灰色聚类法在地下水质量评价中的应用

运用灰色聚类法对淮北市地下水质量进行评价，结果表明淮北市地下水质量达到了Ⅲ类水质标准以上，能够达到当地地下水环境功能区划要求。评价结果与实际情况一致，评价方法较精确，具有科学性。     

金华市奶牛基地的资源利用问题研究

本文论述了奶牛基地发展与红黄壤资源开发利用之间的内在联系。在分析奶牛基地农业资源利用现状基础上,运用灰色系统理论和多元统计的系统聚类分析方法,对影响奶牛基地发展的因素及区域发展对策作了定量分析,并预测奶牛基地发展趋势,提出奶牛生产不同区域的红黄壤资源开发利用的重点,这为合理开发利用红黄壤资源种草养奶牛提供理论和实际参考依据。     

基于DIAHP的旅游环境容量分析及相关对策研究

基于普通层次分析法的基本原理,融入"权变"思想,针对旅游环境容量的动态性,提出了1种动态改进层次分析法(DIAHP),分析旅游环境容量随时间变化的动态波动特征.最后,通过对南戴河国际娱乐中心的实证研究,验证了其可行性和有效性.     

基于模糊聚类的生态功能区若干指标定量划分的研究

以江苏省沿东陇海线地区为研究区域,利用遥感与G IS工具,研究基于模糊聚类与若干定量指标的生态功能区划方法。该方法选取了5个指标用于区划过程,首先将研究区域分成5×5km大小的空间单元,并提取各单元属性数据生成属性矩阵,利用M atlab模糊逻辑工具箱对单元进行模糊聚类。参照聚类结果与其他资料进行了生态功能区划,并根据区划结果对各分区的产业发展与生态保护提出了建议。     

数学统计方法在公路降噪措施中的应用

基于交通噪声影响的复杂性,提出将层次分析法和加权系统聚类法应用于公路环保设计降噪措施的决策中,有效避免了人为主观因素的片面性.选取3个反映交通噪声影响程度的指标并予以量化,采用层次分析法和加权系统聚类法对各个受噪声影响的环境敏感点按影响程度进行排序及归类,在时间及资金有限的情况下,实现降噪措施的优化设计及资金的有效使用.     

一种改进的模糊聚类方法在大气环境质量评价中的应用

一般的模糊聚类分析方法只能解决大气污染状况的顺序问题,或只能得出此地与彼地的大气质量状况的相似程度,而不能同时确定大气质量的具体等级。本文提出的一种改进的模糊聚类分析方法解决了这一问题、该方法的主要改进在于:①以各污染因子的污染程度的分级标准值为聚类中心,待分样本与聚类中心之间的相似系数或待分样本隶属于某一类的隶属函数作为聚类函数;②相似系数设计为几何平均最小型,隶属函数设计为正态分布型。实例计算和比较表明,该方法是大气环境质量评价中的一种较好的方法。     

灰色聚类修正模型在系统总量控制中的应用

基于指数型白化函数的灰色聚类修正模型避免了原模型存在零权重现象的缺陷,使得各指标样本值与非零权重值之间存在对应关系.经验证,修正模型的综合评价结果与原模型相比较,能更准确的反映实际情况.利用该模型可进行系统现状综合评价,也可计算各单项因子的目标总量控制值.实例分析了杭州市环境空气质量现状.结果表明,综合等级为Ⅱ级,已达到该地区大气质量标准.2005年若保持达标水平,需对该市可吸入颗粒物的浓度控制在0.1227mg/m³以下,NO₂浓度最大控制值为0.0848mg/m³.     

聚类问题集合论解法初探

本文主要探讨了应用集合论来解聚类问题的方法即集合聚类法,它是在论域上寻找由等价关系所诱导的划分。     

常熟市昆承湖水质时空变异特征和环境压力分析

采用综合指数法对近年来昆承湖水质变化趋势进行了分析,用模糊数学综合评价法对该湖2005年水质进行评价,用等级聚类法进一步分析了其水质的空间差异,用TSIM方法分析了昆承湖富营养化状况。结果表明:1998-2004年间,昆承湖污染总体上呈加重趋势,2003年污染最严重,在各污染因子中,氮污染负荷最大;昆承湖为Ⅴ类水或劣Ⅴ类水体;由于陆源污染差异和围网养殖的影响,湖区污染程度北湖>湖西区>南湖区;昆承湖TSI_M>70,呈现较严重富营养化。并分析了昆承湖所承受的巨大环境压力。     

Designing environmental monitoring networks to measure extremes

This paper discusses challenges arising in the design of networks for monitoring extreme values over the domain of a random environmental space-time field {X _ij } i = 1, . . . , I denoting site and j = 1, . . . denoting time (e.g. hour). The field of extremes for time span r over site domain i = 1, . . . ,I is given by \(\{Y_{i(r+1)}=\max_{j=k}^{k+n-1} X_{ij}\}\) for k = 1 + rn, r = 0, . . . ,. Such networks must not only measure extremes at the monitored sites but also enable their prediction at the non-monitored ones. Designing such a network poses special challenges that do not seem to have been generally recognized. One of these problems is the loss of spatial dependence between site responses in going from the environmental process to the field of extremes it generates. In particular we show empirically that the intersite covariance Cov(Y _i(r+1),Y _i′(r+1)) can generally decline toward zero as r increases, for site pairs i ≠ i′. Thus the measured extreme values may not predict the unmeasured ones very precisely. Consequently high levels of pollution exposure of a sensitive group (e.g. school children) located between monitored sites may be overlooked. This potential deficiency raises concerns about the adequacy of air pollution monitoring networks whose primary role is the detection of noncompliance with air quality standards based on extremes designed to protect human health. The need to monitor for noncompliance and thereby protect human health, points to other issues. How well do networks designed to monitor the field monitor their fields of extremes? What criterion should be used to select prospective monitoring sites when setting up or adding to a network? As the paper demonstrates by assessing an existing network, the answer to the first question is not well, at least in the case considered. To the second, the paper suggests a variety of plausible answers but shows through a simulation study, that they can lead to different optimum designs. The paper offers an approach that circumvents the dilemma posed by the answer to the second question. That approach models the field of extremes (suitably transformed) by a multivariate Gaussian-Inverse Wishart hierarchical Bayesian distribution. The adequacy of this model is empirically assessed in an application by finding the relative coverage frequency of the predictive credibility ellipsoid implied by its posterior distribution. The favorable results obtained suggest this posterior adequately describes that (transformed) field. Hence it can form the basis for designing an appropriate network. Its use is demonstrated by a hypothetical extension of an existing monitoring network. That foundation in turn enables a network to be designed of sufficient density (relative to cost) to serve its regulatory purpose.