bw2calc Changelog#

2.5.0 (2026-05-16)#

  • #148: Add PartitionedMonteCarloLCA, a Monte Carlo LCA class that pre-solves a static background system once and only resamples the stochastic foreground on each iteration

  • #150: Fix file handle leak in switch_method, switch_normalization, and switch_weighting — empty FilteredDatapackage objects were accumulating in self.packages, keeping old method filesystems (e.g. open zip files) alive across repeated calls

2.4.0 (2026-02-27)#

  • #145: Add a JacobiGMRESLCA class to solve large technosphere matrices. Thanks @romainsacchi

2.3.2 (2026-01-25)#

  • Update MethodConfig for changes in 2.3.1

2.3.1 (2026-01-25)#

  • Handle str and list[str] impact category identifiers

2.3 (2025-12-05)#

  • Factored out FastSupplyArraysMixin to allow for its use in normal LCA classes

  • Move to a src layout

  • Move to absolute imports

  • #141: Clarify error message of LCA.check_demand method related to Activity id

2.2.2 (2025-11-11)#

  • Change import detection of bw2data to avoid timing bug

2.2.1 (2025-10-27)#

  • Handle version checks for bw2data.__version__ being a string or a tuple

  • Add FastScoresOnlyMultiLCA to bw2calc namespace

2.2 (2025-10-17)#

  • #135: Fast scores-only multiLCA

2.1.1 (2025-09-16)#

  • #132: UMFPACK converts to CSC sparse array on factorization and then raises an error

  • #128: Recent fsspec emitting log messages and breaking tests

  • #120: Improve scikit-umfpack import error handling

2.1 (2025-06-23)#

  • #125: Remove Numpy version pin to <2. Thanks @michaelweinold!

2.0.2 (2025-03-21)#

  • Fix #110: Method invert_technosphere_matrix crashes with multi LCA

2.0.1 (2024-12-10)#

  • Update matrix_utils dependency

2.0 (2024-11-27)#

  • Add CachingLCA class which saves supply vector calculations

2.0.DEV23 (2024-09-18)#

  • Allow MethodConfig as an input to MultiLCA

2.0.DEV22 (2024-09-18)#

  • Add MethodConfig to __init__.py for better UX

2.0.DEV21 (2024-09-13)#

  • Fix #108: MultiLCA normalizes and weights in all combinations

2.0.DEV20 (2024-09-02)#

  • Bump dependencies as upstream has breaking changes

2.0.DEV19 (2024-09-02)#

  • Fix #105: MultiLCA class returning identical results under iteration

2.0.DEV18 (2024-07-24)#

  • #101: Convert technosphere_matrix to csr when using pypardiso for speed boost on repeated calculations.

  • #104: Explicitly set array-size in MultiLCA’s lci_calculation(). Solves bug with a single functional unit passed to MultiLCA.

2.0.DEV17 (2024-06-05)#

New MultiLCA implementation#

The new MultiLCA implementation support efficient calculation of multiple functional units, impact categories, normalizations, and weightings, and supports advanced features of bw_processign datapackages across the complete calculation chain, including scenarios and correlated uncertainty. It also adds usability improvements compared to previous implementations.

To perform a MultiLCA, you first need to define your configuration. There is a pydantic validator MethodConfig to make sure it is constructed correctly; see its docs for more info.

Here is an example MultiLCA calculation in action:

import bw2data as bd
import bw2io as bi

# All the LCI and LCIA data exist in the current project
functional_units = {
    "γ": {bd.get_node(name="foo").id: 1},
    "ε": {bd.get_node(name="bar").id: 2},
    "ζ": {bd.get_node(name="baz").id: 3},
}
config = {
    "impact_categories": [
        ("first", "categ."),
        ("second", "categ."),
    ],
    "normalizations": {
        ("am I normal?",): [
            ("first", "categ."),
            ("second", "categ."),
        ]
    },
    "weightings": {("heavy", "weight"): [("am I normal?",)]},
}
data_objs = bd.get_multilca_data_objs(functional_units=functional_units, method_config=config)
mlca = MultiLCA(demands=functional_units, method_config=config, data_objs=data_objs)
mlca.lci()
mlca.lcia()
mlca.normalize()
mlca.weight()
mlca.scores
>>> {
  (("heavy", "weight"), ("am I normal?",), ('first', 'categ.'), 'γ'): 190.31896453011996,
  (("heavy", "weight"), ("am I normal?",), ('first', 'categ.'), 'ε'): 42.504568745060126,
  (("heavy", "weight"), ("am I normal?",), ('first', 'categ.'), 'ζ'): 431.9104810103497,
  (("heavy", "weight"), ("am I normal?",), ('second', 'categ.'), 'γ'): 1205.2506281852936,
  (("heavy", "weight"), ("am I normal?",), ('second', 'categ.'), 'ε'): 269.17264029471556,
  (("heavy", "weight"), ("am I normal?",), ('second', 'categ.'), 'ζ'): 5614.00396766416
}

One big change compared to previous implementations is the labelling of matrices and results, which is used across the MultiLCA class.

The MultiLCA class can be used for both static and Monte Carlo calculations. You can configure which types of uncertainty to use where with selective_use. See the MultiLCA class docs for more information.

Here is an example of a selective use config. In this case, assuming use_distributions=False was passed to the MultiLCA constructor, on the characterization and weighting matrices would draw from the probability distribution information provided.

{
    "characterization_matrix": {"use_distributions": True},
    "weighting_matrix": {"use_distributions": True},
}

Gotcha: In MultiLCA, you need to use the selective use flags characterization_matrix, normalization_matrix, and weighting_matrix, even though all three of these are dictionaries, and called characterization_matrices, normalization_matrices, and weighting_matrices. This is because selective use flags can also be used in the base LCA class, and use the same code. The flags apply to all impact categories, normalizations, and weightings; You would need to write a custom subclass to turn on certain forms of uncertainty only for selected impact categories.

Result matrices (characterized_inventories, normalized_inventories, weighted_inventories) are all dictionaries, with each combination of functional unit and impact category/normalization/weighting possible as described in the MethodConfig. The values in this dictionary are sparse matrices with the same dimensions as the inventory, namely biosphere flows (rows) by processes (columns).

Other changes#

  • Removed graph traversal completely - use bw_graph_tools instead.

  • Migrate from fs to fsspec

  • Start using @ for matrix multiplication

  • Removed MonteCarloLCA (normal LCA class can do Monte Carlo) and added IterativeLCA (different solving strategy)

  • Fix miscellaneous deprecated API calls

2.0.DEV16 (2023-10-22)#

  • Fix #78: Allow for selective use of distributions or arrays depending on matrix label

  • Fix #77: Prevent multiple remap_inventory_dicts calls

  • Fix #71: Poor error message when no LCIA data is supplied

  • Documentation improvements

  • Switch packaging to pyproject.toml

2.0.DEV15 (2023-10-22)#

Never happened, can’t prove anything :)

2.0.DEV14 (2023-08-24)#

  • Packaging updates

2.0.DEV13 (2023-05-07)#

  • CI workflow updates

  • Merge #65: Add PyPI and conda-forge badge

  • Fix hidden dependency on bw2data

2.0.DEV12 (2022-09-19)#

  • Add some backwards compatiblity methods

2.0.DEV11 (2022-08-31)#

2.0.DEV10 (2022-08-19)#

  • Add LCA.to_dataframe, based on work by Ben Portner

2.0.DEV9 (2022-07-07)#

2.0.DEV8 (2022-06-28)#

2.0.DEV7 (2022-05-22)#

  • Add LCA.keep_first_iteration to make iteration simpler

2.0.DEV6 (2022-04-23)#

  • Add an optional warning on LCA instantiation if excluding resources (arrays or distributions) which could be useful

  • Add function stubs to be used by subclasses on iteration

2.0.DEV5 (2021-11-26)#

  • Fix a bug in switch_method if given a bw2data method tuuple instead of a list of datapackages.

2.0.DEV4 (2021-11-03)#

  • Add invert_technosphere_matrix with algo from @haasad

  • Fix switch_method, switch_normalization, switch_weighting

Compatibility changes:

  • LCA.score will return weighted or normalized score, if weighting or normalization has been performed

  • LCA.weighting will now trigger a deprecation warning. Switch to .weight instead.

  • LCA.redo_lci deprecated in favor of LCA.lci(demand); LCA.redo_lcia deprecated in favor of LCA.lcia(demand)

2.0.DEV3 (2021-10-17)#

  • Fix for constructing characterization matrices with semi-regionalized impact categories

2.0.DEV2 (2021-10-01)#

  • More 2.5 work and fixes

2.0.DEV1#

Version 2.0 brings a number of large changes, while maintaining backwards compatibility (except for dropping Py2). The net result of these changes is to prepare for a future where data management is separated from calculations, and where working with large, complicated models is much easier.

Future DEV releases#

Before 2.0 is released, the following features will be added:

  • Presamples will be adapted to use bw_processing

  • Logging will be taken seriously :)

  • ~~LCA results to dataframes~~

Breaking changes#

Simplification of user endpoints#

The structure of this library has been simplified, as the LCA class can now perform static, stochastic (Monte Carlo), iterative (scenario-based), and single-matrix LCA calculations. Matrix building has been moved to the matrix_utils library.

Python 2 compatibility removed#

Removing the Python 2 compatibility layer allows for much cleaner and more compact code, and the use of some components from the in-development Brightway version 3 libraries. Compatible with bw2data version 4.0.

Removal of classes and methods#

  • LCA.rebuild_*_matrix methods are removed. See the TODO notebook for alternatives.

  • DirectSolvingMixin and DirectSolvingMonteCarloLCA are removed, direct solving is now the default

  • ComparativeMonteCarlo is removed, use MultiLCA(use_distributions=True) instead

  • SingleMatrixLCA is remove, use LCA instead. It allows for empty biosphere matrices.

Simplified handling of mapping dictionaries#

Mapping dictionaries map the database identifiers to row and column indices. In 2.5, these mapping dictionaries are only created on demand; avoiding their creation saves a bit of time and memory.

Added a new class (DictionaryManager) and made it simpler reverse, remap, and get the original dictionaries inside an LCA. Here is an example:

LCA.dicts.biosphere[x]
>> y
LCA.dicts.biosphere.original # if remapped with activity keys
LCA.dicts.biosphere.reversed[y]  # (generated on demand)
>> x

The dictionaries in a conventional LCA are:

  • LCA.dicts.product

  • LCA.dicts.activity

  • LCA.dicts.biosphere

~~LCA.reverse_dict is removed; all reversed dictionaries are available at LCA.dicts.{name}.reversed~~.

In 2.5, these mapping dictionaries are not automatically “remapped” to the (database name, activity code) keys. You will need to call .remap_inventory_dicts() after doing an inventory calculation to get mapping dictionaries in this format.

Weighting is a diagonal matrix instead of a single number#

It is easier to have everything in the same mode of operation. This also allows for the use of arrays, distributions, interfaces, etc. in weighting. Implemented in new SingleValueDiagonalMatrix class.

Architectual changes#

Use of bw_processing#

We now use bw_processing to load processed arrays. bw_processing has separate files for the technosphere and biosphere arrays, and explicit indication of . Therefore, the TechnosphereBiosphereMatrixBuilder is no longer necessary, and is removed.

No dependency on bw2data#

bw2data is now an optional install, and even if available only a single utility function is used to prepare input data. bw2calc is primarily intended to be used as an independent library.

Changes in Monte Carlo#

Smaller changes#

New LCA input specification#

The existing input specification is still there, but this release also adds the ability to specify input arguments compatible with Brightway version 3. Previously, we would write LCA({some demand}, method=foo) - this requires bw2calc to use bw2data to figure out the dependent databases of the functional unit in some demand, and then to get the file paths of all the necessary files for both the inventory and impact assessment. The new syntax is LCA({some demand}, data_objs), where some demand is already integer IDs, and data_objects is a lists of data packages (either in memory or on the filesystem).

bw2data has a helper function to prepare arguments in the new syntax: prepare_lca_inputs.

This new input syntax, with consistent column labels for all structured arrays, removes the need for IndependentLCAMixin. This is deleted, and the methods get_vector, get_vector_metadata, and set_vector are added.

More robust matrix building#

More tests were identified, and undefined behaviour is now specified. For example, the previous matrix builders assumed that the values in the provided row or column dictionaries were sequential integers starting from zero - this assumption is now relaxed, and we allow this dictionary values to start with an offset. There are also tests and documentation on what happens under various cases when drop_missing is False, but missing values are present.

1.8.0 (2020-02-27)#

  • Replace .todense with .toarray to satisfy changes in Scipy API

  • Add atol parameter to iterative solver to satisfy changes in Scipy API

  • Fix regression in 1.7.7 which raises errors when no new demand was present (PR #6)

1.7.8 (2019-11-01)#

  • Add check to make sure not all arrays are empty during matrix construction

  • Allow numpy loading pickled data

1.7.7 (2019-10-31)#

Switch lca.demand when running .redo_lci or .redo_lcia. Thanks Aleksandra Kim!

1.7.6 (2019-10-22)#

Fixed #25: Sort array filepaths when loading. Thanks Pedro Anchieta!

1.7.5 (2019-09-19)#

Merged Pull Request #4 to directly pass Numpy or byte arrays instead of filepaths. Thanks Jan Machacek!

1.7.4 (2019-08-23)#

  • Improved support for independent LCA calculations (i.e. without Brightway2 databases, only processed arrays)

  • Added ability to calculate LCAs in a single matrix (for BONSAI)

1.7.3 (2018-10-24)#

Updated Monte Carlo for upstream presamples changes

1.7.2 (2018-08-21)#

Merged Pull Request #3 to fix some attributes in graph traversals. Thanks Bernhard Steubing!

1.7.1 (2018-02-14)#

Compatibility with presamples release version

1.7 (2018-01-18)#

Add compatibility with bw_presamples

1.6.4 (2018-01-12)#

Really fix bug in seed generation for pooled Monte Carlo calculations

1.6.3 (2018-01-11)#

  • JOSS submission

  • Fix bug in MultiMonteCarlo

  • Add some logging to support presamples in the future

1.6.2 (2017-04-17)#

Fix license text

1.6.1 (2017-04-06)#

Simplify indexing

1.6 (2017-04-05)#

Replace bw2speedups indexing with numpy array trickiness which is ~5 times faster

1.5.4 (2017-02-24)#

Remove non-ascii characters from license text, because setuptools

1.5.3 (2016-10-28)#

  • Restructure imports to not depend on bw2data

  • Use io.open in setup.py

1.5.2 (2016-10-28)#

Specify encoding of license file

1.5.1 (2016-09-15)#

Bugfix for broken import statement

1.5 (2016-09-15)#

Merge pull request from Adrian Haas to enable Pardiso solver usage when available.

1.4 (2016-07-14)#

  • Added utility functions for load_calculation_package and save_calculation_package for independent LCAs and cloud computing.

  • Compatibility with bw2data 2.3

1.3.6 (2016-07-01)#

Fixed bugs where RNG and technosphere matrix builder would change values in arrays meant to be static

1.3.5 (2016-07-01)#

Fix bugs and add tests for ParameterVectorLCA

1.3.4 (2016-06-10)#

Changed ParameterVectorLCA: Can no longer be called, split off rebuild_all into a separate method, added tests.

1.3.3 (2016-06-10)#

Better test coverage and Windows comaptibility

1.3.2 (2016-06-08)#

  • FEATURE: Add class and mixin for Monte Carlo using direct solvers

  • CHANGE: Move tests to root directory and add Monte Carlo tests

  • CHANGE: Consistent use of __next__ and next() so that all Monte Carlo iterator classes are Py2/3 compatible and programmed the same way. ParameterVectorLCA.next() will no longer work on Python 3; instead, call next(ParameterVectorLCA). When providing a new vector, call the class itself (after it is instantiated): pv = ParameterVectorLCA(args); pv(new_vector).

1.3.1 (2016-06-06)#

  • CHANGE:Updates for bw2data 2.2

  • BUGFIX: Correctly handle regionalized CFs in site-generic calculations

  • FEATURE: Add contribution methods to LCA classes

1.3 (2016-05-28)#

BUGFIX: Correctly handle project names in multiprocess calculations

1.2.1 (2016-03-14)#

BUGFIX: switch_* was seriously broken due to new handling of processed arrays filepaths

1.2 (2016-03-14)#

  • Feature: Py3 compatibility

  • FEATURE: Independent LCAs which don’t rely on bw2data and the brightway2 ecosystem

  • Feature: Add DenseLCA

  • FEATURE: Added switch_* and to_dataframe methods to LCA class

  • FEATURE: Allow graph traversal to skip links in static databases

  • BUGFIX: Terminate multiprocessing pools after calculations

  • CHANGE: Load data automatically in Monte Carlo

  • CHANGE: Automatically clean dirty databases before starting calculations

Plus lots of small bugfixes, and compatibility with projects.

1.0 (2015-03-08)#

CHANGE: Split activities and products

0.17.1 (2015-02-13)#

BUGFIX: Don’t require substitution in TYPE_DICTIONARY.

0.17 (2015-02-13)#

  • FEATURE: Properly handle substitution type exchanges

  • CHANGE: Compatible with bw2data version 2

  • BUGFIX: Fix handling of nested dependent databases

0.16.1 (2014-12-05)#

  • CHANGE: Better documentation for most code.

  • BUGFIX: Graph traversal handles most coproducts, and raises sensible errors when it can’t.

0.16 (2014-08-03)#

FEATURE: Changes in MatrixBuilder should make normal static LCA calculations about three times faster.

0.15.1 (2014-07-30)#

Update dependencies.

0.15 (2014-06-11)#

BREAKING CHANGE: Use Database.filename for processed data. Requires update to bw2data version 0.16 or greater.

0.13 (2014-04-16)#

  • BREAKING CHANGE: LCA.fix_dictionaries now sets/uses _mapped_dict to determine if fix_dictionaries has been called.

  • BUGFIX: LCA.build_demand_array doesn’t break if fix_dictionaries has been called.

0.12 (2014-02-13)#

BREAKING CHANGE: Matrix builder will only include parameter array rows that are correctly mapped, instead of raising an error when unmapped rows occur. This behaviour can be turned off by passing drop_missing=False.

0.11.1 (2014-01-29)#

BUGFIX: Change column names in method matrix building to be consistent with bw2data 0.11

0.11 (2014-01-26)#

  • BREAKING CHANGE: Graph traversal was reworked, and some functionality for interpreting the output was moved to bw2analyzer.

  • BREAKING CHANGE: Deleted SimpleRegionalizedLCA class. Regionalization will be provided in bw2regional.

  • BREAKING CHANGE: Deleted initial sensitivity work, moved for now to branch, as it was not yet usable.

  • FEATURE: Much better and more thorough documentation.

  • FEATURE: Improved testing and test coverage

On this page
Edit on GitHub

This Page

Support Brightway