I

r
Bicycle System Performance
Measures
BY BRUCE W. LANDIS

day’s trend toward quantitative

analysis puts more pressure than
travel demand, a measure of the rela-
tive amount of bicycle travel that would
factor in bicyclists’ route choice, as well
as a significant factor in whether a bicy-
ever on transportation planners to pro- occur on a road segment if there were cle trip is even initiated} It thus repre-
vide project planning “by the num- no bicycle travel inhibitions caused by sents a measure of facility supply as
bers.” Increasingly, project prioritiza- motor vehicle traffic. It employs a sim- opposed to a crash predictor. Just as
tion for recurring Transportation plified, probabilistic gravity model travel speed and time are level-of-service
Improvement Plans (TIPs), Capital technique to quantify the proximity and measures for motor vehicle facilities, the
Improvement Plans (CIPS), and metro- magnitude of bicycle trip attractors perception of hazard is a level-of-service
politan area Long Range Plans andlor generators. Applied together on measure for bicycle facilities.
(LRTPs) requires a numerical basis to a segment-by-segment basis, the mod- Two recent efforts attempted to sim-
demonstrate that all projects can reach els provide a clear indication of which ulate the perception of hazard: the
measurable objectives. This require- roads should be candidates for bicycle Broward County, Florida, Roadway
ment, coupled with the comparative facilities. Many U.S. metropolitan Condition Index (RCI)2 and the Florida
infancy of bicycle system planning, planning organizations (MPOS) are cur- Bicycle Coordinators’ consensus-
poses a formidable challenge to trans- rently using these models. developed Segment Condition Index
portation planners, who may be faced (SCI) Model. While these modeling
efforts represented a major advance-
with having to provide numerical justi- The IHS Model
ment in quantifying the level of service
fication for a bicycle facility. However,
There has been advancement in of on-road bicycle facilities, their
two recently developed analytical tools,
recent years in modeling the actual haz- methodology used to assign values to
the Interaction Hazard Score and ard of road segments for bicycle use as variables was flawed, particularly in
Latent Demand Score models, enable well as in estimating the perceived haz- estimating pavement and location (or
planners to quantitatively evaluate, test ard to bicyclists of sharing the right-of- adjoining land access intensity). Conse-
and prioritize on-road bicycle projects. way with motorized vehicles. The first quently, if the same person was not
The Interaction Hazard Score (IHS) of these modeling efforts, known as the
and Latent Demand Score (LDS) mod- Davis Model: sought to predict bicycle
els incorporate the two classic factors crashes using variables such as average
used in evaluating transportation facility annual daily traffic, number of travel Bnlcew Lun@
performance—supply and demand. The lanes, pavement and location factors. It P. E., AICP, is
IHS Model provides a supply-side mea- did not meet its goal of predicting Vice President of
sure of the on-road bicycling environ- crashes2 because of its failure to Planning, Trans-
ment. Using existing roadway and traf- 1) incorporate actual bicycle volumes portation and
fic variables, it estimates from the as a normalizing or exposure variable Traffic Engineer-
bicyclists’ perspective the perceived and 2) consider operator error, the ing at Sprinkle
hazard of bicycle/motor vehicle interac- major determinant in crashes. Howev- Consulting Engineers Inc. in Tampa,
tion. Its results are often translated into er, the Davis Model stimulated other Fla. He received a bachelor’s in civil
level-of-service categories for bicycle modeling efforts to estimate bicyclists’ engineering from the Georgia Institute
travel mode planning. perception of the hazard of sharing the of Technology, and a master’s in civil
The LDS Model, on the other hand, roadway with motor vehicles. engineering from the University of
provides an estimate of latent bicycle This perception of hazard is a major South Florida. He is a Member of ITE.

18 ● ITE JOIJRNALO FEBRUARY 1996

used year after year to assess roadway pavement, the more the bicyclist’s
conditions, substantially different attention is diverted to the immediate
model outputs could occur, distorting travel line. Thus, the perceived hazard
project prioritization and other repeti- of sharing the right-of-way is magnified.
tive applications. The IHS Model was In the transverse environment,
developed to avoid these shortcomings. uncontrolled vehicular movement
(adjoining land access activity and on-
Theoretical Basis
street parking maneuvers) presents a
The IHS’S theoretical basis includes “turbulence” or conflict hazard to the
portions of earlier models. The IHS bicyclist sharing the right-of-way.
Model was developed primarily to Indicative roadway features include dri-
remove the subjective aspects of the veways and on-street parking spaces.
RCI and SCI models. The model data
required is standardized and collected
objectively and economically. The IHS
Model quantifies the bicyclist’s percep-
tion of the hazard level of interacting
with motor vehicles. It accounts for the
longitudinal and transverse interactions
within the on-road bicycling environ-
ment.
In the longitudinal roadway environ-
ment, there are several factors that
affect the bicyclist’s perception of inter-
action hazard. They are:
■ Volume of motor vehicle traffic.
■ Speed of the motor vehicle traffic.
■ Traffic characteristics.4
■ Proximity of the bicyclist to motor NEW DENOMINATOR
vehicle traffic. TRAFFIC COUNTERS AND TALLY BOARDS
■ Pavement condition.
Featuring below-the~ounter positioning
As the volume, speed and size of the
of keys, affording the lightest actuating
motor vehicle traffic in the outside travel pressure.
lane increases, so does the bicyclist’s
. Ruggedlzad yet Iightwalght
perception of the hazard of interacting.
cLscqueredmaeonitetmerds-tailor
The narrower the outermost roadway mede for your counting naede
lane, the closer together are the bicycle ● Work h all kinds of weether
and motor vehicle travel paths (see Fig- .. ● Absolute accuracy
Mcst popular beam.
ures 1 and 2). Close proximity of the 16 units on one foot square ● Write for brochureandprice list

bicyclist to motor vehicles results in a lacquered masonite board

high perception of interaction hazard.
Sawing thosewhocountsince 1914
Pavement condition also has some
THE DENOMINATOR CO., INC., WOODBURY, CT 06798
effect on the bicyclist’s perception of
(203) 263=3210
hazard. The more deteriorated the

ITE JOURNAL* FEBRUARY 1996 ● 19

 (ADT) ~ (14~
IHS=— — x [al ~ x (1 + %HV)2
{ L w

+ a2PF] + a,LU X CCF X ~ (1)

}
Where:
ADT = Average daily traffic (ADT data is used as input due to its
widespread availability the calibrated model converts this
volume to an equivalent hourly lane volume)
L = Total number of through lanes
w = Usable width of outside through lane (includes width of any
bike lanes, measured from pavement edge, or gutter pan, to
center of road, yellow stripe or lane line, whichever is less)
LU = Intensity (multiplier) of the land use adjoining the road
segment (simplified to a commercial land use trip genera-
the bicyclist’s assessment of hazard. tion of 15 or a noncommercial rate of 1)
CCF = Uncontrolled vehicular access (such as driveways or on-
30 , street parking spaces) frequency per mile
PF = Pavement factor (the reciprocal of the Federal Highway
Administration’s Highway Performance Monitoring
25 System’s [HPMS] “PAVECON FactoP)
s = Speed limit (a surrogate for average running speed)
HV = Presence of heavy vehicles (such as trucks) expressed in
20 decimal form
al–a3 = Calibration coefficients.
The raw score can be converted to level-of-service cate-
gories A, B, C, D, E and F (See Figures 4 and 5). While at
. present the base model does not include the effect of limited
10
horizontal and vertical sight distance, it has widespread appli-
cation for the vast majority of the collector and arterial roads
5 and streets comprising U.S. metropolitan planning networks.

Equation Adjustment,
0
Sensitivity Analysis and Applications
0 2 4 6 8101214161620222426263032 34363640
Two nonstatistical adjustments were performed during the
Figure 4. Typical lHS Model scores hktogram. initial development of the IHS Model: an interterm and an
overall equation. Based upon consensus group meetings as
well as interviews with bicyclists representing the entire spec-
60
IHS=6-6
trum6 of bicyclists, the calibration coefficients al, az and as

1
70 were adjusted to 0.01, 0.01 and 0.02, respectively. This estab-
IHS=3-5
lished the final speed, pavement condition and transverse
60 terms as comprising 79 percent, 13 percent and 8 percent of
IHS=9-I 1 IHS=12-18 the equation’s value, respectively.
50
Second, a sensitivity analysis of the entire equation was
40 conducted for changes in its variables using the consensus
group and interview technique. This adjustment process also
30 confirmed the values of the equation coefficients al, a2 and a3.
Table 1 shows various traffic and roadway conditions, the cor-
20
responding IHS model value and the percentage change.
10 The IHS Model is used as a planning tool in numerous metro-
politan areas throughout the United States. The Birmingham
0 Regional Planning Commission, several central Florida MPOS
c
I
A B D E F
and the Charlotte-Mecklenburg MPO are among those using the
Level of Service Category IHS level-of-service model in their comprehensive plan and TIP
I I
Figure 5. Typical histogram of bicycle level-of-service cate- processes. The Tampa area MPO uses the model regularly to pri-
gories. oritize its candidate road segments with similar bicycle travel
demands for resurfacing, enhancement and widening projects.
ception of hazard occurs. Magnifying this perception is the For example, Columbus Drive, a minor arterial in the City of
volume of driveway traffic (see Figure 3) and the turnover Tarnpa slated for eventual widening throughout its entire length,
rate of on-street parking. has seven segments. Due to the surrounding demographics and
land uses (see discussion of the LDS Model in the next section),
Model Terms virtually all its segments had similar bicycle travel demands. The
Based upon these bicycle/motor vehicle interaction factors, IHS Model was used for a supply-side, or level-of-service, evahr-
the IHS Model has been developed with the general form: See BICYCLE, page 25

20 “ ITE JOURNAL” FEBRUARY 1996

BICYCLE, from page 20 Table 1. IHS Model Sensitivity Analysis
ation to identify which segments had the
highest priority for bicycle facility construc- IHS ={(lQI) x (.14)2x (a, S x (1+%HV)2 + a,PF) + a,LUxCCF) x 1
tion due to their poor provision for bicycle LW 30 10
travel. A combination of narrow travel lanes
Baseline inputs:
with high traffic volumes and curb cut fre-
quency placed the North Boulevard to ADT = 15,0CKI vpd %HV= O
Nebraska Avenue segment at the top of the L . 2 lanes PF = 0.25 (good condition pavement)
priority list for the Columbus Drive widen- w= 12ft LU = 15 (commercial area)
ing program. s= 45 mph CCF = 42 per mile
and calibration coefficients a, = a2= a,= 0,01, 0.01 and 0,024 respectively

The LDS Model M % Cha ae

Baseline Interaction Hazard Score (IHS) 19,3 NJ;
Unlike travel demand planning for
motor vehicles, which uses a four-step Lane Width Modifications to Segment
(Urban Transportation Planning System
(UTPS)-type) sequential gravity model, W=ll (substandard) 22.7 18% increase
W = 12 (standard) 19.3 no change
bicycle travel demand planning has largely
W = 14 (wide outside lane) 14.5 24% reduction
relied upon vintage 1930s desire line plan- W = 16 (dedicated bike lane) 11.5 40% reduction
ning methods. One such method developed
in the late 1970s, the Area Corridor Analy- Speed Control Measures
sis Technique,’ utilized this approach.
S=55 22.7 18% increase
However, the major drawbacks to this and
S = 45 (baseline value) 19.3 no change
similar techniques are a heavy reliance on S=40 17.6 9% reduction
the judgments of the analyst and—if the S=30 14.2 26% reduction
techniques are to be used for repetitive
planning applications such as CIPS, TIPs Traffic Volume Reduction Measures
and LRTPs-a need for the same person to
20,003 25.3 31% increase
repeat the analysis so that updates have 15,(XXI (typical LOS D volume) 19.3 no change
consistent results. Also, the techniques pro- 10,002 13.4 31% decrease
vide no quantification of travel demand, a 1,OCIJ (typical collector threshold) 3.8 80% decrease
serious drawback many planners encounter
especially when pressed for numerical justi- Pavement Surface Conditions
fication during prioritization of projects. PF = 1.0 (PAVECON = 1.0 very poor) 27,0 40% increase
The LDS Model was developed to pro- PF = 0.5 (PAVECON = 2,0 poor) 21.9 13% increase
vide transportation planners with the ability PF = 0.25 (PAVECON = 4.0 good condition) 19.3 no change
to quantify facility demand. The LDS PF = 0.20 (PAVECON = 5,0 new) 18.8 3% decrease
Model is a probabilistic gravity model. It dif-
fers from the classic four-step travel demand Access (curb cut or on-street parking) Management
model in the following way: Where the CCF= 220 (Continuous Parallel Parking) 25.7 33% increase
UTPS gravity model requires extensive net- CCF= 100 (Typical CBD Condition) 21.3 11% increase
work coding and algorithms to simulate CCF= 42 (Spacing = 125 ft) 19.3 no change
travel between its trip generators and attrac- CCF = 4 (Spacing = 1320 ft) 17.9 7% decrease
tors, the LDS Model quickly estimates the
probability of bicycle travel on individual Truck Route Control
road or street segments based upon the %HV= 10% (truck route-typical) 22.6 17% increase
proximity, frequency and magnitude of %HV= 5% (typical of arterial) 21.0 9% increase
adjacent bicycle trip generators and/or %HV= 2% (typical of collector) 20.0 4%increase
attractors. The steps of the LDS Model are: %HV= O% (typical of local road) 19.3 no change
1. Establish tripmaking thresholds of the
bicycle trip attractors/generators. These
include the home-based work markets,* 2. Geocode andlor map the attractorlgenerators and tabulate
home-based shopping indicators, public parks (stratified into their number for each segment, stratifying according to prox-
minor, staffed and major) and public schools. imity using GIS or CAD software (see Figure 6).
3. Compute the tripmaking probability summation for each net-
* Census tracts with households that have a high level of home-based work segment using the LDS equation (see Figure 7) and devel-
work trips with durations of less than 10 minutes as reported in the oped trip probability vs. distance curves for each trip purpose
1990 Census. As Question 24b of the Long Form asks the travel time share.’
(in automobile travel minutes) between home and work, the relative
The LDS Model uses readily available demographic data,
(among the component block groups) number of respondents with a
travel less than 10 minutes (for example, five miles at 30 mph vehicle employing simple manual geocoding (or GIS-based spatial
travel time) provides an indication of residential areas with a poten- query functions) for spreadsheet-based gravity computations. It
tial for a shift to the bicycle mode. is important to note that the LDS Model estimates, using a

ITE JOURNAL” FEBRUARY 1996 ● 25

 sought for statistical calibration of the
models,” transportation planners
throughout the United States already are
using the consensuslfocus group calibrat-
ed models in the planning and prioriti-
zingof their bicycle systems and facilities.

References
1. Davis, Jeff. Bicycle Safety Evaluation.
Auburn University, 1987.
2. Epperson, Bruce. “Evaluating Suitability
of Roadways for Bicycle Use: Toward a
Cycling Level-of-Service Standard. ”
Transportation Research Record 1438.
Washington, D. C.: TRB, National
Research Council, 1994.
3. Highway Safety Research Center, Univer-
sity of North Carolina and HDR Engi-
neering Inc.. National Bicyclingand Walk-
ing Study: Final Report. FHWA, U.S.
Department of Transportation, 1991.
4. Sorton, Alex. “Bicycle Stress Level as a
Tool to Evaluate Urban and Suburban
Bicycle Compatibility. ” Transportation
Research Record 1438. Washington, D. C.:
TRB, National Research Council, 1994.
5. Department of Transportation. Highway
Performance Monitoring System—Field
Manual. Washington, D. C.: Federal
Highway Administration, 1987.
6. Wilkinson, W. C., A. Clarke, Bruce
Epperson, et al. Selecting Roadway
Design Treatments to Accommodate
Bicycles. Washington, D. C.: Federal
Figure 6. The proximity of bicycle trip-attracting land rises is a factor in the probability Highway Administration, 1992.
7. Kimley-Horn and Associates. Area and
that bicyclists would use the street segment, provided that bikeways were available.
Corridor Analysis Technique (A CA T)
Manual. Tallahassee, Fla.: Florida
nonunit quantity, the relative latent Conclusion Department of Transportation, 1980.
demand of bicycle travel on each seg-
The IHS and LDS models are cost- 8. Landis, Bruce W. “The Bicycle Interac-
ment of a road network. It provides a effective prioritization tools for bicycle tion Hazard Score: A Theoretical
clear indication of the relative level of system planning and implementation. Model.” Transportation Research Record
desired bicycle use should a bicycle facil- These models: 1438. Washington, D. C.: TRB, National
ity be provided on the road segment. ■ Are adaptable to a variety of software.
Research Council, 1994.
I
■ Use data avail- -
Applications able in virtually all
The LDS Model, like the IHS Model, metropolitan areas.
is being applied rapidly as a quantitative
tool in both systems planning and facili-
ties construction programming. The
■ Use objectively
collected field data.
■ Update easily.
.Ds = ~TTsn x
n=l
x gan
1
■ Are used for “fin- n = bicycle trip purpose ( eg, work,
Tampa urban area’s MPO uses the LDS
ger-tip” policy test- personal/business, recreation, school)
Model in conjunction with the IHS
ing of traffic calming lTS = trip purpose share of all bicycle trips
Model to select street segments to “fill number of generators or attractors per trip
or other alternatives. GA =
in the gaps” of its bikeways network. In purpose
■ Generate easily
cases where levels-of-service are equally TG = average trip generation of attractor or
understood results.
as poor, specific funding for bike lanes is generator
Both the IHS and
directed to those segments with the P = effect of travel distance on trip interchange,
LDS models are
highest travel demand as identified by expressed as a probability
rapidly gaining popu-
the LDS Model. The LDS Model also is ga = number of generators or attractors within
larity as bicycle plan-
specified travel distance range
used to prioritize existing bicycle facili- ning prioritization
d = travel distance _ from generator or
ties for signage and pavement marking and implementation
attractor
improvements based upon travel tools. While research
demand. funding is being ‘igure 7. Latent Demand Score Model equation.

26 ● ITE JOIJRNALO FEBRUARY 1996