Advancing Uncertainty Decomposition in Environmental Assessment: The LMDI Method in Life Cycle Analysis of Bioethanol

The study addresses uncertainty in life cycle assessment (LCA), a widely used method for evaluating the environmental impacts of products and processes. Although LCA provides comprehensive insights, it relies on numerous data inputs, assumptions, and methodological choices that introduce uncertainty. Traditional Monte Carlo simulation (MCS) quantifies overall variability but does not clearly identify which inputs contribute most to uncertainty.

To overcome this limitation, the research integrates Monte Carlo simulation with the additive Logarithmic Mean Divisia Index (LMDI) decomposition method. The objective is to systematically break down and attribute uncertainty in the LCA results of bioethanol production from broken rice in India.

Methodological Framework

The study follows ISO 14040/44 standards and conducts a cradle-to-gate LCA of bioethanol production from broken rice. The functional unit is one liter of bioethanol. The system includes rice preparation, milling, liquefaction, saccharification, fermentation, distillation, dehydration, and by-product handling (DDGS).

Foreground data were collected from Indian bioethanol distilleries, while background processes were sourced from the ecoinvent database. Environmental impacts were assessed using the ReCiPe 2016 method across seven midpoint categories:

• Global warming score (GWS)

• Fine particulate matter formation (FPMF)

• Terrestrial acidification potential (TAP)

• Freshwater eutrophication potential (FEP)

• Terrestrial ecotoxicity potential (TEP)

• Human carcinogenic potential (HCP)

• Fossil resource scarcity (FRS)

Uncertainty in input data was quantified using the pedigree matrix approach, which assigns data quality scores to derive lognormal distributions for Monte Carlo simulation.

Monte Carlo Simulation

A total of 1,000 Monte Carlo iterations were performed to propagate uncertainty through the model. This process generated distributions for each impact category, providing measures such as mean, standard deviation, and coefficient of variation (CoV).

All impact categories exhibited relatively low variability, with CoV values below 10%, indicating consistent and stable model performance. However, while MCS quantified overall uncertainty, it did not identify the relative contribution of each parameter to the total variability.

LMDI Decomposition Method

To address this gap, the additive LMDI method was applied. LMDI decomposes the difference between deterministic and simulated results into contributions from:

1. Life cycle inventory (LCI) inputs

2. Characterization factors (CFs)

The additive approach was selected because it avoids residuals and handles zero or near-zero values effectively. For each Monte Carlo iteration, the deviation from the deterministic baseline was decomposed into contributions from individual inventory flows and characterization factors.

This allowed the study to quantify how much each input parameter—such as electricity, steam, or polydimethylsiloxane—contributed to overall uncertainty.

Deterministic LCA Results

The deterministic analysis revealed the following key environmental impacts per liter of bioethanol:

• Global warming: 3.16 kg CO₂ eq

• Fossil resource scarcity: 0.73 kg oil eq

• Human carcinogenic toxicity and terrestrial ecotoxicity were also significant

Steam and electricity consumption were the dominant contributors across most categories. For example:

• Steam accounted for nearly 59% of global warming impacts.

• Electricity significantly influenced particulate formation, eutrophication, acidification, and fossil resource depletion.

• Human carcinogenic toxicity was largely driven by heavy metal emissions associated with energy generation.

Normalized results highlighted human carcinogenic toxicity as the most critical environmental concern in the system.

Contributions from Inventory and Characterization Factors

For most impact categories:

• LCI inputs contributed 70–80% of total uncertainty.

• Characterization factors contributed 20–30%.

Electricity emerged as the most consistent contributor to uncertainty across nearly all categories. It accounted for a substantial portion of both LCI-related and CF-related variability.

Steam also contributed significantly to uncertainty, particularly in global warming impacts.

Key Findings

1. Electricity is the dominant driver of both environmental impact and uncertainty across most categories.

2. Steam significantly contributes to global warming and related uncertainty.

3. Polydimethylsiloxane introduces high uncertainty despite having a minor environmental impact.

4. Most uncertainty originates from inventory data rather than characterization factors.

5. The additive LMDI method effectively decomposes uncertainty without computational complexity or residual errors.

Methodological Contributions

The study demonstrates that combining Monte Carlo simulation with additive LMDI provides:

• Transparent attribution of uncertainty

• Computational efficiency compared to global sensitivity analysis

• Clear identification of priority areas for data improvement

Unlike variance-based sensitivity methods, LMDI does not capture interaction effects but offers simplicity, interpretability, and practical applicability.

Conclusion

The research confirms that the additive LMDI decomposition method is an effective and accessible tool for uncertainty analysis in LCA. By integrating it with Monte Carlo simulation, the study isolates and quantifies contributions from individual inputs and characterization factors.

Electricity and polydimethylsiloxane are identified as major drivers of uncertainty, with electricity also dominating environmental impacts. The findings emphasize that minor inputs can create significant variability when data quality is weak.

Overall, the framework enhances transparency, improves interpretability, and supports more reliable sustainability decision-making in environmental assessment.