NorMITs Demand
In short, NorMITs Demand is Transport for the North's (TfN) mainland GB demand tools. These tools started out as Northern specific models, however they are currently moving towards more flexible zoning to allow them to be applicable to elsewhere in mainland GB too.
We're keen to support sharing these tools where we can, see support for more information.
Contents
Summary
NorMITs Demand is made up of a number of smaller models, each with their own individual purpose. For further information on how these models work, click the links in their names. Currently, the NorMITs Demand models are capable of building the following:
- Northern Trip End Model (NoTEM) - MSOA modelled trip ends (base and future year) based on the provided Land Use and National Transport Survey data.
- Travel Market Synthesiser (TMS) - Builds the synthetic base year matrices, by distributing the trip end data provided by NoTEM.
- External Forecast System (EFS) - Future Year forecast matrices, built on top of the Post-Matrix-Estimation matrices provided, the trip ends from NoTEM, and the more accurate distributions from TMS.
- Elasticity Model - Cost-change adjusted future year forecast matrices, taking exogenous cost changes into account, and adjusts the outputs of the EFS matrices.
- NorMITs Matrix Tools - A collection of tools for manipulating matrices that underpins much of the work done by TMS and EFS.
NorMITs Demand has been built in a modular way, so that each sub-model can be swapped out for any model that uses the same inputs to make the same outputs. For example, if you already have a set of base and future year trip ends that you would like to use in-place of NoTEM, they can be slotted in. How these models interact can be seen below, grey and red boxes indicate where alternate inputs/outputs could be used:
Quick Start Guide!
This section will be updated on more detail once we've produced a front-end to all our models, making the interaction between each much simpler!
Information on how to install Python and the required packages is given in Python environment instructions.
Required Data
To get NorMITs Demand running numerous data inputs are required. Due to the nature of the data required licenses are needed, therefore we are unable to share this data alongside our models here. For more information on how to get hold of this data, please see the sharing section.
As mentioned above NorMITs Demand is modularised, therefore each model requires a different set of inputs. Some models will feed data into others, it will be made clear where this is possible.
NoTEM
NoTEM is the most complex model in terms of inputs. The inputs are described in more detail later on. Briefly, these inputs can be broken down into: Land Use data, trip rates, and splitting factors.
Inputs
Land Use data covers residential (where people live) and non-residential (shops/work/schools) data, currently this needs to be segmented by MSOA for NoTEM to be able to pick it up. We're working on making this more flexible. Land Use data is required for every year that trip ends need to be produced for.
Trip rates are the crucial input for converting Land Use data into productions and attractions. Our trip rates are extracted from analysis of National Transport Survey data.
Splitting factors are the various types of splits that are used by NoTEM to split data into categories such as mode shares and time period splits. Similar to trip rates, our various splitting factors come from analysis of National Transport Survey data.
Outputs
NoTEM outputs 4 sets of trip ends per year: home-based productions, home-based attractions, non-home-based productions, and non-home-based attractions. These trip end outputs can be fed into TMS and EFS. By default there are segmented by:
| Segment | Description | Categories |
|---|---|---|
| Purpose | As NTEM | 14 |
| Mode | As NTEM | 5 (No M4) |
| Gender | As NTEM | 3 |
| Soc | Skill Level Proxy | 4 |
| Ns-SeC | Income Level Proxy | 5 |
| Car Availability | Simplified NTEM | 2 |
| Time Period | As NTEM | 6 |
TMS
TMS is designed to take the base year trip ends and accurately distribute them into a synthetic trip matrix. Once the trip ends are made, the inputs to TMS are comparatively simpler.
Inputs
Firstly, TMS requires a set of trip ends to distribute. Segmentations used here are usually fairly aggregate compared to NoTEM outputs so the sample sizes stay significant.
The zoning system that TMS is to run at needs to be defined, along with an "internal" and an "external" definition of zones. The internal area is the area of focus, whereas the external area would be the rest of mainland GB, and would be more disaggregate.
To aid distribution TMS requires some segmented cost distributions to aim towards while distributing the given trip ends. Alongside this, a set of cost matrices that define the cost between each and every zone is required.
Outputs
TMS outputs a set of Production-Attraction (PA) base year matrices. These matrices will be segmented the same as the input data. The outputs matrices can be used by EFS as the base year distributed demand on which the future year demand is based off of.
EFS
EFS in its simplest form is designed to take the base and future year synthetic trip ends alongside a set of post-matrix-estimation (post-ME) matrices. The synthetic data is then used to grow the post-ME matrices into future years and re-distribute. More detail on what EFS can take into account in future years can be found here.
Inputs
Firstly, EFS requires a set of base and future year trip ends to distribute, these can be supplied by NoTEM. As the post-ME matrices should be quite accurately distributed, the EFS trip ends can be more segmented than the TMS ones.
EFS also requires a set of Post-ME matrices to base its future year matrices off of. These matrices should be as accurate as possible to produce more accurate future year predictions.
Outputs
EFS outputs a set of PA future year matrices. EFS outputs a set of PA future year matrices. The outputs can be further fed into the Elasticity Model to reflect exogenous cost changes on demand.
Elasticity
Lastly, the elasticity model aims to take a set of future year PA matrices alongside a set of future cost changes. The cost changes are then applied to the demand represented in the PA matrices, moving demand between different modes of transport.
Inputs
First, the elasticity requires the PA matrices to adjust. These can be provided by the EFS.
Second, the elasticity requires some cost definitions:
- The desired cost adjustments. Proportional and by cost component.
- A definition of how to calculate Generalised Cost from the cost components.
- A definition of the costs between each and every zone.
Outputs
The elasticity model outputs set of cost adjusted PA future year matrices. These outputs will be in the same format as the input matrices, but with adjustment factors applied.
Documentation
Current code documentation can only be found in the codebase. Planned improvements include hosting this online, so it's easily accessible!
Planned Improvements
This section is a work in progress and will be updated with more detail. Planned updates include:
- Online hosting of documentation
- NoTEM
- Zoning system flexible
- Northern tram infill
- Integration of visitor economy data
- Improvement of attraction modelling
- Multi-Modal TMS running (currently only car and rail)
- Improved multi-cost-distribution gravity model calibration
- Multi-Modal EFS running (currently only car and rail)
Sharing
TfN's Technical Assurance, Modelling and Economics (TAME) team have done a great deal of work to develop TfN’s Analytical Framework. As part of this, we would like to start sharing some of our tools, NorMITs Demand included.
Sharing Models
We've categorised out ways of sharing into 3 different profiles, each with a distinct risk/reward profile.
1. Utilisation of Open Source tools and casual TfN support.
This includes forking our repository and mostly working without TfN support. This profile would be facilitated though submitting issues and TfN clarification supporting where possible.
2. TfN Builds outputs for requester
Data requests will be submitted using the requests process. TfN will then assess the feasibility, and aim to build and hand over the required outputs.
3. TfN Builds outputs with requester
Data requests will be submitted using the requests process. TfN will then assess the feasibility, and a discussion will begin to decide how best to work together to produce the required output.
Sharing Requests
If you are interested in acquiring any of the data, or would like some support in utilising NorMITs Demand, please submit your requests to data.requests@transportforthenorth.com.
All requests should contain the following information:
- Requestor Name
- Requestor Organisation
- Request Date
- Required by date
- Expected deliverables
- Format required, where possible
- Purpose of data
- Will data be published?
- Comments
Please note that the TAME team are in high demand with limited resources so responses to requests may not be immediate. However, the team will endeavour to provide you with an estimate of how long it would take to share the data.
Gory Details
This section talks about how each of the models in the NorMITs Demand suite work in detail. It will provide more insight into the transport methodologies used, and the coding detail that makes it so fast. If you are looking for more of an overview, look here!
This section is currently a work in progress and will be updated. An overview of the points we plan to provide more information on can be found in each section.
Northern Trip End Model
- DVector - how NoTEM runs quickly
- Flexible zoning - how to use
- Flexible segmentation - how to use
- Tram infill
Travel Market Synthesiser
- 3 dimensional Furness
- Gravity Model
- Upper and Lower tier distributions
External Forecast System
- Working from home adjustments
- Bespoke zone handling
NTEM Forecasting
Elasticity Model
- Types of elasticities
- VoT/VoC adjustments
- Generalised Cost approach
Matrix Tools
- Segment Tier Converter
- PA to OD conversion
- OD to PA conversion