Developing a Sketch Plan Model for the Southern California Association of Governments Jim Lam, Caliper Corporation Guoxiong Huang, SCAG Mark Bradley, BB&C Rich Kuzmyak Hong Kim TRB Applications Conference, May 18, 2009
Outline Goal of the Sketch Plan Model Project Sketch Plan Model Simplifications Zonal Aggregation Models Other Sketch Plan Features Sketch Plan Demo Model Run Times Calibration Results Future Work
Goal of the Sketch Plan Model Project Current Regional Model is 4000+ zones and takes 24 hours to run To create a sketch planning version of the regional travel demand model Tool will be a quick response tool for testing a wide range of policies and projects to assess their benefits for the region. Tool will be simpler to run, much faster in execution, but be sufficiently accurate to identify beneficial projects for further detailed analysis with the full model. Keep model code consistent with Regional Model
Sketch Plan Model Simplifications Aggregate TAZs to a simpler zonal structure Option to “skip” major portions of the model, such as transit Reduce input and output files, and reduce user-controllable model parameters Create a simplified interface Reduce feedback loops and one-loop model run Simplify networks* Mode split simplification* Trip purpose reductions* Vehicle ownership category reductions* * Models considered but ultimately not adopted
Zonal Aggregation Models Zones reduced from 4109 to 302 zones, networks remained the same size Network centroid connectors realign to “Superzones” and networks re-skimmed for 302 zone system Extra models necessary to account for aggregation bias and for trip loss during aggregation (Zonal Allocation and Zonal Aggregation Models)
Example of Network Realignment
Zonal Aggregation Models Zonal Allocation Model to determine best “skim” for aggregated OD pair Random choice model Randomly chosen OD pair from Regional skims for consistency with Regional estimations Problem: Need more than one random chosen pair to reduce variability, which reduces performance Randomness causes variability of results in alternatives analysis Weighted zonal allocation model Superzone centroid location is chosen based on HH and Emp weights
Zonal Aggregation Models Intra-Superzone Trips Model To account for underassignment of trips This Trip Becomes an Intrazonal Trip
Zonal Aggregation Models Intra-Superzone Trips Model Divide OD trips into Inter-Superzone and Intra-Superzone trips from original correspondences Expand Intra-Superzone trips into original TAZ trip matrix using base OD matrices and IPF Separately assign Inter-Superzone trips and Intra-Superzone trips 9
Intra-Superzone Assignment Example Aggregate OD Matrix Intra-Superzone Trip Inter-Superzone Trip Correspondence Table Disaggregate Intra-Superzone Trip
Zonal Aggregation Models Superzone Connector Assignment To better assign Inter-Superzone Trips Should really be this This Trip 11
Other Sketch Plan Features Automatically create Sketch Inputs from Regional Inputs Provide a simple interface Simply define alternative scenarios Simply run and compare multiple scenarios 12
Demo of Sketch Plan Model 13
Model Run Times Regional Model – 24 hours Full Sketch Plan Run – 2 hours One-loop highway-only run – 75 minutes 14
Calibration Results 15
Calibration Results 16
Future Work Explore Random allocation of zones for skimming Add transit-only component Integration with TDM and land use models Additional optimization
Thank You