Skip to contentUnited States Department of Transportation - Federal Highway Administration FHWA Home
Research Home
Report
This report is an archived publication and may contain dated technical, contact, and link information
Publication Number: FHWA-RD-98-166
Date: July 1999

Guidebook on Methods to Estimate Non-Motorized Travel: Supporting Documentation

2.4 Pedestrian Sketch Plan Methods

Demand Estimation

Descriptive Criteria: What is It?

Categories:

Empty Box Bicycle Box with an x inside Pedestrian Box with an x inside Facility-Level Empty Box Area-Level

Authors and Development Dates:

Pushkarev and Zupan (1971); Behnam and Patel (1977); Davis, King, and Robertson (1991); Matlick (1996); Ercolano, Olson, and Spring (1997)

Purpose:

A variety of pedestrian sketch-plan methods have been developed to estimate pedestrian volumes under existing and future conditions in a pedestrian activity area. These methods generally use pedestrian counts and regression analysis to predict pedestrian volumes as a function of adjacent land uses (e.g., square feet of office or retail space) and/or indicators of transportation trip generation (parking capacity, transit volumes, traffic movements, etc.). Alternatively, data on surrounding population and employment may be combined with assumed trip generation and mode split rates to estimate levels of pedestrian traffic.

These methods can be used to identify areas of high-pedestrian traffic based on existing land use data, thereby eliminating the need to conduct pedestrian counts on all facilities. They can also be used to forecast changes in pedestrian volumes as a result of future land use or transportation trip generation changes.

Pushkarev and Zupan (1971) and Behnam and Patel (1977) forecast pedestrian volumes in high-density urban areas based on existing land use characteristics and pedestrian volumes for specific locations. Similar studies were performed in other areas in the 1960s and 1970s for the purposes of developing pedestrian demand models (see entry for "Pedestrian Demand Models," Method 2.11)

Ercolano, Olson, and Spring (1997) use existing data routinely collected by most transportation providers (at a minimum, vehicles per hour from traffic counts and local mode shares from the census) to estimate peak pedestrian travel demand in suburban and developing rural activity centers. This sketch-planning method has been applied to help determine the location of pedestrian facility improvements such as pedestrian crossings, sidewalks, and signal retimings.

Matlick (1996) used household population, national transportation survey percentages, and activity center data to calculate potential walking trips in specific corridors. It is a quick method or tool to be used by planners to identify the priority areas for pedestrian facility expenditures.

Davis, King, and Robertson (1991) describe a method to measure and predict pedestrian crosswalk volumes for the evaluation of traffic signal requests and for the compilation of hazard indices. This method of only using short-term counts of 5 to 10 minutes is more cost effective than continuous counts. While this technique does not actually forecast demand, the issues discussed are relevant to the collection of pedestrian data for the other methods described here.

Structure:

Pushkarev/Zupan: Pedestrian volumes were determined on midtown Manhattan surface streets at various times of day using aerial photography. Regression analysis was then used to predict total pedestrian volumes per block. Independent variables included adjacent land uses (square feet of office, retail, and restaurant), distance to transit entrances, and sidewalk and plaza space per block. (A Manhattan-specific factor, whether the walkway was on a street or avenue, was also included.) Flow characteristics by time of day, traffic characteristics, and trip generation characteristics of specific types of buildings were also analyzed.

Behnam/Patel: Similar to Pushkarev/Zupan, regression models were used to estimate the noon-hour and average pedestrian volumes per hour, based on land use data. Behnam/Patel included eight types of land uses as the independent variables (see "Inputs/Data Needs.") Pedestrian volume per hour per block is the dependent variable of the regression. Mid-block sidewalk counts were used to determine pedestrian volumes for estimating the model. Future volumes were then predicted based on forecasts of future land use. Behnam/Patel applied this technique to the Milwaukee central business district.

Ercolano/Olson/Spring: Pedestrian per hour (PPH) values are derived from peak vehicles per hour (VPH) data, transit vehicle/ridership, and non-motorized mode-share estimates. A real- world application is described for a shopping area in Plattsburgh, NY. The steps taken are as follows:

1. Estimate the sources of the pedestrian trips (car/walk-linked trips, walk/bike-only trips, and transit/walk-linked trips). Regarding the trips originating from vehicles, all through traffic including limited-access highway ramp traffic should be eliminated from the analysis (approximately 70 percent of the peak traffic was eliminated in the NY case). A portion of the remaining VPH trips (i.e., turn movements) also should be eliminated because they are assumed to be drive through, truck or drop off/pick up trips (about 20 percent of the trips were eliminated in the NY model case study). For urban areas with fewer than 50,000 residents in the region, walk/bike only and transit trips were considered as part of the remaining peak VPH turning movements.

2. Estimate the average peak pedestrian per hour (PPH) trip rates per person. This is done by combining the vehicle-trip estimates from step 1 with an assumed average vehicle occupancy rate and number of walk trips per hour (assumed to be 1.5 persons/vehicle and five walk trips per person per hour in the Plattsburgh, NY example).

Steps 1 and 2 can be summarized as follows:

Peak PPH = (Peak VPH - Through Movement Trips) =
(VPH Turning Movements x 1.5 Default Average Vehicle Occupancy x 5 Trips Per Person - 20 percent Drive-Through).

3. Distribute and assign PPH trips. Pedestrian trips are categorized into three groups: internal, external, and extended. Internal trips occur within a traffic analysis zone (TAZ); external trips may begin or end in a different zone; and extended trips are longer trips through several zones. To avoid double counting, extended walk-trips need to be weighted per zone using projected peak-hour VPH turning volumes. The adjusted PPH is used to calculate walk trip volumes by season.

Once the PPH trip estimates are assigned and distributed, it is possible to recommend proposed intersection and midblock improvements. When the average hourly pedestrian and vehicle volumes reach a certain level, it is recommended to install crosswalks, pedestrian signals, or refuge islands/medians. For example, the minimum for a crosswalk installation is 200-300 VPH and 25 PPH. When children, elderly or disabled pedestrians are the majority, the minimum is reduced to 100 VPH and 10 PPH.

Matlick: The method uses household population and national transportation survey percentages to calculate potential walking trips. The steps for a corridor-level analysis are as follows:

1. To represent the majority of pedestrian trips, identify a 0.8-km buffer around the selected corridor. GIS possess tools that enable planners to create buffers.

2. Identify traffic generators such as the number of housing units by dwelling type, average persons per unit for each dwelling type, and the average number of trips per person from these locations.

Total Corridor Generated Trips (TCGT) = Population x Trips Per Person

Potential Pedestrian Trips (PPT) = TCGT x (Total All Trips < 0.8 km)

Est. Primary Pedestrian Trips = PPT x (Percent Known Walking Trips < 0.8 km)

OR

Population x Pedestrian Mode Split for the area (if available)

3. Identify traffic attractors such as retail, recreational, social facilities, schools, transit stations, and churches. Since most pedestrian trips are less than 0.4 km, it is important that traffic generators and attractors are in close proximity. Areas with high levels of attractors are likely to have a higher potential for pedestrian activity. When associated with nearby traffic generators, optimal conditions for pedestrians exist.

4. Locate transit, school, and park and ride data to validate the estimated pedestrian trip numbers (refer to "Calibration/Validation Approach" for more details).

Davis/King/Robertson: The technique only requires short-term vehicle counts of 5-, 10-, 15- or 30 minutes over a 1- to 4-hour period. Pedestrian counts that are recorded in the middle of the hour are shown to have greater accuracy as opposed to counts at the beginning or end of an hour. Furthermore, short-term counts taken over 4 hours are more accurate than counts taken over 1 to 3 hours. The method gives detailed instructions for designing a data collection experiment, including (1) selecting the type of application; (2) selecting the count interval; (3) collecting the data; and (4) computing estimated volumes.

Calibration/Validation Approach:

Pushkarev/Zupan and Behnam/Patel: Sidewalk pedestrian counts in the areas analyzed (Manhattan and Milwaukee CBD) were used to develop the quantitative models. If applied to other cities or areas, the models could be re-estimated based on pedestrian counts from the specific area.

Ercolano/Olson/Spring: The results of the case study were cross-referenced with a land-use-based study for the same area done in 1978 by Kagan, Scott, and Avin. Kagan et al. used counts from 215 city sites to develop trip-rate averages by land use type. Predicted volumes were in relatively close agreement, with values from Ercolano, et al.'s access-egress method being 7, 24, and 29 percent lower in the three applicable zones than based on the earlier land use-based method.

Matlick: To ensure the accuracy of the estimated primary pedestrian trips calculated in step 2, compare this number to transit ridership and the number of non-bussed schoolchildren. The number should be about the same as the number obtained in step 2.

Davis/King/Robertson: Four out of the 12 sites were used to validate the expansion model. These sites were located in the same city yet their volume distribution patterns differed. The percent difference between the actual and predicted counts ranges between 11.9 percent and 34.5 percent.

Inputs/Data Needs:

Pushkarev/Zupan: For each city block, required data include:

  • Square m of office, retail, and restaurant space;
  • Square m of sidewalk and plaza space; and
  • Distance to nearest transit station.

Behnam/Patel: Requirements include pedestrian volumes at the four corners of a block and the following land use data:

  • Commercial space;
  • Office space;
  • Cultural and entertainment space;
  • Manufacturing space;
  • Residential space;
  • Parking space;
  • Vacant space; and
  • Storage and maintenance space.

Ercolano/Olson/Spring: The method uses walk-trip counts; if this information is not available, the following data sources could be used:

  • Peak vehicle-per-hour (VPH) turning movements;
  • Transit ridership;
  • Walk/bike only mode shares (based on the U.S. Census);
  • Zoning or land use map;
  • Square meters or feet of new development space; and
  • Aerial photographs and/or specific site, corridor, or subarea block configurations.

Matlick: Desired data include:

  • Land uses;
  • Maps;
  • Transportation mode split information;
  • Generator information: Housing types, density, persons per household unit, and hotels;
  • Attractor information: retail, recreation, social facilities, schools, employment, and churches;
  • Daily transit ridership information;
  • Local school information: number of enrolled children, percentage of bussed and non-bussed students; and
  • Park and ride lot information: lots, size, and occupancy rates.

Davis/King/Robertson: The method requires pedestrian counts over a 1- to 4-hour period for 5-, 10-, 15-, or 30-minute time segments. For traffic signal requests, the analysis requires data from the peak hour. The count data can be collected manually, as suggested by the author, or using new advanced traffic sensors as they become more commonplace.

Potential Data Sources:

Pushkarev/Zupan and Behnam/Patel: Local land use data bases.

Ercolano/Olson/Spring: Vehicle traffic counts, zoning/land use maps, other site or area maps.

Matlick: Traffic Analysis Zones (TAZ), U.S. Census block tracts, regional socioeconomic, and transportation data.

Davis/King/Robertson: Not applicable.

Computational Requirements:

Pushkarev/Zupan and Behnam/Patel: Regression analysis is required.

Ercolano/Olson/Spring and Matlick: The computations can be done using spreadsheets.

Davis/King/Robertson: The methods uses simple equations.

User Skill/Knowledge:

Pushkarev/Zupan and Behnam/Patel: The user should be familiar with localized land use and transportation data and with techniques of regression analysis and traffic counting methods.

Matlick: The user should be familiar with localized land use, socioeconomic, and transportation data.

Ercolano/Olson/Spring: The user should have some knowledge of general modeling assumptions and methods as well as knowledge of the specific site, corridor, or subarea.

Davis/King/Robertson: The user should be familiar with survey and traffic counting methods.

Assumptions:

Pushkarev/Zupan and Behnam/Patel: It is assumed that the land use variables included can adequately predict pedestrian volumes. Other factors that may affect pedestrian trip generation rates, such as pedestrian environment quality, are not analyzed.

Ercolano/Olson/Spring: For urban areas with fewer than 50,000 residents in the region, walk/bike only and transit trips are assumed to be part of peak-vehicle turn movements that are used in the study. For urban areas with regional populations that exceed 50,000, the analyses would have to be separate for pedestrian trips by car, walking/biking only, and transit modes.

Matlick: Using the Nationwide Personal Transportation Survey (NPTS), the method utilizes national travel data when regional- or corridor-level data does not exist.

Facility Design Factors:

Pushkarev/Zupan: Considers sidewalk width and/or total sidewalk and plaza area.

Behnam/Patel: The method does not consider the impact of facility design factors.

Ercolano/Olson/Spring: The method accounts for the following different levels of pedestrian facility designs: non-existent, partial, and complete.

Matlick: The method does not consider the quality of pedestrian facilities.

Davis/King/Robertson: The method considers only existing pedestrian volumes.

Output Types:

Pushkarev/Zupan and Behnam/Patel: Existing pedestrian volumes can be predicted based on land uses if pedestrian counts are not available, and future pedestrian volumes can be predicted as a function of future land uses.

Ercolano/Olson/Spring: The estimated intersection crossing data are categorized according to zone, level of pedestrian facility completeness, and season/climatic condition.

Matlick: The output consists of two estimates, one for traffic generators and the other for attractors. The generator estimate states the number of primary potential pedestrian trips in the corridor while the attractor estimate reveals the number of customers, employees and students in a given area. Planners can use this data when comparing corridors for future pedestrian facility improvement projects.

Davis/King/Robertson: The output of this method is an expanded pedestrian volume for a period from 1 to 4 hours, depending on the number of hours used in the counting procedure.

Pedestrians in an urban area
Figure 2.4: Data on surrounding population and employment may be combined with assumed trip generation and mode split rates to estimate levels of pedestrian traffic.

Real-World Examples:

Pushkarev/Zupan: Models of pedestrian traffic were developed for midtown Manhattan.

Behnam/Patel: The case study for the report was the CBD of Milwaukee.

Ercolano/Olson/Spring: The case study was a suburban growth corridor in Plattsburgh, New York. Some of the specific findings are shown in the above "Structure" section.

Matlick: The case study was a suburban roadway corridor in Seattle, Washington. Results are described under "Performance."

Davis/King/Robertson: The case study was developed from data collected in Washington, DC, which involved over 18,000 5-minute pedestrian count intervals.

Contacts/Source:

Scott Davis: Analysis Group, Inc., 500 E. Morehead Street, Suite 315, Charlotte, NC, 28202.

L. Ellis King and H. Douglas Robertson: Civil Engineering Department, University of North Carolina at Charlotte, Charlotte, NC, 28223.

James Ercolano, Jeffrey Olson, and Douglas Spring: New York State Department of Transportation (Albany, NY)

Jeff Zupan, Regional Plan Association of New York (New York, NY)

Publications:

Behnam, Jahanbakhsh and Bharat G. Patel, A Method for Estimating Pedestrian Volume in a Central Business District, Pedestrian Controls, Bicycle Facilities, Driver Research, and System Safety, Transportation Research Record 629, Washington, DC, 1977.

Davis, Scott E., L. Ellis King and H. Douglas Robertson, Predicting Pedestrian Crosswalk Volumes, Transportation Research Record 1168, Washington, DC, 1991.

Ercolano, James M., Jeffrey S. Olson, Douglas M. Spring, Sketch-Plan Method for Estimating Pedestrian Traffic for Central Business Districts and Suburban Growth Corridors, Transportation Research Record 1578, Washington, DC, 1997.

Matlick, Julie Mercer. If We Build it, Will They Come? #69 Forecasting Pedestrian Use and Flows, Forecasting the Future, Bicycle Federation of America -- Pedestrian Federation of America, Pro Bike/Pro Walk '96, 1996, pp. 315-319.

Pushkarev, Boris and Jeffrey M. Zupan. Pedestrian Travel Demand. Highway Research Record 355, 1971.

Evaluative Criteria: How Does It Work?

Performance:

Pushkarev/Zupan: Similar comments apply as for Behnam/Patel, but the analysis is specific to Midtown Manhattan, and more expensive aerial photography data collection techniques are used.

Behnam/Patel: The method works well for high-density urban areas but has not been applied to low-density areas. The data collection process is not labor intensive and requires only standard information making the method cost-effective. The process takes into consideration the geographical distribution of pedestrians yet is best used at the central business district or facility level, not at a city level.

Ercolano/Olson/Spring: Pedestrian volumes predicted from this sketch planning method compared reasonably well with those predicted based on the trip generation of adjacent land uses. Moudon (see TRR 1578) also provided evidence that "completeness of pedestrian facilities, etc." supports more pedestrian travel and influences mode share.

Matlick: The traffic generator estimate equals 1,378 primary pedestrian trips in the corridor; the validation for the traffic generator estimate is reasonable at 1,133. The attractor information consists of 500 students who attend class on a daily basis at the local college (branch) campus, 1,200 weekday customers at the grocery store, and 3,169 daily transactions at one of the fast food establishments.

Davis/King/Robertson: The method works well for the city in which the study was conducted. Further research needs to be done on its accuracy in other cities. Additional research also would improve the multi-hour estimates since a lower confidence was used for these hourly counts. The expansion model provides an easy and cost-effective method to estimate pedestrian volumes over a 1- to 4-hour period.

Use of Existing Resources:

Pushkarev/Zupan and Behnam/Patel: The method uses land use data that can be obtained from the planning departments of any major city. Pedestrian counts also are required.

Ercolano/Olson/Spring: The method uses vehicle data that is routinely collected at the local level. The vehicle data can be substituted for more specific pedestrian-traffic count data if available. The method also provides a basis for more refined modeling for pedestrian accommodations during the project design to implementation phases.

Matlick: The method uses basic population data along with national transportation trip survey percentages that can be substituted for more site-specific transportation data.

Davis/King/Robertson: Any pedestrian counts using 5-, 10-, 15-, or 30-minute intervals over a 1-, 2-, 3-, or 4-hour period could be used. Manual counts were used in this example yet more high-tech means also could be used such as infrared and videotaping systems when these technologies progress.

Travel Demand Model Integration:

Pushkarev/Zupan and Behnam/Patel: Their methods were meant to assist with facility-level planning not city-wide analysis. The trip generation relationships could be used as inputs to local pedestrian travel demand models, if such models were developed.

Davis/King/Robertson, Ercolano/Olson/Spring, and Matlick: Their methods were not designed for model integration.

Applicability to Diverse Conditions:

Pushkarev/Zupan and Behnam/Patel: The general technique is probably most applicable to high-density CBD areas. The specific models developed are probably applicable only to the city/area in which they were developed.

Ercolano/Olson/Spring: The authors believe that the method can be applied to other areas, and site-specific data can be substituted for default inputs. The method is able to adjust for seasonal variations and for different infrastructure scenarios. The different infrastructure scenarios range from complete (i.e., ADA-compliant sidewalks, medians/refuge islands or pedestrian-oriented crossings) to partial (i.e., limited facility amenities) to non-existent.

Matlick: The method uses national transportation data although site-specific data can be substituted when available.

Davis/King/Robertson: The model is able to accommodate for different sampling procedures such as surveys for different time allotments (i.e., 5-, 10-, 15- and 30-minutes), and for different time periods (i.e., 1 to 4 hours).

Usage in Decision-Making:

Behnam/Patel and Ercolano/Olson/Spring: The methods were developed to help determine the location of pedestrian facility improvements such as pedestrian crossings, sidewalks, and signal retimings.

Matlick: The method was developed as a tool to help planners compare potential corridor-level pedestrian activity.

Davis/King/Robertson: The expansion model was developed to provide planners with a cost-effective method for measuring existing pedestrian volumes for the evaluation of traffic signal warrants and for the establishment of hazard indices.

Ability to Incorporate Changes:

Behnam/Patel: The inputs can be easily updated.

Ercolano/Olson/Spring and Matlick: The vehicle traffic counts and national transportation survey data that are used as the default can be substituted for actual field data.

Davis/King/Robertson: The counts can easily be taken again since the time increments only amount to between 5 and 30 minutes.

Ease-of-Use:

Behnam/Patel, Ercolano/Olson/Spring, and Matlick: The methods are easy to understand since they use basic transportation data as inputs and can be manipulated using spreadsheets.

Davis/King/Robertson: The method is easy to understand and inexpensive to implement.

Comments:

Ercolano/Olson/Spring: The primary purpose of this research was to develop a quick sketch plan method to ensure consideration of pedestrian access and safety during project scoping/initiation. Completeness of pedestrian facilities was also viewed as an important factor in supporting more pedestrian travel and influencing mode share, as evidenced in Moudon, et al. (1997).

Matlick: Matlick's study uses the 1990 NPTS data. The 1995 NPTS is now available and provides more detail on personal travel. The use of the data remains limited by the size of the sample for non-motorized trips (6.3 percent).

 

FHWA-RD-98-166

Previous | Table of Contents | Next

ResearchFHWA
FHWA
United States Department of Transportation - Federal Highway Administration