G-37 (Buildings) / G-35 (Products) Uncertainty analysis for comparative assertion

Aspect G-37 (Buildings) / G-35 (Products) Uncertainty analysis for comparative assertion
Description
An uncertainty analysis measures the variability of the data values: for example, a low variance leads to a high precision. Because of the long supply chain of building products and buildings as a whole, uncertainties are aggregated over the life cycle. Following [Huijbregts 2001], different types of uncertainty can be found in LCA: parameter uncertainty, model uncertainty, uncertainty due to choices, spatial variability, temporal variability, and variability between the sources and the objects. In general, an uncertainty analysis should not be confused with a data quality assessment. Similarly, an uncertainty analysis for a stand-alone LCA should not be confused with an uncertainty analysis for a comparative assertion: in the first case, all the various types of uncertainty mentioned above are present, whereas in the second case, not all of them are present. In a comparative assertion supporting  decision-making, it is important only to ensure that the LCA results remain robust, even if the absolute LCA is uncertain. This is especially true when the differences between two alternatives are less than 20% (a percentage often used by LCA practitioners as the level below which it cannot be stated that A is better than B).In this context, when and how should an uncertainty analysis be performed for comparative assertion in the building sector?

related study objective

  stand-alone LCA comparative assertion

related study phase

goal and scope definition inventory analysis (LCI) impact assessment (LCIA) interpretation reporting

relevant for

new buildings existing buildings building products screening LCA simplified LCA complete LCA
Provisions An uncertainty analysis should be conducted for comparative assertions, and should be considered during the review process. It may be addressed by scenario analyses. For extended studies and in-depth discussions of uncertainty, the rules of the ILCD Handbook should be used. Also, the ongoing scientific discussion may be taken into consideration.
Rules from:
ILCD
Provisions: 6.5.4 LCI modelling provisions for Situations A, B, and C
I.a.vi) Comparative studies, scenarios, uncertainty calculation:

“Uncertainty calculation shall be performed [for comparative studies], unless it has already been used to derive the reasonably best and worst case scenarios.”

“It is recommended to also perform and report such assumption scenarios and uncertainty calculations for non-comparative LCI and LCA studies.”

ANNEX E: Addressing Uncertainties in LCA
16.3 Aggregating uncertainties over the life cycle

“Three main sources of uncertainty have been addressed: stochastic uncertainty, choice uncertainty, lack of knowledge of the studied system”

The ILCD Handbook also provides rules for not mixing the data quality aspects (falling under the term ‘accuracy’) with the other concept of ‘precision’ (or ‘uncertainty’). The graphical illustration in Figure 13 demonstrates this point.

Provisions: 6.5.4 LCI modeling provisions for Situations A, B, and C

I.a.vi) Comparative studies, scenarios, uncertainty calculation:

“Uncertainty calculation shall be performed [for comparative studies], unless it has already been used to derive the reasonably best and worst case scenarios. “

„It is recommended to also perform and report such assumption scenarios and uncertainty calculations for non-comparative LCI and LCA studies.“

ANNEX E: Addressing Uncertainties in LCA

16.3 Aggregating uncertainties over the life cycle

“Three main sources of uncertainty have been addressed: stochastic uncertainty, choice uncertainty, lack of knowledge of the studied system”

The ILCD Handbook also provides rules for not mixing the data quality aspects (falling under the term ‘accuracy’) with the other concept of ‘precision’ (or ‘uncertainty’). The graphical illustration in Figure 13 demonstrates this point.

Figure 15: Concept of precision

Guidance
1) General guidance for product and building LCA studies
The main difference between LCA studies with comparative assertions and stand-alone LCA studies lies in the fact that some parameters or assumptions are the same in both cases (also called ‘conventional parameters’) so that the number of uncertainty sources may be considerably reduced. In this case, the practitioner should check only whether the LCA results remain robust even if the stand-alone LCA is uncertain on several parameters (e.g. epistemic such as like the linear relations between flows, processes and impacts; the life expectancy of a building etc.).Another problem for the LCA practitioner lies in selecting an adequate method to assess the uncertainty. Various quantitative methods exist in the literature to assess uncertainties, such as Monte Carlo simulation, for example. Other methods, such as fuzzy logic and statistical Bayesian methods, have been developed and applied to LCA by researchers.However, most of these sophisticated numerical methods are currently not implemented in user-friendly LCA software for buildings, so the practitioner typically cannot easily use them.Based on the state of the art of building LCA tools, the best way to conduct an uncertainty analysis is to carry out scenario analyses that assess the most significant parameters of the LCA study. It is then possible to check whether one alternative still remains better than another in the various scenario analyses.The main issue is to identify the parameters, and to define their ranges of variation.

More information on scenarios for LCA in construction can be found in a previous report of the LoRe-LCA European project (FP7), available online.

2) Specific guidance for advanced building LCA studies

LCA may often be used as a decision-making tool to support a comparison of alternatives, e.g. during the building design stage. In this case, it is relevant to have guidelines on how to assess whether the choice of the best alternative from an LCA point of view is robust when using uncertain aspects of the building LCA model. Different steps may be conducted for a detailed analysis:

Identification and quantification of uncertainties for the key parameters

Uncertainty sources for building LCA studies can be found for the various contributors:

– building products: the LCA data of the building materials, the reference service life of the building products, the quantity take-off;

– energy consumption: the LCA data of the energy processes, value taken from thermal simulation software;

– water consumption: the LCA data of the water processes and treatment, the value determined with a calculation tool.

In addition, uncertainty can be found in the methodological choices made by the practitioner within the LCA software (data used, calculation rules etc.)

These sources of uncertainty may be assessed by defining distribution curves (e.g. log-normal, normal, triangle, Weibull) and then by deriving the relevant statistical parameters (e.g. the confidence interval at 95% or other statistical parameters if relevant).

This may provide relevant information for building LCA practitioners: for example, for a building product, the share of uncertainties linked to the reference service life (RSL), the LCA data or the quantity take-off (if these three sources of uncertainty are relevant for the study). A similar approach can be applied to the operational energy and water impact values.

Propagation of uncertainties in building LCA results

For a building case study, the relative share of impacts driven by building products, operational energy or water use, etc. can be identified and the corresponding uncertainties assessed.

This approach allows building LCA results to be presented with e.g. a mean value and a standard deviation (as e.g. 10 kg eq-CO2/m²/yr ±1.4 for the GWP indicator of a building).

This is the first step in identifying the sources of uncertainty in building LCA results. If relevant, these uncertainties can be reduced, e.g. by collecting more accurate/precise data.

Use of distribution curves for the key parameters in comparative assertions

The uncertainties for the key parameters can then be used in a comparative assertion for two building case studies (fulfilling the ISO 14044, ILCD and EeBGuide requirements, e.g. the two buildings shall have the same functional equivalent). This allows the practitioner to assess whether alternative A is better than B when taking uncertainties into account. If the standard deviations of alternatives A and B do not overlap, then the comparative LCA results can be considered robust. If they do overlap, it is not possible to state that A is better than B, given the corresponding uncertainties.


Comments are closed.