Comparison of Four Machine Learning Methods for Predicting PM10 Concentrations in Helsinki, Finland

Machine learning methods can offer a practicalalternative to deterministic and statistical methods forpredicting air pollution concentrations. However, for agiven data set, it is often not clear beforehand whichmachine learning method will yield the best predictionperformance. This study compares the variable selection andprediction performance of four machine-learning methods ofdifferent complexity: logistic regression, decision tree,multivariate adaptive regression splines and neuralnetwork. The methods are applied to the task of predictingthe exceedance of the European PM₁₀ daily averageobjective of 50 g m^-3 for a station in Helsinki,Finland. Our study shows that some predictors were selectedby all models but that the different models also pickeddifferent variables. The performance of three of the fourmethods investigated was very similar, however, performanceof the decision tree method was significantly inferior.Performance was sensitive to the learning sample size andtime period used.     

Characterisation of airborne particles collected within and proximal to an opencast coalmine: South Wales,U.K

Airborne particulate matter has been collected from within,and proximal to, an opencast coal mine in south Wales. Thiswork forms the first part of a three year project to collectand characterise, then determine the possible toxicology ofairborne particles in the south Wales region. High-resolutionField Emission SEM has shown that the coal mine dusts consistlargely of an assemblage of mineral grains and vehicle exhaustparticles. SEM-EDX has shown that the mineralogical make-up ofthe PM10 is complex, heterogeneous, and constantly changing.These findings are supported by analytical TEM-EPXMA.However, patterns can be determined relating the mineralogicalcomposition of the airborne particles to collection locationsand mining activities within the opencast. At our studyopencast, Park Slip West, quartz, which has known healtheffects, never exceeded 30% of the total collection mass, andaverage levels were much less. Vehicle exhaust emissions wasthe largest source in terms of particle numbers. The mass ofairborne particulate matter within the pit averagedapproximately twice that of outside the pit: importantlyhowever, this higher mass was due to relatively large, andnon-respirable, mineral grains. This study demonstrates that the physicochemical andmineralogical characterisation of airborne particles frommining and quarrying is essential to quantify the respirablefraction, and to identify potentially hazardous componentswithin the PM10.     

The possibility of prenatal diagnosis by gene dosage: Confirmation of duplication 10q24→qter from Got-l activity in fetal erythrocytes

The activity of four enzymes, including GOT-1, has been investigated in the erythrocytes of a 10q→24qter trisomic fetus. Analyses have been performed on a feto-maternal blood mixture sampled by fetoscopy and on red cells obtained by cardiac puncture, following therapeutic abortion. The demonstration of a 40 percent increase of GOT-1 activity, as compared to normal fetuses of similar gestational age, suggests that gene dosage studies may be a useful confirmatory technique in prenatal diagnosis of unbalanced chromosomal aberrations. Practical application of a similar diagnostic approach is conditioned by (1) precise characterization of fetal chromosome imbalance; (2) confirmed assignment of the gene locus coding for the gene product under investigation; (3) evidence of a linear proportionality between gene dose and concentration of the gene product in patients with the same chromosome imbalance detected in the fetus; (4) knowledge of the range of normal variation at different weeks of gestation of the enzyme activity to be tested in the fetus; (5) safety of fetal sampling procedure.     

Integrated Assessment of Abatement Strategies to Improve Air Quality in Urban Environments, the USIAM Model

Planning effective strategies to combat air pollution in amajor city such as London requires integration ofinformation on atmospheric concentrations and where theyexceed prescribed air quality standards, detailed data onemissions and potential measures to reduce them includingcosts, and a good understanding of the relativecontributions of different emission sources to pollutantconcentrations plus the remaining background. The UrbanScale Integrated Assessment Model (USIAM) is designed as atool to integrate such information, and to explore andassess a variety of potential strategies for improving airquality. It is based on the same principles as theAbatement Strategies Assessment Model (ASAM) that has beenused in the UN Economic Commission for Europe. To startwith the USIAM model is being developed with respect to theparticulate PM₁₀ only, and in particular the primaryparticulate contribution. The secondary particulate istreated as part of the background superimposed on theprimary particulate concentrations; this may need to betreated more specifically at a later stage, particularlywith respect to nitrate formation over the city. The USIAMmodel therefore sets out to examine a selection of severeepisode conditions as well as long-term annual averageconcentrations, and aims to find strategies that aresuccessful in eliminating exceedance of the prescribedtarget concentrations. By ranking different options forabatement of emissions, for example in terms of cost orease of implementation, the USIAM model can also select andprioritise different potential strategies.     

南京市大气PM_(10)浓度的估算模型研究

以南京市为例,利用空气污染指数API、气象数据和TERRA/AQUA卫星气溶胶光学厚度(AOD)产品,分析了南京市PM_(10)浓度的变化规律,在PM_(10)浓度与气象要素进行相关分析的基础上,初步建立了基于气象要素和AOD的PM_(10)浓度估算模型。结果表明,南京市PM_(10)浓度在每年11,12月或1月最高,7,8月最低,季节性变化表现为冬春季浓度最高,秋季其次,夏季最低,PM_(10)浓度有逐年下降的趋势,但年均值仍高于国家II级标准;除了大气混合层高度外,PM_(10)浓度与大气压、风速、气温、相对湿度、水汽压、能见度、气溶胶光学厚度都有较好的相关性;基于气象要素的PM_(10)浓度估算模型的绝对系数R~2为0.510、平均相对误差为26.04%,基于AOD的PM_(10)浓度估算模型以TERRA和AQUA卫星AOD平均值构建的最佳,绝对系数R~2为0.482、平均相对误差26.11%,两种模型对PM_(10)的预测预报具有一定的指示意义。     

Variation of airborne quartz in air of Beijing during the Asia-Pacific Economic Cooperation Economic Leaders' Meeting

Quartz particles are a toxic component of airborne paniculate matter(PM).Quartz concentrations were analyzed by X-ray diffraction in eighty-seven airborne PM samples collected from three locations in Beijing before,during,and after the Asia-Pacific Economic Cooperation(APEC) Leaders' Meeting in 2014.The results showed that the mean concentrations of quartz in PM samples from the two urban sites were considerably higher than those from the rural site.The quartz concentrations in samples collected after the APEC meeting,when the pollution restriction lever was lifted,were higher than those in the samples collected before or during the APEC meeting.The quartz concentrations ranged from 0.97 to 13.2 μg/m~3,which were among the highest values amid those reported from other countries.The highest quartz concentration exceeded the Californian Office of Environmental Health Hazard Assessment reference exposure level and was close to the occupational threshold limit values for occupational settings.Moreover,a correlation analysis showed that quartz concentrations were positively correlated with concentrations of pollution parameters PM_(10),PM_(2.5),SO_2 and NO_x,but were negatively correlated with O_3 concentration.The results suggest that the airborne quartz particles may potentially pose health risks to the general population of Beijing.     

我国施工扬尘排污费征收制度研究

目前,我国的城市化建设正在快速发展,建设施工过程产生的大量施工扬尘,是导致雾霾污染的主要原因之一。为有效抑制施工扬尘的产生,部分地区推行了施工扬尘排污费征收制度。但现行的施工扬尘排污费征收制度,在技术、经济和管理层面上均存在着一些较大的问题,例如排污费征收标准偏低,缺乏必要的激励机制,缺乏季节性调整方案等。这些问题使得现行扬尘排污费征收制度未能有效起到防治扬尘污染的作用。文中对国内现行的施工扬尘排污费征收制度进行了系统的分类,并针对现行制度存在的问题,结合我国的实际情况,提出相应的改进措施,用以完善现有的施工扬尘排污费征收制度,从而有效抑制施工扬尘排放,并最终起到改善空气质量的作用。     

道路交通扬尘排放因子测量系统研发及应用

道路交通扬尘排放是城市大气环境颗粒物(PM_(10)和PM_(2.5))的主要来源之一,对其排放测量研究是进行排放清单建立、环境影响分析和制定控制方案的依据.本研究设计了一种道路交通扬尘排放因子测量系统,通过测量行驶中车辆尾羽不同位置的颗粒物浓度,应用浓度剖面积分的方法计算单车行驶过程中扬尘PM_(10)排放量.在北京市典型道路测量了小汽车和大客车在不同车速下的交通扬尘颗粒物排放因子,结果显示,车辆尾羽的颗粒物浓度特征呈明显的"层状"分布,距离路面越近浓度越高,在车辆行驶方向中心浓度最高,向两侧浓度逐渐降低,车速越快浓度越高.在试验车速范围内,排放因子与车速呈幂函数关系,幂指数为2.7~2.8.排放因子与积尘负荷呈幂函数关系,幂指数为0.85.不同路段或同一路段的不同区域排放因子空间变异性较大,应用排放因子测量系统进行实测的结果更加准确可靠.     

成都市一次重雾霾天气PM10中水溶性离子污染特征

2015年1月10~19日,对成都市青羊区雾霾期间大气PM10样品进行了采集,分析其质量浓度和9种水溶性离子(F-、Cl-、NO-3、SO2-4、NH+4、Ca2+、Mg2+、K+和Na+)的分布特征。结果表明:此次雾霾期间PM10的平均质量浓度高达357.8μg/m3;水溶性离子的总质量浓度在雾霾天较高,二次无机水溶性离子NO-3、SO2-4和NH+4占总水溶性离子的74.53%以上;雾霾和非雾霾天PM10中水溶性离子的质量浓度、含量百分比和各离子比值体现了本次雾霾是由机动车尾气、燃煤、建筑施工和气象等因素造成;PM10中NO-3/SO2-4的比值范围0.755 1~1.410,均值为1.050,说明成都是燃煤和机动车并存的复合型污染;PM10中NH+4和NO-3、SO2-4两两之间均具有较好的相关性,NH+4和NO-3、SO2-4的相关性系数分别为0.845 0、0.966 0,NH+4和NO-3、SO2-4主要以(NH4)2SO4、NH4NO3形式存在于PM10中,Ca2+和Mg2+的相关性系数高达0.913 0。     

城市道路二次扬尘中PM_(10)浓度量化模型实验研究

机动车行驶条件、路况与道路二次扬尘中PM10浓度有密切的关系。通过对它们设计正交试验和多元回归统计分析,得到PM10浓度计量模型:ρ=0.0314W0.8248v0.537q0.7021M0.357,该方程具有99%的置信度,且拟合的较好。经回归系数显著性检验,自变量对因变量的影响均极为显著,其显著性大小依次为:道路路面粉尘负荷、机动车车型、车速和车流量。最后采用实测值与模型计算值间的均方差对模型进行验证,结果表明该模型具有较高的可信度。