Romano Daniela, Bernetti Antonella, De Lauretis Riccardo
APAT, Italian Environmental Protection Agency, Via V. Brancati, 48 00144 Rome, Italy.
Environ Int. 2004 Oct;30(8):1099-107. doi: 10.1016/j.envint.2004.06.006.
Characterization of the uncertainty associated with air emission estimates is of critical importance especially in the compilation of air emission inventories. In this paper, two different theories are discussed and applied to evaluate air emissions uncertainty. In addition to numerical analysis, which is also recommended in the framework of the United Nation Convention on Climate Change guidelines with reference to Monte Carlo and Bootstrap simulation models, fuzzy analysis is also proposed. The methodologies are discussed and applied to an Italian example case study. Air concentration values are measured from two electric power plants: a coal plant, consisting of two boilers and a fuel oil plant, of four boilers; the pollutants considered are sulphur dioxide (SO(2)), nitrogen oxides (NO(X)), carbon monoxide (CO) and particulate matter (PM). Monte Carlo, Bootstrap and fuzzy methods have been applied to estimate uncertainty of these data. Regarding Monte Carlo, the most accurate results apply to Gaussian distributions; a good approximation is also observed for other distributions with almost regular features either positive asymmetrical or negative asymmetrical. Bootstrap, on the other hand, gives a good uncertainty estimation for irregular and asymmetrical distributions. The logic of fuzzy analysis, where data are represented as vague and indefinite in opposition to the traditional conception of neatness, certain classification and exactness of the data, follows a different description. In addition to randomness (stochastic variability) only, fuzzy theory deals with imprecision (vagueness) of data. Fuzzy variance of the data set was calculated; the results cannot be directly compared with empirical data but the overall performance of the theory is analysed. Fuzzy theory may appear more suitable for qualitative reasoning than for a quantitative estimation of uncertainty, but it suits well when little information and few measurements are available and when distributions of data are not properly known.
对与空气排放估算相关的不确定性进行表征至关重要，尤其是在空气排放清单的编制过程中。本文讨论并应用了两种不同的理论来评估空气排放的不确定性。除了数值分析（在《联合国气候变化框架公约》指南框架下针对蒙特卡罗和自助模拟模型也推荐使用）之外，还提出了模糊分析。这些方法在一个意大利的实例研究中进行了讨论和应用。从两个发电厂测量了空气浓度值：一个是由两台锅炉组成的燃煤电厂，另一个是由四台锅炉组成的燃油电厂；所考虑的污染物有二氧化硫（SO₂）、氮氧化物（NOₓ）、一氧化碳（CO）和颗粒物（PM）。已应用蒙特卡罗、自助和模糊方法来估算这些数据的不确定性。关于蒙特卡罗方法，最准确的结果适用于高斯分布；对于其他具有几乎规则特征的分布，无论是正不对称还是负不对称，也观察到了较好的近似值。另一方面，自助法对于不规则和不对称分布能给出良好的不确定性估计。模糊分析的逻辑与传统的数据简洁性、明确分类和精确性的概念相反，将数据表示为模糊和不确定的，遵循不同的描述方式。除了随机性（随机变异性）之外，模糊理论还处理数据的不精确性（模糊性）。计算了数据集的模糊方差；结果不能直接与经验数据进行比较，但对该理论的整体性能进行了分析。模糊理论可能看起来更适合定性推理而非不确定性的定量估计，但当可用信息少、测量次数少且数据分布不明确时，它很适用。
