Traffic Analysis & Safety · Study · PE Civil: Transportation · FE → PE Prep
Traffic Analysis & Safety
6% of exam
Volume and speed studies, trip generation and modal split, traffic forecasting, crash rates and severity, collision analysis, and Highway Safety Manual crash modification factors and predicted crashes.
6 concepts
D. Traffic analysis
Volume, Speed & Delay Studies
Expand short counts to AADT with daily/monthly/axle factors, read the 85th-percentile speed from a spot-speed study, and quantify travel time and delay from runs.
Unlocks with an access pass — one-time payment, no auto-renew. View passes
Pass holders
Trip Generation & Modal Split
Walk the four-step travel model: estimate trips from ITE land-use rates, distribute them with the gravity model, and split them by mode with a logit share.
Unlocks with an access pass — one-time payment, no auto-renew. View passes
Normalize crashes by exposure (MEV at intersections, HMVM on segments), weight by severity with EPDO, and screen high-crash locations against a critical rate.
Safety questions almost never ask 'how many crashes?' — they ask 'how many crashes ·per unit of exposure·?', because a busy interchange with twenty crashes a year can be far safer than a sleepy rural curve with three. Converting a raw count into a rate, weighting that rate by how badly people were hurt, and then comparing the site against its peers is the core skill the PE tests in the safety area. The NCEES PE Civil Reference Handbook (§5.1.4 Accident Analysis) names accident rates and collision diagrams but supplies almost no method, so the working procedure here is drawn from the AASHTO Highway Safety Manual (HSM-1) — cite the HSM, not a handbook page, for everything in this concept.
Two exposure denominators: MEV and HMVM
Exposure is measured differently for a point and for a length. An intersection is a ·spot· — exposure is the number of vehicles that arrive, so the denominator is million entering vehicles (MEV), built from the total approach volume. A segment is a ·length· — exposure is vehicle travel, so the denominator is hundred million vehicle-miles (HMVM), built from AADT×L. Picking the wrong denominator (rating an intersection per HMVM, or a segment per MEV) is the single most common setup error on these problems, so first decide: point or length?
RMEV=365TVC×106,RHMVM=365T(AADT)LC×108
Building the intersection rate
For an intersection, V is the total entering volume in vehicles per day summed over all approaches, T is the study period in years, and C is the crash count over that same period. Total exposure in MEV is 365TV/106
Building the segment rate
For a roadway segment the exposure is vehicle-miles of travel: VMT=365T(AADT)L with L in miles. The HSM and most state DOTs report segment rates per ·hundred· million vehicle-miles, hence the 108 in the numerator. A useful sanity anchor: a two-lane rural road runs on the order of 50
Severity weighting with EPDO
Two sites with identical crash rates are not equally dangerous if one has fatalities and the other only fender-benders. The equivalent-property-damage-only (EPDO) index collapses severity onto a single scale by weighting each crash by an agency-specific factor — typically derived from comprehensive crash costs — relative to a property-damage-only (PDO) crash of weight 1. Fatal/serious-injury weights are large (often hundreds when based on cost), so a single severe crash can dominate the index. Always use the agency's published weights; the numbers below are illustrative only.
EPDO=wFNF+wINI+wPNP
The collision diagram
Before any screening, the collision diagram turns a crash list into a picture: each collision is drawn at its location and approach with a standardized arrow symbol (rear-end, right-angle, sideswipe, head-on, pedestrian, fixed-object), annotated with date, time, severity, and condition (wet, dark). Patterns the table hides jump out — a cluster of wet-night run-off-road crashes points to friction or drainage; a stack of left-turn right-angle crashes points to sight distance or signal phasing. The diagram is qualitative but it drives the countermeasure, so the PE may ask you to read one and name the likely contributing factor.
High-crash-location screening with a critical rate
A site is flagged for investigation when its observed rate exceeds a ·critical rate· built from the average rate of similar facilities plus a statistical allowance for random variation. The critical-rate method adds a confidence term (with K=1.645 for the 95% level) that shrinks as exposure M grows, so low-volume sites need a much higher rate before they are flagged — this protects against over-reacting to a couple of crashes at a quiet location. If Rsite>Rc
Exam strategy
Decide point-vs-length first, then write the matching rate formula with its power of ten (106 for MEV, 108 for HMVM) before plugging numbers. Watch the period T
Problem. A signalized intersection carries a total entering volume of 25,000vpd. Over a 3-year study it experienced 18 reported crashes. Compute the crash rate per MEV.
•Citing the NCEES handbook for the crash-rate or critical-rate ·method· — it lists accident analysis (§5.1.4) but does not give these formulas. The governing source is the AASHTO HSM; cite it.
•Mismatching the denominator: intersections are rated per MEV (entering vehicles), segments per HMVM (vehicle-miles). Using AADT×L for an intersection, or entering volume for a segment, is the most common setup error.
•Dropping the power of ten: 106 for MEV and 108 for HMVM. The number is meaningless without it; if your rate is microscopic or in the millions, the exponent is wrong.
References
AASHTO Highway Safety Manual, 1st Edition (HSM-1) — Part B, network screening and crash-rate/critical-rate methods
NCEES PE Civil Reference Handbook — §5.1.4 Accident Analysis (names accident rates and collision diagrams)
FHWA, Highway Safety Improvement Program (HSIP) Manual — site screening and EPDO weighting
F. Nonmotorized facilities analysis
Nonmotorized (Pedestrian & Bicycle) Facilities
Quantify walkway flow and pedestrian space for LOS, size pedestrian clearance times, and select bicycle facilities and assess shared-use-path level of service.
Unlocks with an access pass — one-time payment, no auto-renew. View passes
Pass holders
G. Traffic forecasts and monitoring
Traffic Forecasting & Growth
Project volumes to a design year with linear and compound growth, accumulate traffic with a growth factor, and convert AADT to the design-hour volume via K and D.
Unlocks with an access pass — one-time payment, no auto-renew. View passes
Predict crashes from a safety performance function, adjust for site features with crash modification factors and a calibration factor, then refine with the empirical-Bayes method.
Unlocks with an access pass — one-time payment, no auto-renew. View passes
Pass holders
, and the rate is simply
C
divided by that exposure. Keep the
106
honest: a rate 'per million entering vehicles' means you divide the count by exposure expressed ·in millions·, so a typical urban signal lands around
0.3
to
1
crash per
MEV
. If your answer is in the hundreds, you forgot the
106
.
to a few hundred crashes per
HMVM
, while a modern freeway is far lower per mile of travel because the exposure denominator is enormous.
, the site is a candidate for a road-safety audit.
Rc=Ra+KMRa+2M1
: multiply
365T
in the denominator, and use the ·same·
T
-year count
C
in the numerator. For EPDO, read the weights from the problem and do not assume
PDO=1
implies the others are small. For screening, compute exposure
M
in
MEV
once and reuse it in both the rate and the critical-rate confidence term — a mismatch there is the classic trap.
C = crashes in study period, T = years, V = total entering AADT (vpd). Result in crashes per million entering vehicles.
N = counts by severity (fatal/injury/PDO), w = agency severity weights relative to PDO =1.
Critical crash rateRc=Ra+KMRa+2M1
Ra = average rate of similar sites, M = site exposure (MEV or HMVM), K=1.645 at 95%. Flag if Rsite>Rc
Crash frequencyf=TC
Crashes per year — raw, un-normalized; use only to compare identical-exposure sites.
Severity indexSI=CEPDO
Average equivalent-PDO weight per crash; higher means more severe crash mix.
.
Rate:
RMEV=27.37518=0.6575crashes/MEV
.
**Answer:
0.658
crashes per MEV.** Sanity check: a value under
1
is typical for an urban signal; if the
106
had been dropped the answer would be a meaningless
6.6×10−7
, confirming the units are handled correctly.
RMEV=365(3)(25,000)18×106=0.658
Segment rate per HMVM with EPDO
Problem. A 2.4-mi rural two-lane segment (AADT=14,500vpd) had 32 crashes over 5 years: 1 fatal, 6 injury, 25 PDO. Using agency weights wF=12, wI=3, wP=1, find the crash rate per HMVM and the EPDO index.
Problem. Similar signals in the region average Ra=0.45 crashes/MEV. For the intersection of Example 1 (M=27.375MEV, observed rate 0.658), apply the critical-rate method at the 95% confidence level (K=1.645). Is it a high-crash location?
Forgetting 365T in the denominator — using a one-year exposure with a multi-year crash count (or vice versa) is a factor-of-T blunder. Count and exposure must span the same period.
•Assuming EPDO weights are small. Cost-based fatal/serious-injury weights are often in the hundreds; a single severe crash can dominate the index, so always read the agency's published weights.
•Computing the critical rate with a different exposure than the site rate. Compute M once and reuse it in both Rsite=C/M and the KRa/M and 1/(2M) terms.
•Confusing crash frequency (crashes per year) with crash rate (per unit exposure). Frequency only compares sites with identical exposure; ranking by frequency alone over-flags high-volume locations.
.
.
EPDO:
12(1)+3(6)+1(25)=12+18+25=55
.
**Answer:
50.4
crashes/HMVM; EPDO
=55
.** Sanity check: rate is in the expected tens-to-hundreds band for a rural two-lane; severity index
SI=55/32=1.72
, reasonably above
1
given the fatal and injury crashes.
.
Probability term:
2M1=54.751=0.0183
.
Rc=0.45+0.2109+0.0183=0.679crashes/MEV
.
Compare:
Rsite=0.658<Rc=0.679
.
**Answer: NOT flagged** — the observed rate is below the critical rate, so the excess over the average is within expected random variation. Sanity check: the confidence allowance (
0.21
) is large relative to the excess over average (
0.658−0.45=0.21
), which is exactly why a borderline site like this does not screen positive.