CSTM Improvements AITPM Canberra 11 May 2017 7 December 2016 1
Introduction Josh Everett SMEC Australia Worked on modelling in Canberra for 11 years Used TransCAD, then EMME, now back to TransCAD Presentation Structure CSTM Updates to the CSTM Calibration/Validation of the CSTM Hopefully time for questions 2
CSTM Structure 4-step strategic transport model 824 Zones 28,000 Links 6 categories 102 Bus Routes (87 AM, 90 PM) plus school buses Four land use types: Population, Employment, Retail Space, Enrolments Based on generalised cost function 3
CSTM The Four Steps Trip Generation Fixed generation rates for each of six purposes based on population Home Based trips are modified by socio-economic factors Trip Distribution Trip totals apportioned by land use Three different gravity models Work, Education, Other Mode Choice Multi-level binomial structure for Work, Education, Other Includes Car, Bus, Bicycle and Park & Ride Traffic Assignment Equilibrium assignment with link-based delay 4
Update: TransCAD TransCAD is a native GIS program More intuitive user interface than EMME Object oriented coding environment Includes all procedures as functions instead of terminal script Uses different database structure, 2-way links CSTM converted to TransCAD 7 model Retains model structure and flow Utilises built-in TransCAD functions where necessary 5
Input Data Update Converted road network to TransCAD format Reviewed and updated 2011 and 2016 road networks Updated 2016 public transport network with actual imported directly from GTFS Included actual bus stop locations from GTFS Updated future road networks with latest capital works projections, including 2041 Included Town Centre Roads with lower speed and capacity to approximate friction and congestion 6
Trip Generation Update Previous CSTM used 2-hour trip rates modified to 1-hour using 65% assumption Updated using trip generation rates calculated from Household Travel Survey New 1-hour rates are 74% of 2-hour rates Updated directional distribution for 1-hour Updated 2011 AM 2011 PM Trip (One hour) (One hour) Purpose Rate Proportion Rate Proportion HBW 0.112 24.5% 0.089 24.7% HBE 0.092 20.2% 0.013 3.6% Other 0.253 55.3% 0.258 71.7% HBS 0.006 1.4% 0.045 12.4% HBO 0.124 27.1% 0.114 31.8% NHBEB 0.005 1.2% 0.006 1.6% NHBO 0.117 25.6% 0.093 25.8% All 0.457 100% 0.359 100% 7
Trip Propensity RMSE Trip Distribution Update 25 Exponential Gravity Curve Calibration Model uses inverse exponential distribution: P ij = 1 = e pck epck Calibrated to survey data using iterative minimisation of RMSE HBW used Census Journey-to-Work data (24 hour) HBE used ACT Government school enrolments (missing private schools, CIT, university) Other used Household Travel Survey Generated K-factor matrices to better fit data 20 15 10 5 0 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 Gravity Curve Coefficient HBW HBE OTH AM OTH PM Calibrated Gravity Curves 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 HBW HBE OTH AM OTH PM 0.2 0.1 0 0 5 10 15 20 25 Trip Generalised Cost 8
Mode Choice Update Previously updated mode choice to include bicycles and Park & Ride Modified generalised cost component weights to match NGTSM Updated fares, including different fares by purpose, and removed the distance-based fare structure Updated parking costs by travel purpose Updated bicycle coefficients Currently working on future PT network Bicycle Person Trips Total Person Trips Car Person Trips Motorised Person Trips Enhanced Transit Trips Bus Person Trips Park & Ride Person Trips Additional Bus Person Trips Additional Car Person Trips 9
Assignment Update No major changes to assignment Conical function for link delay with interrupted/uninterrupted capacity modifiers Uses TransCAD s built-in PT assignment, including Park & Ride Tighter convergence threshold to meet validation criteria (from 2% to 0.1%) 10
PM Peak Model New PM peak model for all years Calculated new trip generation rates based on Household Travel Survey PM peak has 79% of AM Peak trips, lower than expected Main reduction in trips due to education New PM public transport network Work and Education gravity coefficients same as AM, Other uses different coefficient Park & Ride disabled Road network almost the same, except for school zones 11
Validation Criteria Developed new validation criteria in association with EPSDD based on VicRoads, RMS and UK DfT standards Assignment convergence checks that assignment equilibrium has been reached Feedback convergence checks that sufficient model iterations have been used Screenline volumes are total volumes across district screenlines Link volumes are individual volumes at district screenlines Public transport OD matches model trip matrix to ticket data Stage Source Criteria Threshold Assignment Convergence Feedback Convergence Screenline Volumes Link Volumes Public Transport OD VicRoads/ UK DfT VicRoads VicRoads/ RMS NSW VicRoads - RGAP, and < 1% RAAD, or < 1% AAD, or < 1 veh/h %Diff < 5% > 95% %RMSE, or < 1% Max GEH < 2 %Diff ±50V 0.3953 R² > 0.9 %RMSE < 30% Slope 0.9 1.1 GEH < 5* 50% GEH < 10* 80% R² > 0.9 Slope 0.9 1.1 12
Difference between Observed and Modelled Difference between Observed and Modelled Screenline Validation AM Screenline Validation +60% +40% +20% +0% -20% -40% -60% 0 5,000 10,000 15,000 20,000 25,000 30,000 Observed Volume Difference Minimum Maximum PM Screenline Validation +60% +40% +20% +0% -20% -40% -60% 0 5,000 10,000 15,000 20,000 25,000 30,000 Observed Volume Difference Minimum Maximum 13
Modelled Volume Modelled Volume Linkflow Validation 4,500 AM Peak Comparison of Car Volumes on Screenline Links Criterion Target Original 2011 AM Updated 2011 AM 2011 PM R² 0.9 0.83 0.93 0.89 Slope 0.9-1.1 0.94 0.95 0.97 %RMSE < 30% 43% 28% 32% GEH < 5 50% 33% 50% 51% GEH < 10 80% 64% 73% 70% Average GEH - 9.2 6.7 7.1 4,000 3,500 3,000 2,500 2,000 1,500 1,000 500 0 4,500 4,000 3,500 3,000 2,500 2,000 1,500 1,000 500 0 y = 0.948x R² = 0.9255 0 500 1,000 1,500 2,000 2,500 3,000 3,500 4,000 4,500 Observed Volume PM Peak Comparison of Car Volumes on Screenline Links y = 0.9679x R² = 0.8869 0 500 1,000 1,500 2,000 2,500 3,000 3,500 4,000 4,500 Observed Volume 14
Modelled Trips Modelled Trips PT OD Validation Gathered OD data from MyWay ticketing Used trips that have mid-point between 8:00-9:00AM 1400 1200 1000 AM Peak Public Transport Trips by District y = 0.9763x R² = 0.9934 Joined transfers into a single OD trip Compared model and ticket data at a district level 800 600 400 200 0 0 200 400 600 800 1000 1200 1400 Observed Trips PM Peak Public Transport Trips by District 900 800 700 600 500 y = 0.9979x R² = 0.9896 400 300 200 100 0 0 100 200 300 400 500 600 700 800 900 Observed Trips 15
Future Upgrades Native TransCAD functions Travel Data Household Travel Survey Traffic count program Input Data Separated land use Education, Retail, Employment Intersection delay Future Travel Behaviour Light Rail Autonomous vehicles 16
Questions? 17