What does ‘clean’ mean from cradle to grave? Life cycles, water and land
Drag a boundary around a solar farm, a gas plant and a data center, and watch carbon, water and land change with what you count
- Analyst · L2
- 25 min guided · about 30 min practice
- After F11
- Completed
Three objects, two boundaries: on site only, and cradle to grave
| Object and footprint | On site only | Cradle to grave |
|---|---|---|
| Solar farm, carbon | 0 g CO₂e per kWh | 43 g CO₂e per kWh |
| Solar farm, water consumed | 1 gal per MWh | 82 gal per MWh |
| Solar farm, land | 10.3 m² per MWh a year | 10.3 m² per MWh a year (2 stages not counted) |
| Gas plant, carbon | 389 g CO₂e per kWh | 461 g CO₂e per kWh (stages added; whole-life median 486) |
| Gas plant, water consumed | 205 gal per MWh | 222 gal per MWh |
| Gas plant, land | under 0.4 m² per MWh a year | 4.1 m² per MWh a year (2 stages not counted) |
| Data center, carbon | not counted | 340 g CO₂e per kWh (3 stages not counted) |
| Data center, water consumed | 99 gal per MWh | 1,293 gal per MWh (2 stages not counted) |
| Data center, land | not counted | not counted |
The gas plant has a cooling tower; the data center is the 2023 U.S. average, per unit of electricity it uses.
A new solar farm's attributional share is about 43 g per kWh.
If gas plants run less, the consequential change against not building it is 43 − (71 + 389) = −417 g per kWh: avoided, not removed.
With JavaScript on, you drag the boundary stage by stage, pick the cooling, count the land between and switch on a drought.
Medians and averages, not one plant. Each footprint is a published median by stage, or a 2023 U.S. average for data centers; a real plant can sit anywhere in the ranges.
Carbon in g CO₂e per kWh, the same number as kg per MWh: NREL's 2021 medians by life-cycle phase.
Water in U.S. gallons consumed per MWh (1 gallon = 3.785 litres), from two NREL-led reviews. Land in m² held per MWh made in a year.
The data center is a load: Berkeley Lab's 2023 averages for U.S. data centers, per unit of electricity used. Lesson F8's campus draws 240 MW in every hour.
About 25 guided minutes in eight steps.
What you just did, as diagrams
- Boundary
- On site, a solar farm emits 0 g per kWh; from cradle to grave, about 43 g, mostly from making the panels.
- One boundary
- Gas at its stack (389 g) against solar from cradle to grave (43 g) is unfair; on one boundary it is 486 against 43.
- Water
- Withdrawn is not consumed: once-through cooling takes 11,380 gal per MWh and consumes 100.
- Land
- Counting the land between moves gas from 4.1 to 19.0 m² per MWh a year, past solar's illustrative 14.9.
- A load
- A data center's indirect water, 1,194 gal per MWh it uses, is about 12 times its own cooling water (U.S., 2023).
- Two questions
- A new solar farm's share is about 43 g per kWh; what changes depends on what it replaces.
- Drought
- In EIA's drought case, hydropower falls from 15% to 8% of California's summer generation and gas rises from 45% to 50%.
Go deeper · AnalystMedians, ranges and harmonization
- Medians of published studies. NREL screened about 3,000 life-cycle studies and kept fewer than 15% (2021).
- Medians need not add. A median is the middle estimate after sorting. Gas’s phases add to 461 g per kWh, its total median is 486 g.
- Harmonized. Re-running studies on common assumptions narrows their spread (both from the papers’ abstracts).
- Harmonized medians. Crystalline-silicon panels 45 g per kWh (middle half 39 to 49); combined-cycle plants 450 g (420 to 480).
- What moves a solar number most. Sunshine, panel life and efficiency, and the grid that powered the factory.
- What moves a gas number most. The plant’s efficiency, and methane leaked on the way from the well. The harmonized gas studies count methane at its 100-year warming effect (Heath et al., 2014).
Go deeper · ModelerShared burdens and system expansion
- Two products, one plant. When a plant sells both power and heat, its burden is split between them, by energy or by value (ISO 14044).
- System expansion. Instead, the study can be widened to include what the second product replaces (ISO 14044).
- Attributional, by name. IPCC AR6 Working Group III, Annex III (2022): attributional studies use average data for a product’s own impacts.
- Consequential, by name. Consequential studies widen the boundary to the effects of a change, and depend more on their assumptions (the same annex).
- In a model. What responds to a change is an output of a dispatch or planning model: A10 asks it for a day, M6 over years.
Go deeperWhat this lesson leaves out
- Any one real plant: its sunshine, gas field, river and grid.
- Harms other than climate, water and land: air pollution, toxicity, wildlife.
- The land of mines and factories (their water is in Build where the reviews count it), and a data center's servers and buildings.
- Where water is scarce: a gallon consumed in a desert counts the same here as one by a full river.
Check yourself
Myth or fact? “A solar farm has no carbon footprint.”
- Myth
- Fact
Answer
Myth. On site it burns nothing, but from cradle to grave it is about 43 g CO₂e per kWh (NREL's median), mostly from making the panels.
Myth or fact? “A plant that withdraws more water consumes more.”
- Myth
- Fact
Answer
Myth. Once-through cooling withdraws 11,380 gallons per MWh and consumes 100; a cooling tower withdraws 255 and consumes 205 (medians).
Myth or fact? “Most of a data center's water is the water it cools with.”
- Myth
- Fact
Answer
Myth, for the U.S. average in 2023: 99 gallons per MWh on site against 1,194 at the power plants and reservoirs that make its electricity (Berkeley Lab).
A gas plant with a cooling tower makes 1,000 MWh. At the medians, how much water does its cooling consume?
- 255,000 gallons
- 205,000 gallons
- 11,380,000 gallons
- 205 gallons
Answer
205,000 gallons: 205 gallons per MWh × 1,000 MWh.
255,000 is what it withdraws; 11,380,000 would be the withdrawal of once-through cooling.
Numbers, sources and what’s next
Numbers to remember
| Number | What it means | Source |
|---|---|---|
| 43 and 486 g per kWh | Medians of published cradle-to-grave CO₂e of solar panels and of gas power. | NREL, NREL/FS-6A50-80580, Table 1 · September 2021 |
| 255 and 205 gal per MWh | A combined-cycle plant with a cooling tower: water withdrawn and consumed (medians). | Macknick et al., 2012 |
| 4.52 litres per kWh | Water consumed at the power plants that supplied U.S. data centers, per kWh used, 2023. | Berkeley Lab, 2024 report, page 57 · December 2024 |
Next
Sources and further reading
- NREL. Life Cycle Greenhouse Gas Emissions from Electricity Generation: Update (NREL/FS-6A50-80580), September 2021, Table 1 and Figure 2.
- Hsu, D. D. et al. (2012). Life Cycle Greenhouse Gas Emissions of Crystalline Silicon Photovoltaic Electricity Generation: Systematic Review and Harmonization. Journal of Industrial Ecology 16(S1). Abstract.
- O'Donoughue, P. R., Heath, G. A., Dolan, S. L. and Vorum, M. (2014). Life Cycle Greenhouse Gas Emissions of Electricity Generated from Conventionally Produced Natural Gas: Systematic Review and Harmonization. Journal of Industrial Ecology 18(1), 125–144. Abstract.
- IPCC (2022). Annex III: Scenarios and modelling methods. In Climate Change 2022: Mitigation of Climate Change, Working Group III contribution to the Sixth Assessment Report; Part I, Section 1 (life-cycle assessment).
- Meldrum, J., Nettles-Anderson, S., Heath, G. and Macknick, J. (2013). Life cycle water use for electricity generation: a review and harmonization of literature estimates. Environmental Research Letters 8, 015031; Tables 5, 6 and 11.
- Macknick, J., Newmark, R., Heath, G. and Hallett, K. C. (2012). Operational water consumption and withdrawal factors for electricity generating technologies: a review of existing literature. Environmental Research Letters 7, 045802; Tables 1 to 3.
- Bolinger, M. and Bolinger, G. (2022). Land Requirements for Utility-Scale PV: An Empirical Update on Power and Energy Density. IEEE Journal of Photovoltaics; Berkeley Lab summary of 1 February 2022 (2019 medians: 394 MWh a year per acre for tracking plants, 447 for fixed tilt).
- Heath, G. A., O'Donoughue, P., Arent, D. J. and Bazilian, M. (2014). Harmonization of initial estimates of shale gas life cycle greenhouse gas emissions for electric power generation. PNAS 111(31), E3167–E3176.
- Lovering, J., Swain, M., Blomqvist, L. and Hernandez, R. R. (2022). Land-use intensity of electricity production and tomorrow's energy landscape. PLOS ONE 17(7), e0270155.
- Shehabi, A. et al. (2024). 2024 United States Data Center Energy Usage Report. Lawrence Berkeley National Laboratory, LBNL-2001637, December 2024; pages 5, 55 and 57.
- USGS. Thermoelectric Power Water Use, 2020 estimates (Galanter et al., 2023); page read 6 November 2026.
- EIA, Today in Energy: power-sector water efficiency (14 June 2023: combined-cycle plants withdrew 2,803 gallons per MWh on average in 2021); dry cooling (29 August 2018); hydropower in the 2021 drought (30 March 2022).
- EIA. Short-Term Energy Outlook Supplement: Drought Effects on California Electricity Generation and Western Power Markets, May 2022, Summary.
- Earlier lessons: A10 (four emissions questions), F8 (a data center as a load), A9 (droughts and heat on the grid), F4 (how plants make power).
Spotted an error, or stuck? Write to hlli@ksu.edu. Corrections are made in place and dated here.
Text licensed CC BY 4.0, code MIT: teachers and students may reuse and adapt it, with credit to Honglin Li.