水资源可持续利用规划耦合模型与应用

介绍了水资源可持续利用规划耦合模型和应用实例。从数学形式上看,它是多目标优化模型,目标函数选用自由度较高并具竞争性的系统状态指标;约束条件应该包括各相关子系统的状态模拟方程,使模拟与优化实现耦合。在研究实例中,选用了GDP和COD排放量指标来表征研究区的经济发展和环境状况,约束条件主要包括水资源系统模拟模型和宏观经济系统扩大再生产—投入产出耦合模型。由于水资源可持续利用规划决策是半结构化的决策问题,开发决策支持系统辅助决策有助于提高决策的科学性和有效性。     

SPOT卫星影像居民地信息自动提取的决策树方法研究

以南京市江宁县为研究区域，首先分析了该区域居民地的影像特征，然后研究了居民地及其背景地物在SPOT—4(Ⅺ)卫星影像4个波段上的光谱特征，并由此探讨了它们在光谱特征上的可分性。研究发现，除道路在光谱特征上与居民地差异不大而难以完全分开外，其它背影地物均可以依据各波段亮度值的大小关系或适当的阈值与居民地分开。但道路和居民地在形状上存在明显差异，因此可以利用形状指数的差异加以去除。最后分析建立了基于光谱特征和形状特征的简单决策树模型，对研究区域居民地信息进行了提取并对结果进行了精度评价。结果表明，该方法的总体提取效果较好，特别是对于面积大于10000m^2的城镇和集村。其提取精度与通常的监督分类方法相比有了很大的提高，只是在水际交界处和道路两侧有误判现象。因此，利用该模型可以将背景地物类型复杂的江南地区的城镇和集村居民地自动提取出来，并且模型受时相影响较小，只是在域值大小上会存在一些差异。     

基于MODIS数据的决策树分类方法研究与应用

介绍了目前国际上流行的两种决策树算法———CART算法与C4·5算法,并引入了两种机器学习领域里的分类新技术———boosting和bagging技术,为探究这些决策树分类算法与新技术在遥感影像分类方面的潜力,以中国华北地区MODIS250m分辨率影像进行了土地覆盖决策树分类试验与分析。研究结果表明决策树在满足充分训练样本的条件下,相对于传统方法如最大似然法(MLC)能明显提高分类精度,而在样本量不足下决策树分类表现差于MLC;并发现在单一决策树生成中,分类回归树CART算法表现较C4·5算法具有分类精度和树结构优势,分类精度的提高取决于树结构的合理构建与剪枝处理;另外在决策树CART中引入boosting技术,能明显提高那些较难识别类别的分类准确率18·5%到25·6%。     

PPGIS在城市规划决策中的应用

在介绍PPGIS概念的基础上,给出了基于PPGIS城市规划决策系统的框架,分析了基于元数据的分布式空间数据库的数据组织方式,提出了公众参与城市规划决策系统中公众冲突的解决机制,并在传统适宜性分析模型的基础上建立了公众参与的适宜性分析模型。     

空间决策支持系统及其支撑软件的设计与应用

本文讨论了SDSS的概念和基本理论,以决策支持为目标,面向半结构化和非结构化问题,用模型驱动的SDSS,是GIS未来的发展方向。在技术上用C语言实现了SDSS支撑软件的设计,并应用于国家科技攻关课题“三川河流域区域治理与开发信息系统”,取得了良好效果;同时对SDSS开发中存在问题作了讨论。     

像元信息分解和决策树相结合的影像分类方法

该文提出了一种基于像元信息分解和决策树相结合的遥感自动分类方法。选择广州市番禺区作为研究区。用像元信息分解和多变量决策树法把TM影像分为水体、植被、水泥地、土壤4种基本组分,分离成4类树枝;分别以4种基本地物组分作为分类掩膜,采用BP神经网络分类、形状指数提取、光谱特征提取等复合方法进行分枝,并开展野外遥感调查,以提高和验证分类精度。结果表明：该方法保证了分枝时地物的纯洁度,有效地避免了地物提取时多余信息的干扰和影响,提高了分类精度。结合实地调查数据与最大似然分类算法进行对比实验,表明该模型比最大似然总体分类精度高16％。     

基于决策树特征提取的支持向量机在岩性分类中的应用

由于支持向量机属于黑箱模型,因此在进行模型学习时无法直接对特征进行选择,而决策树模型在递归创建的过程中自身具有一定的特征选择能力。针对岩性分类问题,本文将决策树和支持向量机结合,通过决策树的建立,在考虑特征重要性的前提下,利用树节点的高度对特征进行提取,并将具有更高分类能力的特征送入支持向量机进行岩性分类。结果表明:通过决策树的特征提取,减少了支持向量机模型的输入特征,从而有效控制了模型的复杂度,使得模型更加稳定并具有更高的分类精度,测试集精度能够提升10%以上。     

台风风暴潮情景构建与时空推演

因台风风暴潮的突发性、情景演变时间的连续性和路径的不确定性,导致应急决策者在应急救援中难以做出正确决策,针对这一现状,将“情景—应对”应用在台风风暴潮应急决策中。本文在分析台风风暴潮情景、情景要素的概念模型基础上,首先通过资料搜集、属性识别等方法提取关键情景要素,采用框架表示法构建情景;然后分析台风风暴潮情景演变规律及演变路径;其次通过动态贝叶斯网络法构建台风风暴潮动态情景网络;最后利用先验概率与条件概率计算情景状态概率,实现了台风风暴潮的关键情景推演。本文以2018年9月16日11时至17时山竹台风对广东省沿海城市影响为例,演示了台风风暴潮的情景推演流程及关键技术。实证分析结果表明,溃堤、海水倒灌、洪水、滑坡发生的概率分别为85%、81%、74%、54%,验证了情景推演在风暴潮中应用的合理性。     

Modeling the environmental susceptibility of landfill sites in California

??The proper management of solid waste (SW) is a global environmental challenge. A major issue is the proper disposal of SW while balancing a wide range of criteria and working with different spatial data. In this study, we used geographic information system as a tool to perform multi-criteria decision analysis with an analytical hierarchy process to develop an environmental impact susceptibility model (EISM) for landfills. The model was applied to the state of California, USA and results are presented herein. In particular, the EISM considers factors such as geology, pedology, geomorphology, water resources, and climate as represented by 13 associated environmental indicators. The results of the EISM indicate that more than 75% of California’s territory is situated in areas with very low, low, and medium environmental impact susceptibility categories. However, in the remaining 25% of the state’s land, 61 landfills are located in the high and very high categories. These results are alarming because during the period from 2000 to 2015, these 61 landfills received approximately 308 million tons of SW, which corresponds to more than 57% of all SW disposed in California. The model results can be used toward mitigating the environmental impacts of these facilities.     

Automatic selection of weights for GIS-based multicriteria decision analysis: site selection of transmission towers as a case study

Transmission line (TL) siting consists of finding suitable land to build transmission towers. This is just one of the numerous complex geographical problems often solved using GIS-based multicriteria decision analysis (MCDA), which is a set of techniques that weight several geographical features to identify suitable locations. This technique is mostly employed using expert knowledge to identify the correct set of weights; thus adding a certain amount of subjectivity to the analysis, meaning that for the same problem if we change the experts involved, we may reach different results.This research is a first attempt to try and solve this issue. We employed a statistical analysis to quantitatively calculate these weights and we tested our method on a case study about transmission line siting in Switzerland. We compared the distances between each sample in our dataset, in this case study these are location of transmission towers, with each geographical feature, e.g. distance from water features. Then we calculate the same distances but for random points, sampled throughout the study area. The reasoning behind this method is that if samples present a distance from a geographic feature statistically different from the random, it means that the feature played an important role in dictating the location of the sample. In this case for instance, high-voltage transmission towers are purposely built as far away as possible from urban areas. Random points are on the contrary by definition sampled without any constraint. Therefore, when comparing the two datasets, we should find that transmission towers have a larger average distance from urban areas than random points. This allows us to determine that this criterion (i.e. distance from urban centers) is important for planning new TL.The results indicate that this method can successfully weight and rank the most important criteria to be considered for an MCDA analysis, in line with weights proposed in the literature. The advantage of the proposed technique is that it completely excludes human factors, thus potentially increasing the social acceptance of the MCDA results.