edges: Edge-based life cycle impact assessment
edges is a library allowing flexible Life Cycle Impact Assessment (LCIA)
for the brightway2/brightway25 LCA framework.
Unlike traditional LCIA methods that apply characterization factors (CFs) solely to nodes
(e.g., elementary flows), edges applies CFs directly on the edges — the exchanges between
suppliers and consumers — allowing for more precise and context-sensitive impact characterization.
This approach enables LCIA factors to reflect the specific context of each exchange, including parameters such as:
- Geographic region of production and consumption
- Magnitude of flows
- Scenario-based parameters (e.g., changing atmospheric conditions)
The edges Python library offers a novel approach to applying characterization factors
(CFs) during the impact assessment phase of Life Cycle Assessment (LCA).
Unlike conventional methods that uniformly assign CFs to nodes (e.g., processes
like Water, from well in the brightway2 ecosystem), edges shifts the focus to the
edges—the exchanges or relationships between nodes. This allows CFs to be conditioned
based on the specific context of each exchange. Essentially, edges introduces unique
values in the characterization matrix tailored to the characteristics of each edge.
By focusing on edges, the library incorporates contextual information such as the attributes of both the supplier and the consumer (e.g., geographic location, ISIC classification, amount exchanged, etc.). This enables a more detailed and flexible impact characterization, accommodating parameters like the location of the consumer and the magnitude of the exchange.
When necessary, custom CF definitions can also match on unit to distinguish flows
that would otherwise share the same name, location, categories, or
reference product.
Furthermore, edges supports the calculation of weighted CFs for both static regions
(e.g., RER) and dynamic regions (e.g., RoW), enhancing its ability to model complex
and region-specific scenarios.
Key Features
- Edge-based CFs: Assign CFs specifically to individual exchanges between processes.
- Geographic resolution: Supports 346 national and sub-national regions.
- Scenario-based flexibility: Incorporate parameters (e.g., CO₂ atmospheric concentration) directly in CF calculations, enabling dynamic scenario analysis.
- Efficient workflow: Clearly separates expensive exchange-mapping tasks (performed once) from inexpensive scenario-based numeric CF evaluations.
- Transparent reporting: Can expand weighted fallback consumer regions into country-specific rows in the detailed CF table for deterministic runs.
- Mixed supplier matrices: Supports method files that combine
biosphere-technosphereandtechnosphere-technosphereCF rows in one JSON file.
Currently, the library provides regionalized CFs for:
- AWARE 2.0 (water scarcity impacts)
- ImpactWorld+ 2.1
- GeoPolRisk 1.0
- GLAM 1.0 (32 curated practitioner methods from V1.0.2024.10)
- GLAM3 Land use impacts on biodiversity
- IBIF v2 biodiversity intactness (CO2, NH3, NOx, land occupation, roads, and all pressures)
- Swiss Ecological Scarcity 2021 (UBP total and all 20 categories, including Noise)
For UBP comparisons across software, see the SimaPro comparison evidence and modelling differences. Agreement with another implementation is a diagnostic check, not proof of correctness.
The 32 curated GLAM practitioner methods are bundled in edges/data/ and
listed by get_available_methods(). For example, select
("GLAM 1.0", "Terrestrial acidification", "damage") as the method in the
LCIA workflow below. Indicators, scenarios and assessment levels are separate
alternatives with their own output units; do not add their scores together.
The existing GLAM3 biodiversity collection remains separately available.
The methods contain 222,676 rules with exact elementary-flow names, category
paths and input units from the ecoinvent 3.10--3.12 reference catalogs.
Unsupported identities, compartments and unit conversions are excluded.
Each JSON retains a one-sentence indicator description, full author/year/DOI
citations and two-space indentation. Exchanges contain only supplier,
consumer and value; detailed provenance stays in external audit files.
All 288 practitioner validation cases passed across electricity, wheat and copper inventories in ecoinvent cutoff 3.10, 3.11 and 3.12. Independent checks verified direct CF values and exact final exchange sets after geographic fallback, with no missing, extra or duplicate mapped edges. These checks ran sequentially with one numerical thread and a 2 GiB memory guard; moving the unchanged JSON files into the package preserves that evidence.
The GLAM build and coverage audit documents source
conversions, approved mapping assumptions and nine climate substances without
published CFs, whose flows remain uncharacterized. The 195 optional native
research methods remain under dev/GLAM/methods_native/, outside the installed
package, and can be loaded by filepath. Their full runtime validation is
separate from the practitioner release checks.
Installation
You can install the library using pip:
pip install edges
Documentation
Getting Started
Check out the examples' notebook.
Check available methods from edges
from edges import get_available_methods
# Get the available methods
methods = get_available_methods()
print(methods)
Perform edge-based LCIA with edges
import bw2data
from edges import EdgeLCIA
# Select an activity from the LCA database
act = bw2data.Database("ecoinvent-3.10-cutoff").random()
# Define a method
method = ('AWARE 2.0', 'Country', 'unspecified', 'yearly')
# Prospective variant: ('AWARE 2.0 prospective', 'Country', 'all', 'yearly')
# Swiss Ecological Scarcity total: ('Ecological Scarcity 2021', 'total', 'UBP')
# GLAM practitioner: ('GLAM 1.0', 'Terrestrial acidification', 'damage')
# Initialize the LCA object
LCA = EdgeLCIA({act: 1}, method)
LCA.lci()
# Map CFs to exchanges: apply suggested strategies
LCA.apply_strategies()
# or apply these strategies manually
#LCA.map_exchanges()
# If needed, extend the mapping to aggregated and `dynamic` regions (e.g., RoW)
#LCA.map_aggregate_locations()
#LCA.map_dynamic_locations()
#LCA.map_contained_locations()
# add global CFs to exchanges missing a CF
#LCA.map_remaining_locations_to_global()
# Evaluate CFs
LCA.evaluate_cfs()
# Perform the LCIA calculation
LCA.lcia()
print(LCA.score)
# Optional but recommended: print a dataframe with the characterization factors used
# this allows you to check whether exchanges have been given the correct CFs
# include_unmatched=True allows you to see which exchanges were not matched (and if some should have been)
# split_aggregate_consumers=True expands weighted consumer fallback rows
# (e.g. RER, GLO, RoW, RoE) into country rows in deterministic runs
LCA.generate_cf_table(split_aggregate_consumers=True)
For deterministic regionalized runs, generate_cf_table(split_aggregate_consumers=True)
replaces weighted fallback rows for aggregate or dynamic consumer regions with
country-specific rows whose amount and impact sum back to the original
row. The raw split used for each exchange is also available after
evaluate_cfs() via lca.scenario_cfs[*]["reporting_split"].
When a method mixes biosphere- and technosphere-supplier CFs, the exported
table includes supplier matrix and direction columns so you can tell
which contributions came from biosphere-technosphere versus
technosphere-technosphere matches.
Perform parameter-based LCIA
Consider the following LCIA data file (saved under gwp_example.json)`:
{
"name": "Example LCIA Method",
"version": "1.0",
"description": "Example LCIA method for greenhouse gas emissions",
"unit": "kg CO2e",
"exchanges": [
{
"supplier": {
"name": "Carbon dioxide",
"operator": "startswith",
"matrix": "biosphere"
},
"consumer": {
"matrix": "technosphere",
"type": "process"
},
"value": "1.0"
},
{
"supplier": {
"name": "Methane, fossil",
"operator": "contains",
"matrix": "biosphere"
},
"consumer": {
"matrix": "technosphere",
"type": "process"
},
"value": "28 * (1 + 0.001 * (co2ppm - 410))"
},
{
"supplier": {
"name": "Dinitrogen monoxide",
"operator": "equals",
"matrix": "biosphere"
},
"consumer": {
"matrix": "technosphere",
"type": "process"
},
"value": "265 * (1 + 0.0005 * (co2ppm - 410))"
}
]
}
Symbolic CF expressions are intentionally narrow: arithmetic, parameter names,
literal values, and bare allowlisted function calls. Arbitrary Python syntax
such as attribute access, subscripting, comprehensions, imports, or method calls
is rejected. Functions passed through allowed_functions remain trusted Python
code.
We can perform a parameter-based LCIA calculation as follows:
import bw2data
from edges import EdgeLCIA
# Select an activity from the LCA database
bw2data.projects.set_current("ecoinvent-3.10.1-cutoff")
act = bw2data.Database("ecoinvent-3.10.1-cutoff").random()
print(act)
# Define scenario parameters (e.g., atmospheric CO₂ concentration and time horizon)
params = {
"some scenario": {
"co2ppm": {"2020": 410, "2050": 450, "2100": 500}, "h": {"2020": 100, "2050": 100, "2100": 100}
}
}
# Define an LCIA method (symbolic CF expressions stored in JSON)
method = ('GWP', 'scenario-dependent', '100 years')
# Initialize LCIA
lcia = EdgeLCIA(
demand={act: 1},
filepath="lcia_example_3.json",
parameters=params
)
# Perform inventory calculations (once)
lcia.lci()
# Map exchanges to CF entries (once)
lcia.map_exchanges()
# Optionally, resolve geographic overlaps and disaggregations (once)
lcia.map_aggregate_locations()
lcia.map_dynamic_locations()
lcia.map_remaining_locations_to_global()
# Run scenarios efficiently
results = []
for idx in {"2020", "2050", "2100"}:
lcia.evaluate_cfs(idx)
lcia.lcia()
df = lcia.generate_cf_table(split_aggregate_consumers=True)
scenario_result = {
"scenario": idx,
"co2ppm": params["some scenario"]["co2ppm"][idx],
"score": lcia.score,
"CF_table": df
}
results.append(scenario_result)
print(f"Scenario (CO₂ {params['some scenario']['co2ppm'][idx]} ppm): Impact = {lcia.score}")
Scenario-dependent methods can store expressions directly in value, as older
method files do, or keep a numeric baseline value and use value_expression
for the dynamic scenario/year value. Regionalized methods can similarly pair
numeric weight with weight_expression; aggregate fallback shares are
recomputed from evaluated dynamic weights during evaluate_cfs(...).
If the method declares an interpolation policy with method: "linear" and
extrapolation: "nearest", missing numeric years are linearly interpolated and
out-of-range years use the closest available endpoint. Methods without this
metadata keep the legacy exact-or-last parameter fallback.
Data Sources
See Methods from Documentation.
Methodology
See Theory from Documentation.
Contributing
Contributions are welcome! Please follow these steps to contribute:
- Fork the repository.
- Create a new branch for your feature or fix.
- Commit your changes.
- Submit a pull request.
License
This project is licensed under the MIT License. See the LICENSE.md file for more information.
Contact
For any questions or inquiries, please contact the project maintainer at romain.sacchi@psi.ch.
For questions or issues around the use of edges, you can also join the
discussion group at https://premise.groups.io/g/edgese.
Contributors
- Romain Sacchi
- Alvaro Hahn Menacho
- Raphaël Jolivet - contributed to the CLIPSpy-based rule engine implementation.
Acknowledgments
The development of this library was supported by the French agency for Energy ADEME, via the financing of the HySPI project. The HySPI project aims to provide a methodological framework to analyze and quantify, in a systemic and prospective manner, the environmental impacts of the decarbonization strategy of hydrogen production used by the industry in France. We also acknowledge financial support from the Europe Horizon project RAWCLIC as well as the Europe Horizon project PRISMA.
Metadata
Release files for edges 1.5.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| edges-1.5.0.tar.gz | 39.9 MB | Without provenance | Details for edges-1.5.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| edges-1.5.0-py3-none-any.whl | Python 3 | none | any | Details for edges-1.5.0-py3-none-any.whl |
Total release size: 81.2 MB
Release files / edges-1.5.0.tar.gz
| Download URL | edges-1.5.0.tar.gz |
|---|---|
| Size | 39.9 MB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
beec151f2f717b1e931244dee7c018d28cce9937a7106d4091bb533434ca43a6
|
|
BLAKE2b-256 checksum How to use checksums |
f17e63dd85b04202f9be3e1a8a96642ad2eabbd251fcd25c56fc20703e022726
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
Release files / edges-1.5.0-py3-none-any.whl
| Download URL | edges-1.5.0-py3-none-any.whl |
|---|---|
| Size | 41.3 MB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
6aec92d46207bf66d9bebf2afe26266e37e72c7fe07ca664d4d070e7e4118bb2
|
|
BLAKE2b-256 checksum How to use checksums |
3e108c9e6a4184f1ce04917114398f7828614e4eaf984654a362dda41e87c9d1
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
Security for edges 1.5.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.