3. Lake Charles, LA¶

Acknowledgements

Several members of the SimCenter Team assisted with the development and documentation of this testbed: Barbaros Cetiner, Adam Zsarnoczay, Kuanshi Zhong, Chaofeng Wang, Sascha Hornauer, and Wael Elhaddad. The conceptualization of the testbed was guided by Tracy Kijewski-Correa (University of Notre Dame), with implementation supervised by Frank McKenna (University of California, Berkeley). Additional guidance was provided by Greg Deierlein (Stanford University), Andrew Kennedy (University of Notre Dame), and Matt Schoettler (University of California, Berkeley).

The Hazard Characterization utilized models, software, and simulations executed by the following groups, whose collaboration is greatly appreciated:

The wind fields were generated by Applied Research Associates (ARA) under the guidance of Dr. Peter Vickery and made available to the community through a partnership with the National Institute for Standards and Technology (NIST) under the guidance of Dr. Marc Levitan, with long-term curation at DesignSafe-CI.

Validation benefited from special access to ZTRAX data provided by the University of California, Los Angeles (UCLA) and the group of Professor Ertugrul Taciroglu.

Preface

This documentation is intended to introduce the implementation of the SimCenter hurricane regional loss modeling workflow in the context of Lake Charles, Louisiana, during Hurricane Laura. While certain aspects of the workflow remain unchanged in a given application context, this testbed specifically demonstrates how inventory generation capacities and existing heuristic rulesets, grounded in codes/standards and normative construction practices in coastal communities, can be engaged to rapidly generate these loss estimations following a major hurricane. Users are encouraged to read the documentation from start to finish, beginning from Overview, to deepen their understanding of the implementation and the underlying assumptions/limitations. Those already familiar with the concept of hurricane loss estimation can proceed directly to Getting Started for step-by-step instructions on how to execute the testbed and interact with its results.

Versioning

Since this documentation will continue to evolve as features are added and software capabilities improve, the documentation, associated input files, results, and rulesets will be updated over time. The applications are hosted in the NHERI-SimCenter DesignSafe, and the versioned source codes are archived via Zenodo. Other related files are maintained in two types of repositories:

The DesignSafe Project PRJ-3207 contains inputs (inventories, rulesets, and hazard data), outputs (results), and PDFs that provide more details about the inventory data model as well as the basis for the rulesets used to assign various attributes. This will enable ongoing refinement and extension of these products by the SimCenter and its users over time.
GitHub is used as the home for all software, scripts, and applications engaged by the testbed, including Python implementations of the rulesets. A dedicated GitHub repository has been created for the rulesets used in this testbed. See Table 3.2.

Both types of repositories are version-controlled. The version history of the documentation is tracked in Table 3.2, and the versions of all applications and rulesets of the current version of the testbed are reported in Table 3.2. Table 3.3 provides the versions of BRAILS and its various classifiers used in the generation of results herein. Note: When accessing the project in DesignSafe, please ensure that the version drop-down is set to the latest available version.

Table 3.1 Testbed Version History¶
Description	Version	Release Date
Initial release	v1.1	11/2021

The software descriptions and results herein were generated by executing the R2D application using the versions of this application and associated SimCenter software as listed in Table 3.2 and Table 3.3. With the software versions listed, the testbed is currently estimated to execute in 15 minutes using five Skylake (SKX) nodes (96 cores) on Frontera.

Table 3.2 Software and Supporting Files¶
Files	Version	DOI/Release
R2D	v1.1.2
AssetRepresentationRulesets	v1.0.0
DesignSafe Project PRJ-3207	v4	Data Depot

Table 3.3 Versions of AI application (BRAILS) and models Used in the Current Version of Testbed¶
Model	Version
RoofTypeClassifier	0.2
OccupancyClassifier	0.2
NumberOfFloorsDetector	1.0
FacadeParser	1.1
GarageDetector	1.0

Envisioned Use Cases

Given the significance of building inventory generation for this testbed, this documentation was written to cater to two primary audiences/use cases:

Case 1: End users who wish to use the testbed to explore specific research questions such as:

Observing prevailing patterns in the results to identify specific areas or classes of buildings that were significantly affected.
Exploring the potential benefits of various mitigation efforts (changing selected attribute assignments and/or damage/loss descriptions to reduce the losses observed in this event).

Such individuals may not wish to generate their own inventories but require some background in order to meaningfully interpret results. This documentation will enhance their understanding of the various assumptions made in generating these inventories and assigning the attributes required for the adopted loss models. Please see Getting Started for a detailed description of how to conduct a run of this testbed.

Case 2: Users who wish to use this testbed as a validation event can compare the damage/loss estimates from this workflow to direct observations in the field for the purposes of improving/refining the underlying descriptions of damage and loss. This can include taking advantage of the data published by the Structural Extreme Event Reconnaissance (StEER) Network, as well as FEMA. Those doing so are advised to:

Update the heuristic rulesets used for building and attribute assignment to capture the specific regulatory environment and construction practices in this region. These detailed rulesets were introduced in the testbed of Atlantic County, NJ and are linked as DOIs in DesignSafe; their implementation as Python scripts is available in corresponding links embedded in the documentation.
Expand the provided inventory beyond residential construction to encompass other diverse classes of buildings and infrastructure affected by the hurricane (see Asset Description).
Incorporate coastal hazards into the workflow, borrowing the approach developed in Atlantic County, NJ, and extend the inventory down into Cameron Parish and other areas significantly impacted by the storm surge.

The revised damage and loss models or extensions to the heuristic rulesets can be contributed back to the SimCenter to expand the capabilities of this community software.

If you have any feedback or questions, please check out the github discussion page. This forum can also be browsed for additional updates, answers to Frequently Asked Questions, and suggestions from the SimCenter team and user community.