Department of Mathematical Sciences, Faculty of Science and Technology, Universiti Kebangsaan Malaysia, 43600, Bangi, Selangor, Malaysia.
Center for Geospatial Sciences, University of California, Riverside, CA, USA.
Environ Monit Assess. 2020 Oct 21;192(11):719. doi: 10.1007/s10661-020-08666-8.
An environmental problem which is of concern across the globe nowadays is air pollution. The extent of air pollution is often studied based on data on the observed level of air pollution. Although the analysis of air pollution data that is available in the literature is numerous, studies on the dynamics of air pollution with the allowance for spatial interaction effects through the use of the Markov chain model are very limited. Accordingly, this study aims to explore the potential impact of spatial dependence over time and space on the distribution of air pollution based on the spatial Markov chain (SMC) model using the longitudinal air pollution index (API) data. This SMC model is pertinent to be applied since the daily data of API from 2012 to 2014 that have been gathered from 37 different air quality stations in Peninsular Malaysia is found to exhibit the property of spatial autocorrelation. Based on the spatial transition probability matrices found from the SMC model, specific characteristics of air pollution are studied in the regional context. These characteristics are the long-run proportion and the mean first passage time for each state of air pollution. It is found that the probability for a particular station's state to remain good is 0.814 if its neighbors are in a good state of air pollution and 0.7082 if its neighbors are in a moderate state. For a particular station having neighbors in a good state of air pollution, the proportion of time for it to continue being in a good state is 0.6. This proportion reduces to 0.4, 0.01, and 0 for the cell of moderate, unhealthy, and very unhealthy states, respectively. In addition, there exists a significant spatial dependence of API, indicating that air pollution for a particular station is dependent on the states of the neighboring stations.
如今，全球关注的一个环境问题是空气污染。空气污染的程度通常是基于空气污染观测水平的数据进行研究的。尽管文献中存在大量关于空气污染数据的分析，但通过使用马尔可夫链模型考虑空间相互作用效应来研究空气污染动态的研究却非常有限。因此，本研究旨在利用空间马尔可夫链（SMC）模型，基于从马来西亚半岛 37 个不同空气质量站收集的 2012 年至 2014 年的纵向空气污染指数（API）数据，探索时空空间相关性对空气污染分布的潜在影响。由于从马来西亚半岛 37 个不同空气质量站收集的 2012 年至 2014 年的 API 日数据表现出空间自相关的性质，因此该 SMC 模型具有应用相关性。根据 SMC 模型中找到的空间转移概率矩阵，研究了空气污染在区域背景下的特定特征。这些特征是每种空气污染状态的长期比例和平均首次通过时间。结果发现，如果特定站点的邻居处于良好的空气污染状态，则该站点的状态保持良好的概率为 0.814，如果邻居处于中等空气污染状态，则概率为 0.7082。对于处于良好空气污染状态的邻居的特定站点，其继续处于良好状态的时间比例为 0.6。对于处于中等、不健康和非常不健康状态的单元，该比例分别降低至 0.4、0.01 和 0。此外，API 存在显著的空间依赖性，表明特定站点的空气污染取决于相邻站点的状态。
