Mission

Provide access to consistent, and nationally available point based structure inventories with attribution to support evaluation of consequences from natural and man-made hazards.

Vision

To support all federal agencies interested in collaborating on structure inventory data,

Goals

1. Provide access to the data to as many people and agencies as possible.
2. Improve the quality of the data.
3. Improve the ability for the U.S. to respond to disasters.
4. Improve the ability for the U.S. to plan for future disasters.

Purpose

The National Structure Inventory (NSI) is a system of databases containing structure inventories of varying quality and spatial coverage. The purpose of the NSI databases is to facilitate storage and sharing of point based structure inventories used in the assessment and analysis of natural hazards. Flood risk is the primary usage, but sufficient data exists on each structure to compute damages and life safety risk due to other hazard types. The purpose of this document is to describe the NSI data structure and to document the processes utilized to produce the 2022 NSI base data.

Overview of the National Structure Inventory

The National Structure Inventory Base layer was created and is maintained by the U. S. Army Corps of Engineers (USACE). The USACE base data layer was created to simplify the GIS pre-processing workflow for the USACE Modeling Mapping and Consequence center, but the data has gone on to see use in a variety of USACE, FEMA, and other agency applications. The NSI is a repository of point structure inventories with a structured RESTful API service, and the inventory contains a series of required attributes or fields that describe each point in the inventory. 

Accessing the NSI

The NSI is available in GeoPackage and GeoJSON format. State level GeoPackages can be downloaded using the NSI Download web page. GeeJSONs may be downloaded for individual counties or specified bounding boxes using the NSI API.

NSI Public Fields

The NSI application programming interface (API) requires structure inventory attributes to be consistent across all datasets in the NSI databases. These attributes exist to meet the computational constraints of the software consuming the NSI. To successfully upload datasets to the NSI, the datasets must contain the required attributes. The analyst is responsible for giving approximate values for each attribute, and documenting the assumptions in providing those attributes. Currently, only the NSI development team is capable of adding inventories to the NSI and only the base layer is available, however, upload functionality is expected to be expanded in coming months. The NSI attributes available to the general public are:

NSI Private Fields

The NSI restricts access to certain fields to Federal users. While these attributes are generally not critical to the use of common consequence models, they may provide additional context to users attempting to refine the data for their analyses. The available attributes for the NSI are:

USACE-Developed NSI Base Data

This section of the document serves as the metadata for the NSI Base data provided by USACE. The document assumes a familiarity with GIS and terms related to the NSI attributes. The HAZUS (2014) database (website: https://msc.fema.gov/portal/resources/hazus) provided the bulk of the base data included in the NSI-2015 Base layer. The team has shifted to a reliance on multiple input sources in order to create discrete points and attributes to represent structures. This Base quality data is not an exact representation of reality, but rather a pseudo inventory with many homogeneous assumptions. Although there are some accuracy issues, the Base dataset functions as a minimum standard for the United States. Appropriate uses include situations where more accurate data is too costly to produce and cannot be created, or when limited by time constraints. Another general use of the NSI Base dataset is for assessments on a national level, where regional assumptions may introduce bias into the analysis. The following sections describe the processes used to produce the NSI Base data.

2022 Base Quality Level Data Generation

In June 2022 the NSI team created the dataset using the following inputs from numerous input data sources. The two main sources of data are Lightbox parcel files for residential structures and ESRI business layer for non-residential structures. Each data file used contains data on the type of development that exists at a given location. For example, the parcel data often stated whether a structure was single-family residential or a multi-family structure; ESRI data reported the NAICS code for each business. These source data categories were converted to a format consistent with one of 40 different HAZUS Occupancy Type classification. Residential Occupancy types are further revised later in the process based on other structure characteristic assignment, with single family residences’ “RES1” classification being appended with the number of stories and basement status (e.g. “RES1-2SNB”).

Main Data Sources

Structure Placement Refinement

The XY location for each structure is initially provided by the source data, such as the centroid of the parcel or the geo-reference of a business’s address. However, the NSI Generator modifies these initial locations by matching the structures to buildings footprints within the same parcel polygon. Commercial structures located outside of a parcel will be moved to an appropriate parcel based on consideration of distance and use code similarity. If there are multiple footprints within a parcel polygon, structures are placed in the largest footprints first. If there are multiple structure types within a parcel polygon, then structures are paired with footprints in the following order: schools first, then commercial structures, and finally residential structures. Structures are placed in unpaired footprints until all footprints are paired with structures, at which point multiple structures may be stacked within the same footprint. Structures that are not matched with parcels (most commonly NSI-2015 based structures in areas that lack parcel data) are moved to an available footprint within their census block instead.

Structure Aggregation

If structures are stacked within the same location, then the structures may be partially or completely merged together. Residential units stacked at the same location are assumed to be multi-family structures; the number of units will be used later to update the occupancy type of the structure (for instance, more than 50 units would mean that a residential structure would be identified as a RES3F). However, commercial structures are not completely merged unless they share an occupancy type; instead, the NSI generator links the stacked structures so that they share certain characteristics such as number of stories and construction material. Each commercial business within the stack will receive a weighted portion of the square footage which informs the valuation of each structure.

Population Growth and Assignment to Structures

Due to the unreliability of census block level population estimates in the 2020 census. the NSI-2022 attempts to estimate 2020 level populations using 2010 block data and 2020 county data. To accomplish this, the NSI Generator was provided a table that recorded the number of increased persons residing in a county above 2010 population levels (counties that lost population received no adjustment). The NSI estimates block level population growth in an iterative process until the total increased population for the county is depleted. Population is first added to structures that had no housing units in 2010 but now have housing units in the newly generated inventory. Next, population is distributed to blocks whose number of housing units is greater in the NSI than it was in the 2010 census. Finally, population is randomly assigned to census blocks until the population growth is fully distributed.

Commercial worker population was derived from the U. S. Census Bureau’s Longitudinal Employer-Household Dynamics (LEHD) database (website: https://lehd.ces.census.gov/). This database contains counts for the number of residents leaving a census block to work and the number of workers arriving in a census block. Departing workers are subtracted from the residential population; as are enrolled students. 2015 era data, last accessed in 2019, was used to support this version of the NSI.

Once block level population estimates are made, population is assigned to particular structures within the block. Population is assigned from 20 separate pools, reflecting combinations of Day and Night, Over and Under 65 years of age, and for five population categories, including Workers and Residents. Population is assigned from commercial population pools to commercial businesses weighted by number of employees, and from residential population pools to residences weighted by number of housing units. The three remaining population pools: institutional population, nursing home population, and student population are used to adjust or assign population to specific subcategories of non-residential structures.  The assignment process also accounts for the relative likelihood of those over 65 years of age to work or stay at home.

Structure Valuation

Depreciated replacement values were estimated based on an assumed replacement category and a dollar per square footage estimate for that category; these assignments are informed by an analysis of survey data, parcel use types, and other source inputs. All values are in 2021 prices levels. Dollars per square foot estimates are then multiplied by the square footage estimate for each structure to obtain the structure value.

These replacement values for structures are then depreciated in order to obtain depreciated replacement value; each structure is depreciated by 1% per year for the first 20 years, after which it is assumed that routine maintenance would keep structure values at 80% of their replacement values.

Content values are obtained by multiplying structure values against an occupancy type specific structure to content value ratio; these ratios are shown below in the Occupancy Type table. 

Occupancy Type

Occupancy types are used to help determine structure valuation, population, and to define structure damage criteria (for flooding). The occupancy types are based on the FEMA occupancy type definitions with further classification to meet the criteria for USACE economic guidance memorandums. The table of occupancy type names and their descriptions are below. These are utilized to support the base level data and are not required for other datasets.

As previously discussed, Occupancy Types are largely mapped from a parcel use type or available NAICS codes. However, occupancy types are often modified if another data input (specific datasets for Nursing Homes, Hospitals, Mobile Home, or Schools) overwrites the initial use type, or, if the use type is updated based on the estimated number of housing units.

Construction Type

Construction type is mapped from parcel data where available. When this data is not available, a construction type is randomly assigned from HAZUS data. The hzCensusBlock table contains an attribute for building scheme, and this attribute is related to the flGenBldgScheme tables from the MSH.mdb database. The building scheme attribute is used to define structures as Wood, Masonry, Concrete Block, Manufactured, and Steel using random assignment based on the probabilities indicated in the HAZUS table.

Square Footage and Number of Stories

Methods for estimating square footage and number of stories varies between RES1s and other occupancy types. RES1s, the most common occupancy type, generally have more reliable parcel data from which to pull values; there are typically values recorded for square footage and number of stories and there is typically only one structure within each single-family parcel so there is no confusion regarding which structure to apply the recorded values too. Therefore, there is typically little need to estimate these values using other methods.

Nonetheless, certain counties lack this data as do certain structures within counties that generally possess the data. When square footage values are missing for RES1s, they are estimated by taking 86% of the structure's footprint; this attempts to remove unoccupied areas such as garages and porches from the footprint area. This percentage was estimated at a nationwide level using available parcel, survey, and footprint datasets. If no footprint data is available either, then square footage is randomly assigned a value from a distribution that varies by the structure's year built and census region. If number of stories data is not available for RES1s, it is estimated by diving the square footage by the structure's footprint; if the value is greater than 1.25, then a second floor is assumed (or if greater than 2.25, then a third floor is assumed and so on). If not footprint is available, then number of stories is randomly assigned from a distribution that varies by year built and census region.

RES3s and Non-Residential structures generally do not have reliable parcel data from which these values can be assigned. There is often no data reported for these use types or there are multiple structures in each parcel and its not clear which, if any, of the structures are represented by the parcel data.

Instead, square footage is estimated based on an assumed needed square foot per housing unit or employees an the assumed number of housing units or employees. If this estimated demand for square footage is less than the structures footprint, then the footprint's square footage is used and the number of stories is kept at a value of one. If the estimated demand for square footage is greater than the footprint's area then an additional story is added. Number of stories added through this process are capped by the footprint's height field when available.

Foundation Type and Height

Foundation type is mapped from parcel data where available. When this data is not available, a foundation type is randomly assigned using HAZUS data. Based on the information in the hzCensusBlock table for building scheme and the tables in the MSH.mdb database that also contain the building scheme attribute, structures are classified into Slab, Pier, Crawl Space, and Basement. The likelihoods vary with flood zone (e.g. coastal or riverine high risk areas) and the structure's year built (for example, those in coastal flood zones are more likely to have a pile foundation than those in low-risk riverine areas).

Foundation heights (in feet) are mapped to each foundation type (for example, for RES1s slab = 0.5 feet and basement = 2ft). These heights were estimated in a 2021 survey completed by USACE economists with each assumed height closely matching the median value from the survey. 

Vehicles

Vehicle values for each structure are based on the number of housing units for residential structures and the number of employees for commercial structures. These estimates do not vary with vehicle ownership rates or income levels throughout the nation.

Ground Elevations

Ground elevations (feet) are determined using the USGS National Elevation Dataset (NED), based on the structure location (website: https://nationalmap.gov/elevation.html).

Acronyms