Background pollution represents the lowest levels of ambient air pollution to which the population is chronically exposed, but few studies have focused on thoroughly characterizing this regime. This study uses clustering statistical techniques as a modelling approach to characterize this pollution regime while deriving reliable information to be used as estimates of exposure in epidemiological studies. The background levels of four key pollutants in five urban areas of Andalusia (Spain) were characterized over an 11-year period (2005–2015) using four widely-known clustering methods. For each pollutant data set, the first (lowest) cluster representative of the background regime was studied using finite mixture models, agglomerative hierarchical clustering, hidden Markov models (hmm) and k-means. Clustering method hmm outperforms the rest of the techniques used, providing important estimates of exposures related to background pollution as its mean, acuteness and time incidence values in the ambient air for all the air pollutants and sites studied.

背景污染是人口长期暴露在其中的最低水平的环境空气污染，但是很少有研究集中于对这种制度进行全面表征。这项研究使用聚类统计技术作为建模方法来表征这种污染状况，同时得出可靠的信息以用作流行病学研究中的暴露估计。在11年（2005-2015年）内，使用四种广为人知的聚类方法对安达卢西亚（西班牙）五个城市地区的四种主要污染物的背景水平进行了表征。对于每个污染物数据集，使用有限混合模型，聚集层次聚类，隐马尔可夫模型（hmm）和k均值。聚类方法hmm优于其他使用的技术，它提供了与背景污染有关的暴露的重要估计值，包括背景，所有研究的空气污染物和场所在环境空气中的平均值，急性程度和时间发生值。