construction · environment · family: the missing yardstick
different answers for one building's carbon
Embodied carbon measurements vary across LCA databases
Problem statement
There is not yet a consistent, standardized method to measure embodied carbon in buildings. A 2025 systematic review reports that discrepancies in database structure and boundary assumptions can result in significant variations in carbon outcomes, even for identical buildings, and that carbon estimates can fluctuate by up to 40% depending on system boundaries. A NIST review adds that the choice of method matters too: for the building sector in Finland, input-output analysis gave estimates two times higher than process analysis. NIST also cautions that Environmental Product Declarations (EPDs) "cannot be used to compare the environmental performance of two different materials (e.g., concrete and wood)," and that EPDs should be compared only when their impacts were calculated with the same methodologies and life cycle modules for functionally equivalent products. This means the carbon figure reported for a building depends in part on the database, system boundary, and method the assessor chooses.
Why this matters
The building and construction sector was responsible for 37% of annual global carbon emissions in 2022, with 28% coming from building operations and 9% from construction materials. As energy efficiency continues to improve, building materials — their manufacture, transportation, installation, maintenance, and disposal — may become the dominant source of carbon emissions in new construction. Governments have begun to act on embodied carbon: the City of Vancouver's Zero Emissions Building Plan integrates embodied carbon into new construction permitting, and a California law requires the state's Air Resources Board to develop a framework to measure and achieve a 20% reduction in the carbon intensity of new building construction by 2030 and a 40% reduction no later than 2035. But embodied carbon assessments lack consistent system boundaries, agreed functional units, and common benchmarks, which limits the usefulness of carbon data in procurement or regulation. Assessments limited to the product stage distort results and may reward short-term performance gains while neglecting long-term impacts.
What’s been tried and why it hasn’t worked
System boundary inconsistencies plague the field: in one systematic review, product-stage phases appear in 98% of the articles, while only 39% address end-of-life or reuse phases. Use-stage modules are also underrepresented, although recurring embodied impacts during the use phase can range from 9% to over 100% of initial embodied carbon, depending on maintenance frequency and material choices. Sixty percent of the publications originate from only 5 countries (China, the United Kingdom, the United States, India, and Australia), leaving a gap in context-specific knowledge for developing and climate-vulnerable regions such as Sub-Saharan Africa, the Caribbean, and parts of Southeast Asia. The databases have gaps of their own: NIST's comparison of commercial databases notes that Athena has data only for structural and envelope materials and that Ecoinvent and GaBi are not U.S. based. EPDs, while increasingly adopted, vary significantly in format, assumptions, and data transparency, and where they are unavailable practitioners rely on generic databases or literature-derived values, which introduces inconsistencies in accuracy and system boundary assumptions.
What would unlock progress
A unified reference dataset with mandatory system boundary definitions — covering cradle-to-grave including end-of-life — would enable meaningful comparison across studies and databases. The Buildings review calls for regionally representative databases and for national benchmarking platforms that offer typology- and region-specific references, noting that many building designers must rely on generic or international references that fail to reflect local material availability, construction methods, or climatic conditions. Mandatory disclosure of database selection and system boundary assumptions in regulatory compliance reports could make the measurement uncertainty transparent to decision-makers.
Entry points for student teams
A team could select a standard building design (e.g., a 4-story residential building) and calculate its embodied carbon using 3–4 different databases (Ecoinvent and GaBi, which NIST lists as commercial data; ICE; or the publicly available USLCI data and free tools such as the Athena Impact Estimator and EC3), documenting the magnitude and sources of variation across databases for each building component. An architecture team could conduct a case study of an actual building project comparing EPD-based calculations against site-specific data. Relevant disciplines: civil engineering, architecture, environmental science, data science.
Genes
Structural cousins: stuck the same way, in other fields
Sources
Zhang, Y. et al., "Systematic Review of Embodied Carbon Assessment and Reduction in Building Life Cycles," NIST Special Publication 1324, 2024, Bacheva, T.S. and Raposo Grau, J.F., "Embodied Impacts in Buildings: A Systematic Review of Life Cycle Gaps and Sectoral Integration Strategies," Buildings, 15(10):1661, 2025, accessed 2026-02-20 go to source 1 ↗ go to source 2 ↗
Working notes
Notes made while writing and checking this brief, shown as written. They use working shorthand.
NIST Special Publication 1324 is a review of embodied carbon assessment methods, databases, tools, and case studies; the Buildings article is a systematic review of the research literature on embodied impacts. Neither carries out its own measurement of variation across databases; the key passages are the Buildings review's statement that database and boundary discrepancies produce significant variations even for identical buildings, and NIST's call for standardization across tools and databases. Related briefs: construction-3d-printed-concrete-code-void (regulatory gap for new construction technology), energy-building-performance-prediction-gap (parallel measurement-vs-reality gap for operational energy). The disciplinary-silo tag applies because LCA practitioners, structural engineers, architects, and policymakers each use different databases, system boundaries, and assumptions without cross-disciplinary standardization.
Reconciliation 2026-10-01: Fact repair after reading the whole write-up against both sources it names. The NIST report was read in full text (90 pages) from https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=958514, the PDF linked from the NIST publication page on the Source line; both fetched 2026-10-01. The Buildings article was read in full text (31 pages) from the Internet Archive copy of the publisher's PDF (web.archive.org/web/20250520010246/mdpi-res.com/d_attachment/buildings/buildings-15-01661/article_deploy/buildings-15-01661.pdf), fetched 2026-10-01, because mdpi.com refuses automated access; its citation record was read from https://api.crossref.org/works/10.3390/buildings15101661. (1) The central claim was that an analysis of 464 datasets from 20 databases found embodied carbon varying by up to 11.5x for steel buildings and 6.5x for cross-laminated timber. None of these four numbers appears in either source, and neither source describes such an analysis. The claim was removed from the Problem Statement and from the first paragraph of these notes. In its place the text now gives what the sources do say: the Buildings review states that discrepancies in database structure and boundary assumptions can result in significant variations in carbon outcomes, even for identical buildings, and that carbon estimates can fluctuate by up to 40% depending on boundaries; NIST reports that input-output analysis gave estimates two times higher than process analysis for the building sector in Finland. (2) Both titles carried the removed number and were changed. Long title: old "Embodied Carbon Measurements Vary 11.5x Across LCA Databases", new "Embodied Carbon Measurements Vary Across LCA Databases". Display title: old "Eleven Answers for One Building's Carbon", new "Different Answers for One Building's Carbon". (3) The quotation that EPDs are "not designed to be comparable between products" is in neither source. It was replaced with NIST's own words, that EPDs "cannot be used to compare the environmental performance of two different materials (e.g., concrete and wood)," and with NIST's condition for comparing them. The statement that regulators and designers routinely use EPDs for comparison anyway is not in the sources and was removed. (4) "Approximately 40% of global energy-related CO2 emissions" is not in the sources; NIST gives 37% of annual global carbon emissions in 2022 (28% building operations, 9% construction materials), and the sentence now says that. (5) "50–80% of lifecycle emissions in new high-performance buildings" is not in the sources; NIST says building materials may become the dominant source of carbon emissions in new construction, and the sentence now says that. (6) "Cities including Vancouver, New York, and London have adopted or proposed embodied carbon limits": the sources support only Vancouver (its Zero Emissions Building Plan brings embodied carbon into permitting) and say nothing of this kind about New York or London. The sentence now names Vancouver and the California law NIST describes, which requires the state's Air Resources Board to develop a framework to measure and achieve a 20% reduction in the carbon intensity of new building construction by 2030 and a 40% reduction no later than 2035. "Vary by an order of magnitude" was removed. (7) The text tied the 39% figure to "missing 9–100%+" of embodied carbon. In the Buildings review these are two separate findings: end-of-life phases are included in only 39% of articles, and recurring use-stage impacts can range from 9% to over 100% of initial embodied carbon. The sentences now keep them apart, and the second is given in the review's terms (impacts during the use phase, depending on maintenance frequency and material choices). (8) "Leaving the Global South without representative data" was restated in the review's terms: a gap in context-specific knowledge in developing and climate-vulnerable regions such as Sub-Saharan Africa, the Caribbean, and parts of Southeast Asia; the five countries are now named. (9) Three claims found in neither source were removed: that no consistent data exists for fenestration, adhesives, and MEP systems; that site-specific datasets change GWP calculations by up to 95%; and that product-specific EPDs show about 15% lower GWP in Norwegian case studies. The paragraph now gives the database limitations NIST lists (Athena covers only structural and envelope materials; Ecoinvent and GaBi are not U.S. based) and the Buildings review's statement on how EPDs vary. (10) "Regional calibration factors" is not a proposal in either source; the sentence now gives the Buildings review's call for regionally representative databases and national benchmarking platforms, with its remark that many building designers must rely on generic or international references that fail to reflect local conditions. The mandatory-disclosure sentence is the write-up's own proposal and was softened from "would" to "could". (11) The Source line gave the second source's title as "Embodied carbon impacts and lifecycle gaps" with no authors. The record at the cited address gives "Embodied Impacts in Buildings: A Systematic Review of Life Cycle Gaps and Sectoral Integration Strategies" by Bacheva and Raposo Grau; the title and authors were copied from the record, and the link is unchanged. (12) The entry point listed Ecoinvent and GaBi without saying they are commercial; NIST lists them as commercial data, so the text now says that and adds the public and free options NIST names (USLCI, the Athena Impact Estimator, EC3); NIST does not describe licence terms, and it names ICE only in a figure caption. Confirmed in the sources and left as written: the NIST citation (Zhang and four co-authors, Special Publication 1324, September 2024); 60% of publications from five countries; 39% for end-of-life phases; 9% to over 100% for recurring impacts; and the first sentence of What Would Unlock Progress, which matches both sources' calls for consistent system boundaries that include end-of-life stages. Not found in either source and left as the write-up's own framing: the reasoning for the disciplinary-silo tag in the first paragraph of these notes. A second reader went over these repairs against both sources on the same day (2026-10-01): all twelve stand, and five sentences were reworded to stay closer to the sources (the California law, the use-phase range, the regional-data proposal, the description of Ecoinvent and GaBi, and the first paragraph of these notes, which now says neither review carries out its own measurement of variation across databases, since the Buildings review's table of earlier reviews does relay figures from them). Genome tags untouched.