Results
Detailed Count Point and Journey Time Results
We were advised that although overall model accuracy
is important, for small area analysis it is also important to examine the individual results
from individual count points
and for specific journey times. This is what we looked at next.
Count Points
Six
count points with most relevance to the proposed link road were identified for individual examination. These were:

Each count point is bi-directional
giving 12 counts in total. For the morning peak period 4 of the 12 met the calibration criteria. The largest errors were:

For the evening peak period 8 of the 12 met the calibration criteria. In this case the two highest failures were:

Three points
can be made about these results:
First, looking across the whole, for neither period was the 85% standard met
– for the evening it was 67%, for the morning just 33%.
Second, some of the individual errors in the count are extremely high. To be under or over-estimating by around 70% raises real doubts about the accuracy of this model.
Third, the failures at count point 10 are particularly significant.
Count point 10
is part of the existing orbital traffic flow from the A47 through East Park Road and Victoria Park Road to Welford Road. This is a heavily used route, particularly in the morning peak, and traffic passing count point 10 will make up a significant proportion of the traffic using Victoria Park Road. Every modelled value at this count point is below the actual traffic count. The error ranges from minus 21% to minus 71%. Combining the four counts the total modelled count is 43% below the actual traffic count.
These figures suggest the volume of traffic along Victoria Park Road recognised by the modelling will be far below the actual traffic on that road as nearly half the traffic feeding in through Mayfield Road is not counted. To the model Victoria Park Road will appear congestion free with spare capacity.
In consequence the modelling process will re-allocate or redistribute traffic from more congested or slower routes and redirect it along Victoria Park Road until journey times through the different routes are equalised – in the model. This is probably how the claimed reduction in numbers of vehicles using Clarendon Park Road, with an increase in vehicles using Victoria Park Road, is achieved in the model, and it is consistent with the increase in distance travelled by vehicles travelling east - west using the new link road in the morning peak hour.
In the real world the situation will be quite different.
Victoria Park Road
is already heavily congested in the morning peak and in all likelihood will become more so with additional traffic entering through the existing Mayfield Road orbital route. The Mayfield Road roundabout
and junction with the A6 London Road is where the Putney Road link is intended to link to, via Victoria Park Road. With additional congestion along London Road inbound, caused by more congestion at the Mayfield Road roundabout and along Victoria Park Road, the actual impact in the real world will be the reverse of that predicted by the model. More vehicles will choose to divert through Clarendon Park Road and other roads, not fewer, and rat-running will increase, not decline.
This is an
example
of how deficiencies in the modelling produce traffic predictions which don't fit with how the real world road network around Clarendon Park operates. Given the scale of the short-comings in the modelling, there will almost certainly be other examples in the traffic forecasts.
Journey Times
We identified 12 specific journeys
on the main radial roads within the wider Area Of Influence
claimed for this model – London Road, Welford Road, Saffron Lane, and Aylestone Road, and looked at their actual
and modelled
journey times for the morning and evening peaks.

Looking first across all the selected journeys in neither of the time periods was the required 85% standard achieved
- it was 66%
for the morning peak, and 75%
for the evening. Once again the validation criteria are not met
in this sub-sample of route segments
Across both time periods there was a general tendency to understate
the modelled journey times meaning that the actual journeys took longer than the modelling showed. However, when the reverse applied and the modelled journey was longer than the actual one the error tended to be very large
– in the worst three cases by 97%, 68%, and 36% for the morning peak, and 43% and 23% in the evening. For the evening peak there were also large errors the other way at 48%, 33%, and 31%.
One route in particular appeared extremely difficult to model
which was along Welford Road between Victoria Park Road and Almond Road at the southern end of the inner ring. For the morning peak the inbound modelled journey time was very nearly double the actual time, and for the outbound journey over a third higher. For the evening peak the outbound time was amplified by 43% but the inbound understated by 14%.
This section of road is one of the keys to forecasting the impact of the Putney Link Road and getting it right will be essential for it is one of the radial routes from which it is claimed traffic will be diverted and on which it is claimed congestion will be reduced. Using journey time as a proxy measure of congestion this model is greatly exaggerating the level of existing congestion
so any claimed reduction is unlikely to be real – it is more likely to be an artefact created by shortcomings in the modelling.