Introduction

As part of the National Geospatial Technical Operations Center (NGTOC), The National Map Corps (TNMCorps) is a crowd-sourced mapping project that relies on volunteers to assist the U.S. Geological Survey (USGS) National Geospatial Program by collecting and editing built structures data.

Tracing its initial discussions to a 2020 meeting between the Federal Emergency Management Agency (FEMA) and USGS personnel, the St. James Parish pilot project took place from February through March of 2024. The project focused on exploring opportunities to utilize TNMCorps platform for improving data quality and addressing mission needs of FEMA and USGS. TNMCorps partnered with FEMA to use the power of TNMCorps’ volunteer community to improve building classifications in the original FEMA U.S.A. Structures dataset.

The improvement of the U.S.A. Structures dataset is a collaborative effort between FEMA’s Response Geospatial Office, Oak Ridge National Laboratory (ORNL), and the USGS to build and maintain the Nation’s first comprehensive inventory of all built structures larger than 450 square feet (ft²) for use in flood mitigation, emergency preparedness, and response. For a list of key staff involved in this effort, please refer to appendix 1.

The goal of the collaboration is to enhance the U.S.A. Structures dataset’s occupancy type attribution (for example, residential, industrial, and government) by augmenting ORNL’s automated workflow with verification and refinement by TNMCorps volunteers. The success of this pilot project could lead to a more versatile dataset for multiscale mapping, improve emergency decision-making, and foster interagency cooperation. Ultimately, this effort could help to expand the dataset's utility and contribute to future cartographic solutions for the USGS. Figure 1 summarizes key collaborators, goals, process, benefits, and effects of the project.

Project Area

The pilot project focused on St. James Parish, Louisiana (fig. 2). FEMA identified this region as a high-priority area due to its vulnerability to flooding and storms. Selecting this area allowed for testing the project's efficacy in a region with a high concentration of structures that benefited from the use of accurate data.

Methods

The project implements an innovative methodology that combines automated classification with citizen science contributions to enhance structural classification accuracy across the United States. The ORNL has developed an automated multistep classification process that draws from machine learning in combination with 62 distinct source mapping data layers to classify structure occupancy types (Hauser and others, 2023; Yang and others, 2024). When structures lack sufficient source data for classification through traditional data conflation and fusion methods, a machine learning model is used to determine the appropriate classification.

In St. James Parish, out of 12,328 total structures, 750 required machine learning classification due to gaps in availability of the source data layers. Through the help of TNMCorps volunteers, we discovered a 15-percent error rate in the source data. Subsequent crowdsourced improvements eliminated detectable errors in the dataset.

This verification process through TNMCorps has generated valuable data that can be used to refine ORNL’s classification methodology. The two-step process, combining automated classification with human verification, demonstrates the effectiveness of hybrid approaches in achieving highly accurate structural classification data.

To enhance the accuracy of FEMA’s U.S.A. Structures dataset, TNMCorps volunteers, guided by detailed instructions (app. 2, app. 3, and app. 4), thoroughly reviewed and refined the classifications through a customized TNMCorps editing platform (fig. 3). This two-step process integrated automated classification with human verification to improve data quality.

Results

This pilot project indicated that the ORNL manually verified a random sample of 100 volunteer-modified records, confirming a 100 percent accuracy rate for checked structures. During a one-month period, 1,800 edits were made using the editing platform, significantly refining classifications and descriptions within the dataset. Reviewers either agreed with the modified classifications from TNMCorps or found no compelling information to dispute the new classifications.

The ORNL expects to incorporate the feedback on occupancy types in the upcoming dataset update and cite that the validation was performed by the USGS TNMCorps. The modifications volunteers made were particularly helpful in identifying vulnerable areas and structures such as manufactured homes. The validated volunteer feedback was incorporated into the latest dataset deliverable made to FEMA.

Discussion

The sections below summarize project challenges and how they were addressed. They also include a discussion of benefits, future opportunities, and a long-term vision.

Summary

The National Maps Corps—Federal Emergency Management Agency and Oak Ridge National Laboratory pilot project was completed in February and March of 2024. The project highlighted the use of crowdsourcing, interagency collaboration, and continuous improvement in enhancing data quality to inform critical decision-making. By focusing on improving data locally—particularly in coastal areas and in regions with low data quality—the project not only addresses pressing regional needs but also contributes to a broader effect: enhancing overall national data quality through more accurate inputs for modeling. This dual benefit underscores the project’s goal of improving data at a local level while simultaneously helping to address national-scale challenges. The project’s outcomes and the interest in its continuation indicate that this innovative approach can be used across the Nation.