HOME | TRASHIN' THE BOG | TRASHIN' YOUR HOOD | PULLING THE WOOL OVER YOUR EYES | F.A.Q.

Technical Volume 5
Calibration of the Gateway Program Traffic Models

This report is full of jargon and undefined acronyms that render the report virtually incomprehensible, even for a skilled engineer. The diagrams do not even contain a key to assist the reader in understanding what is being shown - for example in the figure (no ID) Gateway Sub-Area Model Gates System - Hwy 1 (A), there are a number of coloured circles - purple, green and red. In the text there is a brief remark about the green and purple ones, but nothing about the red one. The key is basically obscured.

Exhibit 4.3 - AM Peak Other Trip Production Results - page 32

Although the model has a very high correlation coefficient, note that the sub-area with the greatest error in predicting AM Peak Trip Productions is South Delta and North FVRD.

Exhibit 4.18 - RSQ of Observed to Model Fit for Traversal Gates by Project, AM and PM Peak Hours

The graphs show the weakest model correlations for the AM and PM Peak Hours Traversal Gates for the SFPR, although the correlations are still very good at 0.93 and 0.89 respectively.

The table in Appendix I: AM and PM Gate Count Comparison clearly demonstrate the unreliability of the model applied to the Annieville area. Among the largest errors in the entire model, Elevator Road and adjacent routes show errors that are at unacceptable levels:

GEH is an artifact method used to calculate data error that gives less variability than simply using percentage difference. It reflects the relative impact of error prioritizing internally to the amount of impact. Here is a quote from the report itself on Page 33:

"For investment grade studies, a GEH statistic of less than 10 on individual
        screenlines has been recommended by some agencies (e.g. TransFund New Zealand)."

The values for the PM times are not too bad, but note that the AM model data show gross errors for these zones that render the model inapplicable. While these zones have been chosen because they have high error rates (and so are biased selections), it is clear that the overall model is applicable to many other regions in the GVRD. There are numerous examples of accurate zonal predictions in the model for other areas, but clearly, there is an error bias on routes around Annieville.  Better traffic data and adjustment of model parameters in the Annieville area are required, if the predictions about present and future Annieville traffic flows is considered to be important criteria in designing this route.

The models developed for South Delta do not set-up the zones in South Delta to properly evaluate the Hoover-Nass proposal. A very large zone is used in the centre of the farmland so the traffic impact of an interchange at Deltaport Road onto a route following the train tracks is not available. Neither is there the abilty to analyse the impact of  the H&N truck traffic onto Highways 99 and 91 at that interchange. I must conclude that the model has been designed to bias the choice of routes to the north rather than to the east as proposed by Hoover-Nass.

The report itself was done to attempt to rectify problems in the database that are discussed in depth in Appendix 3. In the main , the work appears to have achieved that goal overall. But as the report itself points out in Appendix C, applicability of this modeling approach to sub-areas can be contradictory and in gross error.

From the report (page 5)

In summary, 2001 Census data (population and employment) was allocated to traffic zones, and adjusted to year 2003 based upon an interpolation with future year 2011 forecasts. See Appendix C for further discussion regarding the reconciliation of base year 2003 population and employment estimates at the traffic zone level1. - page 5.

From Appendix C
         Reconciliation of Traffic Zone Level Data
         for the Greater Vancouver and Fraser
         Valley Regional Districts  - by Urban Futures Inc.

The Greater Vancouver Regional District's Livable Region Strategic Plan Growth Management Scenario (GMS), currently in GMS 4.0 format, was originally produced over a decade ago using Census Tract level data from the 1971 to 1986 period, with subsequent revisions to incorporate the preliminary results of the 1991 Census. Since its creation, the base year data has been updated as information became available (usually through subsequent Census releases) to reflect the pattern of population and employment growth and change being realized throughout the Lower Mainland. In addition to updating base data, the number of traffic zones has increased from 435 zones when conceived to just under 1,000 today. This is also a reflection of the changing pattern of regional development with new zones being added where significant level of population or employment did not historically exist.

Combined, the ongoing updating of base data, the addition of new zones to the network, and the visionary nature of the GM Strategy have introduced considerable inconsistencies in the spatial distribution of historical and future transportation generators within the GM Scenario. In terms of the historical development of the data base, while updating the Scenario as new data becomes available is essential, projected values for future years must also be modified to reflect emerging patterns of growth and change. As an example, while a traffic zone may not have been anticipated to grow due to policy objectives outlined in the founding years of the GMS, growth being realized in this zone in the base years would, without updating projected years, result in numerical population decline in this zone over following projection periods. The net result would be a decline in the number of traffic generators over time (either people or jobs), while in actuality an increase in generators would be experienced.

In terms of the refinement of the traffic zone system itself, when traffic zones used in the original modeling were split into smaller zones the process has generally involved simply splitting the values of transportation generators for the old larger zones amongst the new smaller zones, rather than modeling the new values to reflect historical trends and spatial capacities at the finer level of geography. This has introduced a great deal of variance in the patterns of change at the new zonal level where none existed at the original higher level of geography. The inconsistencies introduced by using a vision-based scenario such as the GMS 4.0 are not attributable to the vision per se, but rather to the fact that the reality of development in the region since the articulation of the vision has not always been in accordance with what was envisioned. With no opportunity for reality (as reflected through the Census) to modify both vision and policy, this has resulted, in some instances, in the GM Scenario4.0 values at the zonal level for future dates being below those recorded in the 2001 Census. The combined variance introduced by these inconsistencies results in a significant minority of zones under the GM Scenario 4.0 demonstrating future patterns of growth and change that diverge from either recent data and logic and, in some instances, both.