Stream mesocosm experiment on benthic macroinvertebrate and algal communities

The dataset consists of results from two stream mesocosm experiments that were conducted in the summer-autumn of 1996 and 1997 to distinguish the influence of fine sediment loads and nutrient concentrations on benthic macro-invertebrate and algal communities. 11 biological variables were extracted from the results of this experiment and were standardized for the purpose of training neural networks that could be used to diagnose nutrient and fine sediment impacts in field surveys. The 11 variables were selected according to how well they correlated with the experimental treatment levels (high and low values of both nutrients and fine sediments). The 11 variables were: chlorophyll a (mg/m2), macro-invertebrate familial richness, total abundance, and the abundance of Oligochaeta, Leptoperla varia (Gripopterygidae), Nousia spp. (Leptophlebiidae), Austrophlebioides spp. (Leptophlebiidae), Orthocladiinae, Tanypodinae, Tipulidae and larval Scirtidae. These taxa were abundant within and among the stream mesocosm communities and are common in a wide range of Tasmanian rivers. Values for each of 11 biological response variables were standardized by dividing by their average value observed in the experimental controls mesocosm samples from that year. See Magierowski RH, Read SM, Carter SJB, Warfe DM, Cook LS, Lefroy EC, et al. (2015) Inferring Landscape-Scale Land-Use Impacts on Rivers Using Data from Mesocosm Experiments and Artificial Neural Networks. PLoS ONE 10(3): e0120901. https://doi.org/10.1371/journal.pone.0120901 https://doi.org/10.1371/journal.pone.0120901. This data was collected for the purpose of training artificial neural networks that could diagnose nutrient and sediment impacts in Tasmanian rivers. Each of the 11 variables were standardized by their average value observed in the experimental control samples from that year and some experimental treatment effects (Light) were ignored to simplify the neural network training process. Therefore, these data should not be used to make conclusions about the impacts of fine sediments and nutrients in Tasmanian rivers.

本数据集包含两项溪流中型实验生态系统（stream mesocosm）实验的结果，两项实验于1996年与1997年的夏秋两季开展，旨在厘清细泥沙负荷与营养盐浓度对底栖大型无脊椎动物（benthic macro-invertebrate）及藻类群落的影响。研究从实验结果中提取了11项生物学变量并进行标准化处理，以训练可用于野外调查中营养盐与细泥沙影响评估的神经网络（neural network）。

11项变量的筛选依据为其与实验处理梯度（即营养盐与细泥沙的高低浓度组）的相关性强弱，具体变量包括：叶绿素a（chlorophyll a，单位mg/m²）、大型无脊椎动物科丰富度、总个体丰度，以及寡毛纲（Oligochaeta）、变细扁石蝇*Leptoperla varia*（Gripopterygidae）、Nousia属（Leptophlebiidae）、Austrophlebioides属（Leptophlebiidae）、直突摇蚊亚科（Orthocladiinae）、巨足摇蚊亚科（Tanypodinae）、大蚊科（Tipulidae）与圆泥甲科幼虫（larval Scirtidae）的个体丰度。上述分类群在溪流中型实验生态系统群落内部及群落间均具有较高丰度，且广泛分布于塔斯马尼亚州的各类河流中。

11项生物学响应变量的数值均通过除以对应年份实验对照中型生态系统样本的平均观测值完成标准化处理。详见文献：Magierowski RH、Read SM、Carter SJB、Warfe DM、Cook LS、Lefroy EC 等（2015）《利用中型实验生态系统实验与人工神经网络（artificial neural network）数据推断河流景观尺度土地利用影响》，*PLoS ONE*，10(3): e0120901，DOI: https://doi.org/10.1371/journal.pone.0120901。

本数据集的采集目的为训练可用于评估塔斯马尼亚州河流营养盐与泥沙影响的人工神经网络。11项变量均通过除以对应年份实验对照样本的平均观测值完成标准化，且为简化神经网络训练流程，部分实验处理效应（光照（Light））被排除。因此，本数据集不可用于推导塔斯马尼亚州河流中细泥沙与营养盐的影响结论。