Establishment of Freeway Link Volume Validation Targets based on Traffic Count Distributions in the Dallas-Fort Worth Region Behruz Paschai, Arash Mirzaei,

Establishment of Freeway Link Volume Validation Targets based on

Traffic Count Distributions in the Dallas-Fort Worth Region

Behruz Paschai, Arash Mirzaei, Francisco Torres

North Central Texas Council of Governments (NCTCOG)

15th TRB Conference on

Transportation Planning Applications

Atlantic City, NJ

May 2015

May 2015 15th TRB Conference on Transportation Planning Applications 2

ObjectivesUsing the historical freeway counts in the region and comparing the actual counts versus the actual AAWDTs:

- Evaluate the traffic count variations and compare with the available data from other regions

- Evaluate and recommend appropriate validation measures

- Evaluate the statistical significance built in to the link

volume validation process- Evaluate the errors associated with estimating

the AAWDTs.


Permanent Traffic Count Locations

115,000 < AWDT < 150,000

200,000 < AWDT < 250,000 80,000 < AWDT < 95,000

130,000 < AWDT < 150,000

I35E

I30

I820

US75

70,000 < AWDT < 80,000 I45

70,000 < AWDT < 80,000

I20


Data Cleanup Process

1. Only considered counts on Tue, Wed, and Thu

2. Eliminated counts collected on:

a. 4th of July

b. Thanksgiving weekend

c. December 23rd - January 1st

3. Eliminated counts that were outside 4*Sigma around the AAWDT (not more than 5 winter days in January and February when we probably had a winter storm event)

5

AAWDT Box Plots – I35

May 2015 15th TRB Conference on Transportation Planning Applications

Note : The whiskers indicate the maximum and minimum values since the data outside of the 4*sigma range has been already removed.

AA

WD

T

Year

6

Deviation Measures


1. Many deviation measures exist

2. Each measure has its own limitations:

1. biased towards positive differences

2. cannot be used when observations include zero

3. biased towards larger differences

4. difficult to explain

3. %RMSE is commonly used in the validation process although it has a more complicated definition in comparison to simpler measures such as MAPE.

RMSE

APEMAPE

MAERMdSE

RMSPEGMRAE

MdAPEMdRAE

SMAPEsMdAPE

MdRAE

MASERMdSPE

MSEMdAE

7

Deviation of Counts vs AAWDT


I35EI30

I820 US75

2%

4%

2%

4%

2%

4%

6%

8%

2%

4%

6%

8%

10%

%RMSEMAPEWxMAPE

%RMSEMAPEWxMAPE

%RMSEMAPEWxMAPE

%RMSEMAPEWxMAPE

8

CV vs AAWDT


I35EI30

I820 US75

Average Annual Weekday Traffic Count

Co

effi

cien

t o

f V

aria

nce

I20I45

9

%RMSE vs AAWDT


I35EI30

I820US75


%R

MS

E

I45

I20

10

Average Deviation from AAWDT



Ave

rag

e D

evia

tio

n

I35E

I30

I820US75

I20

I45

11

Deviation of Weekday Freeway Counts around AAWDT (All Locations)



Dev

iati

on

of

We

ekd

ay C

ou

nts

aro

un

d A

AW

DT

+10%

-10%

+5%

-5%

-20%

+20%

12

Percent Weekday Freeway Counts within 5% of AAWDT (All Locations)


Year

Av

era

ge

Pe

rce

nt

Da

ily

Co

un

ts

13

Percent Weekday Freeway Counts within 5% of AAWDT (All Locations)


Month

Av

era

ge

Pe

rce

nt

Da

ily

Co

un

ts


Preliminary Conclusions

1. Similarities can be observed in the daily variation of traffic volumes on freeways across the country.

2. On average 80%-90% of the weekday counts at the studied locations are within 5% of their corresponding AAWDT.

3. The probability of getting a count on the freeways that is within 5% of AAWDT is highest when collected between March and October (excluding the major holidays).


Preliminary Conclusions

4. Prefer to use MAPE over %RMSE as a measure of deviation.

5. For all practical purposes, Tue, Wed, and Thu are similar in total daily traffic and its variation.

6. The provided graphs compare the actual daily counts against the actual AAWDTs. Validation targets need to be set with accounting for errors in calculating the AAWDTs.


Next Steps

1. Evaluate the traffic conditions in the peak hours.

2. Evaluate the accuracy of the AAWDT estimations based on daily counts and calculated factors from previous years.

17

Contact Information

Behruz Paschai [email protected] Arash Mirzaei [email protected] Francisco Torres [email protected]



Background

Daily Roadway Traffic Counts are Random Variables

- TMIP Validation Manual provides examples of validation requirements/recommendations from Florida, Ohio, and Oregon.- Validation measures are typically:

- %RMSE, MAPE, or CV- A function of AADT

- Acceptable ranges could be presented as a confidence intervals around the AADT.


Available Historical Data

1. Traffic counts obtained from permanent count stations

2. Utilized the data from six freeway locations that:

a. had 7-10 years of daily historical data between years 2004 and 2014

b. represented different weekday count ranges or similar ranges on different facility types (AAWDT 70k-250k)

c. represented urban and rural areas


Validation Recommendation Examples

Source: Model Validation and Reasonableness Checking Manual, TMIP/FHWA, February 1997.

Co

effi

cien

t o

f V

aria

tio

n

Var

iati

on

Aro

un

d M

ean

Per

cen

t D

evia

tio

n

Tar

get

%R

MS

E

21

Variations of AAWDT (PDFs)


I35EI30

I820 US75

40%

20%

40%

20%

40%

20%

40%

20%

142,

500

87,5

00

132,

500

60%

222,

5000% 0%

0% 0%

22

Variations of AAWDT (PDFs)


I45

I20

40%

20%

40%

20%77

,500

72,5

00

60%

60%

0%

0%

23

AAWDT Box Plots – US75



AA

WD

T

Year

24




AA

WD

T

Year

25




AA

WD

T

Year

26




AA

WD

T

Year

27




AA

WD

T

Year

28

Deviation of Counts vs AAWDT


I45

I20

2%

4%

6%

8%

2%

4%

6%%RMSEMAPEWxMAPE

%RMSEMAPEWxMAPE

Establishment of Freeway Link Volume Validation Targets based on Traffic Count Distributions in the Dallas-Fort Worth Region Behruz Paschai, Arash Mirzaei,

Documents

rmse15th trb conference

rmsema trb conference

i2015th trb conference

actual counts

thueliminated counts

deviation measuresmay

rmse vs aawdtmay

traffic count distributions