基于机器学习的大气PM2.5中金属浓度预测模型的研究

基于2013—2018年哈尔滨市气象数据、大气污染物数据和细颗粒物（PM_2.5）中金属成分数据,采用机器学习方法探索大气PM_2.5中金属浓度预测模型,并选择最优模型进行污染物浓度预测。结果表明,多元线性回归（MLR）、人工神经网络（BP-ANN）、支持向量机（SVM）和随机森林（RF）4种模型中,RF对大气PM_2.5中5种金属［锑（Sb）、砷（As）、铅（Pb）、镉（Cd）、铊（Tl）］的浓度预测效果最佳,在训练集和测试集中表现均较稳定,其中相关系数（r）均＞0.7, 平均绝对误差（MAE）和均方根误差（RMSE）数值较小。RF在大气PM_2.5中金属浓度预测上具有较好的表现,可在缺乏监测和实验数据的情况下,实现对大气颗粒物中金属浓度的快速预测,为全面了解颗粒物中金属污染特征提供数据基础。     

郑州市PM2.5浓度时空分布特征及预测模型研究

利用统计学原理和GIS技术,对郑州市2013年8月17—12月31日期间PM2.5浓度时空分布特征进行分析,同时结合气象资料与前一日污染数据,建立人工神经网络反向传播算法模型(BP-ANN)和多元线性回归模型用于该市细颗粒物污染的短期预测。结果表明,郑州市PM2.5浓度日变化呈单峰模式,随逆温现象的发生和交通的密集于上午11:00达到峰值,午后逐步下降。在工作日、周末与国庆节的对比中,国庆节期间颗粒物污染浓度高出平日32.8%,表明人为活动的加剧影响PM2.5的排放;周末与工作日期间无显著差异。在空间分布上,金水区、管城回族区污染最为严重,工业燃煤、地铁施工等源排放是造成污染的主要原因;位于远郊的岗里水库,受秸秆焚烧和市区污染输送等影响,PM2.5浓度亦维持较高水平。最后,研究将所构建的BP-ANN预测模型和多元线性回归模型对比,结果发现两模型在建模阶段预测值与真实值的拟合一致性指标分别为0.944、0.918,均方根误差分别为59.788、70.611;验证阶段拟合一致性指标分别为0.854、0.794,平均绝对误差分别为25.298、32.775,表明BP-ANN模型在预测郑州市PM2.5污染过程中更具优势。     

基于复合LSTM模型的PM2.5浓度预测

在对淄博市19个空气质量监测站点监测数据进行分析后,提出了一种基于机器学习的复合模型——灰色关联度分析(GRA)-改进的完备总体经验模态分解(ICEEMD)-长短期记忆网络(LSTM)模型。通过分析淄博市2019年大气污染物和气象数据,选用LSTM模型预测PM_2.5浓度。由于传统单一模块机器学习模型具有训练时间较长和预测精度较低的问题,提出了复合LSTM模型。该模型由3部分组成：GRA,用于PM_2.5浓度影响因素变量筛选;ICEEMD,用于PM_2.5分解、分量筛选和原始大气污染物及气象数据处理;LSTM,用于PM_2.5浓度预测。预测结果表明：淄博市中部丘陵地带PM_2.5浓度高于南部山区和北部平原,东部高于西部;淄博市逐月PM_2.5浓度呈“U”形分布,1月最高,8月最低;淄博市PM_2.5浓度受PM₁₀和CO影响较大,受湿度和温度影响较小。对比单一LSTM模型和GRA-LSTM模型,GRA-ICEEMD-LSTM模型...     

基于小波分析优化PM 2.5浓度预测模型

采用多元线性回归方法(MLR)和BP神经网络方法(BPNN),按1 h、3 h、6 h、12 h、24 h、48 h预测时长对贵港市2015—2018年PM_2.5浓度建模并检验对比模型准确率。结果表明,基于MLR与BPNN都能对PM_2.5浓度作预测,预测效果随着预测时长的增加而下降,MLR、BPNN模型预测结果平均绝对误差(MAE)分别为4.01μg/m³~15.48μg/m³、3.89μg/m³~15.63μg/m³。采用小波分析方法对污染物数据优化并再次建模,结果表明,小波-多元线性回归(W-MLR)模型与小波-神经网络(W-BPNN)模型均得到优化,3 h~24 h预测时长优化效果尤为显著,W-MLR、W-BPNN模型预测结果分别使MAE降低1.6%~13.5%、0.8%~9.8%,且后者预测效果优于前者。     

基于随机森林算法的滁州市空气质量预报研究

利用2017—2019年滁州市6种空气污染物质量浓度和AQI资料、气象要素及ECMWF数值预报作为输入参数,构建基于随机森林算法的污染物质量浓度和AQI预报模型,其中AQI预报按季节划分为4个模型。结果表明：6种污染物中O3的预报效果最好,相关系数为0.84,PM2.5、PM10和NO2相关系数分别为0.76、0.72和0.72,SO2和CO预报效果略差;基于分季节模型AQI的24 h预报结果TS评分为0.77,空报率（FAR）和漏报率（PO）分别为15%和12%,相较于非季节模型预报效果更好;春季和秋季的TS评分分别为0.88和0.86,相较于冬季和夏季预报效果更好。     

长沙市区PM_(2.5)浓度的主要因素分析及预测

针对长沙市区PM2.5浓度的主要影响因素,采用统计学方法对长沙市区空气质量指数AQI包含的5个基本因素进行了相关因素分析,得出了PM2.5浓度与其余4个因素的关联度系数,并根据分析结果提出了几点改善PM2.5浓度的建议。通过分析长沙市区空气中的PM2.5浓度数据,采用ARIMA模型对PM2.5浓度进行了较为精准的预测。所提出的方法为长沙市区PM2.5的防治提供了一定的参考。     

上海市雾霾污染物对人群健康影响的统计研究

通过收集2013—2015年上海市每日雾霾污染物浓度数据、气象监测数据,并统计研究期间就诊于上海市某医院的门急诊量数据,分析PM_(2.5)浓度与门急诊量的暴露-反应关系,进而研究PM_(2.5)对人群健康的影响。采用泊松分布的半参数广义相加模型,以医院每日门急诊量为响应变量,通过样条平滑函数控制长期趋势、季节趋势、星期效应和气象因素等混杂因素,拟合PM_(2.5)浓度与每日门急诊量的暴露-反应关系模型。结果表明,PM_(2.5)浓度升高会引起医院呼吸内科、心血管内科和皮肤科门急诊量的增加。     

2017年厦门金砖会晤期间PM2.5的化学特征及来源

为了探讨厦门金砖会晤期间的排放控制措施以及天气形势对大气颗粒物污染特征的影响,于2017年8月10日至9月10日对厦门气态污染物、细颗粒物(PM2.5)中的水溶性离子以及有机碳(OC)、元素碳(EC)等主要化学成分开展了高时间分辨率的在线监测。根据空气质量管控措施和天气形势将研究期分为6个阶段。管控前、管控期Ⅰ(非台风)和管控期Ⅱ(非台风) PM2.5质量浓度分别为(33. 12±9. 48)、(30. 30±17. 00)、(16. 01±4. 71)μg/m^3。管控期Ⅰ(台风)和管控期Ⅱ(台风) PM2.5质量浓度分别为(12. 40±3. 73)、(12. 45±3. 28)μg/m^3。结果表明:管控期Ⅰ(非台风)阶段受静稳天气的影响,管控效果削弱,PM2.5质量浓度下降幅度小;台风对颗粒物质量浓度下降的影响比管控更显著。管控初期,PM2.5中二次无机离子的质量浓度下降明显;台风对碳质组分质量浓度的影响不如无机组分显著。PMF源解析结果表明,二次无机源是PM2.5主要来源,随着管控措施的实行,扬尘源的贡献从21%降低到6%,而机动车源的贡献降幅不明显。台风期间SO4^2-、NO3^-、SO2、NO2以及硫酸盐氧化比值(SOR)均明显低于非台风期间,氮氧化比值(NOR)反而升高。台风和非台风期间NOR的日变化特征一致,NOR与阳离子的相关性分析结果表明,台风或高风速海风期间NOR与Na^+呈现很强的正相关性,说明海盐粒子可促进NO2非均相反应生成NO3-。     

基于图像感兴趣区域融合特征的PM2.5浓度预测方法

通过图像预测PM_2.5浓度的准确性,在很大程度上取决于模型所选用的特征参数。为丰富特征参数的表达,设计了一种基于图像传统特征与深度特征充分融合的PM_2.5浓度预测方法。首先,根据不同PM_2.5浓度下的成像差异,选定图像感兴趣区域,解决图像尺寸过大导致的模型运算效率较低问题。然后,针对所选取的局部图像,利用传统图像处理方法手动设计并提取图像浅表视觉特征,同时利用卷积神经网络自动提取图像深层语义特征。最后,将两种特征融合,交由卷积神经网络的全连接层实现对PM_2.5浓度的回归预测。预测误差比对结果显示,相比使用单种特征,使用融合特征能够有效提高模型的预测性能。     

NARX神经网络模型在昆明市环境空气质量预测中的应用

神经网络模型作为一种重要的手段被广泛应用于数学计算、物理建模、水文模拟、环境预测、人工智能等研究领域。为验证神经网络模型在高原山地城市环境空气质量预测中的作用,以昆明市环境空气自动监测站气象因子和污染物浓度数据为基础,构建NARX神经网络模型,对污染物浓度进行预测。结果表明,基于NARX神经网络建立的预测模型具有很强的非线性动态描述能力,能够对环境空气6参数做出较为准确的预测,其预测浓度相对误差显著低于CMAQ、NAQPMS空气质量数值模式以及LSTM统计模型预测结果。优化后的NARX神经网络对污染物浓度变化趋势的预测较其他几个模式更为敏感。     

A New Mathematical Approach in Environmental and Life Sciences: Gene–Environment Networks and Their Dynamics

An important research area in life sciences is devoted to modeling, prediction, and dynamics of gene-expression patterns. As clearly understood in these days, this enterprise cannot become satisfactory without acknowledging the role of the environment. To a representation of past, present, and most likely future states, we also encounter measurement errors and uncertainties. This paper surveys and improves recent advances in understanding the foundations and interdisciplinary implications of the newly introduced gene–environment networks, and it integrates the important theme of carbon dioxide emission reduction into the networks and dynamics. We also introduce some operational and managerial issues of practical working and decision making, expressed in terms of sliding windows, quadrants (modules) of parametric effects, and navigating (controlling) between such effects and directing them. Given data from DNA microarray experiments and environmental records, we extract nonlinear ordinary differential equations that contain parameters that have to be determined. For this, we employ modern (Chebychevian) approximation and (generalized semi-infinite) optimization. After this is provided, time- discretized dynamical systems are studied. A combinatorial algorithm with polyhedra sequences allows to detect the region of parametric stability. Finally, we analyze the topological landscape of gene–environment networks with its structural (in)stability. By embedding as a module and investigating CO₂ emission control and figuring out game theoretical aspects, we conclude. This pioneering work is theoretically elaborated, practically devoted to health care, medicine, education, living conditions, and environmental protection, and it invites the readers to future research.      

微波消解-原子吸收光度法测定土壤中铜锌铅镉镍铬

用微波消解－原子吸收光度法测定土壤中铜，锌，铅，镉，镍和铬。通过硝酸－氢氟酸－过氧化氢体系消解液对土壤样品消解，选择出微波最佳消解条件。对硝酸－盐酸－过氧化氢体系消解液和硝酸－氢氟酸－过氧化氢体系消解液进行消解对比试验，发现前者不能将土壤样品完全消解，后者能将样品消解完全，但需将消解液中剩余的酸赶尽，否则测定结果将明显偏低。微波消解土壤与传统电热消解相比，操作简便快速，可提高工作效率。     

Topsoil and Housedust Metal Concentrations in the Vicinity of a Lead Battery Manufacturing Plant

The aim of the study was to determine the extent of heavy metal pollution in the vicinity of a lead battery manufacturing plant in the Czech Republic, both in the general environment and within homes. Topsoils (0–5 cm) were sampled from 100 sites along 4 transects which crossed the battery factory, the town centre and outlying rural areas. Housedust samples were collected from 15 plant-workers' homes and 15 non plant-workers' homes located at various distances from the factory. Samples were analysed for levels of Pb, Zn, Cd, Cu, As and Sb, using ICP-AES. Soil concentrations of Pb, Cd, As and Sb peaked at the perimeter fence of the battery factory with Pb levels of up to 12 650 g g> ^-1. Concentrations of these metals decreased exponentially within 500 m of the plant. Whilst the battery factory does appear to be a source of all the metals with the exception of Zn, emissions do not appear to be transported to any great extent over distances of more than 0.5 km. Topsoil and housedust concentrations of Pb, Zn, Cd, Cu, As and Sb in the general Mlada Boleslav area do not appear to be significantly different from those recorded in a large scale survey of urban soils in the United Kingdom. A comparison of housedust Pb concentrations in homes of plant workers and non-plant workers suggests that there may be a small input of Pb into the home environment by battery plant employees.     

An Overview of Contaminant-Related Issues Identified by Monitoring in San Francisco Bay

The San Francisco Estuary Regional Monitoring Program for Trace Substances (RMP) began in 1993 and is sponsored by 74 local, state, and federal agencies and companies through their discharge or Bay use permits. The RMP monitors water, sediment, toxicity, and bivalve bioaccumulation at 25 sites in the Bay that are considered to represent "background" conditions. Several major environmental issues have been identified by the RMP. Polychlorinated biphenyls and mercury were often above water quality guidelines, and often occurred in fish tissues above U.S. Environmental Protection Agency (U.S. EPA) screening values. Concentrations do not appear to be decreasing, suggesting continuing inputs. Episodes of aquatic toxicity often occurred following runoff events that transport contaminants into the Bay from urbanized and agricultural portions of the watershed. Sediment toxicity occurred throughout the Bay, and has been correlated with concentrations of specific contaminants (chlordanes, polynuclear aromatic hydrocarbons) at some locations; mixtures of contaminants were probably also important. Since the RMP does not monitor all ecosystem components, assessments of the overall condition of the Bay cannot be made. However, in terms of contamination, the RMP samples suggest that the South Bay, and North Bay sites are moderately contaminated.     

Biomagnification factors (fish to Osprey eggs from Willamette River,Oregon, U.S.A.) for PCDDs,PCDFs, PCBs and OC pesticides

A migratory population of 78 pairs of Osprey (Pandion haliaetus) nesting along the Willamette River in westernOregon was studied in 1993. The study was designed to determinecontaminant concentrations in eggs, contaminant concentrationsin fish species predominant in the Ospreys diet, andBiomagnification Factors (BMFs) of contaminants from fish specieseaten to Osprey eggs. Ten Osprey eggs and 25 composite samplesof fish (3 species) were used to evaluate organochlorine (OC)pesticides, polychlorinated biphenyls (PCBs), polychlorinateddibenzo-p-dioxins (PCDDs), and polychlorinated dibenzofurans(PCDFs). Mercury was also analyzed in fish. Geometric meanresidues in Osprey eggs were judged low, e.g., DDE 2.3 g g^-1 wet weight (ww), PCBs 0.69 g g^-1, 2,3,7,8-TCDD 2.3 ng kg^-1, and generally well below known threshold values for adverse effects on productivity, and the population was increasing. Osprey egg residue data presentedby River Mile (RM) are discussed, e.g., higher PCDDs were generally found immediately downstream of paper mills and eggsfrom the Willamette River had significantly elevated PCBs and PCDDs compared to reference eggs collected nearby in the CascadeMountains. Prey remains at nest sites indicated that the Largescale Sucker (Catostomus macrocheilus) and NorthernPikeminnow (Ptychocheilus oregonensis) accounted for an estimated 90.1% of the biomass in the Osprey diet, and composite samples of these two species were collected from different sampling sites throughout the study area for contaminant analyses. With the large percentage of the fishbiomass in the Osprey diet sampled for contaminants (and fisheaten by Ospreys similar in size to those chemically analyzed),and fish contaminant concentrations weighted by biomass intake, a mean BMF was estimated from fish to Osprey eggs for the largeseries of contaminants. BMFs ranged from no biomagnification(0.42) for 2,3,7,8-TCDF to 174 for OCDD. Our findings for themigratory Osprey were compared to BMFs for the resident HerringGull (Larus argentatus), and differences are discussed. Webelieve a BMF approach provides some basic understanding ofrelationships between contaminant burdens in prey species offish-eating birds and contaminants incorporated into their eggs,and may prove useful in understanding sources of contaminants inmigratory species although additional studies are needed.     

An investigation on biological stability of product water generated by lab-scale and pilot-scale distillation systems

Material Disposal Area G is the primary low-levelradioactive waste disposal site at Los Alamos NationalLaboratory, New Mexico, and is adjacent to Pueblo of SanIldefonso lands. Pueblo residents and Los Alamos scientists areconcerned about radiological doses resulting from uptake of AreaG radionuclides by mule deer (Odocoileus hemionus) andRocky Mountain elk (Cervus elaphus), then consumption ofdeer and elk meat by humans. Tissue samples were collected fromdeer and elk accidentally killed near Area G and were analyzedfor ³H, ⁹⁰Sr, total U, ²³⁸Pu, ^{239, 240}Pu,²⁴¹Am, and ¹³⁷Cs. These data were used to estimatehuman doses based on meat consumption of 23 kg y^-1. Humandoses were also modeled using RESRAD, and dose rates to deer andelk were estimated with a screening model. Dose estimates tohumans from tissue consumption were 2.9 × 10^-3 mSv y^-1and 1.6 × 10^-3 mSv y^-1 from deer and elk, respectively,and RESRAD dose estimates were of the same order of magnitude. Estimated dose rates to deer and elk were 2.1 × 10^-4 mGyd^-1 and 4.7 × 10^-4 mGy d^-1, respectively. Allestimated doses were significantly less than established exposurelimits or guidelines.     

The Draft of the New German Guideline VDI 3782/8 to Model Automobile Exhaust Dispersion

The draft of the German guideline to calculate automobile exhaust dispersion is explained. It contains a two-stage-system: For first quick estimates the guideline contains the simple models MLuS and STREET. In case these models are not applicable or their results shows concentration levels close to the air quality standards, the more complex models PROKAS_V and MISKAM are recommended. PROKAS_V is a Gaussian plume model, MISKAM is a 3-dimensional microscale non hydrostatic flow model for built-up areas with an Eulerian dispersion model. The guideline comprises cases in rural areas without or with few adjacent buildings as well as urban areas with buildings near the roads. The contribution gives information about the models, typical results and some of the problems showing up presently.     

Hydrocarbon Utilization in Yeast Isolates Found to Grow in Association with Petroleum in a Polluted Ultisol of Midwestern Nigeria

The population counts of hydrocarbon-utilising yeasts weremonitored at six sampling stations in the Benin City municipalarea over a four-week period. Although the population countswere relatively constant in each locality, the highest countsoccurred in areas heavily polluted by domestic and industrialeffluents. A total of thirty-five hydrocarbon-degrading yeastswere isolated from soil at the sampling stations usingn-hexadecane as sole carbon source. The isolates were identifiedas belonging to the genera Candida (27 strains), Endomycopsis (4 strains). All the organisms grew on long-chainn-alkane, kerosene, diesel oil and crude oil but failed to growon short-chain n-alkane, aromatic and alicyclic hydrocarbons.Measurement of growth attributes of the isolates usingn-hexadecane, diesel oil and crude oil as substrates showed thatthe Candida species were better utilizers of hydrocarbonsubstrates relative to Endomycopsis and Schizosaccharomyces species.     

Hydrologic–Economic Model for Managing Irrigation Intensity in Irrigation Areas under Watertable and Soil Salinity Targets

This paper gives mathematical details and sample applications of SWAGMAN Farm (SWAGMAN, Salt Water and Groundwater Management), a farm-scale hydrologic economic model that integrates agronomic, climatic, irrigation, hydrogeological and economic aspects of irrigated agriculture. The model is capable of determining optimum mix of land use to keep watertable and soil salinity within acceptable limits while maximising the economic returns. Alternatively, the model can simulate water and salt balance and economics of a given cropping preference. Web-based and Geographic Information Systems versions of the model are available for integration with the environmental reporting systems of the irrigation areas.     

Assessing health consequences in an environmental impact assessment: The case of Amsterdam Airport Schiphol

In this article a comprehensive approach for the evaluation of possible health effects in an environmental impact assessment (EIA) is described, illustrated with the example of Amsterdam Airport Schiphol. Unlike many EIAs, we estimated quantitatively the impact of aircraft-related pollution in terms of the number of affected people for aircraft noise annoyance, odour annoyance and hypertension. In addition, an analysis of health registry data on cardiovascular and respiratory diseases and a short survey on annoyance and risk perception were carried out. The scope of a health impact assessment depends on the situation, available knowledge and data, concern in the population about the impact and the number of people concerned. It is important to pay attention to the perception of risks and concerns from all parties involved. Moreover, the results demonstrate that far more people outside the area for which standards apply were affected than inside.