Guidance on The Level of Effort Required to Conduct Traffic Analysis Using Microsimulation

CHAPTER 4. DATA COLLECTION

The true validity of any model is dependent on the quality of the data that goes into it. The data requirements should be evaluated in the early inception stages of a project to get an early estimate of effort required. Quality/variability of existing data will impact sample sizes; therefore, early statistical evaluations of the data (margin of error) can prove highly valuable for subsequent development of effort required. Key calibration performance measures should be identified early to help determine the data needed to estimate these performance measures.

Often, it is possible to use existing data for data collection, but these data may be outdated or from different timeframes for different parts of the network. In that case, resources need to be allocated for new data collection. If there is limited funding, resources need to be spent judiciously to collect sufficient quality data. Table 7 lists example data required for simulation model calibration.

It also is important to plan for documentation of the information. The documentation methods should include a clear file-naming structure, an explanation of the sources of data, and a database structure that can be readily incorporated into model inputs and model reports. Efficiency of managing the data can be further enhanced through the development of automated procedures.

As part of the scope development process, a data plan needs to be developed. The data collection plan includes sources and types of data that will be needed to prepare and calibrate the model.

Table 7. Example data requirements for simulation.

TYPES OF DATA

The types of data required can be categorized in four main areas: travel demand, traffic control, physical geometry, and ITS elements, as highlighted in table 7. Travel demand data include traffic counts, vehicle classification counts, speeds, travel times, congestion, and queuing observations. Travel demand data require the majority of the data collection effort. Traffic control data include signs, signal control, and timing plans. Physical geometry can be obtained from rectified aerial photography and base mapping files that may be prepared as part of the design effort for projects.

The four types of data can be further categorized into the following two areas:

• Data for model development.
  
  
• Additional data for model calibration.

The tenets of data collection and data management that are important for conducting an effective analytical study include the following:

• The type of analysis being conducted should dictate the quantity of data collected.
  
  
• The required accuracy should drive the quantity of data collected.
  
  
• It is important to utilize data that are recent, internally consistent, and relevant because the quality of the analysis and the resulting decisions depend on the data used.
  
  
• Collection of traffic data should capture the temporal variations in demand, and performance and should be finite enough to allow for an appropriate analysis to be conducted.

Helpful tips on data collection include the following:

• Do not solely rely on the veracity of historical data; allocate some resources to verify the data through current field inspection.
  
  
• Measure flows and estimate or forecast demand. Procedures for estimating demand are given in Traffic Analysis Toolbox Volume III: Guidelines for Applying Traffic Microsimulation Modeling Software.⁽¹⁾
  
  
• Check flow conservation; entry flows should match the exit flows.
  
  
• Always check for data quality. Use the most recent geographic information system (GIS) files, maps, and drawings.
  
  
• Use traffic counts from prior to the onset of congestion to after congestion dissipates.
  
  
• Broaden the geographic scope to cover bottlenecks and spatial and temporal extent of queues.

Key questions related to data collection include the following:

• Do you have the data you need to develop the base model?
  
  
• How old are the data?
  
  
• Have you statistically verified the data quality?
  
  
• Were there any field inspections conducted?
  
  
• Did you check the quality of the sensor data?
  
  
• If you need additional data collection, do you have the resources?
  
  
• What did you do to address missing data?
  
  
• Did you make use of data from other regions that have similar demand?
  
  
• Did you have to make any assumptions? What were they and how did you arrive at them? Have you documented the assumptions?
  
  

DATA SOURCES

Travel Demand

The basic demand data needed by most simulation software are the entry volumes (i.e., the travel demand entering the study area) at different points of the network. At intersections, the turning volumes or percentages should be specified.

O-D

O-D trips can be estimated from a combination of TDM trip tables and from traffic counts. O‑D data can be acquired from the local metropolitan planning organization’s (MPO) regional TDM, but these are generally 24-h estimates and, as such, need to be adjusted and refined to produce hourly or peak period estimates for use in simulation models. Typically, these estimates are further disaggregated to represent 5- or 15-min estimates required for simulation. License plate matching surveys can be used to estimate hourly trips, but this is resource-intensive. Depending on project complexity, a cost effective way of estimating O-D trips is to adjust the TDM O-D trip estimates based on field counts.

If the study area has transit, HOVs, and trucks or if there is significant interaction with bicycles and pedestrians, the corresponding demand data would be needed. In addition, vehicle dimensions and vehicle performance characteristics (e.g., maximum acceleration and deceleration) are required.

Even if only the peak periods are being examined, demand data should be collected before the onset of congestion and should continue until after the congestion has dissipated. Also, to capture the temporal variations in demand, it is best not to aggregate demand data to intervals longer than 15 min.

Vehicle Characteristics

Vehicle characteristics data can be obtained from the State transportation departments or air quality management agencies. National data can be obtained from car manufacturers, the Environmental Protection Agency, and FHWA.

Traffic Control

Data from traffic control devices at intersections or junctions are required. Control data refers to the type of control device (e.g., traffic signal, stop sign, ramp meter, etc.), the locations of these control devices, and the signal timing plans. Traffic control data can be obtained from the agencies that operate the traffic control devices in the given study area.

Traffic operations and management data on links are also needed. These include location and type of warning signs. If there are HOV lanes, information on the HOV lane requirement (e.g., HOV-2 versus HOV-3), their hours of operation, and the location of signs are needed. If there are high-occupancy toll lanes, information on the pricing strategy is required.

Operational Conditions

If there are variable message signs (VMSs) in the study area, the type of information that is displayed, the location, and, if possible, the actual messages that were displayed are needed. Most of this information can be obtained from the public agencies that are responsible for operating the VMSs. The types of signs and locations can be obtained from GIS files, aerial photographs, and construction drawings.

Event data can be received from public agencies, such as traffic management center logs. Crash databases should be verified since data may not always be recent and may not be for the specific study area. The data should be from concurrent timeframes.

Transit Data

Transit data can be obtained from the local and regional transit operators. These data can include schedules and stop locations. Calibration data could include transit automatic vehicle locator data, boarding and alighting data, and dwell time at stops.

Mobile Source Data

Mobile source data include data derived from mobile phones, Bluetooth^® devices, and other mobile sources. Mobile source data are relatively new sources of data that can be used to augment other data collected. The primary types of data obtained from mobile sources are speed and travel time. In most cases, the mobile source techniques use samples of vehicles in the traffic stream, but they may not be reliable sources for traffic counts and vehicle composition.

The mobile source data are typically obtained, stored, and sold by private vendors. Before purchasing these types of data, it is a good idea to have a demonstration of the data and a means to compare the data supplied by the vendor with a real observation of what is happening in the field. Consideration of how the data are provided (format, structure, and software) should be given and should take into account how they will be used for traffic modeling and other purposes.

The following list provides a brief description of techniques and technologies that are all probe vehicle-based applications:

• Toll tag readers: Toll tags that are typically used for electronic toll collection can be used by toll tag readers deployed at various points on a roadway network to obtain average travel time and speed information. There are four components in a toll tag travel time system: electronic tags, antennas, readers, and a central computing and communication facility. As a vehicle with an electronic tag passes near a toll tag reader, the time and toll tag identification number are recorded. If the same vehicle passes the next reader location, the travel time and average speed between the two locations can be determined. The toll tag identification number can be encrypted to protect privacy.
  
  
• Bluetooth^® device matching: Many computers, car radios, navigation devices, personal digital assistants, cell phones, headsets, and other personal devices are Bluetooth^® enabled to allow wireless communication between devices. Generally, manufacturers assign unique median access control (MAC) addresses to Bluetooth^® equipped devices. Bluetooth^®-based travel time measurement involves identifying and matching the MAC address of Bluetooth^®-enabled devices carried by motorists as they pass a detector location.
  
  
• Wireless location technology: Wireless location technology involves using signaling information from cell phones (i.e., cell tower handoffs) to anonymously track wireless devices as they traverse a roadway network. Cell tower handoffs occur when calls are transferred from one tower to another, during which the cellular phone is assigned a new frequency. By tracking the handoffs between towers, it is possible to determine which road a vehicle is traveling on and derive average travel time and speed information for that roadway segment.
  
  
• Crowd sourcing: Crowd sourcing involves obtaining real-time traffic congestion information from a driver’s Global Positioning System-enabled mobile phone. Crowd sourcing relies on a large number of users and is primarily supported by personal navigation systems and mobile applications.

CHALLENGES WITH DATA

Systematically collecting the critical data, verifying data quality, and documenting any assumptions are important to justify the results of a study to decisionmakers and the public. A statistical analysis of collected and previously available data can be helpful to determine the statistical data variability and the margin of error contained in the data. Data variability can have a significant effect on the number of model runs required to represent the model’s replication of observed traffic conditions. This report provides helpful suggestions on data types, sources, and challenges. Development of a more comprehensive data quality guide is under consideration by FHWA.

Data Comprehensiveness

Comprehensive data cover different performance measures (i.e., volumes, speeds, bottlenecks, queuing, and congestion data) across freeways and arterial streets, as well as transit data and incident data. Traffic counts should be taken at key locations in the study area. Key locations include major facilities (i.e., freeway segments, major intersections and interchanges, and major on- and off-ramps). If possible, this should be done simultaneously at all key locations. Otherwise, the counts should be taken during similar timeframes with similar demand patterns and weather conditions.

If a model that was calibrated several years ago is being used, it needs to be recalibrated to reflect more current field conditions. Therefore, not only would data be needed to estimate capacity and the calibration performance measures, but demand data need to be collected. Furthermore, the analyst must verify the accuracy of geometric data, traffic control data, traffic operations, and management data.

Challenges

Due to the innovative nature of many operational strategies, collecting relevant data to support such analyses includes the following challenges:

• Automated data sources are often best used for collecting long-term data necessary to develop and calibrate simulation models; however, many existing automated data collection systems lack the robustness or reliability to effectively compile relevant data sets. A thorough assessment of the data quality from all sources is recommended to identify any potential problems early on in the process and establish methods to address any deficiencies. Automated data availability can generally be classified in the following three categories:
  
  
  • Fully automated: Freeway and arterial data are available for the study area network.
    
    
  • Semi-automated: Only freeway mainline data are available.
    
    
  • Manual: There are no automated data for the study area network.
    
    
• Data quality from automated data sources (e.g., roadway loop detectors) may sometimes be insufficient for modeling purposes. Sample datasets should be obtained early in the data collection process and analyzed to assess data quality. The data collection plan should specify data quality procedures and minimum data quality requirements for this purpose. Further, the analysis manager should discuss any data quality issues with operations personnel familiar with the data source during the development of the data collection plan in order to understand and anticipate any problems with data source reliability, data accuracy, or other condition-specific issues (e.g., inaccurate speeds recorded during high-volume periods).
  
  
• Accuracy is defined as the measure of the degree of agreement between data values and a source assumed to be correct. It is also defined as a qualitative assessment of error. It is important to have accurate, internally consistent, and recent data. If information on future traffic conditions is not available, it could be helpful to study data from other regions that may have the bottlenecks and traffic patterns that are envisioned for the study area. This is critical since the goal of model calibration is to not only see if the model can represent observed conditions but also to examine if the model can handle future congestion. With respect to data filtering and fusion, it is crucial to adopt standard ways to accept or reject field data and to address data gaps and missing data.
  
  
• A small margin of error in the collected data is required to increase the validity of simulation results. It is necessary to collect multiple data points for each performance measure (i.e., volumes, speeds, etc.) so that the sample is an accurate representation of the mean and standard deviation of the performance measure. Depending on the mean to standard deviation ratio and the desired margin of error, the required sample size may vary greatly. Table 8 shows different sample sizes based on different mean to standard deviation ratios and different margins of error.

Table 8. Margin of error for different standard deviation to mean ratios.

DEVELOPMENT OF A DATA COLLECTION PLAN

This section describes a typical data collection plan for traffic analysis. The data collection plan will guide the collection, compilation, analysis, and archiving of data.

Approach and Work Steps

This section presents an overview of the subtasks and work steps related to the development and implementation of a data collection plan. The data collection plan should retain sufficient flexibility so that lessons learned in the compilation of data sources may be incorporated as part of the continuous improvement of the analysis effort.

Specific steps in the development of the data collection plan include the following:

1. Research and identify available data for the study area: Existing data sources and data requirements should be used to identify available data for the analysis area. The data collection plan should also identify those individuals/‌stakeholders responsible for compiling the data. The analysis manager should work closely with stakeholders to compile the data. If possible, the analysis manager should obtain samples of the datasets prior to full collection to view the content and format of the data and adjust collection plans as necessary.
   
   
2. Identify information/data gaps and recommend an approach to filling those gaps: Once available data sources have been investigated and dataset samples have been reviewed, the analysis manager should assess the appropriateness of the available data for use in the analysis and identify any critical gaps in data availability. Potential approaches to filling data gaps should be investigated, and recommended approaches should be documented in the data collection plan.
   
   
3. Identify data management strategies: Procedures for conducting data quality control and data archiving should be identified. Any required thresholds for minimum data quality should be identified as well as high-level descriptions of processes for addressing data shortcomings. Plans for archiving the data should also be identified. Responsibilities for data quality testing and data archiving should be clearly defined.
   
   
4. Develop data collection plan: The data collection plan should document all of the information listed in these steps and detail data elements to be obtained and their respective data sources. The data collection plan should outline data collection methodologies and contain budget and schedule estimates to fill data gaps.

An example outline for the data collection plan is as follows:

• Introduction and background.
  
  
• Data requirements.
  
  
• Available data.
  
  
• Data gaps.
  
  
• Data collection methodology.
  
  
• Data collection locations.
  
  
• Data collection budget and schedule.

Once the data collection plan is developed, the required data should be collected in accordance with the plan. Generally, implementing the plan includes the following activities:

• Assembling/collecting data on physical infrastructure and geometrics. Much of these data are likely to be available in existing models and regional GIS.
  
  
• Assembling/collecting existing traffic performance data within the study area.
  
  
• Gathering available information from existing studies. These studies include those currently underway as well as those that have been recently completed. Example studies include, but are not limited to, existing conditions analyses, environmental impact studies, and lists of projects and strategies that have been planned or programmed.
  
  
• Conducting field reviews within the study area.
  
  
• Collecting new data as specified in the data collection plan. All data collected in this effort should be analyzed and archived according to the data management procedures documented in the data collection plan. Any identified problems with data quality or the successful archiving of data should be immediately communicated to the analysis managers as part of the continual improvement process.

Deliverables

Major deliverables under this task include the data collection plan and the archived datasets.

Schedule

The time required to complete this task is dependent on the types, quantity, and quality of data required; the data collection methods; and the amount of readily available archived data from automated sources. The schedule for developing the data collection plan is estimated to be approximately 2–4 months. Completing the collection of data is extremely variable and may take approximately an additional 2–6 months depending on the data required.