MODELING THE IMPACT OF TRAFFIC CONDITIONS ON THE VARIABILITY OF

MID-BLOCK ROADSIDE PM2.5 ON AN URBAN ARTERIAL

Adam Moore Miguel Figliozzi Alexander Bigazzi

Friday Transpo Seminar

17 January 2014

INTRODUCTION

2

Portland State University has been studying the Powell Boulevard corridor in southeast Portland – Busy arterial linking downtown to suburbs

•  Investigating variations in PM levels •  Incorporating many data sources –  Traffic, air quality, meteorology

•  Utilizing statistical analyses to control for many factors

BACKGROUND

Exposure to Air Pollution on Roadways

Vehicle Public Transportation Bicyclist/Pedestrian Public Transportation Bicyclist/Pedestrian

What factors affect exposure to air pollution?

3

Vehicle Activity

Built Environment

Meteorological Conditions

Very High Resolution Data Collection

Vehicle Activity

Meteorological Conditions

FINE PARTICULATE MATTER (PM2.5)

4

•  Cancer •  Heart disease •  Increased mortality rates •  Increased incidence of

respiratory disease (asthma, etc.)

image: epa.gov

Vehicle emissions, brake wear, tire wear

STUDY LOCATION

5

Downtown

map: bing.com/maps

Powell Boulevard

N  

31,500 vehicles daily 1,500-1,800 peak hours

STUDY SITES

6 map: google.com/maps

Intersection

“Mid-block”

May 1, 2013 7:00-9:00am

N  

MID-BLOCK STUDY SITE

7

Parking Lot

Parking Lot

Powell City Park

Monitoring Location

SE 24th Ave

N

(looking south)

EQUIPMENT

8

9

Fine Particulate Matter (PM2.5)

TSI DustTrak DRX 8533

EQUIPMENT

photo: www.tsi.com

10

Temperature Relative Humidity

Onset HOBO U12

EQUIPMENT

photo: www.onsetcomp.com

Wind Speed Wind Direction

RM Young Ultrasonic Anemometer 81000

photo: www.youngusa.com

11

Video Reference

CountingCars CountCam System

EQUIPMENT

photo: www.countingcars.com

12

Vehicle Volume, Speed, Classification,

Lane Occupancy

Wavetronix SmartSensor HD

EQUIPMENT

Veh

icle

s/m

in

1030

502:00am 7:00am 12:00pm 5:00pm 10:00pm

1030

50

Spe

ed (m

ph)

020

4060

Occ

upan

cy (%

)

2:00am 7:00am 12:00pm 5:00pm 10:00pm

Westbound Eastbound

TRAFFIC ACTIVITY

13

NEARBY SIGNALS

14

123 sec

125 sec

Can we see vehicle platooning?

map: google.com/maps

VEHICLE PLATOONING

15

One lag = 10 seconds

-0.2

0.0

0.2

0.4

Westbound

ACF

Eastbound5 10 15 20

5 10 15 20Lag

-0.2

0.0

0.2

0.4

PACF

130 seconds

Autocorrelation Function

Partial Autocorrelation

Function

PLATOONING EFFECT ON PM2.5

16

-0.1

00.

000.

10

Wes

tbou

nd C

CF

-20 -10 0 10 20

Lag

-0.1

00.

000.

10

East

boun

d CC

F

•  Westbound platooning did not have a significant effect on PM2.5 concentrations

•  Eastbound platooning did have a significant effect

Cross-correlation Function

One lag = 10 seconds

EXAMINING CONGESTION

17

050

0010

000

1500

0

Westbound

N(t) N(t)

v0t, v0 = 11.27mph

Eastbound7:30am 8:00am 8:30am

N(t)v0t, v0 = 24.23mph

7:30am 8:00am 8:30am

-500

050

015

00

N(t)-

v 0t

Pre-congestion Congestion

“Active” queuing

Inactive Active

67

89

1011

12

Pre-congestion

mg/m

3

Queuing

Inactive Active

Congestion

mg/m

3

Condition

•  Pre-congestion, active queuing associated with lower PM2.5 concentrations

•  During congestion, active queuing associated with raised PM2.5 concentrations

18

CONGESTION EFFECT ON PM2.5

0.35% decrease

10.8% increase

REGRESSION FINDINGS

19

Vehicles passing when wind blows across roadway towards monitoring station

–  Current and previous observations Relative humidity increases Congestion worsens

Wind blows from the background as vehicles pass

Wind speed increases

𝑅↑2   .6589  𝑅↓𝑎𝑑𝑗↑2   .6533  

PM2.

5 c

on

ce

ntr

atio

ns

VEHICLE SEMI-ELASTICITIES

20

Variable…   …lagged…   …changed  PM2.5  by…   …per…  

Occupancy   0  sec   .05%   %  occupancy  increase  

When  wind  was  blowing  across  roadway  towards  monitoring  sta<on:  

EB  Passenger  Vehicle   (80,  110,  200)  sec   (.49%,  .46%,  .45%)   Addi<onal  vehicle  

EB  Heavy  Vehicle   0  sec   2.45%   Addi<onal  vehicle  

When  wind  was  blowing  from  background  neighborhoods:  

WB  Passenger  Vehicle   0  sec   -­‐.40%   Addi<onal  vehicle  

CONCLUSIONS

Exposure to Roadside Fine Particulate Matter

21

Data Availability

Congestion Platooning

Model Calibration

Policy Implications to Minimize Exposure

QUESTIONS? Adam Moore adam.moore@pdx.edu Dr. Miguel Figliozzi Alexander Bigazzi

Acknowledgements

Moore, A.; Figliozzi, M.; Bigazzi, A., Modeling the Impact of Traffic Conditions on the Variability of Mid-block Roadside Fine Particulate Matter Concentrations on an Urban Arterial. Proceedings of the 93rd Meeting of the Transportation Research Board, January 2014. Paper #14-4970.

REGRESSION MODEL

23

Dependent variable: ln(PM2.5) R2 (R2

adjusted) .6589 (.6533) Residual standard error .06138 on 600 degrees of freedom AICa –1663.29

Variable Lag Coefficient SE p Unit increase effects on dependent variable (Semi-elasticity)

Intercept 0 .63261 .05693 <.001

ln(PM2.5) 1 .63829 .03013 <.001 Traffic Conditions

WB Occupancy 0 .00046 .00013 <.001 Increase by .05% per percentage point occupancy increase

Meteorological Conditions

Relative Humidity 0 .00138 .00020 <.001 Increase by .14% per percentage point RH increase

Wind Speed 1 –.01274 .00368 .001 Decrease by 1.27% per 1m/s increase 2 .00917 .00368 .013 Increase by .92% per 1m/s increase Vehicle Volume × Wind

Wind towards monitoring station EB Passenger Veh 8 .00488 .00214 .023 Increase by .49% per additional vehicle 11 .00455 .00215 .035 Increase by .46% per additional vehicle 20 .00448 .00214 .037 Increase by .45% per additional vehicle

EB Heavy Veh 0 .02418 .01075 .025 Increase by 2.45% per additional vehicle

Wind away from monitoring station

WB Passenger Veh 0 –.00400 .00144 .006 Decrease by .40% per additional vehicle

aAkaike Information Criterion 1