Powerpoint Sl from San Diego ITE Mtg

Uploaded from authorPOINTLite
Views:
 
Category: Entertainment
     
 

Presentation Description

No description available.

Comments

Presentation Transcript

Predicting the Duration of Congestion: 

Predicting the Duration of Congestion on Bay Area Freeways Steven B. Colman, AICP Dowling Associates, Inc.

Motivation: 

Motivation Develop a Simple Model to Predict the Duration of Congestion: Discern important variables Practical– use available data Performance measure usable when peak hour volume > physical capacity

Freeway Congestion: 

Freeway Congestion Definition and Characteristics: When speeds < 35 mph for >15 minutes (Caltrans) Collected using floating cars (2002) 369 route miles (almost 60%) “congested”

San Francisco Bay Area Overview: 

San Francisco Bay Area Overview 9 counties (rural to very urban) 7 million residents, 3.5 million jobs ~620 route miles of freeway Among worst congestion of any metro area in USA (72 h/resident/year) Topographic constraints

PM Congestion 2002: 

PM Congestion 2002 San Francisco

Why Use Congestion Duration as a Performance Measure?: 

Why Use Congestion Duration as a Performance Measure? Readily understood by public/ decisionmakers Measures “convenience” of travel (must I engage in time-shifting behavior?) May be a better metric than LOS to measure improvement to congested urban freeway Works where peak demand>capacity

Excluded Areas: 

Excluded Areas Toll plazas Tunnels Single-lane freeways Freeway-freeway connectors (immediate vicinity) San Francisco County Variable lane configurations Uncongested segments

Variables Used: 

Variables Used Dependent variable: typical weekday hours congested at 46 locations Independent data set: AADT per lane Interchange spacing Percent heavy trucks Grade (flat/hilly/mountainous) Percent of weekday trips that are HBW HOV lane (dummy)

Simple Congestion Duration at a Bottleneck: 

Simple Congestion Duration at a Bottleneck Congestion persists even after V<C Shape (slope) of demand curve will affect congestion duration

The Data Set: 

The Data Set

Regression Approaches: 

Regression Approaches Linear model Log-log model

Linear vs. Log-log model: 

Linear vs. Log-log model Linear model says ADTLN, PCTTRK, & GRADE most significant Congestion begins at 13,600 ADT/LN Log model more intuitive Log model says ADT, Lanes, PCTTRK, GRADE, and HOV dummy significant Residuals somewhat better with logs

Residuals Comparison: 

Residuals Comparison Rowena to prepare

Conclusions: 

Conclusions Variable signs all intuitive– more traffic= longer congestion; more lanes= shorter Percent of HBW trips & interchange spacing don’t appear significant Reliability of method ~ +/- 1 hour Cautious application advisable

Limitations & Sources of Error: 

Limitations & Sources of Error Method based on AADT, not AWDT Traffic volumes vary within sections Congestion duration measured on only a few days Auxiliary lanes ignored Directionality (D) ignored Estimation of truck percentages

Availability of Data: 

Availability of Data Go to www.dowlinginc.com