Athena Sustainable Materials Institute

The Many Dimensions of Uncertainty Analysis in LCA (PDF)

- By:

Courtesy of Courtesy of Athena Sustainable Materials Institute

Uncertainty comes into play when analysts, databases, or reports make statements or assertions about something in the real world based on Life Cycle Assessment methods and data. These statements or assertions can be either quantitative or qualitative. An example of a quantitative statement within LCA might be, “the total ‘embodied’ carbon dioxide emissions caused by generating a kilo-watt hour in the northeastern liS are x lbs per kWh.” An example of a qualitative assertion might be, “the life cycle particulate emissions of recycled newsprint are lower than those of newsprint made from virgin fibers.”
Why do we care about uncertainty in LCA? Because the assertions or statements we make about the world based on LCA may be wrong they are uncertain. Wc presumably use the results of LCA to help us, or others, decide future courses of action. Future decisions based (at least in part) on our LCA analysis will be better decisions ifthey take into account not only our results and conclusions, but also the uncertainty in these results and conclusions. This point was made rather forcefully by Wilson et al [1985j: “A decision made without taking uncertainty into account is barely worth calling a decision.”
Usefril information about uncertainty in LCA will tell us something about the likelihood or probability that our statements and conclusions are right or wrong. Ideally, it will also establish “confidence bounds” on our results “wiggle room” around our results within which the true values have an estimated likelihood of falling.
For example, consider our example LCA-based assertion that, “life cycle particulate emissions of recycled newsprint are lower than those of newsprint made from virgin fibers.” This statement may be right or wrong. How re/jable is it? How con/idem’ are we that it indeed holds true? The terms confidence and reliability go hand in hand. That is, when we assert that our conclusion is 90° o reliable, this is the same as asserting that we can have 90° o confidence in our results.
In classical statistics, “90°c confidence” means that ifwe were to repeat our analysis very many times, each time using new and equally probable, randomly selected values for our uncertain quantities, our conclusions would be correct 90° o of the time. Wc cannot use all the methods of classical statistics in LCA, primarily because our underlying LCA data are not based on random samples, and we are not strictly dealing with “random variables” which follow some known or unknown frequency distribution. However, we are dealing with uncertain quantities, about which we may develop or obtain subjective probability distributions, so we can use methods from the well-developed and active fields that fall under the broad heading of uncertainty analysis.
Subjective probability distributions are an alternative to simple point estimates. Rather than use in our analysis a single number to estimate some real-world quantity (say, tons of sheet steel required to make one ton of product), we instead use a range within which we expect the true value to lic, and optionally a description of the relative likelihoods (probabilities) that the true value lies within certain portions of this range.
IF we use or develop subjective probability distributions in this manner, we can then use the methods of uncertainty analysis in LCA and truly take uncertainty into account in our results, conclusions, and decisions. THEIV, when we make an LCA-based statement such as, “life cycle particulate emissions of recycled newsprint are lower than those of newsprint made from virgin fibers,” we will also be able to understand, communicate, and take into account the reliability of the statement. Ifthe reliability of the conclusions is not sufficient for our decision-making needs, uncertainty analysis will also help us identif which data uncertainties are most influential. It can thrther help us determine the levels of reduction in data uncertainty required to reach a specified level of results reliability or confidence, and or it can tell us what the results reliability consequences will be of pre-specified reductions in data uncertainty. In short, we will no longer be “flying blind” about uncertainty and reliability in LCA.

Customer comments

No comments were found for The Many Dimensions of Uncertainty Analysis in LCA (PDF). Be the first to comment!