基于MINE算法的大型底栖生物对溢油污染响应机制研究

The responses of marine macrobenthos to oil spill have been concerned in oil pollution research. Due to the interaction of multiple factors, it is difficult to obtain accurate results of the correlationship between macrobenthos and oil spill pollutants by the linear correlation analysis. MINE (Maximal information-based nonparametric exploration), a new machine-learning algorithm has been carried on R open source platform, being considered a better tool for uncovering novel relationships than linear correlation. MINE provides a nonparametric statistical method for the research of potential relationships under complicated influences. Here, we are hoping to manipulate a large of data in monitoring the Dalian “7.16” oil spill over the past 5 years and aim at providing a technical reference for the ecological assessment and sediment restoration and a new idea for the analysis of macrobenthos data sets under oil spill. We intend to elucidate the relationships between oil spill and macrobenthos by using MINE algorithm to further explore the mechanism of macrobenthos responses to the oil pollution from the perspective of nonparametric exploration.

溢油污染条件下出现的大型底栖生物种类及其出现的原因是溢油污染研究的重要内容。由于大型底栖生物受多重因素的交互作用，通过线性相关分析大型底栖生物与溢油污染因子往往难以得到准确结果。MINE算法作为近年来开发的探索连续型变量间关系的新算法，不局限于线性关系，提供了一种可在复杂因素影响的数据集中发现潜在重要关系的非参数统计方法。本研究以7.16溢油5年监测数据为基础，拟运用MINE算法研究大型底栖生物种类与溢油污染间的关系，从非参数研究的角度探索底栖生物群落对溢油污染的响应机制，以期为溢油污染下的沉积物生态评价与生态修复提供技术参考，也为大型底栖生物数据集的分析处理提供新思路。

2010年7月16日，辽宁省大连市发生一起原油泄漏事故。本研究在6年连续监测的基础上，利用Apriori算法和挖掘方法，寻找指示类群和污染物与底栖生物之间的非线性相关性。结果表明，石油泄漏对污染严重区域的大型底栖动物群落有明显的影响，截至2016年群落结构未得到恢复。Terebellidae和Lumbrineidae是污染区域的耐受类群。结果显示Maldanidae和Glyceridae的同时出现情况可以指示溢油污染。七个分类群被归类为可能的溢油指示物。我们的研究表明Apriori可以从共现频率的新角度来评估生态位重叠，MINE算法为探索污染物与类群之间的非线性相关性提供了一种快速的方法。最后，我们观察到了一些与其他研究不同的分类群的耐受性/敏感性，并重申了在不同的地点建立单独的指示生物清单的必要性。