以系統性機率模式鑑定量化與整合生命週期評估之不確定性

Abstract

傳統生命週期評估通常未進行不確定性量化分析，然而，缺少不確定性資訊，將無法了解評估結果的可靠度。不確定性資訊亦可提供決策者了解生命週期評估之限制，及作為決定是否增加數據收集或研究，以減低其不確定性；因此，發展生命週期評估之不確定性分析有其必要性。本研究之重點主要在發展系統性機率不確定性分析，研究目的區分為三層面，第一是鑑定生命週期評估不確定性之型式及來源，了解不確定性資訊之重要性；第二是發展機率不確定性分析，量化衝擊評估模式之參數不確定性，以及結合專家判斷資訊及適當新收集數據減低其不確定性；第三是進行不同型式不確定性之整合分析，比較結構不確定性(由不同決策及模式假設選擇產生)與參數不確定性之相對重要性。 本研究之主要貢獻為建立系統性機率不確定性分析方法，以解決傳統生命週期評估缺少不確定性資訊之缺點。研究方法以機率分析(蒙地卡羅模擬)為基礎，分別結合敏感度分析、貝氏推論及整合分析架構等理論，有系統地鑑定、量化、減低及整合生命週期評估之不同型式不確定性。本研究以一般廢棄物管理之生命週期評估為研究個案，選擇溫暖化潛勢作為衝擊類別之代表指標，評估不確定性對替代廢棄物處理之影響。 有關生命週期評估之不確定性分類，以定性分析區分為參數不確定性與變異性、模式不確定性、及情境(決策選擇)不確定性等三類。在量化分析方面，以機率分析將傳統生命週期評估轉換成機率式模式，以量化不確定性。結果顯示：機率模式較傳統點估計方法，提供決策者更多不確定性資訊，如均值、標準偏差、完整機率分布等特徵，可能與傳統點估計方法形成不同之決策；此外，結合蒙地卡羅模擬與敏感度分析，以級相關係數鑑定不確定性貢獻高之重要參數。結果顯示：決策者可依兩替代方案之評估結果機率分布之重疊大小，判斷不確定性對生命週期評估結果之影響程度。 其次，在不確定性減低方面，本研究以貝氏蒙地卡羅模擬更新重要參數及評估結果之不確定性，事後機率分布由各參數之事前機率分布與新收集數據之機率分布權重更新。結果顯示：4個重要參數之事前機率分布(IPCC準則專家判斷)結合新收集之統計及焚化場址特定數據貝氏更新後，其事後機率分布之變異係數(CV值)較事前機率分布呈現下降趨勢，評估結果之總不確定性明顯降低。 在不同型式不確定性整合方面，本研究以結合樹狀架構與蒙地卡羅模擬，將決策與模式選擇等結構不確定性整合至機率不確定性分析中，以衡量整合不同型式不確定性之合併效應及相對重要性。結果顯示：結構不確定性之貢獻確實可能影響評估結果，甚至造成評估結果逆轉，因此，此法可鑑定不同型式不確定性對生命週期評估結果之影響，避免不完整之不確定性資訊造成錯誤決策。 最後，本研究在考量不確定性下，進行廢棄物管理決策之情境分析。結果顯示：整合分析可提供替代方案評估結果之完整不確定性資訊，提昇生命週期評估之應用。另外，針對系統性機率不確定性分析方法之應用原則，提供不確定性資訊之考量時機、型式與方法。

The traditional life cycle assessment (LCA) does not perform quantitative uncertainty analysis. Without characterizing the associated uncertainty, however, the reliability of assessment results cannot be ascertained. The uncertainty analysis also provides useful information to assess the reliability of LCA-based decisions and to determine the need of adding data collection or research toward reducing uncertainty. This study focuses on developing the systematic approach of probabilistic uncertainty analysis of the LCA. The purpose of the study is threefold: first, to identify types and sources of uncertainty in LCA in order to understand the importance of uncertainty issues; second, to develop probabilistic uncertainty analysis that quantifies uncertainty in impact assessment model of LCA and reduce the uncertainty of LCA results with statistic and sites-specific information; lastly, to perform integrated analysis that quantifies combined results of different uncertainties (due to model and decision choices) and identify the relative importance of comparing parameter uncertainty. In this study, the main contribution was to establish the probabilistic uncertainty analysis method that was capable of improving the drawback of lack uncertainty information in the traditional LCA. The method was based on the probability analysis to identify, quantify, reduce and integrate the uncertainty in LCA that was in combination with the method of sensitivity analysis, Bayesian inference, and integrated framework, respectively. A case study of applying the method to the comparison of alternative waste treatment options in terms of global warming potential due to greenhouse gas emissions was presented. In the case study, the classification of uncertainty was qualitatively divided into three types including parameter uncertainty, model uncertainty and scenario uncertainty. First of all, in the quantities analysis, the traditional LCA was converted to probabilistic model by incorporating the probabilistic analysis to quantify uncertainty. The results indicated that the incorporation of quantitative uncertainty analysis into LCA revealed more information, such as mean value, standard deviation, and complete probability distribution than the deterministic LCA method. The resulting decision may thus be different. In addition, the sensitivity analysis in combination with the Monte Carlo simulation, calculations of the rank correlation coefficients facilitated the identification of important parameters that had major influences to LCA results. The results indicate that the overlaps of probability density functions (pdf) were used to judge the influence on LCA results between alternatives. Second, in respective of uncertainty reduction, the Bayesian method in combination with the Monte Carlo technique was used to quantify and update the uncertainty in LCA results. In the case study, the prior distributions of the parameters used for estimating emission inventory and environmental impact in LCA were based on the expert judgment from the Intergovernmental Panel on Climate Change (IPCC) guideline and then updated with using the likelihood distributions resulting from both national statistic and site-specific data. The posterior uncertainty distribution of the LCA results was generated using Monte Carlo simulations with posterior parameter probability distributions. The results indicated that by using national statistic data and site-specific information to update the prior uncertainty distribution, the resultant uncertainty (Coefficient of variation) associated with the LCA results was significantly reduced. Third, in respective of integration of uncertainties, the integrated framework in combination with the Monte Carlo technique was used to identify the importance of structural uncertainties due to model and decision choices, and then to evaluate the combined effect and relative importance of different types of uncertainty. The results indicated that the resultant uncertainty associated with structural uncertainties of the LCA results might be different or reversed. Therefore, the integrated analysis could understand the importance of structural uncertainties to avoid incorrect decision-making with incomplete uncertainty information. Finally, the scenario analysis of alternative waste management decision was performed under uncertainty. The results indicated that the integrated analysis revealed complete uncertainty information to enhance the application of LCA-based decision. In addition, the guideline of this method could be used to determine the timing, types and method to use the uncertainty information.