Matrix Adjustments – How to build better matrices
- 1. Insert Heading Matrix Adjustments – How to build better matrices. Plus a National Update on the TMN 11 November 2015
- 2. Topics • Matrix Adjustments / Matrix Estimation – approaches available to create matrices (base and forecast year) – common mistakes – mitigating these mistakes – reporting • Transport Modellers Network – an update on the TMN following the recent national meeting; – an interactive session on the direction of local TMN branch: o broadening the reach of TMN o objectives for TMN locally / nationally o how best the TMN can service the local members o hot topics that should be discussed o other issues/ideas
- 3. Matrix Adjustments / Matrix Estimation • Derivation of demand matrices that can be used in Transport Models – Utilise traffic count data to adjust matrices – Used to derive base year demand i.e. require existing count information • Reasons for use – Most often: lack of data – Fast efficient process – Refining existing matrices to improve accuracy – Requires little data
- 4. Approaches • Preferred approach – – Trip Generation to derive origins/destinations – Trip Distribution to generate OD demand – Mode Choice/Time Period Splicing to derive demand to be assigned • In the absence of information, or where not suitable to apply, three main approaches to deriving matrices – Matrix Estimation/Adjustment <-Base Year – Furnessing/Fratar <-Base Year/Forecast Year – Manual Adjustment <-Base Year/Forecast Year
- 5. Approaches • Base Year Inputs – – Preferably a prior matrix – At a min should have origin and destination totals – No input should be treated very carefully – Bluetooth data? – Traffic Count data • Approach should be assessed based on information available, data quality and timelines of project
- 6. Common Mistakes • Matrix Estimation is a mathematical process – It does not think – It does not question outputs – It attempts to produce the best statistical fit – It is only good as the inputs provided • So what does that mean? – A good r-squared is not necessarily a good indication of a fit-for-purpose matrix – Don’t trust the results on face value – But a good fit may indeed be a good result
- 7. Common Mistakes • Common Mistakes made include – Not checking that demand has been assigned to illogical OD routes – Checking that the key movements still contain the majority of the demand – Confirming that the matrix has not been altered significantly – Assuming that a good r-squared confirms that the demand is suitable – Assuming that errors in the fit are attributed to the demand – Software
- 8. Mitigating Mistakes • Illogical OD Routes – Banned Turn – Demand assigned to OD Pair 1->2 – Resultant trip 1 2 6 5 3 4 1 2 6 5 3 4
- 9. Mitigating Mistakes • Illogical OD Routes – Trip recorded against multiple counts – For simple networks • Check prior matrix for OD pairs that are not sensible to be included • Ensure no large allocation applied before or after estimation – For complex/large networks • Note possible OD pairs that would ideally use turn • Check quantum • Review path data i.e. select-link 1 2 6 5 3 4
- 10. Mitigating Mistakes • Fixing Demand – The performance of the matrix may vary by location – Certain cells or row/column totals may be known – Good Matrix Estimation Tools will allow you to fix cell values, rows / column totals – Provides flexibility and helps remove potential distortion / compensating errors – Allows for new land developments not to distort process Matrix 1 2 3 4 5 6 1 2 3 4 5 6
- 11. Mitigating Mistakes • Reviewing Distribution – Important to understand changes to matrix – Need to understand if large/small multipliers have been used – Large/small adjustments imply poor fit of prior matrix – Need to understand if adjustments valid or masking an underlying issue – What is impact of changes
- 12. Mitigating Mistakes • Reviewing Distribution – Calculate factors applied on OD level – Adjustment Factor = Estimated Matrix / Prior Matrix – Plot as a distribution – Should follow normal distribution if good fit of prior – Quick and easy check 0.0% 2.0% 4.0% 6.0% 8.0% 10.0% 12.0% 14.0% 16.0% 18.0% 20.0% 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 ProportionofODCells ME Adjustment Factors
- 13. Mitigating Mistakes • Reviewing Distribution – Important to review prior and post estimation matrix statistics • Total Demand • Vehicle Kilometres Travelled • Vehicle Hours Travelled • Trip Ends Profile / Trip Distribution profile – Large changes indicate change in travel patterns and a need to investigate further
- 14. Mitigating Mistakes • Network, Counts and Assignment – Don’t just review the demand matrices – Is network correctly coded? • Capacities of links? • Junction layouts? • Phasing? • Turn bans? – How reliable is the count information supplied? • Is it current year? • What type of count? • Do counts provide screenline to assess demand? 1 2 6 5 3 4
- 15. Mitigating Mistakes • Network, Counts and Assignment – What do the travel paths look like? • Select link • Key corridors • Bandwidth plots – Count performance by location? • Systematic/geographical bias/issues? – Centroid loadings • Suitable location • Representative of network access/egress
- 16. Reporting • TELL A STORY! – If the changes or the results seem counter-intuitive are they correct? – Report on the process adopted • Document assumptions – where professional judgement applies • Refer to guidelines for validation/confirmation of suitable process adopted • No of iterations, constraints, checks, source of prior (inc development) – Report on the level of performance and provide key reporting statistics • Change in network stats • Distribution of factors • Modelled vs observed (scatter plot) • Screenline demands (bar charts)
- 17. TMN National Meeting • AITPM TMN Vision: – Better connecting people and organisations within the discipline and “outside” the discipline – Manage modelling stream of National Conference and Forum annually (as a focus) in partnership with the State Committee – Hosting State based events – Enabling national and international links – Doing all this within the umbrella of AITPM
- 18. TMN In South Australia • How do we broaden the reach of TMN? • What should be the objectives for TMN locally / nationally? • How best can the TMN can service the local members? • What are the hot topics that should be discussed? • Any other issues or ideas?