11 th national planning applications conference topic: statewide modeling validation measures and...
TRANSCRIPT
11th National Planning Applications Conference
Topic: Statewide Modeling Validation Measures and Issues
Authors: Dave Powers, Anne Reyner, Tom Williams, Paul Hershkowitz, Rob Bostrom
Date: May 9, 2007
Modeling Issues
• Statewide Modeling Objectives• Forecast major intercity volumes and roadway
deficiencies• Estimate urban external-internal and through
volumes• Freight transport growth and potential for
intermodal diversion• Evaluate major travel corridor improvements
– Urban bypass– Intercity corridor
Modeling Issues
• Off- Model Traffic• Urban intrazonal / local private vehicle• Rural local private vehicle• Freight intra-county • Commercial vehicles
Modeling Issues
• Observed Traffic Counts• Primary independent validation criterion• Vehicle class not universally available• Estimated traffic volumes• Geographically extrapolated traffic volumes
Modeling Issues
• Travel Demand Survey Data - Person• National Samples – not necessarily representative
at the state level• Not comparable to traffic counts• Add-on samples can expand usefulness• Only source of intercity trips
Modeling Issues
• Travel Demand Inventory - Freight• Inter-county, large vehicle only• Requires extensive review for reasonableness• Requires conversion of tonnage to vehicles• Requires estimation of empty vehicle movements• Comparison to counts problematic
Modeling Issues
• Data Checking – Traffic Counts• Focus on rural higher functional classes• Eliminate urban influence• Review spatial and temporal consistency• Eliminate inconsistent and/or estimated/transferred
counts
2003 vs 2006 Counts
Counts Comparision 2003 - 2006
-60%
-40%
-20%
0%
20%
40%
60%
10,000 20,000 30,000 40,000 50,000 60,000 70,000 80,000 90,000
2003 Count
Gro
wth
20
03
- 2
00
6
Modeling Issues
• Data Checking – Person Travel Surveys• Review sample size by household categories• Compare trip rates, distribution characteristics
across household categories• Supplement with similar state data if necessary
Modeling Issues
• Data Checking – Freight Flow Inventory• Compare flow origins/destinations with
development inventories• If available, compare multiple year data• Major flow desire lines - review with knowledgeable
state staff• Review load factors• Review implied empty backhaul factors• Assign and compare VMT and traffic on major
roadways
Modeling Issues
• Validation Reasonableness Tests – Person Travel
• Statistical correlation of model dependent and independent variables
• Household trip rates• Proportion of trips by purpose / length• Average trip length / trip length distribution• Compared against model development data
Modeling Issues
• Validation Objective Tests – Person Travel• VMT by major functional roadway class• Volumes across major rural screenlines• Volumes at major state boundary crossings• Volumes along Interstate and other major
roadways• RMSE and individual count deviation analysis• Identification and explanation of outliers
Validation Screenlines - Mississippi
Validation Screenlines - Kentucky
Modeling Issues• Validation Reasonableness Tests–Freight
• Statistical correlation of model dependent and independent variables
• Tonnage by commodity type, and internal/external orientation
• Replication of special generator origins/destinations
• Inter-district flows• Average trip length / trip length distribution• Load factors / empty backhaul rates• Compared against model development data
Modeling Issues
• Validation Objective Tests – Freight • VMT by major functional roadway class• Volumes across major rural screenlines• Volumes at major state boundary crossings• Volumes along Interstate and other major
roadways• RMSE and individual count deviation analysis• Identification and explanation of outliers
Percent Deviation for Validation Links - Mississippi
Figure VII-2Percent Deviation For Rural Links with Counts
MS Statewide 2000 Transportation ModelN=1751
0.0%
10.0%
20.0%
30.0%
40.0%
50.0%
60.0%
70.0%
80.0%
90.0%
100.0%
- 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0
Traffic Counts (Thousands)
Per
cen
t D
evia
tio
n
Maximum Desirable Deviation
Percent Deviation for Validation Links - Kentucky
0.0%
10.0%
20.0%
30.0%
40.0%
50.0%
60.0%
70.0%
80.0%
90.0%
100.0%
- 10.0 20.0 30.0 40.0 50.0 60.0 Counts (Thousands)
Per
cen
t D
evia
tio
n
Maximum Desirable Deviation
Modeling Issues
• Validation Expectations• Recognize limitations of source data• Focus on rural areas / higher functional classes• Quantitative comparisons will be less accurate
than experienced with urban models• Validation quality may vary depending on model
complexity, state development/road system complexity and traffic counting system
Modeling Issues
• Lessons Learned• Spend time and effort to fully understand
assumptions and estimation procedures present in traffic counts
• Spend time and effort to get input person and freight data as free from logical inconsistencies as possible
• Focus model calibration and objective validation tests on facility types/volume groups where a statewide model can be reasonably expected to perform accurately