Quantifying uncertainty in carbon emission estimation: Metrics and methodologies

Carbon emissions are a significant driver of global climate change in today's world. A central concern in discussing carbon emissions is the level of uncertainty associated with them. This study aims to assess the feasibility of using the Mean Squared Error (MSE) as a point estimation measure and confidence interval (CI) as an interval estimation measure to quantify the uncertainty surrounding carbon emissions. To achieve this goal, the bootstrap and Markov Chain Monte Carlo (MCMC) sampling methods were used, utilizing both classical and Bayesian inference methods to uncover the true parameter value. In the context of Bayesian inference, a 2-chain MCMC method proved to be the optimal choice for generating posterior distributions and accurately estimating the true parameter of the distribution θ. The analysis also shows that, while a CI is valuable as an evaluative measure, it does not inherently provide a quantified form of uncertainty and should not be used for quantitative uncertainty assessment. Instead, the Relative Standard Error (RSE) emerges as a promising quantitative measure for capturing the uncertainty of θ, while the percentage uncertainty offers a qualitative perspective. These combined measures provide a comprehensive toolkit for computing and communicating the uncertainty in carbon emission and emission factor values. This reinforces a move towards incorporating transparent uncertainty metrics, ensuring that stakeholders have access to reliable and accurate carbon emission values.