Effect of On-Ramp Demand on Capacity at Merge Bottleneck Locations

Abstract

Past research acknowledged the impact of ramp vehicles on the occurrence of a breakdown event, but little has been done to quantify the effect of the ramp vehicles on the resulting bottleneck capacity. The objective of this research is to explore the relationship between ramp flow and capacity and to recommend capacity values for merge bottleneck locations. To explore the relationship between ramp flow and capacity, a capacity model has been developed using linear regression. The entire freeway network in Kansas City area was considered for the analysis. All locations which experienced “true breakdown”, breakdown because of merging operations and not due to downstream spillback, were selected for the analysis. Detector data at the six selected locations, were downloaded from KC Scout Portal from January 1, 2016 to June 30, 2016. Per lane and average speed, volume and occupancy data at 5-minute intervals were chosen for the analysis so as to detect breakdowns and find independent breakdown locations. Incident data and bad weather data were also collected for the same period and all the days with incidents and bad weather were removed from the analysis. For this research, free flow speed was defined as the average of flows when speeds were more than 50 miles per hour and the flows were less than 800 vehicles/hour. Breakdown was said to occur when speeds drop more than 25% of the free flow speed and the reduced speeds are maintained for at least 15 minutes, i.e., three 5-minute intervals (TRB, 2016). The breakdown capacities ranged from 3,900 to 8,500 vehicles per hour (veh/h) and when averaged across all the lanes, 1,150 to 2,200 vehicles per hour per lane (veh/h/ln). The upstream breakdown flows (demand) ranged from 3,400 to 8,400 veh/h and when averaged across all the lanes, 1,050-2,100 veh/h/ln. The ramp breakdown flows (ramp demand) ranged from 150 to 2,700 veh/h and when averaged across all the lanes, 150 to 1,500 veh/h/ln. Various variables such as freeway demand, ramp demand, free flow speed, number of lanes, ramp to freeway demand ratio, outer two-lane flow, shoulder lane flow, and remaining lane flow were considered for developing the model. Interactions between the variables were also considered. The final model was developed using 70% of the data, which were randomly selected, and the remaining 30% of the data was set aside for validation. A final model with an R2 value of 0.689 was developed. The high R2 value indicates that the developed model is a good predictor of capacity and this was also proven through the validation test for the developed model. The regression model was used to predict capacity values for different ramp demand and freeway demand. By observing the calculated capacity values it was concluded that the capacity was decreasing as the ramp demand, outer two lanes flows were increasing and the capacity per lane was decreasing as the number of lanes was decreasing, which is consistent with past literature (Lu & Elefeteriadou, 2013; Kondyli et al. 2016).