demonstrates a scalable path toward global, low-cost, automated structural monitoring and also addresses current scientific and engineering constraints for operational deployment at very large scale.

Automated Building Seismic Monitoring and Event Processing (Taiwan)

This work introduces the Quake Structural Integrity System (QSIS), a scalable and cost-effective platform for in-situ seismic and structural monitoring of buildings. QSIS integrates low-cost MEMS accelerometers with cloud-native data workflows to support continuous monitoring and rapid structural assessment.

The system architecture includes local data-capture clients, centralized aggregation and preprocessing services, and web-based interfaces for device management and visualization. A device-level time-synchronization protocol is used to mitigate microcontroller clock drift and preserve timing accuracy for seismic analysis.

QSIS is designed for real-time transmission, efficient archiving, and automated processing, while prioritizing data privacy and security for deployment in occupied and critical facilities. Its modular framework supports multiple sensor types and different installation scales.

The platform enables several automated applications: ambient-vibration tracking of building natural frequencies, array-based dynamic response analysis, and post-earthquake impact assessment. Early deployments across diverse structural environments show that QSIS can deliver actionable seismic insights for long-term monitoring, risk assessment, and resilient urban planning.

Yellowstone Mid-Mantle Seismic Imaging

The origin and magmatic nature of the Yellowstone hotspot remain actively debated. Previous tomography studies suggested a narrow plume-like low-velocity conduit from deep mantle levels, but mid-mantle resolution has been limited.

This project develops a high-resolution 3D elastic model beneath Yellowstone using a box tomography framework. We couple a global 3D solver (SPECFEM3D GLOBE) for wavefield and Green’s-function computation outside the model box with a regional 3D solver (RegSEM) inside the box. This strategy enables efficient use of teleseismic body-wave information needed to illuminate mid- and lower-mantle structure.

The starting reference model is the global radially anisotropic SEMUCB_WM1 model. We further constrain shallow structure with earthquake and ambient-noise surface-wave dispersion data down to a 16 s period. Global wavefields are computed once in the reference model (to 20 s), recorded at box boundaries, and then used in iterative in-box inversions while progressively shortening the cutoff period from 40 s to 20 s.

Most model updates use a Gauss-Newton strategy with a physics-based Hessian from asymptotic normal-mode coupling theory (NACT), followed by final refinement iterations using adjoint-based inversion. This workflow improves imaging of Yellowstone’s mid-mantle plume geometry and continuity to depths near 2200 km.

Selected Publications

1. Kumar, U., Marcou, S., & Allen, R. M. (2025). Ambient vibration analysis of high-rise buildings using MyShake smartphone data. Journal of Building Engineering, 106, 112496. https://doi.org/10.1016/j.jobe.2025.112496
2. Patel, S. C., Gunay, S., Marcou, S., Gou, Y., Kumar, U., & Allen, R. M. (2023). Toward Structural Health Monitoring with the MyShake Smartphone Network. Sensors, 23(21), 8668. https://doi.org/10.3390/s23218668
3. Kumar, U., Legendre, C. P., Zhao, L., & Chao, B. F. (2022). Dynamic Time Warping as an Alternative to Windowed Cross Correlation in Seismological Applications. Seismological Research Letters. https://doi.org/10.1785/0220210288
4. Kumar, U., Legendre, C. P., Lee, J.-C., Zhao, L., & Chao, B. F. (2022). On analyzing GNSS displacement field variability of Taiwan: Hierarchical Agglomerative Clustering based on Dynamic Time Warping technique. Computers & Geosciences, 169, 105243. https://doi.org/10.1016/j.cageo.2022.105243
5. Kumar, U., & Legendre, C. P. (2022). Crust-mantle decoupling beneath Afar revealed by Rayleigh-wave tomography. Scientific Reports, 12(1), 17036. https://doi.org/10.1038/s41598-022-20890-5
6. Kumar, U., Legendre, C. P., & Huang, B. S. (2021). Crustal structure and upper mantle anisotropy of the Afar triple junction. Earth, Planets and Space, 73(1), 166. https://doi.org/10.1186/s40623-021-01495-0
7. Kumar, U., & Legendre, C. P. (2021). STADIUM-Py: Python command-line interface for automated receiver functions and shear-wave splitting measurements. Zenodo. https://doi.org/10.5281/zenodo.4686103
8. Kumar, U., Chao, B. F., & Chang, E. T.-Y. (2020). What Causes the Common-Mode Error in Array GPS Displacement Fields: Case Study for Taiwan in Relation to Atmospheric Mass Loading. Earth and Space Science. https://doi.org/10.1029/2020EA001159
9. Kumar, U., Chao, B. F., Hsieh, Y., & Chang, E. T. Y. (2017). A meteor shockwave event recorded at seismic and infrasound stations in northern Taiwan. Geoscience Letters. https://doi.org/10.1186/s40562-017-0079

Technology Transfer and Invention Disclosures

• Real-Time Earthquake and Building Structural Health Monitoring System - Cloud-Based Automated Seismic Data Processing and Structural Health Analysis
  Invention disclosure filed through the Academia Sinica Department of Intellectual Property and Technology Transfer.
  Covers the QSIS (Quake Structural Integrity System), a cloud-based system for real-time acquisition, automated processing, and analysis of in-building seismic and acceleration data, integrating structural health monitoring with earthquake detection, machine-learning-based event and phase analysis, H/V spectral monitoring, and tracking of building modal frequency evolution for rapid structural safety assessment.
  Academia Sinica disclosure: 04T-1140410, know-how

• System and Architecture for Building-Embedded Seismic Monitoring
  Invention disclosure filed through the Academia Sinica Department of Intellectual Property and Technology Transfer.
  Covers a modular client-server and edge-cloud system architecture for continuous building-embedded seismic acquisition, real-time data transmission, automated event detection, scalable cloud processing, and integration of heterogeneous sensor networks for structural and seismic monitoring in occupied buildings.
  Academia Sinica disclosure: 04T-1140411, know-how

• Smartphone Ambient-Vibration Structural Health Monitoring
  Software and methods disclosure filed with the University of California Office of Technology Licensing, covering the use of ambient smartphone accelerometer recordings to infer building dynamic properties, including fundamental frequency estimation and temporal tracking, enabling scalable structural health monitoring using distributed consumer devices.
  UC disclosure: NCD 34035

Open Software and Data Systems

• SeismoAlert: real-time earthquake monitoring toolkit with CI/CD, automated tests, semantic versioning, and documentation
• STADIUM-Py: automated receiver function and shear-wave splitting workflow
• QSIS: low-cost MEMS-based structural and seismic monitoring platform with cloud archiving and real-time processing www.qsisglobal.com

Recent Conference Activity

• SSA 2026: Session Organizer and Co-Convener, Earthquake Ground Motions and Structural Response: Emerging Tools and Applications
• SSA 2025: Oral and poster presentations on MyShake structural monitoring and QSIS building dynamics
• AGU 2024: Poster on Yellowstone mid-mantle imaging with full-waveform inversion

Awards and Recognition (Selected)

• NSF I-Corps Hub Northwest (2025), technical lead for the MyShake structural health monitoring team
• IMV Innovation Market Value Competition, Taiwan (2024), First Prize (Team QSIS)
• NASA Space Apps Challenge, Taipei (2018), Third Place and IBM Taiwan Best Innovation Award
• AOGS Best Poster Award, Solid Earth Section (2018)

Last updated: February 2026