Assessing Spatial, Temporal, and Analytical Variation of Groundwater Chemistry in a Large Nuclear Complex, USA

Statistical analyses were applied at the Hanford Site, USA, to assess groundwater contamination problems that included (1) determining local backgrounds to ascertain whether a facility is affecting the groundwater quality and (2) determining a ‘pre-Hanford' groundwater background to allow formulation of background-based cleanup standards. The primary purpose of this paper is to extend the random effects models for (1) assessing the spatial, temporal, and analytical variability of groundwater background measurements; (2) demonstrating that the usual variance estimate s ², which ignores the variance components, is a biased estimator; (3) providing formulas for calculating the amount of bias; and (4) recommending monitoring strategies to reduce the uncertainty in estimating the average background concentrations. A case study is provided. Results indicate that (1) without considering spatial and temporal variability, there is a high probability of false positives, resulting in unnecessary remediation and/or monitoring expenses; (2) the most effective way to reduce the uncertainty in estimating the average background, and enhance the power of the statistical tests in general, is to increase the number of background wells; and (3) background for a specific constituent should be considered as a statistical distribution, not as a single value or threshold. The methods and the related analysis of variance tables discussed in this paper can be used as diagnostic tools in documenting the extent of inherent spatial and/or temporal variation and to help select an appropriate statistical method for testing purposes.     

Spatial modelling of the probability of obtaining a detectable concentration of ammonia in Moreton Bay, Australia

Ammonia is an important water quality variable, which in excess, can be detrimental to waterways and their ecosystems. In the Ecosystem Health Monitoring Program in South-east Queensland ammonia is monitored monthly, however, often more than 50% of the ammonia observations in Moreton Bay are below detection limit, making it difficult to draw useful inferences. In this paper a clipped Gaussian random field is used to spatially model and map the probability of detectable concentration of ammonia. The methodology is applied to the Moreton Bay samples collected in February 2005. The results suggest that for this month the oceanic impacted areas have higher probability of detectable ammonia concentration than the areas closer to the main sources of anthropogenic inputs.     

基于RF-LSTM的鸡舍恶臭气体预测研究

以鸡舍氨气为研究对象,对鸡舍氨气预测模型进行了研究.首先,利用随机森林算法（RF）对影响鸡舍氨气浓度的环境变量进行重要性排序,选取温度、湿度、光照、气象温度、降雨量作为模型的输入变量;在此基础上,构建了基于长短时记忆神经网络（LSTM）的鸡舍氨气浓度预测模型,并将提出的预测模型应用于江苏省宜兴市某养鸡场的氨气浓度预测中,并与LSTM模型、RF-Elman模型和RF-BP模型进行了对比实验,结果表明,基于RF-LSTM模型的预测效果最好,其平均绝对误差（MAE）、平均绝对百分误差（MAPE）和均方根误差（RMSE）分别为0.9183、4.9637%和1.4262;同时,为了验证该模型的性能,本文还实现了不同时间尺度的鸡舍氨气浓度预测,提前2h、3h、4h、5h氨气预测的平均绝对误差（MAE）分别为1.6218、2.1991、2.8553和3.0677.本文提出的预测模型提高了鸡舍氨气浓度的预测精度,可为减少鸡舍恶臭气体排放提供科学依据.     

随机振动响应峰值因子的预计理论研究

简要介绍了超越概率理论、超越频次理论、损伤等效理论和功率谱密度(PSD)的时域拟合理论等4种常见的峰值因子预计理论,并基于三角级数提出了一种新的预计理论。结合试飞加速度数据样本,对比分析超越频次理论、PSD时域拟合理论和三角级数理论的预估精度。研究表明,上述4种常见的预计理论本质上属于统计学理论;PSD时域拟合理论预计的峰值因子波动较大,峰值因子与归一化次数满足高斯分布;三角级数理论的预估精度较高,但缺乏离散峰个数的合理判据。     

建筑物尾流区气流与污染物扩散的数值计算

采用了细网格非静力能量闭合边界层模式和随机游动模拟方法建立了一套分析建筑物尾流流场和污染物扩散的数值模拟系统．作为应用研究的例子，在对某城市地下交通隧道排废气的风井塔尾流区流场和浓度场进行风洞流体物理实验的基础上，利用所建的模拟系统对风井塔尾流区气流和污染物扩散特征进行研究．结果表明，所建的模拟系统的模拟结果与风洞试验的结果吻合较好，对一些小尺度带有明显湍流不均匀性的流场和局地空气污染物散布的情况有较好的模拟效果和良好的应用前景．     

Random walk modeling of wake dispersion for the exhaust tower of an underground tunnel in urban area

ＩｎｔｒｏｄｕｃｔｉｏｎＲｅｃｅｎｔｌｙ，ｓｏｍｅｉｎｓｔａｌｌａｔｉｏｎｓｗｉｔｈｓｐｅｃｉａｌｃｏｎｓｔｒｕｃｔｉｏｎｆｏｒｅｘｔｒａｃｔｉｎｇｃｏｎｔａｍｉｎａｎｔｇａｓ，ｓｕｃｈａｓａｎｅｘｈａｕｓｔｔｏｗｅｒｏｆｔｈｅｕｎｄｅｒｇｒｏｕｎｄｔｕｎｎｅｌ，ｅｔｃ．ｗｅｒｅｓｅｔｕｐｉｎｕｒｂａｎａｎｄｉｎｄｕｓｔｒｉａｌｃｏｍｐｌｅｘａｒｅａｓ．Ｔｈｅｅｆｆｅｃｔｓｏｆｔｈｅｓｅｃｏｎｓｔｒｕｃｔｉｏｎｓｏｎｔｈｅｆｌｏｗａｎｄｔｈｅｒｅｂｙｏｎｔｈｅｌｏｃａｌｄｉｓｐｅｒｓｉｏｎｏｆｔ…     

南海重点海洋倾倒区表层沉积物富集特征及其潜在生态风险评价

根据2005南海三个重点海洋倾倒区,即九澳岛重点海洋倾倒区、黄茅岛重点海洋倾倒区、汕头重点海洋倾倒区的监测,分析其表层沉积物重金属的监测资料,采用富集系数法和Hakanson潜在生态危害系数法对其富集度和潜在生态危害程度进行了分析和评价.结果表明:在三个倾倒区表层沉积物所监测的重金属中,Cu和Pb元素的富集度相对最高,Hg和Cd元素的富集度较低,从空间上看,潜在生态危害系数评价结果显示,南海三个重点海洋倾倒区表层沉积物中的重金属对海洋生态系统的潜在生态危害非常轻微,均属于轻微潜在生态危害范畴,其中轻微潜在生态危害程度相对较重的是Cd和Hg元素,Cu和Pb元素的轻微生态危害程度较轻.     

基于随机森林的突发事件分级模型研究

突发事件的分级是快速响应和有效应对的基础，为了解决目前突发事件分级宽泛、主观性强和动态适应性弱等问题，将多分类器集成引入突发事件的分级过程中，借助随机森林组合分类器，针对突发事件特征，构建突发事件分级的随机森林模型，形成突发事件分级过程，进而对事前的应急预案制定、事中的应急资源配置和应急决策提供有力支撑；最后，以2014-2016年洪涝灾害突发事件为例检验构建的模型和过程。研究结果表明:与支持向量机方法相比，通过检验随机森林组合分类器分类结果准确率达到97.56%，在突发事件分级的应用上是可行的，进而可为突发事件的快速响应和应急决策提供信息支撑和参考依据。     

Equity in neighbourhood walkability? A comparative analysis of three large U.S. cities

This study examines the spatial and statistical relationships between social vulnerability (SV) and neighbourhood walkability across three large U.S. cities with different urban typologies and development patterns: Charlotte, NC (a low-density, fast-growing “Sunbelt” city); Pittsburgh, PA (a moderate density, shrinking “Rust Belt” city); and Portland, OR (a progressive West Coast city known for its sprawl-containment policies). Binary logistic regression, independent-samples t-tests, and mapping techniques are employed to determine whether neighbourhoods with high SV (i.e. older populations, higher poverty rates, more service occupations, lower educational attainment, and a higher proportion of minorities) are as likely as those with low SV to exhibit a high degree of walkability. The publically available Walk Score® metric, based on proximity to amenities, street network connectivity, and density, was used as a proxy for neighbourhood walkability. The results indicate significant variability among cities, with Charlotte exhibiting the greatest potential for inequitable access to walkable urban environments and the most prominent concentration of “walk-vulnerable” block groups (BGs) with high SV and low walkability. Both Portland and Pittsburgh exhibited more equitable access when comparing BGs with high and low SV; however, they each presented unique spatial patterns, visualised using a series of maps.     

基于随机抽样推测法的LPG储槽2次爆炸事故的模型研究

在借鉴LPG储槽2次爆炸事故后果不确定分析成果的基础上,对事故过程中的不确定参数进行重新分析与选择,将孔洞上方液位高度h0,气云的TNT当量系数α,泄漏开始到点火源出现之前的持续时间t亦作为不确定分析的参数,并利用随机抽样推测的不确定分析方法,对VCE与BLEVE 2次事故后果进行重新分析,获得了与前人研究成果差异较大的结果,并由此分析这些参数对于事故后果影响的显著性。同时对2次爆炸事故的伤害距离进行了研究与分析,由于LPG的闪蒸以及参数α的影响,本案例中LPG泄漏量为总量的80%-90%以上时,2次事故的死亡半径相等且达到最小。