GEER - Marshall Fire Reconnaissance 2022 (PRJ-3379)

Project Overview

NSF NHERI / GEER Post-Disaster Reconnaissance of the December 2021 Marshall Fire in Boulder County, Colorado (PRJ-3379). This comprehensive dataset documents multi-hazard structural damage, progressive debris clearing, and infrastructure recovery across diverse built environments.

Geographical Collections (5 Target Sectors)

  1. Town of Marshall: Multi-temporal residential damage, commercial structures, and local power facilities.
  2. RT-36 Commercial Corridor: Detailed documentation of commercial destructions (e.g., Element Hotel, Cherokee Restaurant).
  3. Arapahoe & Mulberry Neighborhoods: Multi-epoch residential devastation and spatial clearing.
  4. Spanish Hills: High-end residential structural impacts.
  5. Deployment Documents: Ground Control Points (GCPs), flight logs, and camera calibration parameters.

Multi-Modal Sensor Suites

  • Terrestrial Laser Scanning (TLS): Survey-grade Leica RTC360 and Leica ScanStation P50 millimeter-precision point clouds (ideal for geometric ground truth).
  • UAS Photogrammetry: DJI Mavic 2 Pro high-resolution aerial imagery series.
  • Derived 3D Products: Multi-temporal Structure-from-Motion (SfM) point clouds (.laz) and registered 3D meshes (.ply).

🎯 Our Research Focus & Planned Usage (4D Resilience Project)

Within our research team, this dataset serves as the primary benchmark to develop and validate our multi-temporal 3D computer vision pipeline: 1. Multi-Temporal 3DGS Change Detection: - Utilize the raw UAV imagery sequences (e.g., Town of Marshall) across different temporal epochs to train baseline and recovery 3D Gaussian Splatting (3DGS) models. - Benchmark image-based (Render-to-Real) vs. primitive-based (3D Gaussian attribute comparison) change detection algorithms. 2. Cross-Modal Ground Truth Benchmarking (Responding to Geometric Accuracy): - Leverage the survey-grade TLS point clouds (RT-36 Corridor) as the absolute physical ground truth. - Quantify geometric errors and spatial drift of UAV-derived 3DGS reconstructions against millimeter-level laser scanning. 3. Automated TACC HPC Integration: - Serve as the testbed dataset for establishing automated data fetching, preprocessing, and training workflows via the Tapis API on TACC computing systems (Lonestar6/Frontera).

Data and Resources

  • DesignSafe Project Browser (PRJ-3379)HTML

    Official NHERI / GEER published dataset landing page and web file browser.

    Explore
    • 🔬 DYNAMO Analysis
    ×

    Choose Analysis Type:

  • Town of Marshall - Multi-Temporal SfM Point Cloudslaz

    Contains multi-epoch 3D SfM point clouds: Feb 15, 2022 baseline (6.7 GB) and...

    Explore
    • 🔬 DYNAMO Analysis
    ×

    Choose Analysis Type:

  • RT-36 Commercial Corridor - Millimeter-Level...laz

    Survey-grade Terrestrial Laser Scanning (Leica RTC360 & P50) of commercial...

    Explore
    • 🔬 DYNAMO Analysis
    ×

    Choose Analysis Type:

  • Raw UAV Multi-View Imagery (DJI Mavic 2 Pro)JPEG

    Raw high-resolution aerial photographs across multiple flight dates and...

    Explore
    • 🔬 DYNAMO Analysis
    ×

    Choose Analysis Type:

  • Marshall Town Baseline Point Cloud (2022-02-15)laz

    link to the actual file

    Explore
    • 🔬 DYNAMO Analysis
    ×

    Choose Analysis Type:

Additional Info

Field Value
Source https://www.designsafe-ci.org/data/browser/public/designsafe.storage.published/PRJ-3379
Version 1.0
Author Brad Wham (CU Boulder), Erica Fischer (Oregon State), Shideh Dashti (CU Boulder)
Author Email Brad Wham (CU Boulder), Erica Fischer (Oregon State), Shideh Dashti (CU Boulder)
Maintainer
Maintainer Email
Temporal Coverage Start
Temporal Coverage End
Spatial Coverage