Consensus Elementary Flow Contexts

We built our consensus common set of flow contexts following the lead of Edelen et al, who had a tree of attributes:

We are designing a system which allows for harmonisation across a large number of elementary flow lists while attempting to minimise complexity. This means that we only include context attributes when there are differing characterisation factors provided for the different attributes values. We also need to provide reasonable matches for major flow contexts from systems like SimaPro, EF, and ecoinvent.

We made the following changes:

• In our scheme, resources are limited to accounting concepts like "mass fraction of zinc" or "energy, geothermal, converted", and have the Dimension Resource. A specific type of ore like Sphalerite is a tangible substance present in the environment, so is treated the same as other substances like carbon dioxide or nitrates - it has the Dimension Environmental and a Media description. We assume that these minerals will only be consumed, but that is not required.
• This different approach to resources means that directionality is not a part of context - direction instead comes from inventory model itself. Substances like carbon dioxide can be extracted from or emitted to the atmosphere, but this is the same substance and flow object definition.
• There is no bright line between the economy and the environment, and sometimes we want to characterize flows which don't have a direct environmental impact, including economic factors like value added or circular economy indicators like waste to recycling ratios. We therefore add an additional organising layer on top of Edelen's typology related to the Dimension of the elementary flow, and only require a Media where it is sensible.
• We add the Product media, for substances applied to or present in a product which can have direct effects on health, such as fire-retardant chemicals applied to mattresses.
• We expand the types of Indoor emission classes to include the differentiation given in GLAM.
• Our labels for the secondary context attributes are more explicit, such as "Rural (<1000 people/square mile)", "Medium stack, <150 meters", and "Confined aquifer with fossil groundwater".

Here is our set of contexts:

This typology will change as we apply this theory to real world data and users, and as LCIA methods continue to develop.

Resources take a Media value but not any of the context subclasses.

The nodes shown in italics are classes with their trees of possible values. Here are the possible values for these classes:

Indoor Air Class: Differentiation of indoor air emission classes. Based on GLAM indoor air emissions.

• Unknown
• Near-person
• Residential
• Industrial

Vertical Strata: Vertical strata for the Environmental domain and the Air media.

• Unknown (unspecified outdoor air — no strata information available)
• Ground level
• Low stack, <25 meters
• Medium stack, <150 meters
• High stack, >150 meters
• Aircraft cruise height
• Long-term — not a stratum. A temporal qualifier riding on this field, because strata is the mandatory mutually-exclusive discriminator air already has. It cannot be combined with a real stratum or a population density, so a source that knows both has to give one up.

Population Density: Only applies to the Environmental domain and the Air media. Not included for aircraft or long-term emissions.

• Unknown
• Urban (>1000 people/square mile)
• Rural (<1000 people/square mile)

Geography: Only applies to the Environmental domain and the Ground media. Follows Edelen et al. with the addition of Silvicultural and Unknown to better align with EF 3.1.

• Unknown
• Industrial — a works, a yard, a contaminated site. Both lists that ship an industrial soil compartment file it here. EF 3.1 has four flows here and characterises none of them, so an emission published here carries no factors — see Known limitations.
• Residential
• Commercial
• Agricultural
• Silvicultural — land being worked for wood. This is where a source's forestry compartment goes: a pesticide sprayed on a plantation or the chain oil a harvester loses is an emission from the practice, and the practice is what the compartment names. EF 3.1 has no such compartment, so nothing characterises a flow filed here and the emission is published without factors — see Known limitations.
• Wetland
• Barren land
• Snow and Ice
• Grassland
• Shrubland
• Forest — tree cover, whether or not anybody works it. Pick this only when a source says where the land is, not what is being done to it; a compartment spelled forestry is Silvicultural.
• Non-agricultural — deprecated on arrival. Not a land cover class but EF 3.1's complement of Emissions to agricultural soil, present only so those two EF contexts stop collapsing onto one another. Do not map new sources onto it: prefer a real land cover class, or Unknown.

Water Body: Applies to the Water media on both the Environmental and the Resource domain, and is required on each. It means the body an emission went to, or the body a withdrawal was taken from.

• Unknown
• Surface water — the broader value: lakes and rivers are kinds of surface water, and this is the one to pick when the body is not narrowed to either.
• Lake
• River
• Ocean
• Unconfined aquifer
• Confined aquifer with fossil groundwater
• Waste-water treatment plant — emissions only
• Secondary water treatment — emissions only
• Long-term — emissions only. Not a water body: a temporal qualifier riding on this field, the water-side counterpart of the Long-term vertical strata. A withdrawal has no >100-year counterpart, so the three values above are prohibited on the Resource domain.

Discussion Points

• We could add Space to the vertical strata.
• Three source lists share a compartment that means "in the countryside or from a tall stack", and no single value here can hold both. It is published as Medium stack, <150 meters → Rural, which is neither reading: it is the compartment EF 3.1's own characterisation factors treat it as. That is an inference from factors rather than anything a source states — see Known limitations.

Allowed Context List

The set of allowed contexts is given as a list of JSON objects - for example:

{
  "dimension": "Environmental",
  "context_iri": "https://vocab.brightway.one/flow-contexts/envi-air-grle-ur10pesq",
  "media": "Air",
  "strata": "Ground level",
  "population_density": "Urban (>1000 people/square mile)"
}

There are currently 59 allowed contexts, plus one deprecated. The list below is the whole of it, generated from the data file rather than kept in step by hand; tests/test_documentation.py fails if the two disagree.

Contexts as a List of Strings

Because the set of context attributes is structured as a taxonomy, we can express each possible context as a list of strings, compatible with existing LCA data format such as ILCD. In this formulation, the specification of "Unknown" subclasses is optional, as "Unknown" is assumed if a value is missing.

We note that this way of expressing contexts can be less ambiguous now that indoor air contexts are prefixed with Indoor — for example, ['Environmental', 'Air', 'Indoor', 'Near-person'] vs ['Environmental', 'Air', 'Low stack, <25 meters'].

Here are all contexts as a list of strings; a mapping from the context IRIs to these values is also provided as JSON.

• ['Economic']
• ['Environmental', 'Air']
• ['Environmental', 'Air', 'Aircraft cruise height']
• ['Environmental', 'Air', 'Ground level']
• ['Environmental', 'Air', 'Ground level', 'Rural (<1000 people/square mile)']
• ['Environmental', 'Air', 'Ground level', 'Urban (>1000 people/square mile)']
• ['Environmental', 'Air', 'High stack, >150 meters']
• ['Environmental', 'Air', 'High stack, >150 meters', 'Rural (<1000 people/square mile)']
• ['Environmental', 'Air', 'High stack, >150 meters', 'Urban (>1000 people/square mile)']
• ['Environmental', 'Air', 'Indoor']
• ['Environmental', 'Air', 'Indoor', 'Industrial']
• ['Environmental', 'Air', 'Indoor', 'Near-person']
• ['Environmental', 'Air', 'Indoor', 'Residential']
• ['Environmental', 'Air', 'Long-term']
• ['Environmental', 'Air', 'Low stack, <25 meters']
• ['Environmental', 'Air', 'Low stack, <25 meters', 'Rural (<1000 people/square mile)']
• ['Environmental', 'Air', 'Low stack, <25 meters', 'Urban (>1000 people/square mile)']
• ['Environmental', 'Air', 'Medium stack, <150 meters']
• ['Environmental', 'Air', 'Medium stack, <150 meters', 'Rural (<1000 people/square mile)']
• ['Environmental', 'Air', 'Medium stack, <150 meters', 'Urban (>1000 people/square mile)']
• ['Environmental', 'Biotic']
• ['Environmental', 'Ground']
• ['Environmental', 'Ground', 'Agricultural']
• ['Environmental', 'Ground', 'Barren land']
• ['Environmental', 'Ground', 'Commercial']
• ['Environmental', 'Ground', 'Forest']
• ['Environmental', 'Ground', 'Grassland']
• ['Environmental', 'Ground', 'Industrial']
• ['Environmental', 'Ground', 'Non-agricultural'] — deprecated
• ['Environmental', 'Ground', 'Residential']
• ['Environmental', 'Ground', 'Shrubland']
• ['Environmental', 'Ground', 'Silvicultural']
• ['Environmental', 'Ground', 'Snow and Ice']
• ['Environmental', 'Ground', 'Wetland']
• ['Environmental', 'Other']
• ['Environmental', 'Product']
• ['Environmental', 'Water']
• ['Environmental', 'Water', 'Confined aquifer with fossil groundwater']
• ['Environmental', 'Water', 'Lake']
• ['Environmental', 'Water', 'Long-term']
• ['Environmental', 'Water', 'Ocean']
• ['Environmental', 'Water', 'River']
• ['Environmental', 'Water', 'Secondary water treatment']
• ['Environmental', 'Water', 'Surface water']
• ['Environmental', 'Water', 'Unconfined aquifer']
• ['Environmental', 'Water', 'Waste-water treatment plant']
• ['Inventory Indicator']
• ['Land Use', 'Occupation']
• ['Land Use', 'Transformation']
• ['Resource', 'Air']
• ['Resource', 'Biotic']
• ['Resource', 'Ground']
• ['Resource', 'Water']
• ['Resource', 'Water', 'Confined aquifer with fossil groundwater']
• ['Resource', 'Water', 'Lake']
• ['Resource', 'Water', 'Ocean']
• ['Resource', 'Water', 'River']
• ['Resource', 'Water', 'Surface water']
• ['Resource', 'Water', 'Unconfined aquifer']
• ['Social']

Mapping to Other Flow List Contexts

Every mapping from another list's context strings onto these consensus contexts lives in one file, context-manual-mapping.json, whose default_context_mappings rows are keyed by source. It currently covers EF 3.1, ecoinvent-3.8 and ecoinvent-3.12.

One file, and both stages read it: the default_context_mapping transformer during the transform, and the merge — filtered to the list being merged — through consensus_flow_list.context_mapping. So a rule added here places a row in both stages or in neither.

The merge used to read a per-source projection of this file, written by consensus-flow-list contexts into package data. build never ran that command, so editing the master moved the transform and left the merge on whatever the projection said the last time somebody ran it by hand. The projections are gone. To map a new list's compartments, add rows here under its source string.

When a compartment is not enough

Some lists put two consensus contexts in one compartment, and then say which is which somewhere else. The file has a rule for each way that happens, and the more specific rule wins:

ecoinvent files land occupation and land transformation in one compartment, natural resource / land, and distinguishes them in the name: Occupation, annual crop against Transformation, from annual crop. Land use / Occupation and Land use / Transformation are siblings here, so the compartment rule had to pick one and be wrong about the other — it said occupation, and 122 transformation flows per version were filed as occupations until a name rule read them.

A name rule is a claim to have read the whole compartment: if a flow lands in one and its name matches no prefix, the build stops and names the flow rather than falling back to the compartment's answer. A name the list has started using is a curation decision, not a default.

EF 3.1 gives all six of its water bodies one compartment and no prefix separates them, which is what the per-flow rows are for.