1. TRIMODE model & rail O/D data
Paolo BOLSI,
DG MOVE A3, Economic Analysis & Better Regulation
Eurostat Rail Transport Statistics Working Group
Luxembourg, June 2017

2. Summary 1. TRIMODE model overview 2. Freight data overview – all modes
3. Rail freight data – gaps, issues and actions
4. Rail passenger data - gaps, issues and actions
5. Why TRIMODE needs freight segmentation by NST?

3. TRIMODE Model Overview
Model covering all modes of passenger and freight transport for all of Europe
Includes integrated sub-models: transport demand; network assignment; fleet; energy; economy.
NUTS3 level zoning, plus detailed modal networks
Base year 2010, validation year 2015
- 5-year-steps projections up to 2050
Phase 1 (2017) development and testing of 6 countries: AT, DE, IT, NL, SI, UK
Phase 2 (2019) will extend to cover remaining European (and neighbouring) countries
Access to good data is crucial to success:
- in the model development and testing currently underway
- based on Eurostat data to the maximum extent feasible
- to simplify updating of the model in future years by EC DGs

4. Overview: OD Data Availability & Completeness
Requirement for 2010 is a complete OD tonnes matrix
segmented by NUTS3 x NUTS3 x NST2007 (type of goods) x Mode
Eurostat data availability in 2010 for Phase 1 countries is
OD matrix – Origin Destination matrix of tonnes transported
Road will use 2011 NUTS3xNUTS3 data to spatially disaggregate 2010 control totals
IWW port locations will be used to disaggregate from NUTS2 to NUTS3xNUTS3 level.
Air initially allocate all tonnes to NST class 15: mail and parcels. Later after further analysis, reallocate some from NST15 to other consumer goods NSTs.
Maritime needs some processing but the basic ingredients needed are available.
Due to the major gaps in Eurostat tables we will need to rely on national data sources for pipeline tonnages.
However, pipeline network is of limited size and has just NST 2: crude oil & NST 7: refined petroleum products - so this task appears feasible.
Rail freight data is the main issue to be discussed in later slides.
We already have UK rail OD matrices for 2010 and 2015 from within the team (NUTS3xNUTS3xNST).

5. Overview: National Data Coverage & Consistency
Requirement is that the tonnes transported are defined in a fully consistent fashion across all combinations of: Countries – Modes – Years Otherwise, if the model matched well to the 2010 pattern it would be guaranteed to be wrong for 2015 Eurostat consistency of tonnes for Phase 1 countries is:
Mode
Yearly total
Country NST
Yearly NST
Road OK
DE, NL, SI, UK have some issues
IWW
OK except IT: 2011 &15
IT, UK have issues – small flows
Air
OK – NST:15 - freight & mail
Maritime
Cargo type - OK
Pipeline
Gaps
Rail
Gaps except 2012 &14
Some years OK
IT, NL, UK have issues
Road - NST consistency issues are mainly in the earlier years and in general can be worked around
IWW – Consistency seems good throughout for the major flows
Air and Maritime - both appear consistent throughout
Rail – has some major gaps and issues

6. Rail Freight Data Issues – OD matrices, 2010
Only 12 European countries match to the total tonne-km aggregate within 1% of difference
No OD data at all for Austria, Belgium, Romania
For other Phase 1 countries
major missing OD tonnage data for Italy (-24%) and Netherlands (-14%)
smaller missing tonnages for Slovenia (-5%) and UK (-7%)
Germany matches fully
Source: comparison of Eurostat tables
for annual national totals <rail_ga_typeall>
for 2010 OD based totals <tran_r_rago>

7. Comparison of Eurostat 2010 total rail tonnes (000s):
national <rail_ga_typeall>
OD based <tran_r_rago>

8. Rail Freight Data Issues - NST
No NST segmentation at all in the rail OD dataset
NST only available at country level - at best!
not zonally even at NUTS1 level
Half of all European countries have data gaps in their total national tonnes
For Phase 1 countries:
AT (83%), IT (64%) & NL (31%) of expected totals
But DE (100%), UK (100%) and SI (96%) do match OK in 2010
Source: comparison of Eurostat tables
for annual national totals <rail_go_typeall>
for NST based totals <rail_go_grpgood>
Major issues need to be resolved so to use rail OD data segmented by NST for base year

9. Comparison of Eurostat 2010 total rail tonnes (000s):
national <rail_ga_typeall>
NST based <rail_go_grpgood>

10. NST national consistency over time, Phase 1 countries – Rail
A constant 12.5% spread of colours over 8 years & 20 NSTs would suggest consistent data for country
AT, DE and SI appear broadly OK
IT is missing some or all NSTs in 2008, 2010 & 2013
NL is missing some or all national and transit traffic
UK shows inconsistencies between early and later years for various NSTs
Source: Eurostat NUTS0 table ‘rail_go_grpgood’

11. Freight data - National sources
Requirement is:
complete (calibration) and (validation) OD rail tonnes matrices
segmented by NUTS3 x NUTS3 x NST2007 (20 types of goods)
otherwise a close approximation to these segments
plus meta-data on the source of observations and processing
Observed data is needed – not the output from a model

12. Freight OD matrix availability with NST - 1
Phase 1 countries
DE (2010 only) and UK – matrices are already available
AT – potentially available from
Directorate Business Statistics – Transport
NL – potentially available from
inputs to BasGoed model
SI - contact
Slovenian Ministry of Infrastructure and the Slovenian statistical office
IT – possibly available from:
Italian Ministry of Transport – inputs to SIMPT model
RFI - inputs to SAVEF model

13. Freight OD matrix availability with NST - 2
Phase 2 countries
BE– potentially available from
Belgian Federal Planning Bureau – inputs to PLANET model
SE – potentially available from
inputs to SAMGODS model
NO – potentially available from
inputs to national model
Other countries
Hopefully similar national sources can be made available

14. Passenger data - National sources
Requirement is
complete (calibration) and (validation) OD rail passenger matrices
segmented by NUTS3 x NUTS3
otherwise a close approximation to these segments
plus meta-data on the source of observations and processing
Observed data is needed – not the output from a model

15. Comparison of Eurostat 2010 total rail passengers (000s):
national <rail_pa_total>
OD based <tran_r_rapa>

16. Passenger OD matrix availability
Phase 1 countries
UK – matrices are already available
though require significant processing
AT – possibly available from Statistik Austria
DE - possibly available from DESTATIS
NL – to fill major gaps and to obtain OViN mobility data, contact
CBS - statistical agency and Dutch Ministry of Infrastructure and Environment
SI - contact
Slovenian Ministry of Infrastructure and the Slovenian statistical office
IT – Unlikely to be further data available
Phase 2 countries
Hopefully similar national sources can be made available

17. Tonne kilometres (bn) by NST by reporting country, 2010: (note difference in scales and in pattern across NSTs)
Rail
Road
Why is it important to have our freight matrices segmented by NST / type of goods in a consistent fashion for each mode across years?
tkm across Europe differ in relative importance by NST class between road and rail:
Road: NST4 – food products; NST1 - agricultural goods have highest tkms
Rail: NST19 – containers; NST7 - refined petroleum products have highest tkms
But note total tkm for NST7 is around 50 billion for both road and rail – difference in scale.
The model needs to segment by NST in order to pick up these important differences in competitiveness between modes.

18. EU total by distance class & NST07 class, road, 2013
Tonnes (million) lifted
Vehicle-kilometers (billions)
Looking just at different units of measurement provide different insights for policy making:
NST3 – ores and quarry products have the largest tonnes lifted
But they are mainly loaded on large trucks (heavy bulk products) moving short trip lengths (blue<50kms) so generate lower impacts on traffic congestion (vkms) than might otherwise be expected.
NST4 – food products are moved long distances with low tonnes per truck (light goods – individually packaged sometimes in smaller trucks used for local deliveries) often into urban areas and so have much larger impacts on road congestion than is indicated just by their tonnes lifted.
NST18 – grouped goods is similar to NST4

19. Conclusions
TRIMODE is an important modelling project for the European Commission, which will support the analytical work accompanying the policy initiatives in the transport field Network-based models are data-intense and require O/D detailed data for development and calibration. Using high quality detailed data is a precondition for reliable results! DG MOVE looks forward any detailed O/D data in rail transport, possibly at NUTS3 level, and by type of goods.

20. Thank you for your attention!