About Calibration of SIDRA capacity model using Delay Data
You know how important model calibration is in assessing the ability of your traffic model to replicate existing traffic conditions so that you can trust the intersection design and operational solutions you create. This is a discussion of the dilemma that exists when calibrating SIDRA capacity and performance models using observed delay or queue length data for oversaturated lanes.
Let’s assume that the survey method collected observed delays that match the definition of delay in the SIDRA delay model:
- the method measured travel time between “a point on the approach section upstream of the back of queue” and “a point on the exit section after vehicles departing from the approach queue negotiate the intersection area”, and
- to calculate average delay, “the average travel time for unqueued vehicles to travel between the same two points (same travel distance)” was subtracted from “the average travel time for all vehicles queued and unqueued”.
The survey method counted volumes at the stop line. It was observed that the lane was oversaturated (overflow queues persisted throughout the survey period). Under these circumstances, the departure flow rate at the stop line would be the capacity value (degree of saturation of 1.0).
Let’s assume that, for the survey lane which was at a signalised intersection:
- SIDRA estimated a degree of saturation which is very close to 1.0, therefore estimated a capacity value which is very close to the observed capacity,
- however, the observed delay was much larger than the SIDRA delay estimate (this was in spite of using volumes and signal timings observed at the same time as the delay surveys).
Now, let’s look at what happens with model calibration using the following simple example.
Observation of Delay and Capacity for Oversaturated Conditions
The departure flow measured at the stop line was 1430 veh/h which was input in SIDRA as the arrival flow. SIDRA used a basic saturation flow of 1950 tcu/h and estimated the lane capacity as 1443 veh/h. The degree of saturation was 1430 / 1443 = 0.991 which indicated that the capacity estimate was satisfactory. On the other hand, the estimated average delay was 38.2 s whereas the observed delay was 76.0 sec (an error of 50%).
The method recommended for capacity calibration for traffic signals is to adjust the basic saturation flow. To apply this adjustment to increase the estimated delay, the basic saturation flow was reduced to 1761 tcu/h giving estimated capacity = 1303 veh/h. The degree of saturation was increased to 1430 / 1303 = 1.097 and estimated delay = observed delay =76.0 sec was obtained.
However, this calibration method resulted in underestimation of capacity and the true demand flow. This could cause problems with the estimation of other performance statistics (see below).
This calibration dilemma was resolved by estimating an increased demand flow to match the observed delay rather than decreasing capacity (the demand flow rate was underestimated since the departure flow rate based on vehicle volumes counted at the stopline of an oversaturated lane was used as the arrival flow input in SIDRA). The demand flow rate = 1590 veh/h was found to give estimated average delay = observed delay =76.0 sec with no change in the estimated capacity. The estimated degree of saturation increased to 1590 / 1443 = 1.102.
While the same average delay value (76.0 sec) was found by increasing the demand flow or by decreasing the capacity, the latter resulted in underestimation of such performance statistics as total delay (veh-h/h), total travel time (veh-h/h), total travel distance (veh-km/h), total CO2 emissions (kg/year), back of queue (metres) and the probability of upstream lane blockage in a network in the range 8 to 10 per cent.
Dr Rahmi Akçelik
22 March 2024
#trafficmodel #calibration #analytical #intersections #interchanges #networks #trafficmanagement #trafficcontrol #design #operations #software #SIDRA #SIDRAINTERSECTION #trafficsignals #roundabouts #signcontrol #pedestrians #capacity #delay #queue #CO2 #congestion #oversaturated