Life-Cycle Assessment: Counting Everything, or Trying To
A wind turbine emits nothing while running. A solar panel emits nothing while running. Saying so and stopping there would be useless, because both had to be manufactured, transported, installed and eventually disposed of. Life-cycle assessment is the discipline of not stopping there, and it is why comparing energy sources is possible at all.
What Gets Counted
An assessment follows a product from cradle to grave. For a wind turbine that means iron ore mined and smelted, steel rolled into a tower, resin and glass fibre made into blades, concrete poured for a foundation, components shipped, a crane assembling the machine, decades of maintenance visits, and finally decommissioning and whatever recycling is achievable.
Each of those steps is converted to a common unit, usually kilograms of carbon dioxide equivalent, in which methane and nitrous oxide are weighted by their warming effect relative to carbon dioxide. The total is then divided by all the electricity the machine will produce over its life - which is why the capacity factor and assumed lifetime matter as much to an LCA result as the manufacturing itself.
The standard defines four stages: goal and scope definition, inventory analysis where the physical flows are compiled, impact assessment where flows are converted to effects, and interpretation. The international standards ISO 14040 and 14044 govern the procedure, which is why credible assessments state their assumptions explicitly and why one that does not should be read with suspicion.
The resulting figures are remarkably stable across independent studies for mature technologies. The IPCC's harmonised medians are the usual reference: coal around 820 grams per kilowatt-hour, gas 490, solar photovoltaics 41, hydropower 24, nuclear 12 and wind 11. The remarkable part is not the ranking but that studies using different methods converge.
Why Boundaries Decide the Answer
The single largest source of variation between assessments is not measurement error but where the analyst drew the line. Does a solar panel's assessment include the transmission line built to connect the farm? The storage needed to make its output usable? The gas plant held in reserve for still, dark evenings?
Each of those inclusions is defensible and each changes the number substantially. A study that counts only the panel gets 41 grams. A study that allocates a share of system-level balancing to the panel gets considerably more. Neither is dishonest; they answer different questions, and the disagreement between advocates often turns out to be a disagreement about scope rather than about physics.
The manufacturing grid is the second large lever. A solar panel made in a province running on coal carries far more embodied carbon than the same panel made where the factory grid is hydro. Since most panels are made in a handful of places, this single assumption can shift a global average by a third - and it improves on its own as those grids decarbonise, which means published figures age in a predictable direction.
Allocation is the third. When a process yields two products - a refinery producing both diesel and petrochemicals, or a plant producing electricity and district heat - the emissions have to be divided between them, and there is no physically correct way to do it. The standards permit several approaches and require the analyst to say which was used.
Two Ways to Build the Inventory
Process-based assessment works upward from physical reality. The analyst lists every input - so many tonnes of steel, so many kilowatt-hours of factory electricity - and attaches an emission factor to each. It is precise about what it includes and systematically blind to what it forgets, and the forgetting is unavoidable: a full supply chain branches without limit, so every study truncates somewhere.
Input-output assessment works downward from economics. National accounts record what every industrial sector buys from every other, and environmental accounts record what each sector emits. Tracing a euro of spending through that matrix captures the entire economy by construction, with nothing truncated. The cost is resolution: it knows what the average euro spent on machinery emits, not what this particular turbine emits.
Hybrid approaches use process data for the important, well-understood foreground and input-output data to catch the long tail of the background. Most serious modern assessments are hybrid for this reason.
Comparisons of the two methods consistently find that pure process-based studies underestimate, typically by 10 to 30 percent, because of truncation. That is worth knowing when reading a manufacturer's published figure, which is almost always process-based.
Reading One Critically
Four questions expose most of what matters. What is the functional unit - a kilowatt-hour delivered to the grid, or generated at the plant, which differ by transmission losses? What lifetime and capacity factor were assumed, since both sit in the denominator? Where does the boundary stop? And who paid for the study?
That last question is not cynicism. Industry-funded assessments are not automatically wrong and are frequently the most detailed available, because the sponsor has the process data nobody else has. But scope choices are judgement calls, and judgement calls made by an interested party deserve the same scrutiny as any other.
Carbon is also only one output. An assessment typically reports water consumption, land occupation, human toxicity, eutrophication and mineral resource depletion alongside greenhouse gases, and the rankings differ. Solar photovoltaics scores well on carbon and less well on mineral depletion and toxicity because of the metals involved. Hydropower scores well on carbon and can score badly on land use and biodiversity. Reporting only the carbon column is a choice, and usually an unstated one.
Used carefully, LCA is the only tool that allows technologies to be compared at all, and the convergence of independent studies on the same broad ranking is a genuine result. Used carelessly, it produces a single number that travels into headlines detached from the assumptions that generated it - and most of the arguing about energy emissions is arguing about those assumptions without saying so.
Frequently asked questions
What is life-cycle assessment?
A standardised method for counting the environmental effects of a product across its whole life - raw material extraction, manufacture, transport, use and disposal. For electricity it produces a figure in grams of carbon dioxide equivalent per kilowatt-hour, which is what makes comparison between technologies possible.
What are the typical carbon figures for electricity?
Harmonised medians in grams of carbon dioxide equivalent per kilowatt-hour: coal around 820, natural gas 490, solar photovoltaics 41, hydropower 24, nuclear 12 and wind 11. Independent studies using different methods converge on this ranking, which is itself a meaningful result.
Why do studies disagree?
Mostly because of system boundaries rather than measurement. Whether transmission, storage or backup capacity is allocated to a source can change its result by a factor of two. The manufacturing grid matters too: a panel made on a coal grid carries far more embodied carbon than the same panel made on a hydro grid.
What is the difference between process-based and input-output LCA?
Process-based builds up from physical inputs and is precise about what it includes but inevitably truncates the supply chain somewhere. Input-output works down from national economic accounts and captures everything but at low resolution. Comparisons find process-based studies underestimate by typically 10 to 30 percent.
Does LCA only measure carbon?
No. A full assessment reports water use, land occupation, toxicity, eutrophication and mineral depletion alongside greenhouse gases, and technologies rank differently on each. Solar scores well on carbon and less well on mineral depletion; hydropower scores well on carbon and can score badly on land and biodiversity.
