ITS - Intelligent Transportation Systems Report ITS Home Page
Table of Contents

4.0 Model Calibration

Accurate calibration is a necessary first step for proper simulation modeling. Before modeling ICM strategies, model calibration ensures that base scenarios represent reality, creating confidence in the scenario comparison. It should be noted that even the most detailed microsimulation model still contains only a portion of all of the variables that affect real-world traffic conditions, which can affect the accuracy of the calibration. Details of the methodology that will be used for data calibration are provided below.

4.1 Simulation Model Calibration

Each simulation software program has a set of user-adjustable parameters that enable the practitioner to calibrate the software to better match specific local conditions.  These parameter adjustments are necessary because no simulation model can include all of the possible factors (both on- and off-street) that might affect capacity and traffic operations.  The calibration process accounts for the impact of these “unmodeled” site-specific factors through the adjustment of the calibration parameters included in the software for this specific purpose.  Therefore, model calibration involves the selection of a few parameters for calibration and the repeated operation of the model to identify the best values for those parameters.  Calibration improves the ability of the model to accurately reproduce local traffic conditions. The key issues in calibration are:

4.2 Calibration Approach

Available data on bottleneck locations, traffic flows, and travel times will be used for calibrating the simulation models for the analysis of the Test Corridor. The Test Corridor calibration strategy will be based on the three-step strategy recommended in the FHWA Guidelines for Applying Traffic Microsimulation Modeling Software (Dowling, R., A. Skabardonis, and V. Alexiadis, Traffic Analysis Toolbox Volume III: Guidelines for Applying Traffic Microsimulation Modeling Software, FHWA-HRT-04-040, Federal Highway Administration, July 2004):

  1. Capacity calibration – An initial calibration will be performed to identify the values for the capacity adjustment parameters that cause the model to best reproduce observed traffic capacities in the field.  A global calibration is performed first, followed by link-specific fine-tuning.
  2. Route choice calibration – The Test Corridor has parallel arterial streets, making route choice calibration important.  A second calibration process will be performed with the route choice parameters.  A global calibration is performed first, followed by link-specific fine-tuning.
  3. System performance calibration – Finally, the overall model estimates of system performance (travel times and queues) are compared to the field measurements for travel times and queues.  Fine-tuning adjustments are made to enable the model to better match the field measurements.

Calibration Criteria

Calibration criteria presented in Table 4.1 will be applied for the test corridor simulation, subject to the budget and schedule constraints for the Test Corridor AMS.

Table 4.1 Calibration Criteria for the Test Corridor AMS

Calibration Criteria and Measures

 Calibration Acceptance Targets

Traffic flows within 15% of observed volumes for links with peak-period volumes greater than 2,000

For 85% of cases for links with peak-period volumes greater than 2,000

Sum of all link flows

Within 5% of sum of all link counts

Travel times within 15%

Greater than 85% of cases

Visual Audits
Individual Link Speeds: Visually Acceptable
Speed-Flow Relationship

To analyst’s satisfaction

Visual Audits
Bottlenecks: Visually Acceptable Queuing

To analyst’s satisfaction

4.3 Preliminary Model Calibration Results

Preliminary results for the calibration of the Test Corridor DynaSmart‑P model are presented in this section.

Figure 4.1 shows observed versus simulated traffic volumes across 7 consecutive iterations in the calibration process. Progressively the calibration results have become better from Iteration 1 to Iteration 7. For most links with high-traffic volumes (greater than peak-period traffic flows of 2,000), simulated volumes fall within the 15 percent error range (dotted lines in Figure 4.1).

Figure 4.1 DynaSmart-P Calibration – Observed versus Simulated Volumes

Figure 4.1 is a point plot of the DynaSmart-P calibration at both iteration 1 and iteration 7. The observed volume is along the x-axis and the simulated volume is along the y-axis, both range from 0 to 18,000. There are three trend lines: one depicting symmetry between simulated and observed and two bordering the 15 percent error range. The data points from iteration 7 fit much more closely within the error range than do the data points from iteration 1; this is especially true for the data points above 2,000 vehicles.

Figure 4.2 shows percent error between observed and simulated traffic volumes for the calibrated DynaSmart-P baseline simulation. Again, for most links with high-traffic volumes (greater than peak-period traffic flows of 2,000), simulated volumes fall within the 15 percent error range (dotted lines in Figure 4.1).

Figure 4.2 DynaSmart-P Calibration – Error in Link Volumes

Figure 4.2 is a point plot showing the percent error between observed and simulated traffic volumes for the calibrated DynaSmart-P baseline simulation. The observed volume is located on the x-axis, ranging from 0 to 18,000, and the percentage error is located on the y-axis, ranging from -100 percent to 200 percent. The error is varied for links up to about 4,000 vehicles at which point the error decreases and a majority of the links with volumes above 4,000 are with in the 15 percent error range.

Table 4.2 shows observed, simulated, and free-flow average travel times at 6 segments of the Test Corridor in the southbound direction. Overall, the simulated corridor travel time is within the 15 percent error range.

Table 4.2 Travel Time Calibration – Southbound – AM

Segment

Start

End

Observed Time
(in Minutes)

Simulated Calibrated
(in Minutes)

Free-Flow
Travel Time
(in Minutes)

1

North end

29th Avenue

4.90

5.12

4.30

2

29th Avenue

98th Avenue

4.15

4.02

3.60

3

98th Avenue

I‑238

8.10

8.79

3.70

4

I‑238

SR 92

7.40

10.29

3.90

5

SR 92

SR 84

6.97

10.22

5.73

6

SR 84

Auto mall

6.92

5.51

5.01

Total

No value

No value

38.44

43.95

26.24

Table 4.3 shows observed, simulated, and free-flow average travel times at 7 segments of the Test Corridor in the northbound direction. Overall, the simulated corridor travel time is within a 20 percent error range.

Table 4.3 Travel Time Calibration – Northbound – AM

Segment

Start

End

Observed Time
(in Minutes)

Simulated Calibrated
(in Minutes)

Free-Flow
Travel Time
(in Minutes)

1

Sorth end

Auto mall

6.35

5.72

5.94

2

Auto mall

SR 84

5.12

12.18

5.04

3

SR 84

SR 92

11.27

16.37

5.91

4

SR 92

I‑238

3.90

5.86

3.30

5

I‑238

98th Avenue

6.45

3.97

3.90

6

98th Avenue

29th Avenue

5.37

3.58

3.63

7

29th Avenue

North end

5.45

5.18

3.93

Total

No value

No value

43.91

52.86

31.65

Some of these calibration results do not meet the calibration criteria presented in Table 4.1 primarily because meeting these criteria was assigned a lower priority given the resource constraints for the Test Corridor AMS. In a real-corridor ICM application continued calibration is needed, including a) calibration for average and non-average conditions, b) meeting calibration targets for aggregate travel times and delay, and c) more explicitly taking into account bottleneck flows including temporal variation throughout the modeled time period.

next | previous