DEVELOPMENT OF OD MATRICES USING MOBILE PHONE DATA FOR M25 JUNCTION 30 LOWER THAMES CROSSING TRANSPORT MODEL



DEVELOPMENT OF OD MATRICES USING MOBILE PHONE DATA FOR M25 JUNCTION 30 LOWER THAMES CROSSING TRANSPORT MODEL

Authors

Shaleen Srivastava, Jacobs UK Limited

Description

DEVELOPMENT OF OD MATRICES USING MOBILE PHONE DATA FOR M25 JUNCTION 30 LOWER THAMES CROSSING TRANSPORT MODEL

Abstract

Jacobs’ were commissioned by the Highways Agency to develop a new traffic model for the M25 Junction 30 Congestion Relief Scheme and to assess Lower Thames Crossing scheme options. Jacobs’ proposed the use of anonymous mobile phones data processed using its data fusion process ‘TDiF’ for the development of base year demand matrices for this model. It was agreed with the Highways Agency (and the Department for Transport) that this is the most appropriate data to build demand matrices, knowing the size of the model, accuracy required and the available time constraints. However, it was further agreed that an independent test of the base year demand matrices derived from the data fusion process using anonymous mobile phone data would be carried out, to verify the quality of the output.

The data fusion process processed 26 weekdays of anonymous mobile phone signal movement data to obtain origin and destination information (in pre-allocated zones) of the signal movements. The fusion involved processing around 90 million rows of data and 8 columns across. Effectively, there were 720 million cells of data for the whole operation. This had to be dealt within the SQL database and applying T-SQL logics.

To construct a highway traffic matrix, we processed all captured mobile phone data and converted it into equivalent vehicular movements. This was then expanded based on the expansion factors derived from the data fusion process to estimate all of the traffic within the modelled area, how it traverses the zones and then converts it into traffic matrices. The expanded trip matrices were then segregated into the desired journey purposes using DfT’s Valuation Office data, DfT’s NTS data and Census 2011 data.

A range of roadside interview surveys (RSI) were available for a number of sites in South Essex in the vicinity of M25 J30. A new postcard-based survey was also undertaken in June 2013 to observed northbound traffic patterns at the M25 Dartford Crossing. These RSI data were used as independent sources to verify the trip patterns and trip length profiles of the trip matrices derived from Jacobs’ processing of the mobile phone data.

The trip length distribution of movements from the collected Dartford Crossing RSI sample was compared with that derived from the processed mobile phone data trip matrices.

Traffic patterns at the M25 Dartford Crossing were compared between the RSI survey collected on June 2013 and the processed mobile phone data in terms of trip ends and trip distributions between internal and external sectors. Overall, the results showed a reasonable level of consistency between the two sources of data where the differences in the proportion of trip ends were not statistically significant and the differences in the proportion of inter-sector trips were within the range which could be explained by the differences between the two sources of data and the expected errors within each data source.

Trip lengths of internal to internal trips were compared between the processed mobile phone data, the Dartford Crossing RSI sample, and the processed LATS data. While a close match was found between the processed mobile phone data and LATS data, the RSI data suggested slightly longer internal to internal trips. As discussed earlier, the low number of records for internal to internal trips in the RSI sample suggests a considerable level of uncertainty in the data. Furthermore, the processed mobile phone data tend to include some short trips that might not be included in the Dartford Crossing RSI sample.
Trip distribution patterns were also compared between the processed mobile phone data and the processed RSI samples, collected in 2006 and 2010, for selected inter-sector movements for which estimates could be made using the available RSI data. Overall, given a number of inconsistencies between the two sources of data and uncertainties in the estimates, the results showed a reasonable level of correlation between partial sector-based trip ends and inter-sector movements.

In the absence of any processed planning data, total 12-hour trip ends from NTEM were used to verify trip ends from the processed mobile phone data 12-hour trip matrix in district level within the study area. Despite the inconsistencies between the two sources of data, a strong positive correlation was found between trip ends (both origins and destinations) as well as between the processed mobile phone data trip ends and the 2011 census population.

Publisher

Association for European Transport