SM5002 Chap.4 Interpretation, Sensitivity and Uncertainty
Interpretation, Sensitivity and Uncertainty
Interpretation is where an LCA earns trust
The fourth phase of a life cycle assessment takes the indicator results from impact assessment and decides what they mean. ISO 14044 describes it as a systematic procedure to identify, qualify, check and evaluate the findings and then present conclusions. It is not a summary written at the end.
Interpretation feeds back into the earlier phases: if the checks reveal that the data are too weak or the boundary is drawn in the wrong place, the scope is revised and the model is run again. That iterative character is why the standard draws the four phases as a loop rather than a line.
The phase has three parts. First come the significant issues, the items that have a substantial influence on the overall result.
Second comes evaluation, made up of a completeness check, a sensitivity check and a consistency check.
Third come conclusions, limitations and recommendations written for the intended audience and application named in the goal.
Finding what drives the result
Significant issues can be inventory flows such as energy or waste, impact categories such as resource use or climate change, or life cycle stages and unit processes such as transport. Several techniques help locate them.
Contribution analysis expresses each stage or process group as a share of the total. Dominance analysis ranks contributions using statistical or qualitative tools, including ABC ranking. Influence analysis asks which issues a decision-maker could actually change. Anomaly assessment compares the results with experience and flags anything unexpected, which often turns out to be a data entry or unit error.
Gravity analysis, of which the Pareto curve is the usual form, sorts items by contribution and plots the cumulative share so that the few items responsible for most of the impact stand out.
Studies of common operations have used this to show that a small subset of the products used in theatre accounts for the bulk of the footprint, which tells procurement teams where to focus.
Three checks before any conclusion
The completeness check asks whether the information from every phase is sufficient to reach conclusions consistent with the goal and scope.
Missing information is not automatically fatal, but its necessity must be judged and stated.
The consistency check asks whether assumptions, methods and data have been applied evenly across the study: whether differences in data quality between compared systems fit the goal, whether regional and temporal differences have been treated the same way, and whether allocation rules, boundaries and impact assessment elements have been applied to all product systems alike.
The sensitivity check assesses how far results and conclusions are affected by uncertainty in data, allocation methods or the calculation of indicators, and it must report any sensitivity or uncertainty analysis that was carried out.
Where uncertainty comes from
The GHG Protocol product standard sorts uncertainty into three kinds.
Parameter uncertainty concerns whether a value used in the inventory or in characterisation truly represents the activity; direct emissions data, activity data, emission factors and impact assessment factors all carry it. Even global warming potentials for the main greenhouse gases are estimated to be uncertain by about thirty-five percent either way at a ninety percent confidence interval.
Scenario uncertainty comes from methodological choices such as system boundaries, allocation, use and end-of-life assumptions and the characterisation model. Model uncertainty reflects the limits of any model in representing the real world, and only part of it can be quantified through the other two routes.
Two families of analysis answer these.
Sensitivity or scenario analysis changes one or a few inputs by fixed amounts, for instance by a set percentage or between low, medium and high cases, and shows which inputs move the outputs most; it guides data collection and model improvement.
Uncertainty analysis describes inputs as distributions, such as normal or log-normal, propagates them through the model by random sampling or analytical approximation, and reports a range or probability distribution for the final result; it supports statements about robustness.
The two terms are often used loosely, so name which one you ran.
Parameters make sensitivity testing practical
In openLCA, parameters are named variables that replace hard-coded numbers. They avoid duplicating processes for every variant, document which assumptions matter, apply one assumption consistently across many processes and allow quick reruns.
A local parameter belongs to a single process; a global parameter is shared by every process that references it. The workshop applies this to reusable surgical scissors.
An audit found sterilisation energy ranged from 1.1 to 1.5 kWh per instrument set; running both cases moved the footprint per use by 0.0084 kg CO2e, about seven percent, which is noticeable but leaves the reusable pair far below the single-use pair at 0.62 kg CO2e.
A second test used a global packaging switch: packaging raised the single-use result by nineteen percent and the reusable result by only four percent, because the reusable pair spreads its packaging across many uses.
A switch parameter can also select between electricity providers, so the same product system can be run under different national grids.
The optional bonus exercise turns parameters into distributions and runs a Monte Carlo simulation of one thousand iterations, reading off the mean and the fifth and ninety-fifth percentiles as a ninety percent interval.
Your written assignment does not require a sensitivity analysis, but it does require interpretation that names the largest contributors and discusses uncertainty and limitations, and the group appraisal will expect you to judge whether a published study tested its key assumptions.
What this chapter covers
- 01
The three parts of interpretation under ISO 14044
- 02
Contribution, dominance, influence, anomaly and Pareto analysis
- 03
Completeness, sensitivity and consistency checks
- 04
Parameter, scenario and model uncertainty
- 05
Scenario analysis compared with probabilistic uncertainty analysis
- 06
Local and global parameters in openLCA, with an optional Monte Carlo run
Worked example · free
Testing the scissors result against packaging
- 2Compute the absolute and relative change for each product: single-use rises by 0.091 kg, which is 19 percent of 0.477; reusable rises by 0.005 kg, about 4 percent of 0.128.
- 2Explain the difference: single-use scissors consume 15 g of film every use, while the reusable pair's 30 g is shared across about 40 uses, near 0.75 g each.
- 2Test the ranking: single-use is 3.7 times higher without packaging and 4.3 times higher with it, so the conclusion is unchanged.
Key terms
- Contribution Analysis
- Expressing each life cycle stage or process group as a percentage of the total result to find the main drivers.
- Sensitivity Check
- The ISO 14044 step assessing how far results and conclusions shift with changes in data, methods or assumptions.
- Scenario Uncertainty
- Variation in results caused by methodological choices such as boundaries, allocation or end-of-life assumptions.
- Global Parameter
- An openLCA variable shared by every process that references it, so one change updates all of them.
- Monte Carlo Simulation
- Repeated random sampling from input distributions to produce a range of possible results.
Interpretation, Sensitivity and Uncertainty FAQ
How large does a change have to be before a result counts as sensitive?
There is no fixed threshold. The useful test is whether the conclusion changes: a seven percent shift that leaves the ranking of alternatives intact is worth reporting as a limitation, while a smaller shift that reverses the ranking is decisive.
Do I need to run a Monte Carlo analysis for the written assignment?
No. The workshop states that sensitivity analysis is not expected in the final individual piece, and the Monte Carlo exercise is a bonus. You are still expected to identify major contributors and to discuss uncertainty and limitations in words.
What is the difference between a completeness check and a consistency check?
Completeness asks whether enough information exists to support conclusions matching the goal. Consistency asks whether data quality, regional choices, allocation and boundaries have been treated the same way across every product system that is compared.
Why use parameters instead of copying a process and editing the number?
Copies multiply quickly and invite input mistakes. A parameter records the assumption once, documents it, applies it everywhere it is referenced and lets you rerun alternatives inside one project for direct side-by-side reading.
Assessment move
Take any published healthcare LCA from the reading list and list its significant issues, then mark which of its assumptions were tested by scenario analysis and which were not. For each untested one, write a sentence on whether a plausible change could reverse the headline conclusion.
