Department of Biology and Ecology, Faculty of Sciences and Mathematics, University of Niš, Visegradska 33, 18000 Nis, Serbia.
Department of Mathematics, Faculty of Sciences and Mathematics, University of Nis, Visegradska 33, 18000 Nis, Serbia.
Sci Total Environ. 2018 Mar;616-617:472-479. doi: 10.1016/j.scitotenv.2017.10.262. Epub 2017 Nov 9.
The chironomid community in non-wadeable lotic systems was tested as a source of information in the construction of biological metrics which could be used into the bioassessment protocols of large rivers. In order to achieve this, we simultaneously patterned the chironomid community structure and environmental factors along the catchment of the Danube and Sava River. The Self organizing map (SOM) recognized and visualized three different structural types of chironomid community for different environmental properties, described by means of 7 significant abiotic factors (a multi-stressor gradient). Indicator species analysis revealed that the chironomid taxa most responsible for structural changes significantly varied in their abundance and frequency along the established environmental gradients. Out of 40 biological metrics based on the chironomid community, the multilayer perceptron (MLP), an supervised type of artificial neural network, derived 5 models in which the abundance of Paratrichocladius rufiventis, Orthocladiinae, Cricotopus spp., Cricotopus triannulatus agg. and Cricotopus/Orthocladius ratio achieved a significant relationship (the r Pearson's linear correlation coefficient>0.7) with the multi stressor environment. The sensitivity analysis "partial derivatives" (PaD) method showed that all 5 biological metrics within the multi-stressor gradient were mostly influenced by dissolved organic carbon (DOC). Despite short and monotonous environmental gradients and the absence of reference conditions, the chironomid community structure and biological metrics predictably changed along the multistress range, showing a great potential for the bioassessment of large rivers.
非可涉流水系统中的摇蚊群落可作为构建生物指标的信息源，用于大河生物评估方案。为实现这一目标，我们同时对多瑙河和萨瓦河流域的摇蚊群落结构和环境因素进行了模式识别。自组织映射（SOM）识别并可视化了三种不同的摇蚊群落结构类型，这些结构类型由 7 个显著的非生物因素（多胁迫梯度）描述。指示种分析表明，在既定的环境梯度上，对结构变化最具影响的摇蚊类群在丰度和频率上差异显著。在所提出的 40 种基于摇蚊群落的生物指标中，多层感知机（MLP）——一种监督类型的人工神经网络，得出了 5 个模型，其中 Paratrichocladius rufiventis、Orthocladiinae、Cricotopus spp.、Cricotopus triannulatus agg. 和 Cricotopus/Orthocladius 比值的丰度与多胁迫环境呈显著相关（r Pearson 线性相关系数>0.7）。敏感性分析“偏导数”（PaD）方法表明，多胁迫梯度内的所有 5 种生物指标主要受溶解有机碳（DOC）影响。尽管环境梯度较短且单调，且缺乏参考条件，但在多胁迫范围内，摇蚊群落结构和生物指标仍可预测性地发生变化，为大河生物评估提供了巨大潜力。
