Configuration and input data

This page describes the structure and format of inputs required for building MESSAGEix-Transport.

Both input data and configuration are stored in files under message_ix_models/data/transport/.

In most cases, these files are read from a subdirectory like …/data/transport/nodes/, where nodes denotes the message_ix_models node code list—for instance, “R12”—for which MESSAGEix-Transport will be built. This value is retrieved from the model.Config.regions setting on a Context object.

  • If the file data or configuration settings have a node (\(n\)) dimension, the file must be placed in such a subdirectory. Data for one node list is not usable for base models using a different node list.

  • For other data, a node list–specific file may be used. If none exists, the file of the same name in …/data/transport/ is used as a default. For example, …/data/transport/R12/set.yaml is used if it exists; if not, then …/data/transport/set.yaml is used.

Configuration

General (config.yaml, required)

The contents of this configuration file exactly map to the attributes of the class transport.Config. The class stores all the settings understood by the code for building, solving, and reporting MESSAGEix-Transport. The class also defines the default values for each setting (there is no file …/data/transport/config.yaml containing defaults.) It also has methods for reading the configuration from file; see the detailed description of Config.from_context().

The following is the configuration file for a base model with R12 nodes:

→ View data/transport/R12/config.yaml on GitHub

Technology code list (technology.yaml)

This file gives the code list for the MESSAGE technology concept/set/dimension. Some annotations (iea-eweb-flow, input, report) and the child hierarchy give information about technologies’ grouping according to transport modes.

→ View message_ix_models/data/transport/technology.yaml on GitHub

Code lists for other MESSAGE sets (set.yaml)

This file gives code lists for other MESSAGE concepts/sets/dimensions.

→ View message_ix_models/data/transport/set.yaml on GitHub

Input data files

transport.files.FILES gives a list of all data files. Through transport.build.main() (ultimately, transport.build.add_exogenous_data()), each of these files is connected to a genno.Computer used for model-building, and its contents appear at the key given in the list below.

Example

Contents of the file freight-mode-share-ref.csv are available at the key freight mode share:n-t:ref. The indicates the dimensionality of this quantity is \((n, t)\). The file has column headers “node”, “technology”, and “value”.

Not all files are currently or always used in model-building computations. Some submodules of transport use additional data files via other mechanisms. Most of the files have a header comment including a precise description of the quantity, source of the data, and units of measurement. In some cases extended information is below (where a header comment would be too long). The git history of files, or the GitHub “blame” view can also be used to inspect the edit history of each file, line by line.

ldv-activity.csvldv activity:n:exo

Measure

Activity (driving distance) per light-duty vehicle

Units

kilometre / year

Notes

node = R12_AFR [1]

Obtained from literature, based on estimates from South Africa. The reported value for South Africa is lower (18000 km/year, source) than the one for Kenya (22000 km/year, source).

node = R12_FSU [1]

Based on Russia estimates (source).

node = R12_NAM [1]

Based on US estimates (source <https://afdc.energy.gov/data/10309>`__), Canada estimates tend to [be] lower in general.

node = R12_PAO [1]

Estimates for AU is 11000 in 2020, it’s a sharp decrease from 12600 in 2018 (maybe a Covid effect?). Whereas JP is 8532 (source) in 2016.

node = R12_PAS [1]

Based on Singapore by Chong et al. (2018).

node = R12_SAS [1]

Based on India, mainly Delhi estimate by Goel et al. (2015).

ldv-fix_cost.csv, ldv-inv_cost.csv, ldv-fuel-economy.csv

Data on costs and efficiencies of LDV technologies.

Formerly this data was read from ldv-cost-efficiency.xlsx, a highly-structured spreadsheet that performs some input calculations. The function get_USTIMES_MA3T() reads data from multiple sheets in this file. To understand the sheet names and cell layout expected, see the code for that function.

As the name implies, the data for MESSAGE (V)-Transport was derived from the US-TIMES and MA³T models.

mode-share/default.csv

Measure

Share of each mode in passenger transport activity in the model base year

Dimensions

\((n, t)\) with transport modes expressed in the \(t\) dimensions.

Units

dimensionless

Notes

node = R12_AFR [2]

These new estimates are mainly based on IMAGE regional estimates (average of EA, WA, and SA) after discussion with Jarmo as well as an additional literature search + guesstimates from vehicle count etc. Still, no comprehensive source to validate these. Only broad qualitative impressions formed from the literature. More details in [other] notes.

node = R12_CHN [3]

Based on the total pdt and mode share breakdown from 2017^ as reported in Liu, et al. 2022, and extrapolating to 2020 (assuming the mode share in 2020 is the same as the one in 2017).

Subtracting Waterways from PDT. RAIL includes both urban PT & RAIL. BUS includes both local buses and COACH.

node = R12_MEA [2]

These new estimates are mainly based on IMAGE regional estimates (average of ME & NA) guesstimates from vehicle count etc. Same as [R12_AFR].

node = R12_PAO [2]

Estimated from weighing Japan (0.80) & Aus/NZ (0.2) by population. JP source is ATO, Statistics Japan, IEA. AU source is BITRE 2021. Motorcycle share is guess-timate based on no. of motorbikes, load factor, and comparing it to cars. BUS estimate for Japan is based on ATO data, less certain as source is missing. More details in [other] notes.

node = R12_SAS [2]

Estimated from India ATO & OECD sources. Rest of SA is likely to have lower RAIL share. 2W share also includes 3W (Auto rickshaw).

pdt-cap-ref.csvpdt:n:capita+ref

Measure

Passenger distance travelled per capita in the model base year

Dimensions

\((n)\)

Units

km / year

Notes

node = R12_CHN [4]

Based on the vehicle activity method Liu, et al. 2022 estimate the total PDT for R12_CHN for year (2017) is 9406 billion pkm. This is the latest corrected estimate available from Liu, et al. 2022. Based on similar estimates for 2013 & 2015, I estimate the average growth of PDT to be 8% per year. Using the growth rate and 2017 estimate, the total PDT for year (2020) comes out to be 11848.9 billion pkm.

R12_CHN population estimate from IMAGE: 1.4483 billion

the PDT/capita = 11848.9/1.4483

pdt-elasticity.csvpdt elasticity:scenario-n:exo

Measure

“Elasticity” or multiplier for GDP PPP per capita

Dimensions

\((n, \text{scenario})\). ‘scenario’ identifiers are partial URNs for codes in the SSP_2024 code list.

Units

dimensionless

Where/how used

pdt_per_capita().

Other files

  • demand-scale.csvdemand scale:n-y:exo

  • disutility.csvdisutility:n-cg-t-y:per vehicle

  • energy-other.csvenergy:c-n:transport other

  • freight-activity.csvfreight activity:n:ref

  • freight-mode-share-ref.csvfreight mode share:n-t:ref

  • fuel-emi-intensity.csvfuel emi intensity:c-e:exo

  • ikarus/availability.csvikarus availability:source-t-c-y:exo

  • ikarus/fix_cost.csvikarus fix_cost:source-t-c-y:exo

  • ikarus/input.csvikarus input:source-t-c-y:exo

  • ikarus/inv_cost.csvikarus inv_cost:source-t-c-y:exo

  • ikarus/technical_lifetime.csvikarus technical_lifetime:source-t-c-y:exo

  • ikarus/var_cost.csvikarus var_cost:source-t-c-y:exo

  • input-base.csvinput:t-c-h:base

  • ldv-class.csvldv class:n-vehicle_class:exo

  • ldv-new-capacity.csvcap_new:nl-t-yv:ldv+exo

  • load-factor-ldv.csvload factor ldv:n:exo

  • load-factor-nonldv.csvload factor nonldv:t:exo

  • ma3t/attitude.csvma3t attitude:attitude:exo

  • ma3t/driver.csvma3t driver:census_division-area_type-driver_type:exo

  • ma3t/population.csvma3t population:census_division-area_type:exo

  • mer-to-ppp.csvmer to ppp:n-y:exo

  • population-suburb-share.csvpopulation suburb share:n-y:exo

Other data sources

transport makes use of the tools.exo_data mechanism to retrieve data from common (not transport-specific) sources. DataSourceConfig, transport.Config.ssp, and other settings determine which sources and quantities are used.

These include:

  • GDP and population from the project.ssp data sources or other sources including the ADVANCE project, the Global Energy Assessment project, the SHAPE project, etc.

    Note

    Formerly, file gdp.csv was used.

    This is no longer supported; instead, use databases via exo_data.prepare_computer() or introduce quantities with the same dimensions and units into the Computer used for model building/reporting.

  • Energy from the IEA Extended World Energy Balances.

  • IEA_Future_of_Trucks.

  • MERtoPPP.