Comparing the Quality of OD Matrices in Time and Between Data Sources
Tim Pollard, Mott MacDonald, Nick Taylor, Mott MacDonald, Tom van Vuren, Mott MacDonald
In times of austerity it is useful to be able to use old survey data with confidence. To establish its value, a comparison between an old OD data set and a (partial) new one should be carried out. We explore (statistical) problems when trying to do so and explore alternative ways to compare full or partial trip matrices
Essential but expensive to carry out, roadside interviews are important building blocks for travel demand models. As a rule of thumb, they cost approx. 10 euros per trip record, and capture around 10% of the total passing traffic – expansion factors of 10 are not uncommon. Many (highway) assignment models rely heavily on these types of survey.
The UK Department for Transport requires data in models to be not older than 6 years. In a period of austerity, it is not unreasonable to query whether older roadside interviews will suffice for a model application, or at least to try and use older RSIs, appropriately weighted, in OD matrix estimation. Alternatively, other sources for observed travel patterns are considerably cheaper, but somehow the profession demands that their value is proven by comparison with traditional roadside interviews. An interesting paper by Potter et al raises concerns about the dependence on roadside interviews in modern transport models.
In the re-estimation of the PRISM model for greater Birmingham we are in the fortunate situation to have access to roadside interviews from 2001 and 2011. We also have access to select link analyses from the actual demand model applied to base year planning data, plus observed trips at RSI sites from GPS-enabled satnav devices. This allows us to compare RSIs with a rich set of alternative sources, individually or aggregated in screenlines and cordons. In our paper we discuss the following:
- The complexities in comparing OD matrices at roadside interview sites and in general, using standard statistical tests
- Alternative comparisons using both GIS techniques and data aggregation
- Examples using different data sources available to us
It is our intention to develop in the paper advice on whether the quality of old OD matrices obtained from RSIs or other surveys can be assessed quantitatively and objectively, challenging the potentially wasteful practice of repeating surveys every 6 years or so. In addition, we hope to add to the knowledge base on the value of GPS-based trip matrices as an additional or even replacement data source to roadside interviews.
Association for European Transport