Estimating Percentile Speeds with Probe Vehicle Data on Non-Freeways

Accurate estimation of percentile operating speeds on arterial roads is crucial for calibrating crash-prediction models, evaluating eligibility conditions for traffic safety countermeasures, and informing speed-management decisions. Existing percentile-speed estimation models were developed for specific regions, so their transferability to other geographical jurisdictions needs to be evaluated. To fill this gap, this study develops and validates an 85th-percentile speed ( V 85 ) model for non-freeway arterials using probe vehicles, field surveys, and roadway data. A dataset comprising sixty spot-speed surveys (forty-two urban and eighteen rural) collected in Maryland from 2019 to 2025 was matched to corresponding INRIX segment speeds, roadway geometric attributes, and traffic volumes. Ordinary-least-squares regression analyses were conducted for urban and rural settings, incorporating key variables including INRIX segment speed, posted speed limit, directional annual average daily traffic, segment length, lane width, access density, signal density, and functional-class indicators. The proposed model demonstrated high predictive accuracy for both urban and rural segments, achieving substantial error reductions compared with the baseline Texas A&M Transportation Institute model and a locally calibrated model. The findings also show that the posted speed limit is essential in rural V 85 estimation and remains useful in urban contexts. The model also supports network screening for segments that may warrant speed management or safety review.