Ask a biomanufacturing site how much carbon it emits and you will get a number with a method behind it. Ask the same site for its water footprint and you will usually get a meter reading, which is a different thing entirely. Water is the largest single mass flow in almost every bioprocess, the smallest line on almost every utility bill, and the input whose environmental significance depends most violently on where the plant happens to sit.
This article works the water side of a bioprocess properly: what a water footprint is under ISO 14046, where the litres actually go, the energy term that most inventories miss, and how the answer changes depending on which metric you are being judged by. Every figure below is generated from the bioprocess LCA calculator on one modelled 2000 L CHO fed-batch, the same batch used across this cluster, so the results can be checked rather than taken on trust.
What a water footprint actually measures
Three distinct quantities get called a water footprint, and conflating them is the most common error in the field.
- Withdrawal is every litre taken from the source. It is what your meter reads and what most sites report.
- Consumption is the portion that does not return to the same watershed in the same period and quality. Evaporative losses from cooling towers, water incorporated into product, and water lost to reject streams all count. Effluent that goes to a treatment works and back to the river does not.
- The scarcity-weighted water footprint multiplies the consumptive volume by a regional characterisation factor, so that a cubic metre in a stressed basin counts for more than a cubic metre in a wet one.
ISO 14046 governs the third and requires you to state which you are reporting. That requirement is not bureaucratic. As section 5 shows, the three numbers for a single batch can differ by three orders of magnitude, so a water footprint quoted without its basis is not a result at all, it is a rumour.
Kounina and colleagues catalogued the water footprint methods in circulation and found them differing not just in factors but in what they claim to model, from simple volumetric inventories to full damage pathways. Pfister, Koehler and Hellweg gave the field its first widely used stress-based characterisation, and AWARE later replaced it as the consensus midpoint. All three are in the references.
Run your own water inventory
Enter process and CIP volumes, generation energy and grid region. The calculator returns your water footprint in litres per kilogram, its PMI contribution and the carbon it actually carries.
Where the water goes in a bioprocess
The modelled batch is a 2000 L CHO fed-batch at 80% working volume, 3 g/L titer, 70% downstream yield and 90% batch success, giving 3.024 kg of product. Its water footprint splits into two lines: 58,000 L of process water and 2,000 L of cleaning water, for 60,000 L per batch.
The 58,000 L figure surprises people who think of a bioreactor as the water consumer. It is not. The bioreactor holds 1,600 L. Everything else is downstream: buffer preparation for chromatography and filtration commonly runs 30 to 60 litres per litre of harvest, and that ratio, not the vessel, sets the water footprint. If you want to see where those buffer volumes come from, the buffer preparation calculator builds them from column volumes and step counts.
Cleaning water is the line that separates process formats. The matched stainless preset in the calculator uses 26,000 L of CIP water against 2,000 L for the single-use case, pushing its total to 84,000 L per batch and its intensity to 27,778 L/kg. That is a 40% larger water footprint for identical product, and it is the single largest environmental difference between the two formats. The full term-by-term decomposition, including why the carbon footprints nonetheless land within 0.2% of each other, is in single-use vs stainless environmental impact. If cleaning volumes are your lever, the cleaning validation calculator sizes rinse volumes against carryover limits.
The forgotten term: generation energy
Most water footprint inventories stop at two emission factors: one for supply, one for effluent treatment. In the factor set used here those are 0.149 and 0.708 kg CO2e per cubic metre, both from the UK Government GHG conversion factors under the Open Government Licence, for a combined 0.857 kg CO2e/m³. Applied to 60 m³ that gives 51.4 kg CO2e per batch, which is 0.43% of the batch total and looks entirely ignorable.
It is ignorable. The term that is not ignorable never appears in the water line at all, because it appears in the electricity line: the energy to make pharmaceutical-grade water. At the calculator's default of 15 kWh/m³, 60 m³ draws 900 kWh, which is 3.13% of the batch's 28,759 kWh. On a 400 g CO2e/kWh grid that is 360 kg CO2e.
The 7× result, and when it fails
Generation carbon divided by supply-plus-treatment carbon is a closed form with no process dependence at all:
ratio = generation [kWh/m³] × grid [g CO2e/kWh] ÷ 1000 ÷ 0.857 [kg CO2e/m³]
At 15 kWh/m³ and 400 g/kWh that is 7.00. Volume cancels, so the ratio is identical for a 200 L pilot and a 10,000 L production train. It falls to 1 only when the grid reaches 57.1 g CO2e/kWh, or when generation energy falls to 2.14 kWh/m³. Across the 26 US eGRID subregions the ratio runs from 2.35× in NPCC Upstate New York (134.3 g/kWh) to 13.33× in MRO East (761.8 g/kWh). Zero of the 26 fall below 1. On any US grid, the carbon in a bioprocess water footprint is the carbon of making the water, not of buying or disposing of it.
This is why generation energy deserves its own line even though it is physically electricity. Put it in the electricity total and it disappears into a number dominated by cleanroom HVAC; put it in the water line and the water footprint stops being a rounding error. The engineering detail behind that 15 kWh/m³ figure, and the distribution loop that keeps the water at temperature afterwards, is covered in WFI and purified water systems for bioprocessing.
The number is also a real lever rather than a fixed cost of doing business. The calculator's own guidance puts multi-effect distillation at roughly 10 to 25 kWh/m³ and reverse osmosis with hot-water sanitisation at roughly 2 to 6. Moving from the top of the first band to the middle of the second is a design decision, not an accounting one, and European regulators have permitted non-distillation routes to water for injection since 2019.
| Generation energy | kWh/batch | Batch electricity | kg CO2e/kg product | vs 15 kWh/m³ |
|---|---|---|---|---|
| 0 (municipal, no polishing) | 0 | 27,859 | 3,861 | −2.99% |
| 3 kWh/m³ | 180 | 28,039 | 3,885 | −2.39% |
| 6 kWh/m³ (RO / EDI) | 360 | 28,219 | 3,909 | −1.79% |
| 15 kWh/m³ (default) | 900 | 28,759 | 3,980 | base |
| 25 kWh/m³ | 1,500 | 29,359 | 4,060 | +1.99% |
| 45 kWh/m³ (single-effect) | 2,700 | 30,559 | 4,218 | +5.98% |
The whole sweep, from free municipal water to single-effect distillation, moves the batch footprint by 9.0%, all of it inside the water line. That is larger than the entire single-use plastic inventory and smaller than one step of grid decarbonisation, which is roughly the right mental filing place for it.
99.8% of PMI, 3.4% of the carbon, 0.07% of the cost
Here is the result that makes a water footprint genuinely awkward to manage. The same 60,000 L, scored three ways on the same batch:
| Metric | Total | Water's share | What it implies |
|---|---|---|---|
| Process mass intensity | 19,884 kg/kg | 99.78% | Water is essentially the entire problem |
| Carbon footprint | 3,980 kg CO2e/kg | 3.42% | Water is a minor contributor |
| Batch cost | $468,000 | ≈0.07% | Water is invisible |
PMI counts kilograms in, and a litre of water weighs about a kilogram, so it swamps the 15.9 kg/kg of media and 27.1 kg/kg of consumables that make up the rest. Carbon counts emissions, and water is cheap to move and moderately cheap to make. Cost counts dollars, and at the life cycle costing page's assumed $0.004 per litre for purified water plus $0.12 per kWh for the generation electricity, the entire water bill is about $348 against a $468,000 batch.
A mass metric and a carbon metric therefore disagree about the same water footprint by a factor of roughly 29. Neither is wrong. They answer different questions, which is exactly why ISO 14044 refuses to let a public comparative assertion collapse impact categories into one weighted score. The process mass intensity page works the mass side in full, and life cycle impact assessment covers why category selection, not method vintage, is the choice that changes conclusions.
Water's share of three metrics on one batch. Log scale, because a linear axis cannot show 0.07% and 99.78% together.
Scarcity weighting and the 1,000× problem
Everything so far treats a litre as a litre. Any water footprint intended to say something about environmental impact cannot do that, because a cubic metre drawn from a stressed basin and a cubic metre drawn from upstate New York are not the same environmental event.
AWARE, the WULCA consensus method for the scarcity-weighted water footprint described by Boulay and colleagues, is the current midpoint answer. It reports Available WAter REmaining per unit area in a watershed after human and ecosystem demand has been met, normalised against the world average and inverted. WULCA bounds the resulting characterisation factor between 0.1 and 100, with 1 representing the world-average watershed. It also publishes consumption-weighted world averages: 20 for non-agricultural use, 43 for unknown use, 46 for agricultural.
Applying that scale to our water footprint, as an upper bound in which every litre is treated as consumptive:
| AWARE factor | Represents | Scarcity-weighted result |
|---|---|---|
| 0.1 | Method floor, water-abundant basin | 2 m³ world-eq/kg |
| 1 | World-average watershed | 20 m³ world-eq/kg |
| 20 | WULCA world average, non-agricultural use | 397 m³ world-eq/kg |
| 43 | WULCA world average, use unknown | 853 m³ world-eq/kg |
| 100 | Method ceiling, severely stressed basin | 1,984 m³ world-eq/kg |
Siting moves the scarcity-weighted water footprint by a factor of 1,000 and the carbon result by nothing. That asymmetry is the strongest argument in this cluster for reporting the water footprint as its own impact category rather than folding it into a single sustainability score, and it mirrors the finding in attributional vs consequential LCA that siting is worth 5.1× under one method and exactly zero under another.
Two honest caveats. First, AWARE applies to consumptive use, and most of a bioprocess's water leaves as treated effluent rather than as vapour or product. The table above is therefore a ceiling; scale it linearly by whatever consumptive fraction your site can defend, and state that fraction. Second, WULCA warns explicitly that a country-level aggregate does not represent an average picture of that country, because it is weighted toward where and when water is actually consumed. Use basin-level factors if you have them.
Compare two sites before you commit
Swap grid region and process volumes in the calculator to see how much of a siting decision is carbon and how much is water footprint.
Six process archetypes compared
Running all six calculator presets at a 400 g CO2e/kWh grid gives the spread of the water footprint across formats, and one genuinely counter-intuitive result.
| Process | Water/batch | L per kg product | Generation kWh | Share of electricity | Water share of carbon |
|---|---|---|---|---|---|
| Pichia fed-batch 5,000 L | 123,000 L | 5,423 | 1,845 | 12.8% | 12.55% |
| E. coli fed-batch 10,000 L | 195,000 L | 6,423 | 2,925 | 28.6% | 23.31% |
| CHO perfusion 500 L | 93,200 L | 10,153 | 1,398 | 4.0% | 4.29% |
| Microbial pilot 200 L | 5,400 L | 16,941 | 32 | 2.3% | 2.62% |
| CHO fed-batch 2,000 L, single-use | 60,000 L | 19,841 | 900 | 3.1% | 3.42% |
| CHO fed-batch 2,000 L, stainless | 84,000 L | 27,778 | 1,260 | 4.3% | 4.80% |
The intensity range is 5.1-fold, and the ordering has almost nothing to do with water discipline. The E. coli case uses more than three times the absolute volume of the CHO case and still lands at a third of the intensity, because it makes 30.36 kg of product against 3.024. A water footprint expressed per kilogram is a productivity metric wearing a sustainability costume.
The counter-intuitive result is in the last two columns. Water carries 3.42% of the CHO batch's carbon but 23.31% of the E. coli batch's, and generation is 28.6% of its electricity against 3.1%. Nothing about the water changed. What changed is the denominator: cleanroom HVAC is 93.5% of the CHO batch's electricity because that batch occupies a suite for 14 days, and only 21.1% of the E. coli batch's because that one runs for 1.5. The water footprint matters most where the cleanroom term does not. Short microbial campaigns are where it is a real carbon lever; long mammalian ones are where it is noise.
Perfusion sits in between, at 10,153 L/kg. It uses 93,200 L per campaign, far more than fed-batch in absolute terms, but spread over 9.18 kg of product. The perfusion calculator models the media exchange rates that drive that volume, and the media estimator converts them into consumption per campaign.
Which levers actually move the number
Four candidate interventions on the modelled batch, each scored against all three metrics:
| Lever | L per kg | PMI | kg CO2e/kg |
|---|---|---|---|
| Base case | 19,841 | 19,884 | 3,980 |
| Halve process water (buffer concentrates, in-line dilution) | 10,251 (−48.3%) | 10,294 (−48.2%) | 3,914 (−1.65%) |
| Generation 15 → 6 kWh/m³ | 19,841 (0%) | 19,884 (0%) | 3,909 (−1.79%) |
| Eliminate CIP water entirely (2,000 → 0 L) | 19,180 (−3.3%) | 19,223 (−3.3%) | 3,976 (−0.11%) |
| Titer 3 → 5 g/L | 11,905 (−40.0%) | 11,931 (−40.0%) | 2,388 (−40.0%) |
Three structural conclusions come out of that table.
- The two headline water footprint levers are almost orthogonal. Halving process water is the biggest volumetric win available and is worth 1.65% of carbon. Cutting generation energy is a comparable carbon win and is worth zero litres. A site under a volumetric target and a site under a carbon target should do different things.
- Titer beats both, on both. Raising titer from 3 to 5 g/L cuts every intensity metric by exactly 40%, because it only changes the denominator. Volume per kilogram is exactly inverse in titer: 1 g/L gives 59,524 L/kg and 10 g/L gives 5,952. The marginal abatement cost table reaches the same conclusion from the cost side, where titer is a negative-cost lever at −$30,907 per tonne of CO2e avoided while water is only −$846.
- Eliminating cleaning water entirely is worth 0.11% of the carbon footprint. If a single-use conversion is being justified on environmental grounds, the CIP saving is not the argument. Say so before someone in review says it for you.
Volume levers and carbon levers are different levers. Only titer is both.
None of this argues against reducing a water footprint. At 16 batches a year, the throughput assumed in the costing page, the single-use case draws 960 m³ a year and the stainless case 1,344, a difference of 384 m³. In a stressed basin that is a permitting conversation regardless of what it does to a carbon number, and it is precisely the case where the scarcity factor in section 5 earns its keep.
Five mistakes that break a water footprint
- Reporting withdrawal and calling it a water footprint. A meter reading is an inventory line, not an impact result. State the basis, and if you have not estimated the consumptive fraction, say that too.
- Leaving generation energy in the electricity total. It is 3.1% of electricity on the modelled CHO batch and 28.6% on the E. coli one. Buried in a line dominated by HVAC it is invisible, and the water footprint looks seven times smaller than it is.
- Using a country-level scarcity factor for a specific site. WULCA warns that country aggregates are consumption-weighted and can exclude most of a country's land area. Basin-level factors exist; use them.
- Comparing water footprints across processes without normalising to product. The E. coli case uses 3.25× the water of the CHO case and has a third of the intensity. Absolute volumes rank facilities; intensities rank processes.
- Assuming the single-use route wins on water because it skips CIP. It does use 29% less water per batch, but removing the CIP line entirely from the single-use case changes its carbon footprint by 0.11%. The volumetric win is real and the carbon win is not, and conflating them is how a sustainability claim becomes indefensible in review.
For the wider picture of where a fermentation footprint actually comes from, and where the water footprint sits inside it, the pillar page on the carbon footprint of fermentation puts the water term in context alongside grid, HVAC and materials, and the worked LCA example runs all four ISO phases on this same batch.
Frequently asked questions
What is the water footprint of biomanufacturing?
On a modelled 2000 L CHO fed-batch producing 3.024 kg of product, the calculator returns 60,000 litres of process and cleaning water per batch, which is 19,841 litres per kilogram of product. Across six process archetypes the figure spans 5,423 to 27,778 litres per kilogram, a 5.1-fold range driven almost entirely by titer and downstream yield rather than by water discipline.
Is a water footprint the same as water consumption?
No. Three different quantities get called the same thing. Withdrawal is every litre taken from the source. Consumption is the part that does not return to the same watershed, because it evaporated, was incorporated into product or left as a different quality. The scarcity-weighted footprint multiplies the consumptive volume by a regional characterisation factor. ISO 14046 requires you to say which one you are reporting, because they can differ by orders of magnitude on the same process.
Why does WFI generation energy matter more than the water itself?
Because making pharmaceutical-grade water is an energy process, not a water process. On the modelled batch at 15 kWh per cubic metre and a 400 g CO2e/kWh grid, generation contributes 360 kg CO2e while supply and effluent treatment together contribute 51.4 kg. That is a ratio of exactly 7.0. The ratio equals generation energy times grid intensity divided by the combined supply and treatment factor, so it only falls below 1 on a grid cleaner than 57 g CO2e per kWh. None of the 26 US eGRID subregions is that clean.
How much of process mass intensity is water?
99.78% on the modelled batch. Total PMI is 19,884 kg per kg of product, of which water is 19,841, media 15.9 and single-use consumables 27.1. Yet that same water is only 3.42% of the batch carbon footprint and about 0.07% of batch cost. A mass-based metric and a carbon-based metric disagree about water by a factor of roughly 29.
What is AWARE and do I need it?
AWARE is the WULCA consensus method for water scarcity footprints. It reports Available WAter REmaining per area in a watershed after human and ecosystem demand, expressed as a characterisation factor bounded between 0.1 and 100 where 1 is the world average. WULCA gives a consumption-weighted world average of 20 for non-agricultural use and 43 for unknown use. You need it whenever you compare sites in different watersheds, because the same litre can be worth a thousand times more in one basin than another while the carbon of that litre does not change at all.
Which levers actually reduce a water footprint?
It depends which number you are being measured on. Halving process water through buffer concentrates cuts litres per kilogram by 48.3% but the carbon footprint by only 1.65%. Moving WFI generation from 15 to 6 kWh per cubic metre cuts carbon by 1.79% and litres by nothing at all. Raising titer from 3 to 5 g/L cuts both by 40% because it changes the denominator. Titer is the only lever that moves every metric at once.
References
- Boulay A.-M., Bare J., Benini L., Berger M., Lathuillière M.J., Manzardo A., Margni M., Motoshita M., Núñez M., Pastor A.V., Ridoutt B., Oki T., Worbe S. & Pfister S. (2018). The WULCA consensus characterization model for water scarcity footprints: assessing impacts of water consumption based on available water remaining (AWARE). The International Journal of Life Cycle Assessment 23, 368–378. doi:10.1007/s11367-017-1333-8
- Pfister S., Koehler A. & Hellweg S. (2009). Assessing the environmental impacts of freshwater consumption in LCA. Environmental Science & Technology 43, 4098–4104. doi:10.1021/es802423e
- Kounina A., Margni M., Bayart J.-B., Boulay A.-M., Berger M., Bulle C., Frischknecht R., Koehler A., Milà i Canals L., Motoshita M., Núñez M., Peters G., Pfister S., Ridoutt B., van Zelm R., Verones F. & Humbert S. (2013). Review of methods addressing freshwater use in life cycle inventory and impact assessment. The International Journal of Life Cycle Assessment 18, 707–721. doi:10.1007/s11367-012-0519-3
- Budzinski K., Blewis M., Dahlin P., D'Aquila D., Esparza J., Gavin J., Ho S.V., Hutchens C., Kahn D., Koenig S.G., Kottmeier R., Millard J., Snyder M., Stanard B. & Sun L. (2019). Introduction of a process mass intensity metric for biologics. New Biotechnology 49, 37–42. doi:10.1016/j.nbt.2018.07.005
- Pietrzykowski M., Flanagan W., Pizzi V., Brown A., Sinclair A. & Monge M. (2013). An environmental life cycle assessment comparison of single-use and conventional process technology for the production of monoclonal antibodies. Journal of Cleaner Production 41, 150–162. doi:10.1016/j.jclepro.2012.09.048