Life cycle impact assessment is the step where an inventory stops being a spreadsheet of kilograms and starts being an answer. Everything before it is bookkeeping: how much electricity, how much water, how many kilograms of polymer. Life cycle impact assessment is the modelling layer that says what those flows mean environmentally, and it is where most of the methodological argument in LCA happens.
This article covers what LCIA does, which of its elements ISO 14044 makes mandatory, how characterisation factors work, and the difference between midpoint and endpoint indicators. It then answers a question the textbooks tend to duck: how much does any of it change a real bioprocess result? Every number below is generated from the bioprocess LCA calculator on one modelled 2000 L CHO fed-batch, so the sensitivities are measured rather than asserted.
What life cycle impact assessment is
Life cycle impact assessment (LCIA) is the third phase of an LCA under ISO 14040 and ISO 14044. It converts life cycle inventory flows into environmental impact scores by assigning each flow to an impact category and multiplying it by a characterisation factor.
The four phases of LCA run goal and scope, inventory, impact assessment, then interpretation. Phase 2 hands LCIA a list of elementary flows: substances crossing the boundary between the technosphere and the environment. Phase 3 hands interpretation a much shorter list of category indicator results. That compression is the whole point, and it is also where information is deliberately thrown away.
Two things life cycle impact assessment is not, both of which cause trouble in review:
- It is not measurement. Characterisation factors come from atmospheric chemistry, fate and exposure models and toxicology. They are modelled quantities with their own uncertainty, layered on top of the inventory's uncertainty.
- It is not a ranking. LCIA produces one number per impact category, not one number per product. Turning several category results into a single score requires weighting, which ISO treats as an optional and heavily restricted step.
The mandatory and optional elements
ISO 14044 splits life cycle impact assessment into six elements and makes only the first three mandatory. Knowing which is which settles most arguments about whether a study is compliant.
Classification is the assignment of each inventory flow to every impact category it contributes to. It is not one-to-one. Nitrogen oxides from a boiler are classified into acidification, terrestrial eutrophication and photochemical ozone formation simultaneously, and the same kilogram is counted in full in each, because the categories measure different things rather than shares of one thing.
Characterisation is the arithmetic: multiply each classified flow by its characterisation factor and sum within the category. The output is a category indicator result, such as kg CO2e for climate change or kg SO2e for acidification.
Characterisation factors and the horizon choice
A characterisation factor converts one unit of an inventory flow into the common unit of an impact category. Global warming potential is the one every engineer already knows: it expresses each greenhouse gas relative to carbon dioxide over a chosen time horizon.
Two things about that sentence get skipped. The horizon is a choice, not a constant. And the factors are revised with each IPCC assessment report, so a 2013 study and a 2024 study both reporting "GWP100" are not using the same numbers.
| Gas | AR4 100-yr | AR5 100-yr | AR6 100-yr | AR6 20-yr | Max ÷ min |
|---|---|---|---|---|---|
| Carbon dioxide (CO2) | 1 | 1 | 1 | 1 | 1.00× |
| Methane (CH4) | 25 | 30 | 27.9 | 81.2 | 3.25× |
| Nitrous oxide (N2O) | 298 | 265 | 273 | 273 | 1.12× |
| Sulfur hexafluoride (SF6) | 22,800 | 23,500 | 25,200 | 18,300 | 1.38× |
Methane is the gas that makes the horizon argument worth having. Its atmospheric lifetime is short enough that a 20-year window catches most of its forcing while a 100-year window averages it away, so its characterisation factor triples between the two. Nitrous oxide, which persists for over a century, has the same factor on both horizons.
The practical consequence is a threshold. Because carbon dioxide has a factor of 1 on every horizon, switching from AR4 100-year to AR6 20-year only moves a result in proportion to how much methane is in it. Working from the factors above, the spread equals the methane share of your AR6 100-year total multiplied by about 2.01.
Threshold — when does the horizon choice matter?
Spread from AR4-100 to AR6-20 = (methane share of AR6-100 CO2e) × (81.2 − 25) ÷ 27.9
- To move the result 5%, methane must be 2.5% of your CO2e
- To move it 10%, methane must be 5.0%
- To move it 20%, methane must be 9.9%
- To move it 50%, methane must be 24.8%
Across all 26 US eGRID subregions in the calculator's factor table, methane is between 0.10% and 0.44% of grid CO2e. An anaerobic digester or a landfill-gas boiler would clear these thresholds easily. A grid-electricity bioprocess does not come close.
Does the LCIA method actually change the answer?
For the climate change category in a bioprocess, almost never. We re-characterised the electricity inventory of one modelled 2000 L CHO fed-batch under all four IPCC method vintages, and the three 100-year variants agree to eight thousandths of a percent.
The batch draws 28,759 kWh. Taking the ERCOT subregion, whose factor table carries a per-gas split rather than a pre-aggregated CO2e value, that electricity carries 13,163.1 kg of CO2, 0.995 kg of CH4 and 0.140 kg of N2O.
Worked example — one gas inventory, four LCIA methods
Inventory: CO2 13,163.1 kg · CH4 0.995 kg · N2O 0.1404 kg
- AR4-100: 13,163.1 + (0.995 × 25) + (0.1404 × 298) = 13,163.1 + 24.9 + 41.8 = 13,229.8 kg CO2e
- AR5-100: 13,163.1 + (0.995 × 30) + (0.1404 × 265) = 13,163.1 + 29.9 + 37.2 = 13,230.2 kg CO2e
- AR6-100: 13,163.1 + (0.995 × 27.9) + (0.1404 × 273) = 13,163.1 + 27.8 + 38.3 = 13,229.2 kg CO2e
- AR6-20: 13,163.1 + (0.995 × 81.2) + (0.1404 × 273) = 13,163.1 + 80.8 + 38.3 = 13,282.2 kg CO2e
Spread across the three 100-year vintages: 1.0 kg out of 13,229, or 0.008%. Full spread including the 20-year horizon: 52.4 kg, or 0.40%. Components are rounded for display, so totals computed from unrounded gas masses differ by well under a kilogram.
The cancellation is not a coincidence. AR4 to AR6 raises methane from 25 to 27.9, up 11.6%, and lowers nitrous oxide from 298 to 273, down 8.4%. In a combustion inventory the two trace gases contribute comparable amounts, so the changes offset.
Set that against the choices made earlier in the study. The same calculator, same batch, varying one decision at a time:
| Choice | Phase | Range tested | Change in result |
|---|---|---|---|
| LCIA method vintage | Impact assessment | AR4-100 → AR6-100 | −0.005% |
| GWP time horizon | Impact assessment | 100-yr → 20-yr | +0.385% |
| System boundary | Goal and scope | with → without cleanroom HVAC | −89.3% |
| Titer | Process | 1 → 10 g/L | −90.0% |
| Grid region | Inventory | NYUP (134) → MROE (762) g/kWh | +410.7% |
None of this makes life cycle impact assessment unimportant. It makes one part of it unimportant for one category. The part that does change conclusions is category selection, covered further down.
Run the sensitivities yourself
The calculator carries 26 US grid subregions, four IPCC method vintages and a full per-gas factor table. Change one input and watch which ones move the answer.
Midpoint vs endpoint indicators
A midpoint indicator stops partway along the cause-effect chain at a physical quantity. An endpoint indicator continues to the damage that quantity eventually causes. Climate change reported as kg CO2e is a midpoint; the same emission reported as disability-adjusted life years is an endpoint.
The trade-off is straightforward. Midpoint results are defensible, because radiative forcing is physics and the characterisation factors have narrow uncertainty bands. Endpoint results are interpretable, because a board understands "years of healthy life lost" better than "kg SO2e", but they inherit the uncertainty of the fate model, the exposure model, the dose-response model and the severity weighting, compounded.
For biomanufacturing, midpoints are the right default. Cleanroom-dominated processes are driven by electricity, and electricity's climate midpoint is both the dominant term and the best characterised. Reaching for endpoints adds a great deal of model uncertainty to a result whose ranking was already clear.
The main LCIA method families
An LCIA method is a package: a chosen set of impact categories, a characterisation model behind each one, and the resulting factor tables. Choosing a method is choosing all of that at once, which is why studies state the method and version in goal and scope.
| Method | Origin | Level | Geography | Typical use |
|---|---|---|---|---|
| CML-IA | CML, Leiden University | Midpoint only | Global / Europe | The classic problem-oriented baseline, still common in EPDs |
| TRACI 2.1 | US EPA | Midpoint only | United States | US studies and US regulatory contexts |
| ReCiPe 2016 | RIVM, Radboud, PRé and partners | Midpoint and endpoint | Global, some regionalisation | The most widely used general method; three cultural perspectives |
| EF 3.1 | European Commission (JRC) | Midpoint | Europe | Required for Product and Organisation Environmental Footprint studies |
| IMPACT World+ | CIRAIG, Quantis and partners | Midpoint and endpoint | Globally regionalised | Studies with supply chains on several continents |
| USEtox | UNEP–SETAC consensus model | Midpoint | Global | Not a full method; supplies toxicity and ecotoxicity factors to the others |
ReCiPe 2016 is the one most bioprocess practitioners will meet, because it is the default in several of the packages compared in our LCA software guide. It reports eighteen midpoint categories which condense into three endpoint areas of protection: human health, ecosystem quality and resource scarcity.
Worth being blunt about a licensing reality. Not every method is freely redistributable. Our own factor table ships IPCC global warming potentials, US eGRID electricity data and open government emission factors, all under licences that permit redistribution inside an interactive tool. It deliberately does not ship a full ReCiPe factor set, because the terms for embedding one in a public calculator are unverified. If you need a full multi-category method, that is what the desktop packages are for.
Normalisation, grouping and weighting
These are the three optional elements, and they are optional for good reason. Each adds a layer of value judgement on top of the characterised results.
- Normalisation divides each category result by a reference value, typically the total impact of an average person in a region in a year. It answers "is this a big number?" and makes categories visually comparable. It also imports whatever assumptions sit inside the reference inventory, which for some categories is poorly characterised.
- Grouping sorts categories into sets, for example global versus regional, or ranks them by priority. Ranking is explicitly a value choice and ISO requires it be reported as such.
- Weighting multiplies normalised results by importance factors and sums them into a single score. This is the step that produces the one number everyone wants and no standard will endorse.
ISO 14044 states that weighting shall not be used in comparative assertions intended to be disclosed to the public. The reason is not bureaucratic. Deciding that one unit of climate change is worth some number of units of freshwater ecotoxicity is an ethical judgement, and burying it inside an aggregate score presented to consumers hides the judgement rather than making it. A study may weight internally for its own decision-making; it may not publish the weighted score as a public comparison.
This matters directly for the marketing claims biomanufacturers are asked to support. "Our process is 30% greener" is a weighted statement wearing a percentage. It is not a life cycle impact assessment result.
Which categories a bioprocess actually needs
Category selection is the LCIA choice with real consequences, and the evidence is easy to generate. Take the matched comparison from our single-use versus stainless analysis, same volume, titer, downstream yield, batch length and cleanroom, and read it through different indicators.
| Indicator | Single-use | Stainless | Difference | Verdict |
|---|---|---|---|---|
| Climate change (kg CO2e/kg) | 3,980 | 3,971 | −0.2% | Indistinguishable |
| Cumulative energy (kWh/kg) | 9,510 | 9,629 | +1.3% | Indistinguishable |
| Water use (L/kg) | 19,841 | 27,778 | +40.0% | Single-use wins |
| Process mass intensity (kg/kg) | 19,884 | 27,797 | +39.8% | Single-use wins |
| Consumable mass (kg/kg) | 27.1 | 3.1 | −88.4% | Stainless wins |
This is the concrete case for ISO's position on weighting. There is no objective exchange rate that converts 7,900 litres of water into 24 kilograms of plastic. Any single score reconciling those bars encodes somebody's preference, and the honest presentation is the one above: several numbers, stated separately, with the trade-off visible.
For a cleanroom-dominated biologics process, three categories carry most of the decision-relevant information:
- Climate change, the reporting requirement, and the one your Scope 1, 2 and 3 inventory already feeds.
- Cumulative energy demand, strictly not an impact category, but it strips out grid choice and exposes the underlying process efficiency. Two sites with identical energy and different grids are the same process.
- Water use, the term that separates cleaning regimes, and the one climate change is least sensitive to. Purified-water generation is also an electricity term, which is why the WFI and purified water systems question shows up in both categories.
Process mass intensity is worth tracking alongside these as a process metric rather than an LCIA result. It needs no characterisation factors at all, which makes it fast and auditable, and also completely blind to what the mass actually does once released. Media mass is one of its larger controllable terms, so a media requirement estimate feeds it directly.
Cost sits on the same inventory
The flows that LCIA multiplies by characterisation factors are the same ones a cost model multiplies by prices. Run both and the abatement levers rank themselves.
Five mistakes that break an LCIA
These are the failures that survive review and then fall apart when someone reproduces the study.
- Characterising twice. Most published emission factors are already CO2e, meaning they have been through characterisation. Multiplying a 400 g CO2e/kWh grid factor by a global warming potential again inflates the result by whatever factor you applied. If a factor's unit contains "e", the impact assessment has already happened.
- Mixing method vintages. Pulling grid factors from an AR5 source and refrigerant factors from an AR6 table produces a result that belongs to no method. The error is small for combustion and larger for the trace gases, where the 100-year factors differ by up to 20% between AR4, AR5 and AR6 on the gases in the table above.
- Reporting a single score as a comparison. Covered above, and the most common compliance failure in marketing-adjacent studies.
- Selecting categories after seeing the results. ISO requires category selection in goal and scope, before the numbers exist. Choosing afterwards is how a study ends up reporting only the three categories where the product happened to win.
- Treating characterisation uncertainty as the dominant uncertainty. It rarely is. In the sensitivity table above, characterisation contributes under half a percent while the grid region contributes 411%. Effort spent arguing about method vintage is effort not spent on the electricity data.
The last one is the general lesson. Life cycle impact assessment is a well-characterised, standardised, largely settled layer sitting on top of an inventory that is usually none of those things. When an LCA result is wrong, the inventory is the reason far more often than the impact assessment is.
Build the inventory first
Energy, water, media and consumables per batch, with the grid region and characterisation factors already wired in and every factor's source and licence shown.
Frequently asked questions
What is life cycle impact assessment (LCIA)?
Life cycle impact assessment is the third phase of an LCA under ISO 14040 and 14044. It converts the inventory, kilograms of gas, litres of water and megajoules of fuel, into environmental impact scores by assigning each flow to an impact category and multiplying it by a characterisation factor. ISO 14044 makes three elements mandatory: selecting impact categories and models, classification, and characterisation.
What is the difference between midpoint and endpoint LCIA?
A midpoint indicator stops partway along the cause-effect chain at a physical quantity such as radiative forcing, reported as kg CO2e. An endpoint indicator continues to the damage itself, reported in disability-adjusted life years, species lost per year, or dollars. Midpoints carry less model uncertainty; endpoints are easier for a non-specialist to interpret but add the uncertainty of every extra modelling step.
Does the choice of LCIA method change the result?
For the climate change category in a bioprocess, barely. On a modelled 2000 L CHO fed-batch, moving the global warming potentials from IPCC AR4 to AR6 changes the footprint by 0.005% and switching from a 100-year to a 20-year horizon changes it by 0.4%, because combustion emissions are overwhelmingly carbon dioxide. Which impact categories you report matters far more than which method vintage supplies the factors.
What are characterisation factors?
A characterisation factor converts one unit of an inventory flow into the common unit of an impact category. Global warming potential is the best known example: under IPCC AR6 on a 100-year horizon, methane has a factor of 27.9 and nitrous oxide 273, so 1 kg of methane counts as 27.9 kg CO2e. Every impact category has its own factor set and its own reference substance.
Is normalisation or weighting required in an LCA?
No. ISO 14044 lists normalisation, grouping and weighting as optional elements. Weighting is also explicitly barred from comparative assertions intended to be disclosed to the public, because collapsing several impact categories into one score requires a value judgement that the standard will not let you hide inside a number.
Which impact categories should a bioprocess LCA report?
At minimum climate change, cumulative energy demand and water use, because in a cleanroom-dominated process these three behave differently enough to change a conclusion. On a matched single-use versus stainless comparison, climate change differs by 0.2% while water use differs by 40% and consumable mass by 88% in the opposite direction. Reporting climate alone would hide both results.
References
- Huijbregts M.A.J., Steinmann Z.J.N., Elshout P.M.F., Stam G., Verones F., Vieira M., Zijp M., Hollander A. & van Zelm R. (2017). ReCiPe2016: a harmonised life cycle impact assessment method at midpoint and endpoint level. The International Journal of Life Cycle Assessment 22, 138–147. doi:10.1007/s11367-016-1246-y
- Bare J. (2011). TRACI 2.0: the tool for the reduction and assessment of chemical and other environmental impacts 2.0. Clean Technologies and Environmental Policy 13, 687–696. doi:10.1007/s10098-010-0338-9
- Hauschild M.Z., Goedkoop M., Guinée J., Heijungs R., Huijbregts M., Jolliet O., Margni M., De Schryver A., Humbert S., Laurent A., Sala S. & Pant R. (2013). Identifying best existing practice for characterization modeling in life cycle impact assessment. The International Journal of Life Cycle Assessment 18, 683–697. doi:10.1007/s11367-012-0489-5
- Rosenbaum R.K., Bachmann T.M., Gold L.S., Huijbregts M.A.J., Jolliet O., Juraske R., Koehler A., Larsen H.F., MacLeod M., Margni M., McKone T.E., Payet J., Schuhmacher M., van de Meent D. & Hauschild M.Z. (2008). USEtox, the UNEP-SETAC toxicity model: recommended characterisation factors for human toxicity and freshwater ecotoxicity in life cycle impact assessment. The International Journal of Life Cycle Assessment 13, 532–546. doi:10.1007/s11367-008-0038-4
- Bulle C. et al. (2019). IMPACT World+: a globally regionalized life cycle impact assessment method. The International Journal of Life Cycle Assessment 24, 1653–1674. doi:10.1007/s11367-019-01583-0
- Budzinski K. et al. (2019). Introduction of a process mass intensity metric for biologics. New Biotechnology 49, 37–42. doi:10.1016/j.nbt.2018.07.005