Quantifying Percent Traffic Congestion (pTC) and Mobility Bottleneck Dynamics at Atlanta’s Spaghetti Junction

Highway interchanges are vulnerable components of transport networks, often prone to congestion and crashes. Traditional monitoring methods like loop detectors or travel time queries often fail to capture the granular spatiotemporal distribution of bottlenecks in detail. To address this gap, this study introduces a new approach to quantify congestion and analyze bottleneck dynamics at Atlanta’s Tom Moreland Interchange, one of the nation’s most congested sites. A percent Traffic Congestion (pTC) metric was developed from the Google Maps Traffic Layer for twelve directional routes and validated against observed travel times obtained independently through the Google Maps Routes API. Traffic imagery collected every ten minutes for four months and 746 crash records were analyzed. Findings reveal distinct spatial patterns and temporal dynamics of congestion, with northbound I-85 and eastbound I-285 most affected during afternoon peaks. A quadratic model provided the best fit between pTC and travel times (R2 = 0.85), confirming pTC as a reliable congestion indicator. An LSTM model using pTC time series also accurately predicted mobility trends at the I-285 west to I-85 north bottleneck. Additionally, Seasonal-Trend decomposition using LOESS (STL) identified congestion anomalies, and their association was analyzed with crashes. The proposed methodology offers transportation agencies a cost-effective framework for monitoring, measuring, and understanding congestion in complex interchanges.