Robert A. Johnston is a Professor of Environmental Planning at the University of California at Davis. He has been an expert witness in several important land use cases and has published over 60 refereed articles on land use planning, impact assessment, travel modeling and land use modeling. He has been a member of several transportation advisory bodies, and is currently a member of the National Academy of Sciences, Transportation Research Board, Committee on Transportation and Land Development.

COMMENTS ON THE DEIS

The proposed project is a four-lane new freeway segment, 13 miles long, running northward from the N. Salt Lake City area to the Farmington area. This is a very linear urban region, with only one interstate facility throughout, and so accessibility affects land development very closely and predictably.

1. The Utah DOT level of service standard is D for the peak hour, a very stringent standard for an urban region. Most such regions now use a standard of E for the peak period, meaning two- or three-hour peak periods, two per day. To use the D standard for a single hour will result in over-expenditure of funds and a large degree of induced travel and decentralized land development, due to the free-flowing freeway access.

2. The travel demand model uses the same population and employment projections in both the Build and NoBuild cases, which is not theoretically possible. These projections were made with an implicit assumption that there would be no significant roadway congestion in the region and so logically can only apply to the Build case (the proposed project). The proponent must provide a second set of population and employment projections for the No Build case, since the congestion projected would substantially alter the location of new land developments. Using one set of population and employment projections has been ruled invalid in a Federal case in the Chicago region and violates good modeling practice as described in EPA regulations regarding conformity analysis in serious nonattainment regions. We discuss the effects of correcting this error, below.

3. In addition, the travel demand models ignore the substantial shifts that would occur in the No Build case in household locations and in employment locations. That is, even within the existing stock of buildings, workers would change where they work or where they live to reduce the distance to work, in the congested No Build case. These shifts occur quite readily and must be taken into account in modeling travel. The documentation provided to me earlier concerning the RTP and the travel models show that this effect on trip distribution is not properly represented in this model set. Again, this violates good practice and the EPA regulations outlining good modeling protocols.

4. Figure 1-11 summarizes the text and shows a total demand of 18,500 vehicles and an unmet travel demand in the corridor of about 3,400 vehicles per hour (in the peak direction) in the design year, 2020.

5. The transit shares in 2020 are projected with a completely inadequate method, in which the coefficients in the mode choice equations for passenger rail and bus together are the same as in the current bus mode choice model, which represents only the bus mode. These bus mode choice equations were based on the travel survey done in 1993, which contains only bus transit trips. The DOT and modeling agencies should have borrowed mode choice coefficients from a region with a mixed system of rail and busses, such as Seattle. This is standard practice in all urban regions modeling the rail mode without rail trips in the data set on which the local models are estimated. As an example of this flaw, take the coefficient for in-vehicle time cost. It is well-known that traveler time costs on board rail are lower than on board buses, due to the better ride. Time cost is the largest cost in the mode choice model. The models used underproject transit ridership in 2020 substantially. It is very likely that the transit ridership will be high enough to eliminate the 3,400 trips of unmet need. This becomes especially obvious when one considers that many households and businesses will move to be near to the rail stations as the two lines are completed. The population and employment projections used did not account for the new rail lines and station areas.

6. Even more importantly, in the No Build case substantial land use changes will occur, from the year 2000 to 2020. New development in the north corridor and north from there will be suppressed and development in other subareas of the region will increase. Also, as mentioned above, workers will change place of residence and work, to reduce their commute distance and to be able to ride rail. From my literature reviews and from my modeling in the Sacramento, California region with two different urban models (that include both travel and land development models), I conclude that reductions in new growth of up to 30% in the corridor and 50% north of there, would be very likely. These households and businesses would locate nearer to the CBD and in other subareas in the region. The average reduction in new growth of 40%, applied to the increment of demand from 1993 to 2020 (7,600 peak-hour vehicle trips) gives a reduction in trips of about 3,000. This reduction is in addition to that due to more realistic transit projections.

7. Last, I wish to comment on the use of fixed peak-hour traffic factors. The authors of the DEIS used a fixed 9% of daily trips assigned to the peak hour (1-26). The use of a fixed factor is not correct, as the peak will "spread" due to the higher levels of congestion, especially in the No Build case. A lower factor, such as 7%, would be much more reasonable. Such a correction (going from 9% to 7%) is a 22% reduction in the peak-hour volumes. This correction, alone, lowers the peak-hour design volume from 18,500 to 14,430. This reduction is larger than the asserted unmet demand in Figure 1-11 (3,400).

8. To summarize, the modeling used several completely unrealistic assumptions about population and employment growth, future transit shares, household and employment location (and trip lengths), and peak-hour factoring. Corrections in any one of these four erroneous methods elimates the unmet travel needs.