Introduction

The National Structure Inventory (NSI) 2026 version is a significant update to the NSI base level dataset and is the first update to the base layer since the release of the 2022 version.

The purpose of the NSI databases is to facilitate storage and sharing of point-based structure inventories used in the assessment and analysis of consequences of natural and technological hazards. Flood risk analysis is the primary usage, but sufficient data exists on each structure to compute damages and life safety risk due to other hazard types. The updated NSI includes several important features that enhance the accuracy of previously available fields while also offering new fields. Notable examples of improvements include:

1. New Fields. Several new fields related to vehicle availability, the structures’ full non-depreciated replacement cost, building height, and the Census’s Community Resilience Estimate for each structure’s Census tract.
2. Improved estimates of existing fields. New datasets from HIFLD and other sources have been used to increase accuracy of population estimates of college campuses, prisons, and hotels. Updates have also been made to fields such as depreciated structure value, number of stories, and building material.
3. Updated Data Vintage. Other key datasets saw their own version updates, including parcel data, businesses databases, and decennial and ACS Census data.
4. Updated Error Handlers. Numerous quality control checks have been added to improve the identification and removal of erroneous input data; this includes duplicate residential unit counts, implausible square footage estimates, and the false positive reporting of buildings from source datasets.

User Resources

The NSI-2026 is live through our API alongside numerous user guides, technical documentation and other resources:

1. Technical Overview Documentation.
2. FAQ and Common Issues
3. API User Guide
4. Discourse NSI User Group.

Other resources, such as a detailed technical report and user videos, are intended to be created and made available to users soon.

NSI Download Tool

The NSI download tool is being updated to use the 2026 data and will be released after testing. In the mean time the current download tool still serves 2022 data. Leverage the API or use the LSRI, DST, LST, or LifeSim to download inventories with the 2026 data until the Download Tool is updated.

Major Changes from 2022

Population Modeling

Census Tract Primacy

Previous versions of the NSI included a wide variety of Census data inputs, including block level populations estimates and county level estimates of fields such as disability rates. With the release of NSI-2026, most fields are now based on tract level data. The NSI team believes that this resolution provides the greatest compromise between the benefits of spatial resolution while minimizing the random influence of large margins of error caused by small sample sizes.

This change was also partially necessitated by updates to block level estimates due to the Census’s differential privacy methodology. Census products would, on rare occasions, appear to concentrate populations in locations without structures or, in the LEHD, report that workers are in neighboring census blocks to the large employer’s true location. The switch to tract-based population pools reduces the likelihood of erroneous data in residential and commercial populations estimates being concentrated in small neighboring structures.

Previous versions of the NSI largely relied on using business level employee estimates to help weight block-level worker estimates to individual buildings. NSI 2026 relies more on the square footage estimates to weight tract-level populations to individual structures. Updated tract-level student enrollment data is also used to decrease residential population during the day.

Population Estimates for Unique Occupancy Types

Group Quarters (GQ) include buildings such as prisons, college dormitories, or nursing homes, where populations may be in the care of a government or other organization. Residents do not respond directly to ACS surveys and the buildings they occupy are often marked as “Tax Exempt” and provide minimal information to tax assessors.

NSI 2022 relied on data from HIFLD to inform estimates of nursing home locations and their bed counts, but populations were assigned from the general population pool of its census block. NSI 2026 improves use type designations of these structures through the use of additional HIFLD datasets on prison and university boundaries. It further improves the population assignment of these structures by creating separate population pools based on the residing tract’s GQ population estimates indicated in the ACS.

Student enrollment by college campus data and prisoner counts by facility further help refine day and night population estimates in these specialized structures. Hospitals continue to have their populations increased by a proportion of their bed count. RES4s now have population added based on their square footage; dramatically increasing their night population estimate over previous estimates.

Structure Attributes

Number of Stories and Square Footage Estimates

While single-family residences are still largely based on structure level authoritative data, multi-family and non-residential structures are now largely based on height estimates from available footprint data (Microsoft Building Footprints and USA Structures). This is a significant change from previous versions that primarily estimated number of stories indirectly from the assumed occupancy needs of the structure. This methodology will also improve square footage estimates that are based on these imputed number of stories values.

Foundation Type and Height

The NSI 2026 continues to rely on a combination of county tax assessor data and regional distributions to assign foundation types. However, adjustments were made to the regional distributions to account for the increasing prevalence of slab and the relative concentration of basement foundations in colder climates.

Foundation height assignments were more modest. Slab heights now vary between residential (0.75ft) and non-residential (0.25ft) use types. The representative crawl space height increased from 1.5ft to 2.0ft. The height of pier and pile foundations now vary by occupancy type and flood zone.

Input Error Checks

NSI 2019 and NSI 2022 included a significant number of error checks when generating structures based on parcel and business data.

Numerous additional error checks were added for the NSI 2026 release. This includes upper limits on square footage values (varying by use type), upper limits on housing unit counts (varying by census tract and square footage), improved processes for merging condo units into a single high-rise, and increased skepticism that buildings of certain use types (“Farm Homes” etc.) exist when a building footprint is absent in the parcel’s polygon.

Spatial Placement

Decreased False Negatives

Previous versions of the NSI struggled with large parcels where the number of buildings was larger than the number of businesses or where residential apartments were coded as commercial by parcel data. The NSI 2026 will generally populate each footprint within a parcel with an occupancy type matching available business, assessor, or Census data. Presumed detached garages in single-family parcels are still intentionally not labeled with unique record; the garage’s value is implicitly captured at the location of the assumed primary residence.

Decreased False Positives

The 2022 and 2019 versions of the NSI would ensure that populations indicated by block level Census data had buildings to reside in, even in the absence of indications of building footprints within the block. Those versions accomplished this by pulling structures from legacy HAZUS based pre-2019 versions of the NSI. This often resulted in structures in odd locations.

The NSI 2026 largely avoids this issue by its switch to tract-based population estimates and greater reliance on building footprint data. The supplementation of parcel and business data for occupancy type identification is generally limited to HIFLD datasets on high priority use types and Census data on the number of housing units within an area.

Valuation Updates

Structure Value

The methodology for depreciated replacement value saw multiple updates. New national averages by occupancy type were assumed, based on a combination of publicly available data, industry references, and expert-informed assumptions. Likewise, county level adjustments received an update to be based on construction labor wage ratios. Depreciation is not solely based on the structure’s age, but instead also varies with demographic estimates of the census tract in which the structure resides.

Content Value

The NSI estimates content value by assuming a content-to-structure value ratio. These ratios were updated for all occupancy types other than RES1s (Single-Family Residences). These adjustments incorporate a wider range of sources: expert elicitations, survey data, and analyses of claims data. Mobile homes and industrial facilities saw the larges increases, while multi-family structures and office buildings saw largest decreases.

Vehicle Value

Previous versions of the NSI estimated a single nation-wide household and worker averages for vehicle values at homes and non-residential structures. The updated methods account for the Tract level estimates of vehicle ownership and income as well as state level data on vehicle types. These impact estimates of the number of vehicles present at structures and implicit estimates of vehicle ages and models.

Additional Fields

The previous NSI field list was developed to address the needs of its original user base, which was primarily dam and levee safety consequence analysts and other Flood Risk Management analysts within the US Army Corps of Engineers. However, as the use of the NSI has spread to other hazards and outside USACE, the potential utility of additional fields has increased. These new fields are discussed below.

Values without Depreciation and Likely Depreciation Indicator

USACE guidance encourages the use of depreciated replacement value for its benefit-cost studies; however, other users have expressed interest in the full replacement value of structures. To aid these users, the full structure and content value has been added to the public field list.

Depreciation was previously estimated based solely on the age of the structure, with every year equating to a 1% loss of the structure’s value, up to a 20% loss. Depreciation methodology has now been updated to be based both on age and a depreciation index; the depreciation index is also written as an available field. Depreciated replacement value as a percentage of full replacement value can now be as low as 50% or as high as 84% for a 50+ year old structure.

Vehicle Availability and Type Estimates

Two fields have been added from ACS tract data: the percentage of households that have no vehicles and the average number of vehicles present per household. The same vehicle likelihood values are written for every structure within the same tract. A state level proportion of low-clearance (sedans, coups, etc.) and high-clearance (SUVs, pickup trucks, etc.) vehicles have also been added.

Footprint Fields

Most NSI structures are paired with a building footprint from a publicly available dataset (Microsoft and USA Structures). The square footage of the source footprint and the height of the footprint (where available) are now reported. When the NSI Structure was paired with a USA Structure footprint, the population estimates for the building from ORNL’s LandScan program are also joined where available.

Flood Zone

A spatial join was made for each structure to the FIRM zone’s available as of December 2025. The 2026 release also includes the zone sub type and static base flood elevation where available.

Social Vulnerability

The Community Resilience Estimates is a Census product that indicates the level of social vulnerability in a tract. Components include eight measures of topics such as poverty, age, and disability. The NSI 2026 reports the 2024 CRE tract estimates of the percentage of households with three or more factors as well as a percentile ranking of the tract’s three measure percentage.

Impact on Risk Analyses

Because structure inventories represent the at-risk people and property in consequence assessments, the changes to the NSI base layer may influence risk analyses and characterizations. However, it is not possible to say, in the abstract whether the changes will increase or decrease consequence outcomes. At the national level, exposure has generally increased from the NSI 2022. This is partly because of price level changes and population growth, but it also stems from the NSI having a higher success rate of identifying at-risk buildings and occupants.

However, many counties saw a decrease in exposed structure values and populations. This reflects population loss and degrowth in some cases, but it also represents a decrease in false positives and duplicated square footage estimates.

Users can view the descriptive statistics and analysis of county, state, and national level changes in the NSI Statistical Summary page.

Caveats

Comparing with previous NSI Versions

Due to the numerous methodological changes, it is generally inappropriate to infer that differences between NSI 2026 and earlier versions is due to growth or decline of a region. Time series analyses with the NSI is not recommended.

Remaining Limitations

Known issues and limitations remain with the NSI 2026. These include unavoidable issues, such as data inputs lagging reality, simplifying assumptions, and inconsistent coverage of structure level information as well as methodological issues that will be addressed in future versions. These issues are discussed further in the FAQ and Common Issues Documentation.

NSI is not intended for legal or regulatory determinations or for analyses requiring precise structure-level accuracy. It is designed as a baseline inventory for screening-level and regional analyses. Users are expected to review, validate, and refine the data as needed for their specific application. NSI estimates are generally more reliable at larger spatial scales, where regional assumptions and distributions better reflect aggregate conditions.