人工神经网络在化学和环境科学中应用的若干问题

讨论了人工神经网络中最常用的多层前馈网络 ( BP网络 )及误差反向传播算法应用于化学和环境科学时要考虑的几个问题 :网络的输入与数据的归一化 ;隐含层数、隐含层节数和学习速率 ;训练集与监控集 ;网络误差 ;初始权重     

新乡市镉污染土壤细菌群落组成及其对镉固定效果

重金属污染土壤中具有丰富的微生物群落组成,为生物修复提供了微生物资源.本研究以新乡市某电池厂周边Cd污染土壤为研究对象,采用高通量测序和平板分离方法分析其细菌群落组成.传统培养方法表明新乡市重金属污染土壤细菌由厚壁菌门、放线菌门、变形菌门和拟杆菌门这4个门,芽孢杆菌属、节杆菌属和根瘤菌属等30个属组成;高通量测序表明,其由变形菌门、放线菌门和酸杆菌门等25门及400属组成.相较于培养方法,高通量测序群落组成更为丰富.基于高通量测序的分子生态网路分析表明其关键细菌分别为Arthrobacter、Marmoricola、Nocardioides、Ferruginibacter、Flavitalea、Nitrospira和Lysobacter等组成.对分离的159株可培养菌株进行Cd摇瓶吸附实验,结果表明Aneurinibacillus、Arthrobacter和Bacillus等11个属的30株菌具有较好地固定效果.效果验证实验表明,Ochrobactrum sp. 1-6、Bacillus sp.2-11和Pseudomonas sp.1-9等6株高效菌株能提高青菜(鸡毛菜)生物量、降低...     

A methodological framework for improving air quality monitoring network layout. Applications to environment management

This work aims to provide a methodology framework which allows to improve the performance and efficiency of an air quality monitoring network (AQMN). It requires to be constituted by a minimum and reliable number of measurement sites. Nevertheless, the AQMN efficiency should be assessed over time, as a consequence of the possible emergence of new emission sources of air pollutants, which could lead to variations on their spatial distribution within the target area. PM₁₀ particles data monitored by the Community of Madrid's (Spain) AQMN between 2008 and 2017 were used to develop a methodology to optimize the AQMN performance. The annual spatial distribution of average PM₁₀ levels over the studied period monitored by all current stations vs those more representative was provided by a geographic information system (GIS), and the percentage of similarity between both postulates was quantified using simple linear regression (> 95%). As one innovative tool of this study, the practical application of the proposed methodology was validated using PM₁₀ particles data measured by AQMN during 2007 and 2018, reaching a similitude degree higher than 95%. The influence of temporal variation on the proposed methodological framework was around 20%.The proposed methodology sets criteria for identifying non-redundant stations within AQMN, it is also able to appropriately assess the representativeness of fixed monitoring sites within an AQMN and it complements the guidelines set by European legislation on air pollutants monitoring at fixed stations, which could help to tackle efforts to improve the air quality management.     

区域空气自动监测数据审核方法与制度的探索

区域空气自动监测网络地理分布范围大、系统仪器设备众多、监测数据以海量计,如何对数据进行审核是一个值得关注的问题.以粤港珠江三角洲区域空气监测网络的实践为例,探讨了区域空气自动监测的数据审核方法与制度.     

基于L-M神经网络的道路交通噪声预测研究

神经网络具有很强的预测功能.根据石家庄公路交通噪声的实测数据,利用L-M优化算法的多层神经网络预测模型进行道路交通噪声的预测,经检验,计算值与实测值接近,预测精度令人满意.     

Direct protein-protein interaction network for insecticide resistance based on subcellular localization analysis in Drosophila melanogaster

Abstract

In present study, we constructed the direct protein-protein interaction network of insecticide resistance based on subcellular localization analysis. Totally 177 of 528 resistance proteins were identified and they were located in 11 subcellular localizations. We further analyzed topological properties of the network and the biological characteristics of resistance proteins, such as k-core, neighborhood connectivity, instability index and aliphatic index. They can be used to predict the key proteins and potential mechanisms from macro-perspective. The problem of resistance has not been solved fundamentally, because the development of new insecticides can’t keep pace with the development speed of resistance, and the lack of understanding of molecular mechanism of resistance. As the further analysis to reduce data noise, we constructed the direct protein-protein interaction network of insecticide resistance based on subcellular localization analysis. The interaction between proteins located at the same subcellular location belongs to direct interactions, thus eliminating indirect interaction. Totally 177 of 528 resistance proteins were identified and they were located in 11 subcellular localizations. We further analyzed topological properties of the network and the biological characteristics of resistance proteins, such as k-core, neighborhood connectivity, instability index and aliphatic index. They can be used to predict the hub proteins and potential mechanisms from macro-perspective. This is the first study to explore the insecticide resistance molecular mechanism of Drosophila melanogaster based on subcellular localization analysis. It can provide the bioinformatics foundation for further understanding the mechanisms of insecticide resistance. It also provides a reference for the study of molecular mechanism of insecticide resistance of other insects.     

Identifying key needs for the integration of social–ecological outcomes in arctic wildlife monitoring

For effective monitoring in social–ecological systems to meet needs for biodiversity, science, and humans, desired outcomes must be clearly defined and routes from direct to derived outcomes understood. The Arctic is undergoing rapid climatic, ecological, social, and economic changes and requires effective wildlife monitoring to meet diverse stakeholder needs. To identify stakeholder priorities concerning desired outcomes of arctic wildlife monitoring, we conducted in-depth interviews with 29 arctic scientists, policy and decision makers, and representatives of indigenous organizations and nongovernmental organizations. Using qualitative content analysis, we identified and defined desired outcomes and documented links between outcomes. Using network analysis, we investigated the structure of perceived links between desired outcomes. We identified 18 desired outcomes from monitoring and classified them as either driven by monitoring information, monitoring process, or a combination of both. Highly cited outcomes were make decisions, conserve, detect change, disseminate, and secure food. These reflect key foci of arctic monitoring. Infrequently cited outcomes (e.g., govern) were emerging themes. Three modules comprised our outcome network. The modularity highlighted the low strength of perceived links between outcomes that were primarily information driven or more derived (e.g., detect change, make decisions, conserve, or secure food) and outcomes that were primarily process driven or more derived (e.g., cooperate, learn, educate). The outcomes expand monitoring community and disseminate created connections between these modules. Key desired outcomes are widely applicable to social–ecological systems within and outside the Arctic, particularly those with wildlife subsistence economies. Attributes and motivations associated with outcomes can guide development of integrated monitoring goals for biodiversity conservation and human needs. Our results demonstrated the disconnect between information- and process-driven goals and how expansion of the monitoring community and improved integration of monitoring stakeholders will help connect information- and process-derived outcomes for effective ecosystem stewardship.     

基于链路预测的安全隐患管理研究*

为探寻安全隐患的发生规律，揭示安全隐患的演化特点，为管理人员提供建议和参考。以潞安集团司马煤业有限公司2009—2015年安全隐患记录为数据源，建立无向加权安全隐患共词网络，利用链路预测技术，寻找安全隐患的内在联系及发展趋势。结果表明:相比于共同邻居、资源分配等基于局部信息的相似性指标，资源分配指标能够更好地预测安全隐患各个关键词之间的变化联系；通过分析关键词共现关系，证实预测的准确性，预测出部分未来存在但当前未产生的节点关系。安全管理人员可基于此方法采取有针对性的预防措施，减少相关隐患的发生。     

Quantitative sensitivity and reliability analysis of sensor networks for well kick detection based on dynamic Bayesian networks and Markov chain

Kick is considered as an early warning sign to the blowout that is among the most undesired and feared accidents during drilling operations. Kick detection system is commonly used to timely identify the occurrence of a kick. The method commonly used for kick detection relies on the proper selection of monitoring indicators. A kick detection system should not only have very high accuracy but also maintain reliable over a long time. Different from the existing studies focusing on improving the detection accuracy, this paper presents a frame emphasizing on quantitatively analyzing and enhancing the reliability of the kick detection sensor networks. The dynamic Bayesian network (DBN) for the sensor networks is established that employs Markov chain to obtain the reliability degradation of measurement sensors over time. The proposed method is applied and evaluated by case studies to conduct reliability and sensitivity analysis for kick detection sensor networks. The reliability analysis results demonstrate that the proposed method can quantitatively analyze the reliability of a kick detection sensor networks consisting of various sensors over given time periods. The sensitivity analysis results indicate that the proposed method is effective in identifying the critical sensors that have the greatest effect on the reliability of one certain kick detection system. Based on the analysis results, optimized logical combination of sensors of a kick detection system can be achieved. An improved sensor network for the unreliable case was proposed and evaluated.     

Possible Use of Ash Residues for the Construction of Artificial Reefs at Sea

Utilization of ash residues, including coal ash, oil ash, and municipal solid waste combustion ash, for the construction of artificial reefs at sea has been investigated by many researchers throughout the world for nearly 20 years. Both laboratory and field studies have shown that an artificial reef made of stabillized ash-concrete (SAC) has had no adverse effect on the marine environment. Indeed, published studies have shown that fish counts increase around SAC reefs owing to an abundance of colonizing organisms and to protection provided by cavities within the reef structures. However, public and regulatory resistance to the use of SAC for artificial reef construction at sea is still very strong in certain countries due to concern for possible negative environmental impacts, primarily in the area of bioaccumulation of elements or compounds originating from the ash component of SAC.

In this paper, technological feasibility of using ash residues for artificial reef construction is presented, based on the available 20 years of scientific data, including engineering workability, physical integrity, chemical leaching potential and biological effects. More important, we also identify conceptual barriers for the acceptance of using SAC for the construction of ocean reefs and suggest approaches to overcome these barriers.