During my eighteen months (2023-2025) in the Office of Policy Coordination and Development at the United States Department of Transportation (US DOT), I was technical lead for the development of the Department’s Transportation Communities (TC) Explorer. The TC Explorer, which we released in January 2025, was an improved and updated version of the Equity Transportation Communities (ETC) Explorer, US DOT’s tool for identifying disadvantaged communities as part of the Biden Administration Justice40 Initiative.
In developing the TC Explorer, I was technical lead on designing and implementing (with R scripts) an improved methodology for identifying Census tracts experiencing transportation disadvantage. I also designed and built an intuitive, user-friendly web tool in ESRI ArcGIS Online to display public-facing data.
While the ETC Explorer was taken down from the US DOT website when the Biden Administration’s Justice40 Initiative was shut down in January 2025, I have provided a summary of my work, as well as links to the publicly available data from the project.
Contents
- TC Explorer Index
- TC Explorer Data Sources
- TC Explorer Data Processing
- TC Explorer Web Tool
- TC Explorer Data for Download
- Census Tract Travelsheds
- Transportation Insecurity Analysis Tool
TC Explorer Index
The TC Explorer Index was an updated version of US DOT’s original ETC Explorer Index and, like the ETC Explorer and the Biden administration’s Climate and Economic Justice Screening Tool (CEJST), was intended to identify US Census tracts as “disadvantaged communities” to receive priority in the distribution of Federal funds under the Justice40 initiative. However, while the CEJST tool was designed to be used across the Federal government, the ETC Explorer and TC Explorer were designed by US DOT to focus particularly on disadvantage related to transportation.
Like the ETC Explorer Index, the TC Explorer Index was based on a large number of indicator variables characterizing the built environment, natural environment, demographics, and economic conditions of Census tracts, which were then combined to calculate subcomponent and component scores that characterized the types of disadvantage experienced by a particular Census tract. Census tracts were considered to experience disadvantage if their overall disadvantage score was above the 65th percentile nationally, and to experience a component of disadvantage if a given component score was above the 65th percentile nationally.
However, the underlying indicators, as well as the method for combining them to calculate subcomponent, component, and overall disadvantage scores were completely overhauled as part of the development of the TC Explorer Index, an effort on which I was the technical lead. While the ETC Explorer Index was based on five components—transportation insecurity, environmental burden, social vulnerability, health vulnerability, and climate and disaster risk burden—with one of the components double-weighted, the TC Explorer Index condensed these into three equally-weighted components of overall disadvantage:
- Transportation Insecurity – The disadvantage experienced when people are unable to get to where they need to go to meet the needs of their daily life regularly, reliably, affordably, and safely.
- Place-Based Burden – The disadvantage inherent in a location and experienced by all residents of the location. The subcomponents and indicators in this component are important because they provide transportation decision makers the information needed to develop transportation plans and make funding decisions that ensure a community’s transportation infrastructure is resilient and minimizes negative health and economic impacts.
- Population-Based Vulnerability – The disadvantage experienced by a population due to demographic and socioeconomic traits that make them particularly vulnerable.
TC Explorer Data Sources
Along with developing the overall structure of the TC Explorer Index, I was responsible for identifying potential data sources for use as indicators measuring the three components of overall disadvantage: transportation insecurity, place-based burden, and transportation insecurity. I prioritized datasets produced by the US Federal government and selected indicators available for all fifty states and DC that exhibited sufficient variation with low enough noise levels when measured for Census tracts to provide reliable measures.
Transportation Insecurity
The transportation insecurity component score was calculated from four subcomponents intended to measure communities’ vulnerability to isolation due to insufficient access to transportation and the burden they experience from the financial and safety costs of the transportation network:
- Destination Access Vulnerability – Vulnerability due to limited availability of essential destinations, such as public transit, medical facilities, education, groceries, and jobs within a thirty-minute trip by driving, cycling, and walking. The driving, cycling, and walking travelsheds for each Census tract used to calculate this vulnerability are available for download.
- Vehicle Access Vulnerability – Vulnerability due to lack of access to vehicles or likely inability to drive, as estimated based on the share of the population under 18, over 65, or with disabilities as reported by the American Community Survey.
- Transportation Cost Burden – Burden from the cost of daily transportation, as estimated by the Bureau of Transportation Statistics (BTS)‘s Transportation Insecurity Analysis Tool (TIAT).
- Traffic Fatality Burden – Burden from traffic violence, as measured by the rate of traffic fatalities over the past five years, as reported by the National Highway Transportation Safety Administration’s (NHTSA) Fatality Analysis Reporting System (FARS) data for 2018–2022.
Place-Based Burden
The place-based burden component score was calculated from four subcomponents intended to measure the harms communities experience from climate change, proximity to burdensome transportation infrastructure, and elevated levels of air and surface pollution:
- Extreme Weather Hazard – Burden due to predicted change in climate-related hazards by 2050.
- The estimated number of high-temperature days, days with extreme precipitation, days with freeze-thaw cycles, and increase in days without precipitation were taken from NOAA’s CMRA based on the RCP 8.5 climate model.
- The current level of impervious surface coverage was sourced from the Multi-Resolution Land Characteristics (MLRC) Consortium National Land Cover Database (NLCD) and National Oceanographic and Atmospheric Administration (NOAA) Coastal Change Analysis Program (C-CAP) High-Resolution Land Cover rasters.
- The average annual wildfire burn probability was taken from US Forest Service Wildfire Risk Components Rasters, 3rd Edition.
- The share of land area predicted to be inundated by simultaneous 100-year riverine and coastal flooding events in 2050 based on the RCP 8.5 climate model were taken from World Resources Institute (WRI) Aqueduct Flood Risk Rasters.
- Infrastructure Proximity – Burden due to proximity to freeways, high-traffic roads, railways, airports, and ports, which often result in higher rates of air and noise pollution and can divide communities.
- Air Pollution Burden – Burden due to higher levels of criteria air pollutants (diesel particulates and nitrogen dioxide) and hazardous air pollutants (also known as air toxics), as reported by the US Environmental Protection Agency (EPA) Environmental Justice Screening and Mapping Tool (EJScreen) versions 2.2 and 2.3.
- Surface Pollution Burden – Burden due to proximity to land and surface water pollutants, including hazardous waste biennial reporters, toxic release inventory and risk management plan sites, hazardous waste storage, transfer, and disposal sites, leaking underground fuel storage tanks, and active mines.
Population-Based Vulnerability
The population-based vulnerability component score was calculated from five subcomponents intended to measure demographic and socioeconomic traits that can make populations particularly susceptible to the harms measured by the other two component scores:
- Communication Vulnerability – Vulnerability due to inability to reliably receive information about the transportation system, in terms of the share of households without internet access or with limited English proficiency, as reported by the American Community Survey.
- Employment Vulnerability – Vulnerability due to lack of access to employment, in terms of the shares of the adult population who are not employed and have not graduated from high school, as reported by the American Community Survey.
- Income Vulnerability – Vulnerability due to poverty, in terms of the shares of the population in poverty and of households with median household incomes below the regional average, as reported by the American Community Survey.
- Housing Vulnerability – Vulnerability due to housing conditions, in terms of the shares of households that are income-burdened or that experience overcrowding, lack of indoor plumbing, and lack of kitchens, as reported by the American Community Survey.
- Health Vulnerability – Vulnerability due to higher prevalences of health conditions that may result from or be exacerbated by exposure to pollutants, poor walkability, car dependency, and long commute times, as reported by the Centers for Disease Control and Prevention (CDC) Population Level Analysis and Community Estimates (PLACES) public health dataset.
TC Explorer Data Processing
In addition to acting as technical lead on the selection of indicators and overall development of the TC Explorer Index’ model’s model of overall disadvantage, I developed the data processing algorithms and implemented them as a set of R scripts. These scripts imported the raw data, including using population-weighted interpolation to convert data tabulated for 2010 Census tracts to 2020 Census tracts and tabulating values for raster-based data into Census tracts, as well as identifying population- and area-weighted centroids of the tracts and tracts’ nearest neighbors.
Interpolation for Missing Data
The centroid and nearest-neighbor data was used to provide interpolated values of indicator variables for tracts where the variable would otherwise not be available. This was necessary because political leadership considered it essential that disadvantage scores be calculated for every Census tract, to enable disadvantaged community determinations for all proposed discretionary grants, including those in parks and other areas where data might otherwise be unavailable.
Two types of missing data were considered. First, some data sources were missing values for specific tracts: often, but not always, tracts without population. Second, for data sources based on population characteristics measured by the American Community Survey–the population-based vulnerability indicators, as well as vehicle access vulnerability and transportation cost burden–tracts with fewer than 100 residents or households were treated as having no data because the small sample size of the data source makes any reported values for such tracts unreliable.
In both cases, the missing or suppressed value was replaced with the average of the values for the three nearest tracts, measured as distance between population or land area centroids depending on the type of variable, without missing values. (Interpolated values were never used to to calculate interpolated values for other tracts.)
Normalizing Indicators
In addition to interpolating for missing values, I developed a method for normalizing the raw indicator values that avoided one of the major issues with the ETC Explorer Index. A number of the indicators in the ETC Explorer Index were of limited value because, rather than having a normal distribution or being boolean (with almost all values either 0% or 100%), they had a small number of tracts with extreme values orders of magnitude larger than the range over most tracts varied.
The presence of a few tracts with extreme values in ETC Explorer indicators meant that the variation between all tracts except for these extreme values was effectively “washed out,” with all tracts except for the extreme ones having normalized values of effectively zero. Besides making the indicator useless for comparing disadvantage between the vast majority of the country, this effect was particularly problematic since such extreme values are often due to measurement errors in the underlying data sources or situations where the measured values are not a good representation of the actual situation.
To avoid this issue, instead of a simple min-max normalization, I calculated normalized indicators for the TC Explorer with a modified procedure, setting the 1st percentile tract to a value of 0 and the 99th percentile tract to a value of 1. Tracts with values lower than the 1st percentile tract were set to 0 as well, while tracts with values higher than the the 99th percentile tract were set to 1. Tracts in between these values were given normalized values equal to the difference between their value and the 1st percentile value divided by the difference between the 99th percentile value and the 1st percentile value.
The result of this approach is very similar to or identical to a traditional min-max normalization for indicators where extreme outliers are not present. However, when extreme outliers are present, it produces a broader measure of variation between most tracts while maintaining the status of outlier values at the edges of the distribution.
Calculation of Disadvantage Scores
Once the normalized indicators were calculated, they were used to calculate the subcomponent scores, which were in turn used to calculate component scores and the overall disadvantage score. Two separate sets of subcomponent, component, and overall disadvantage scores were calculated: one based on percentile ranks across the fifty states and DC and one based on percentile ranks within a given state. In both cases, the normalized indicator values were averaged together and the average values were percentile-ranked.
Component scores were calculated from subcomponent scores by averaging the subcomponent scores within a given component, and then percentile-ranking the resulting averages. The same approach was used to calculate overall disadvantage, but with the three component scores rather than subcomponent scores as inputs.
Certain “isolated Alaska tracts,” Census tracts in Alaska where most residents do not have year-round road access to an urban area, were treated as a special case, because our transportation insecurity measures break down in places where most travel does not involve the road network. In these places, transportation costs cannot be estimated based on public transportation and driving expenses: the major transportation cost experienced by residents is the need to fly to cities for medical care, specialized shopping, and other purposes. (The standard transportation cost burden measure neglects long-distance travel costs, including all flights.) Likewise, the traffic fatality burden measure is based on fatalities involving road vehicles, and so does not accurately describe the safety hazards in places where road travel does not occur, but safety risks from off-road snowmobile use and regular flights by small aircraft are present. For the “isolated Alaska tracts,” the transportation cost burden and traffic fatality burden subcomponent scores were set to 1 (maximum disadvantage), and this value was propagated in the calculation of the transportation insecurity component and overall disadvantage scores.
Identification of Disadvantaged Communities
As in the original ETC Explorer tool, Census tracts with overall disadvantage scores above the 65th percentile were identified as disadvantage. Tracts with subcomponent or component scores above the 65th percentile were also identified as experiencing the corresponding type of disadvantage. Additionally, all tracts in the US territories, including Puerto Rico, were classified as disadvantaged communities, as were Native American / Alaskan Native communities regardless of their locations.
Project areas consisting of multiple Census tracts or portions of multiple Census tracts posed an additional challenge, since it was necessary to identify them as disadvantaged or non-disadvantaged based on an algorithm that could be calculated “on-the-fly” in the web tool described below. Project areas were identified as disadvantaged communities if either or both of the following conditions were met: the majority of the Census tracts intersecting with the project area were disadvantaged communities or the majority of the population of the Census tracts intersecting with the project area were disadvantaged communities
TC Explorer Web Tool
Since one of the US DOT’s major equity goals during the Biden administration was to ensure that smaller and less experienced applicants—including those with limited technical resources—were able to apply to the Department’s discretionary grant programs, leadership considered it essential that the TC Explorer Index and datasets be paired with an easy-to-use web tool. While a web tool cannot fully compensate for an applicant not having access to GIS software and people trained to use it, the ETC Explorer web tool was intended to give applicants with limited resources an easy way to make use of the data we compiled.
Although my background is in GIS and spatial data analysis, not the development of web applications, I was put in charge of developing an updated web tool for the TC Explorer to make use of our new data and to correct issues in the older ETC Explorer web tool. I built the tool in ESRI ArcGIS Online’s web tool environment, using Experience Builder for the overall structure of the tool and Dashboards to create displays for our national and state-level results. A link to the Bureau of Transportation Statistics’s Transportation Insecurity Analysis Tool (which I did not build) was also included as a tab.
Since the TC Explorer web tool was taken down as part of the Trump administration’s purge of equity-related information from Department of Transportation websites in January 2025, it is no longer available online. However, I have included below a gallery showing screenshots of the tool:
The TC Explorer web tool’s front page. 
A selection of Census tracts with overall disadvantage scores shown. 
A selection of Census tracts with transportation insecurity scores shown. 
A selection of Census tracts with place-based burden scores shown. 
A collection of Census tracts with population-based vulnerability scores shown. 
A collection of Census tracts with raw data shown.
Along with the introductory text on the “TC Explorer – Home” and download links on the “Data and Methodology Download” tabs, I built the “TC Explorer – National Results” and “TC Explorer – State Results” tabs of the tool. These tabs were essentially mirrors of each other, with the former presenting disadvantage scores calculated from percentile rankings of all Census tracts in the US and the latter presenting disadvantage scores calculated from percentile rankings of all Census tracts in a given state.
Both tabs allowed the user to select a group of Census tracts and, on a set of sub-tabs, see disadvantage scores and subscores calculated for the selected tracts in one of four categories: overall disadvantage, transportation insecurity, place-based burden, and population-based vulnerability. In addition, a fifth tab presented a number of raw data values calculated for the selected tracts. These variables were selected to provide information commonly required as part of US DOT discretionary grant applications.
TC Explorer Data for Download
Before the TC Explorer data and documentation was removed from the US DOT website as part of the Trump administration’s purge of equity and Justice40 information from Federal government websites, I downloaded the publicly available documentation and datasets. This data and the associated documentation, as publications of the United States Government, are in the public domain and can be shared and used freely.
Here, you can find the 73-page technical methodology document for the TC Explorer (largely written by me), along with a Microsoft Excel file containing the data dictionary. The data itself is available at the Census tract level (using 2020 Census tract boundaries and 2023 tract FIPS codes) both as a CSV file and a file geodatabase. In addition, a counties-and-county-equivalents file geodatabase with data is provided, as are the file geodatabases of the population and land area centroids of all Census tracts. This last file also includes a CSV file of the nearest neighbors of each Census tract based on population and land area centroid distance.
- TC Explorer Technical Methodology (PDF)
- TC Explorer Data Dictionary (XLSX)
- TC Explorer Census Tract Data (CSV)
- TC Explorer Census Tract Data (File Geodatabase)
- TC Explorer Counties Data (File Geodatabase)
- TC Explorer Centroids (File Geodatabase and CSV)
I’ve also included for download three additional files that may be of interest: the definitions glossary we provided for download (which is largely a subset of definitions in the technical methodology glossary), and the user tips and quick guide to determining a project area, both of which were intended for users of the web tool.
- TC Explorer Definitions (PDF)
- TC Explorer Quick Guide to Determining a Project Area (PDF)
- TC Explorer User Tips (PDF)
Census Tract Travelsheds
In order to calculate the destination access vulnerability subcomponent score of the TC Explorer index, I used ESRI ArcGIS Pro Desktop with ESRI’s proprietary routing network and the Service Area Analyst tool to generate pedestrian, cyclist, and motorist travelsheds for the population centroids of each Census tract in the United States. These travelsheds were posted on the TC Explorer website before it was taken down and, as publications of the United States Government, are in the public domain and can be shared and used freely:
- Pedestrian Travelsheds (File Geodatabase)
- Cyclist Travelsheds (File Geodatabase)
- Motorist Travelsheds (File Geodatabase)
All three sets of travelsheds are intended to approximate the area reachable in 30 minutes from the population centroid of each Census tract. The full methodology for the generation of these travelsheds can be found in Section 4.1 (pages 26-29) of the TC Explorer Technical Methodology.
The pedestrian travelsheds have radii of 1 mile (approximating a walking speed of 2 miles per hour) while the cyclist travelsheds have radii of 5 miles (approximating a walking speed of 10 miles per hour). Both were generated excluding ferries, limited access roads, roads where walking is prohibited, and roads marked “unsuitable for pedestrians,” as ESRI does not offer separate settings for cyclists.
The motorist travelsheds have radii of 30 minutes, using the ESRI routing network’s estimated travel speeds at 8:30am on a typical Wednesday for travel away from the tract centroids. These travelsheds were generated excluding carpool-only and tolled-express lanes and roads, roads closed for construction, and roads where driving an automobile is prohibited. However, ferries were permitted.
Transportation Insecurity Analysis Tool
One of the major data sources for the TC Explorer was the Transportation Insecurity Analysis Tool (TIAT), developed by US DOT’s Bureau of Transportation Statistics (BTS). Of particular interest in the TIAT data are BTS’s detailed small-area estimates of transportation costs for households at the Census tract level, based on 2019 and 2021 data.
The TIAT was released to the public at the same time as the updated TC Explorer, and was also purged from the US DOT’s website as part of the Trump administration’s purge of equity-related projects. However, the TIAT documentation and data were posted on the TC Explorer website before it was taken down and, as publications of the United States Government, are in the public domain and can be shared and used freely. While I was not involved in the production of this documentation and data, I believe it is an important resource and am hosting copies of it downloaded from the TC Explorer website before it was taken down.