Screening Highway–Rail Grade Crossings by Concordance Between Modeled Freight Importance and Reported Train Activity

Highway–rail crossing inventories report train activity but do not characterize the freight dependence associated with each location. This study developed an auditable framework to compare modeled freight importance with reported through-train activity across the contiguous United States. The framework routed domestic interzonal freight analysis framework (FAF) rail flows (4,911 directed origin–destination pairs; 846.802 million short tons; $370.293 billion annually) along minimum-length network paths. An 11 m association limit and positive-exposure requirement retained 50,643 crossings, represented by 49,854 analytical localities. A crossing freight importance index combined percentile ranks of tonnage, freight value, and origin–destination reach. Rail-network neighborhoods compared focal importance with neighboring activity through descriptive spatial statistics and extreme-quintile classes. Direct association was modest (Spearman ρ = 0.311), while the bivariate Moran statistic was positive (0.275). Across four weighting schemes and three neighborhood sizes, 3,213 localities remained high–high, and 1,516 of 2,013 baseline disagreement localities retained their class. Among common localities, median high–high Jaccard overlap with baseline was 0.859 under independent ±10% routing-cost perturbations and 0.614 under regional ±25% perturbations; overlap was 0.806 under alternative access selection. Positive aggregate correspondence persisted, but screening locations depended on routing and access assumptions. The resulting catalog links classifications to original crossings and processing diagnostics for targeted verification. The findings describe concordance within the retained population; they do not establish operational routes, crossing-level freight accuracy, or safety risk.