How a Life-Cycle Assessment of a Fuel Is Built
A life-cycle assessment (LCA) of a fuel is structured bookkeeping. Researchers pick a unit to measure, such as 1 MJ of fuel or 1 kg of hydrogen, draw a boundary around the steps to count, add up the emissions at each step, convert every gas into one common unit, and decide how to treat by-products. The final number depends on those choices as much as on the fuel itself.
That is why two credible studies of the same fuel can disagree. This article follows an LCA through its choices, using Argonne National Laboratory's GREET model and published studies of ethanol and hydrogen from EPA, Argonne and the National Energy Technology Laboratory (NETL). What the finished numbers mean next to a burning-only factor is the job of combustion vs life-cycle emissions. You can view the sourced properties of both fuels in the Comparator for ethanol and hydrogen.
Four stages and one model
EPA's introductory guide to LCA names four stages: goal and scope definition, inventory analysis, impact assessment and improvement analysis. In a fuel study these become four jobs. First, state the purpose and the unit. Second, list every input and output at each step of the chain. Third, convert them into one measure. Fourth, test and refine the result.
The unit is called the functional unit. EPA's tables for renewable fuels use kg of CO2e per million Btu of fuel, and EPA compares each pathway with the petroleum fuel it would displace. Argonne's methanol study uses 1 MJ. NETL's hydrogen study uses 1 lb of hydrogen at the plant gate, reported as lb of CO2e per lb of hydrogen, which is numerically the same as kg per kg.
One widely used tool for this bookkeeping is GREET, which stands for Greenhouse gases, Regulated Emissions, and Energy use in Technologies. The Department of Energy says it has co-funded Argonne National Laboratory's model since 1995, that it covers hundreds of fuel production pathways plus the manufacture of vehicle powertrains, and that it is free on two platforms, GREET.net and GREET Excel. DOE reports more than 40,000 registered users worldwide.
Drawing the system boundary
The boundary decides which steps count. Studies use shorthand names for it.
| Label | What it covers | Example |
|---|---|---|
| Well-to-wheels | Extracting or growing the feedstock through burning the fuel in a vehicle | GREET's fuel pathways, per DOE |
| Cradle-to-gate | Raw material through the plant gate, without delivery or use | NETL hydrogen: natural gas extraction to hydrogen at the plant gate |
| Cradle-to-grave | Cradle-to-gate plus delivery, dispensing and consumption | Argonne methanol study |
| Vehicle cycle | Vehicle materials through end-of-life recycling | GREET's vehicle analysis, per DOE |
The label can change a number a great deal. A cradle-to-gate hydrogen result leaves out delivery and end use, so it cannot be set beside a well-to-wheels gasoline figure. Rules about exclusions matter too. EPA's renewable fuel analysis counts significant indirect effects such as land use change, but leaves out infrastructure emissions such as building the plant and employee commuting. It also builds the use stage from average vehicle-fleet emissions.
Adding up the stages and converting the gases
Each stage becomes a line in the inventory. NETL's approach shows the logic. It took each hydrogen plant's engineering results to find the upstream energy inputs, then applied emission factors to those inputs and added the emissions at the plant itself.
| Case | Design | Result |
|---|---|---|
| 1A | Steam methane reforming (SMR), no capture, no steam credit | 12 |
| 1B | SMR, no capture, with steam credit | 10 |
| 2 | SMR with CO2 capture | 4.6 |
| 3 | Autothermal reforming with CO2 capture | 5.7 |
For Case 1A, NETL says about 75% of the total is CO2 from the plant stack, 22% comes from natural gas procurement and 3% from grid electricity. Case 1A is the 12 kg CO2e per kg of hydrogen that the Comparator records for hydrogen from natural gas without capture.
Ethanol, described in ethanol fuel explained, is an addition problem too. EPA's table for corn starch ethanol from a natural-gas dry mill (2022 average case) lists five stages, in kg CO2e per million Btu: agriculture 16.5, land use change 27.8, feedstock transport 3.0, fuel production 27.9, and distribution and use 2.1. Their sum is 77.3, while the table's total is 77.2; the gap comes from rounding.
The last conversion is from separate gases to CO2e. Each gas is multiplied by a global warming potential (GWP), and studies do not all pick the same values.
| Source | Methane (CH4) | Nitrous oxide (N2O) | Note |
|---|---|---|---|
| EPA GHG Emission Factors Hub, 2025 | 28 | 265 | 100-year, IPCC Fifth Assessment Report values |
| Argonne methanol study | 30 | 265 | Used with the GREET 2020 model |
| NETL hydrogen study | 36 | 298 | 100-year with atmospheric carbon feedback, fossil values |
One kilogram of methane therefore counts as 28, 30 or 36 kg of CO2e, depending on the study. Methane from the natural gas supply chain feeds directly into the hydrogen result above, and the NETL study calls upstream natural gas the main driver of its variability.
Co-products, uncertainty and why studies differ
Many fuel plants make more than one product, so the bookkeeping must decide who carries which emissions. EPA models farm co-products such as distillers grains, a livestock feed from ethanol plants, inside its agricultural analysis, and judges other co-products by the emissions impacts of their most likely uses. Argonne's integrated methanol systems divide emissions among their outputs by energy content. NETL used system expansion: Case 1B credits the excess steam a plant exports, assuming it replaces steam from a natural gas boiler, which lowers the result from 12 to 10.
Every input carries uncertainty. NETL ran a Monte Carlo simulation with 5,000 iterations and reports the 5th to 95th percentile range. Its upstream natural gas results span regions, from Alaskan offshore gas at the low end to San Juan conventional gas at the high end, and the reforming cases vary by about 25%. The data represent 2016 to 2017.
Studies can also report different answers for the same fuel. EPA's table puts corn ethanol from a natural-gas dry mill 21% below its gasoline baseline. DOE's Alternative Fuels Data Center cites an Argonne analysis that finds an average reduction of 40% for corn-based ethanol. Differences of this kind can come from models, base years and assumptions, which is why the boundary and year must be read before comparing. In EPA's table, land use change alone is 27.8 of the 77.2 total, so one assumption can move the result a long way. For hydrogen made another way, the electricity input needs its own figure, covered in carbon intensity of electricity, and the color labels are explained in gray, blue and green hydrogen.
Common mistakes
- Treating an LCA result as a property of the fuel. It is the output of a model, shaped by a unit, boundary and method. Density is a property. A life-cycle value is not.
- Comparing different functional units. A figure per MJ, per million Btu and per kg of hydrogen cannot be set side by side without conversion and a shared basis.
- Ignoring the boundary label. Cradle-to-gate leaves out delivery and use. Cradle-to-grave and well-to-wheels include them.
- Assuming CO2e is a fixed conversion. Methane counts as 28, 30 or 36 in the three sources above.
- Forgetting co-products. A steam credit moved NETL's hydrogen result from 12 to 10 without changing the plant's fuel use.
- Reading one value as a measurement. LCAs are modeled, and NETL publishes error bars for that reason.
Frequently asked questions
What is GREET?
GREET, short for Greenhouse gases, Regulated Emissions, and Energy use in Technologies, is a model from Argonne National Laboratory, co-funded by the Department of Energy since 1995. It estimates life-cycle emissions and energy use for hundreds of fuel pathways and for vehicle manufacturing, and it is free on two platforms.
What is a functional unit in a fuel LCA?
It is the quantity that results are reported against, such as 1 MJ of methanol, 1 lb of hydrogen at the plant gate or 1 million Btu of ethanol. Without a common functional unit, two results cannot be compared. Always check the unit before reading a number.
What is the difference between cradle-to-gate and cradle-to-grave?
Cradle-to-gate counts raw material extraction through the plant gate, where the fuel is made. Cradle-to-grave adds delivery, dispensing and use, according to Argonne's methanol study. A cradle-to-gate figure is therefore smaller in scope and should not be compared with a figure that includes burning.
Why do two studies of the same fuel give different numbers?
They may use different boundaries, years, plant types, global warming potentials and co-product methods. EPA's 21% reduction for natural-gas corn ethanol and Argonne's 40% average cited by DOE show the effect. The EPA figure includes a land use change estimate, which is a large share of its total.
Sources
- Life Cycle Assessment: Principles and Practice, EPA/600/R-06/060: U.S. Environmental Protection Agency. Accessed November 13, 2026.
- Lifecycle Analysis of Greenhouse Gas Emissions under the Renewable Fuel Standard: U.S. Environmental Protection Agency. Accessed November 13, 2026.
- Lifecycle Greenhouse Gas Emissions for Select Pathways (kg CO2e per mmBtu): U.S. EPA Office of Transportation and Air Quality. Accessed November 13, 2026.
- GHG Emission Factors Hub: U.S. Environmental Protection Agency. Accessed November 13, 2026.
- Life Cycle Emissions Analysis: U.S. Department of Energy, Hydrogen Program. Accessed November 13, 2026.
- Comparison of Commercial, State-of-the-Art, Fossil-Based Hydrogen Production Technologies, DOE/NETL-2022/3241: National Energy Technology Laboratory. Accessed November 13, 2026.
- Technoeconomic and Life Cycle Analysis of Synthetic Methanol Production from Hydrogen and Industrial Byproduct CO2: Argonne National Laboratory, via OSTI. Accessed November 13, 2026.
- Ethanol Vehicle Emissions: U.S. Department of Energy, Alternative Fuels Data Center. Accessed November 13, 2026.
Spotted something wrong or out of date? Tell us through the corrections page. We answer within 10 business days.