Reproducibility code for inter-county distances along U.S. highway and railway networks

Abstract

This software package contains the reproducibility code used to construct and validate the dataset of inter-county distances along the real-world highway and railway networks in the contiguous United States. The workflow implements the QGIS–RStudio crosswalk described in the accompanying study, integrating geographic information system processing with network-based analysis to transform transportation infrastructure shapefiles into spatial networks, select representative network nodes for counties, and compute shortest-path distances between counties.The QGIS–RStudio crosswalk follows four sequential steps: (1) preprocessing of transportation shapefiles in QGIS; (2) construction of highway and railway networks in RStudio; (3) centroid computation and network-node assignment in QGIS; and (4) representative-node selection and shortest-path distance computation in RStudio. The highway and railway line features are transformed into undirected weighted network graphs, with segment lengths used as edge weights. Adjusted geometric centroids are used as spatial references for selecting representative network nodes within counties, while the inter-county shortest-path distances are computed between the selected representative network nodes rather than between county centroids directly.The software package also includes code for the principal Technical Validation analyses reported in the accompanying study. These include FAF-scale empirical comparisons using Freight Analysis Framework Version 5.7.1, evaluation under alternative centroid choices and representative-node selections, county-level sensitivity analyses of centroid choice and representative-node selection, and independent county-level comparisons with external highway and railway routing systems. Population-weighted centroids and second-nearest qualifying network nodes are used as alternative specifications in the sensitivity analyses.For the independent external validation, 150 origin–destination county pairs are sampled for each transportation mode using stratified random sampling across the four U.S. Census regions and three mode-specific distance strata. Highway routing distances are obtained using the HERE Routing API v8 with truck transport mode and short routing mode. Railway routing distances are obtained using OpenRailRouting with the non_tgv routing profile. Routes are calculated in both the origin-to-destination and destination-to-origin directions and averaged for comparison with the symmetric inter-county distances generated from the undirected network graphs.Detailed instructions for required input data, QGIS processing, RStudio-based network construction, representative-node selection, shortest-path distance computation, expected outputs, checkpoint reuse, and validation procedures are provided in the included README.md. The workflow was implemented using QGIS (version 4.2.1) and R (version 4.6.0), consistent with the accompanying manuscript.The associated dataset is available separately on Figshare at DOI: 10.6084/m9.figshare.32211066.