基于实测数据与遥感影像的鄱阳湖水体光学分类

鄱阳湖是我国最大的淡水湖泊,受季风气候影响其水体空间动态变化大,且广阔的水域内部差异也较大,因此湖泊水体光学分类对反演湖泊水质参数及监测湖泊水质有着重要意义。以鄱阳湖为研究区,根据实测的反射光谱数据形态特征将鄱阳湖的水体分为4类:特别浑浊、中等浑浊、轻度浑浊和清水区,并分别对每一类结果进行分析。考虑到实测光谱数据局限于湖区某些离散点的情况,不足以观测整个鄱阳湖区域内所有不同水体类型的空间分布和动态变化,从而将该方法利用于遥感影像以便观测整个湖区水体类型。在Landsat OLI遥感影像上任意选取采样点,根据其波谱形态建立基于斜率的分类算法,并应用决策树模型把鄱阳湖水体分为5类:特别浑浊、中等浑浊、轻度浑浊、清水区和特别清澈,影像的分类结果图与实地考察的情况相一致。把模型应用于其他时期的遥感影像进行鄱阳湖水体分类,对比影像的分类结果图表明:2002、2005和2009年鄱阳湖区分别出现3种、4种和4种不同的水体类型,且水体浑浊范围呈现出动态变化。研究表明水体光学类型分类可以更好的监测湖泊水质的时空变异性。     

基于BERT-BILSTM-CRF模型的电力行业事故文本智能分析*

为解决电力行业事故报告文本较长、语义复杂,难以进行有效文本识别问题,提出1种以BERT作为底层的预训练模型,并设计1种双重注意力机制编码器,结合BILSTM-CRF深度挖掘事故文本语义特征,从而实现文本智能分析。首先构建电力词典,通过对BERT预训练,进行BIO标注,然后引入BILSTM-CRF模型实现对文本标签智能分类,最后将该模型与现行其他4种深度学习模型进行对比。研究结果表明:该模型智能识别精确率、召回率及F1值(查准率)均达到约97%,较其他4种模型中效果最好的模型分别提高0.02,0.03,0.02。研究结果可为电力行业事故报告文本分析提供1种新思路。     

Initial Screening of Contaminated Land: A Comparison of US and Swedish Methods

Preliminary surveys are used to prioritize between contaminated sites to select those to be investigated more thoroughly. The data-gathering steps are almost identical between countries; however, the assessment procedures differ significantly. In this study, we have investigated 21 contaminated sites assessed as belonging to the high-risk or the very high-risk class using the Swedish Methods for Inventories of Contaminated Sites (MICS). We then applied the US Preliminary Assessment (PA) method to the same sites and compared the results and conclusions from the two screening procedures. In both cases, all sites were recommended for further investigation and the two approaches seem to corroborate one another; however, the PA assessment scores and the preliminary MICS classifications did not correlate. The results obtained with the PA method were easier to explain than the final MICS classification. The PA method also seems more transparent and easier to standardize, although objections could be made regarding the weighting scheme, because the outcome in this study was entirely dependent on the surface exposure pathway. However, to examine this in greater detail, it is necessary to include sites with less contamination: The importance of preliminary surveys in the overall risk management process gives a strong motivation for such an evaluation. Generally, the lack of research and scientific support for the various assessment procedures in use suggests that there is a need for method development, standardization, and validation.     

基于无人机影像的城市植被精细分类

针对传统植被资源调查方法工作量大、成本高、效率低的问题,利用高分辨率无人机遥感影像,联合地物光谱-纹理-空间信息,构建了一种适用于描述城市不同植被种类的多维特征空间,在此基础上对三种应用广泛的分类算法(基于像素的、面向对象的支持向量机及深度学习Mobile-Unet语义分割模型)开展了对比分析研究.结果表明:本文提出的联合地物光谱-纹理-空间信息的特征空间构建方法能够有效地描述城市不同类型植被的特征差异,提升影像分割、植被分类的精度;在分类精度上,基于像素和面向对象的支持向量机分类结果的总体精度均超过90%,深度学习方法的总体分类精度为84%;在算法效率上,传统机器学习方法也优于深度学习方法.因此,得出结论针对城市小区域、小样本的植被精细分类,传统机器学习分类方法比深度学习方法效果更好.     

历史抗生素胁迫改变磺胺甲唑和甲氧苄啶对活性污泥的影响:ARGs及其潜在宿主

抗生素共暴露对污水处理厂抗生素抗性基因（ARGs）及微生物群落聚集过程有着重要影响.然而,历史抗生素接触胁迫差异（遗留效应）是否能决定微生物和ARGs对抗生素复合污染的响应尚不清楚.通过选择高浓度（30 mg ·L^-1）磺胺甲唑（SMX）和甲氧苄啶（TMP）作为历史接触胁迫条件,探究SMX和TMP复合污染对ARGs、细菌群落及其交互作用的短期影响.基于高通量荧光定量PCR,共检测出13种ARGs,它们的绝对丰度介于2.21~5.42 copies ·μL^-1（对数值,以DNA计,下同）,其中sul2、ermB、mefA和tetM-01是样品中最主要的亚型,绝对丰度在2.95~5.40 copies ·μL^-1之间;SMX和TMP复合污染会引起ARGs和可移动基因元件（MGEs）的富集,但其对各亚型的影响不同,且SMX的历史遗留效应高于TMP;不同接触史的复合污染下,ARGs间的共现与互斥模式均存在;MGEs （尤其是intI-1）与磺胺类（sul1、sul2）、四环素类［tet （32）］和大环内酯-林可酰胺-链霉菌素类抗性基因（ermB）均呈显著正相关性;基于微生物全尺度分类发现各类群微生物群落结构对复合污染的响应不同,且条件丰富菌属（CAT）得到了显著富集;陶厄氏菌属（Thauera）、假黄单胞菌属（Pseudoxanthomonas）和副球菌属（Paracoccus）是优势的抗性菌属;网络分析共发现31个ARGs的潜在宿主,以条件稀有菌属（CRT）为主,尤其是Candidatus_Alysiosphaera和纺锤状菌属（Fusibacter）与多种ARGs呈现正相关,是多种ARGs的潜在公共细菌宿主;部分稀有菌属［RT,变异杆菌属（Variibacter）、气单胞菌属（Aeromonas）和管道杆菌属（Cloacibacterium）等］是转座子IS 613的潜在宿主,在ARGs的增殖传播中发挥重要作用.综上,研究揭示了污水处理系统中抗生素历史胁迫对ARGs赋存特征和及其宿主细胞的遗留效应,可为削减抗生素复合污染引起的ARGs污染提供新思路与理论依据.     

基于信息量的ALI影像最佳波段选择及融合研究

分析了EO-1 ALI多光谱遥感影像各个波段的信息特征,根据信息量大小和波段间的相关性,联合偏度—峰度指数JSKF,提出针对ALI图像新的最佳波段组合方法。在最佳波段组合下对ALI多光谱数据与全色数据分别进行IHS、HPF、PCA和Wavelet变换融合,从信息量角度出发,按照定量指标对融合前后的遥感影像进行评价,并对融合前后影像进行分类应用。     

Use of watershed characteristics to select control streams for estimating effects of metal mining wastes on extensively disturbed streams

Impacts of sediments and heavy metals on the biota of streams in the copper-mining district of southwestern Montana were examined by comparing aquatic communities of impacted streams with those of control streams. Control streams were chosen through the use of a technique that identifies similar streams based on similarities in their watershed characteristics. Significant differences between impacted and control sites existed for surface substrate, riparian vegetation, and the number of macroinvertebrate taxa. These results revealed that: (a) chemical and physical habitats at the impacted sites were disrupted, (b) the presence of trout was an inadequate measure of ecological integrity for these sites, and (c) watershed classification based on a combination of mapped terrestrial characteristics provided a reasonable method to select control sites where potential control sites upstream and downstream were unsuitable.     

A critical review of the water quality classification system in Turkey: A case study on Meric Basin

An 11-year period of water quality data, collected by the Directorate of Sate Water Works of Turkey are thoroughly analyzed for the purpose of implementing water quality classes to water resources in the Meric Basin, located on the European land mass of Turkey. Water quality parameters are divided into four groups as physical, organic, inorganic, and bacteriological. The quality class of each group is evaluated by taking into account the poorest quality of any parameter in the group, after which a quality rank is assigned to the sampling station and the waterbody in question. This method of water quality classification imposed by the Turkish Water Quality Act, is then criticized with respect to a statistical approach.     

Pipeline leakage detection and isolation: An integrated approach of statistical and wavelet feature extraction with multi-layer perceptron neural network (MLPNN)

Leakage diagnosis of hydrocarbon pipelines can prevent environmental and financial losses. This work proposes a novel method that not only detects the occurrence of a leakage fault, but also suggests its location and severity. The OLGA software is employed to provide the pipeline inlet pressure and outlet flow rates as the training data for the Fault Detection and Isolation (FDI) system. The FDI system is comprised of a Multi-Layer Perceptron Neural Network (MLPNN) classifier with various feature extraction methods including the statistical techniques, wavelet transform, and a fusion of both methods. Once different leakage scenarios are considered and the preprocessing methods are done, the proposed FDI system is applied to a 20-km pipeline in southern Iran (Goldkari-Binak pipeline) and a promising severity and location detectability (a correct classification rate of 92%) and a low False Alarm Rate (FAR) were achieved.