地理研究  2019 , 38 (5): 1092-1102 https://doi.org/10.11821/dlyj020170890

专栏:城市研究

空间微观模拟方法及在城市研究中的应用

马静

北京师范大学地理科学学部 环境遥感与数字城市北京市重点实验室,北京 100875

A review of spatial microsimulation approach and its application in urban research

MA Jing

Beijing Key Laboratory for Remote Sensing of Environment and Digital Cities, Faculty of Geographical Science, Beijing Normal University, Beijing 100875, China

收稿日期: 2017-09-25

修回日期:  2018-04-11

网络出版日期:  2019-05-13

版权声明:  2019 《地理研究》编辑部 《地理研究》编辑部

基金资助:  国家自然科学基金项目(41601148)广东省城市化与地理环境空间模拟重点实验室开放基金项目(2014B030301032)

作者简介:

作者简介:马静(1986-),女,河南郑州人,博士,副教授,研究方向为城市地理学与行为地理学。E-mail: majingbnu@163.com

展开

摘要

基于活动主体的城市系统微观模拟可能在未来城市研究中发挥重要作用,但其通常受到微观个体数据稀缺的限制。空间微观模拟方法(spatial microsimulation)主要基于家庭、个人等微观分析单元,通过整合不同层面的数据源,如宏观汇总层面的人口普查统计表以及微观层面的家庭活动日志调查等,合成大样本微观个体数据集,可以在精细化空间尺度上对微观个体行为进行模拟研究。该方法在城市系统微观模拟、空间分析以及政策评估等方面具有一定优势,在西方国家城市研究中的应用逐渐增多,但在国内较为缺乏。本文尝试对空间微观模拟方法的起源、三种核心算法,包括条件概率(conditional probability) 、确定性加权(deterministic reweighting)以及模拟退火(simulated annealing)进行介绍,并从国际层面综述该方法在城市研究,如收入与贫困、交通出行、健康等领域中的应用,为我国相关研究的开展提供借鉴。

关键词: 空间微观模拟 ; 城市研究 ; 交通出行 ; 微观数据 ; 人口普查

Abstract

Individual agent-based microsimulation might be an important research direction for future urban modeling. However, possibly due to confidentiality issues, the spatially detailed microdata sets with a wide range of individual or household characteristics are usually not publicly accessible in many countries. There is a strong demand for the development of small area estimates of socio-demographics and the potential effects of policy changes, which could help the government acquire detailed information on population’s attributes at a fine geographic scale, better allocate the limited resources to the most needed places, and evaluate the potential impacts of policy decisions. Using individuals or households as the basic analytical unit, spatial microsimulation can synthesize much individual-level spatial microdata for large populations through combining different data sources, such as household activity diary survey and aggregate population census tabulates. Spatial microsimulation can simulate the virtual populations in a spatial setting, and it involves three major procedures, including the construction of small area microdata, static what-if simulations for one time point and dynamic microsimulation over a period. This approach can simulate the synthetic population’s behavior at fine geographic resolution, and perform different what-if simulations to explore the impacts of policy scenarios. In general, spatial microsimulation has multiple advantages for urban research, spatial analysis and policy evaluation, and thus has been increasingly applied in the fields of geography, transport, and social sciences, particularly in developed countries. However, in China, microsimulation studies has been very scarce to date, possibly due to the fact that the microsimulation development is challenging requiring a high level of programming skills, there is little publicly available software suited to microsimulation models, and there is a lack of data at an appropriate scale. This paper aims to first provide a comprehensive review of spatial microsimulation techniques, including conditional probability, deterministic reweighting, and simulated annealing, which have been widely used for creating synthetic populations in microsimulation studies. Further, this paper also reviews the recent applications of spatial microsimulation approach in urban research worldwide, focusing on income distribution and deprivation evaluation, travel behavior and transport carbon emission, and health behavior and outcomes. The paper ends with the discussion and conclusion.

Keywords: spatial microsimulation ; urban research ; travel behavior ; survey microdata ; population census

0

PDF (1066KB) 元数据 多维度评价 相关文章 收藏文章

本文引用格式 导出 EndNote Ris Bibtex

马静. 空间微观模拟方法及在城市研究中的应用[J]. 地理研究, 2019, 38(5): 1092-1102 https://doi.org/10.11821/dlyj020170890

MA Jing. A review of spatial microsimulation approach and its application in urban research[J]. Geographical Research, 2019, 38(5): 1092-1102 https://doi.org/10.11821/dlyj020170890

1 引言

基于计算机的微观模拟或微观仿真技术起源于20世纪50年代,美国经济学家Orcutt通过构建经济系统中的微观个体单元并对其行为随着时间的变化进行模拟研究[1]。微观模拟主要区别于传统自上而下的宏观模拟,后者主要基于汇总数据研究宏观尺度不同变量之间的相关关系或利用平均值进行模拟研究[2],而前者更多基于微观个体数据通过重复性的随机试验对微观分析单元,例如个人、家庭、企业等的属性或状态进行模拟研究,是典型的自下而上的过程[3,4]。早期微观模拟方法主要应用于人口预测、贫困分析,税收政策等经济学领域[5,6],也有部分研究应用于药品影响分析等医学领域[7],但是这些研究更多基于国家或区域等宏观尺度,重点关注人群而非空间的政策影响,对微观尺度的空间要素缺乏深入考虑[8]。空间微观模拟(spatial microsimulation)本质上可以在不同空间尺度上针对微观活动主体进行模拟研究,但更多应用于微观区域估算。这种方法主要发展于20世纪80年代[9],代表性研究是Birkin等[10,11]利用英国宏观汇总层面的人口普查统计表以及小样本微观个体数据,通过Iterative Proportional Fitting (IPF)技术合成利兹大都市区的虚拟人口微观数据集,并对其收入的空间分布进行模拟研究。随后,空间微观模拟方法在地理学、社会科学以及交通等领域中的应用逐渐增多[12,13,14]

区别于元胞自动机主要用来模拟城市系统的空间复杂性如土地利用的演变[15,16]、人工智能体主要模拟不同agent以及agent与环境之间的相互作用机理[17],空间微观模拟方法主要基于家庭、个人等微观分析单元,通过整合不同层面的数据源,如宏观汇总层面的人口普查统计表以及微观层面的家庭活动日志调查等,合成大样本空间微观数据集,并对微观个体行为进行模拟研究,可以用来重现或预测一个动态复杂的系统[2,9]。例如,Ballas 等[18]利用1991年英国人口普查小区统计数据和家庭调查面板数据,通过空间微观模拟技术对英国整体人口至2021年的空间分布情况进行动态模拟和预测。相比其他城市系统模拟方法,空间微观模拟技术可以用来在精细化空间尺度上探索大样本微观活动主体属性特征的空间分布以及政策影响的空间效应,在微观空间分析、政策评估等方面具有一定优势,同时也可以用来进行长期预测和动态模拟[19,20]。该方法在不同空间尺度转换方面比较灵活,在数据存储、更新以及预测等方面比较有效[3];但是随着空间精度的增加,该方法在运算过程中的复杂性也会随之增加[21]

总体而言,空间微观模拟方法主要包括三大部分: ①整合不同层面的数据源,模拟生成大样本空间微观数据集,通常也被称为population synthesis,合成的微观个体数据集要与真实世界的人口属性空间分布特征尽可能接近; ②静态微观模拟(static microsimulation),针对某一时间截面的人口属性特征进行微观空间模拟或政策分析; ③动态微观模拟(dynamic microsimulation),将微观个体数据集的属性特征,如年龄、受教育程度、婚姻状况、职业类型等基于统计模型或条件转移概率在设定的时间段内进行动态模拟,然后设置不同情境进行空间分析或政策评估[18]。由于动态微观模拟需要大量数据获取不同属性的条件转移概率不断进行动态更新,情况较为复杂,目前国际上关于空间微观模拟方法的应用更多集中于前两部分的研究 [9]。相比而言,国内关于城市系统空间模拟方法的应用主要集中于元胞自动机和人工智能体模拟[16,17],应用空间微观模拟方法的研究较为缺乏。随着我国多元化、精细化数据的可获得性逐渐增强,整合不同数据源的需求逐渐增加,运用空间微观模拟方法探索精细化空间尺度城市微观活动主体的模拟研究具有一定的应用价值与实践意义[22]。因此,本文尝试对空间微观模拟方法进行比较系统的介绍,并从国际层面综述该方法在城市研究中的应用进展,为我国相关研究的开展提供借鉴。

2 空间微观模拟方法

对于空间微观模拟方法而言,首要步骤是在给定的地理空间范围内生成大样本基于微观个体的虚拟人口数据集,使得该数据集的人口属性空间分布特征与真实世界尽可能地接近[7]。一般而言,用来合成空间微观数据的技术主要包括两大类[23]:合成重构(synthetic reconstruction)和加权(reweighting)。前者通常指基于条件概率(conditional probability)或蒙泰卡罗抽样(Monte Carlo sampling),依次对不同空间单元的家庭或个人属性特征如性别、年龄、受教育程度等进行虚拟合成。该方法在早期城市系统模拟研究中应用较为广泛,20世纪90年代以后加权技术逐渐成为主流方法[21]。与条件概率不同,加权技术主要通过对实际调查的微观个体样本进行不断赋予权重的过程,计算家庭或个人新的权重使其与不同空间单元的人口属性分布特征相匹配以模拟目标人群[24]。基于不同算法生成新的权重,可以进一步将加权技术划分为确定性加权(deterministic reweighting)和概率加权(probabilistic reweighting),其中模拟退火(simulated annealing)算法可能是最为著名的一种概率加权方法[23]。本文对这三种常用的空间微观模拟技术依次进行介绍。

2.1 条件概率

作为早期城市研究中最为常用的一种空间微观模拟算法,关于条件概率的介绍见于Birkin等的研究[10],利用传统汇总层面的人口普查统计表,通过获取条件概率在英国利兹大都市区合成大样本微观人口空间数据集。总体而言,条件概率技术主要基于不同社会经济属性的条件概率分布情况,通过计算机生成大量随机数,与不同空间单元、不同人群属性特征的条件概率进行对比,依次合成大样本虚拟人口的社会经济属性特征,例如性别、年龄、受教育程度、婚姻状况等。该技术包含一个次序随机分布过程(sequentially random distribution process),通常也被称为蒙泰卡罗抽样。

通常情况下,针对不同的地域空间单元,依据其人口普查统计分布特征,运用条件概率技术可以合成大样本的虚拟人口空间数据集。基本理念是每次合成一个属性特征,每一个虚拟人口的属性特征主要基于条件概率依次进行赋值[10]

px=px1×px2x1×px3x2,x1××pxmxm-1,,x1(1)

式中: px代表某一属性特征的概率,如上所示,随着所要合成的属性或变量逐渐增加,条件概率会变得更为复杂。由于后一个变量的条件概率依赖于前一个变量的概率分布,因此在使用该技术的时候需要对约束条件或约束变量(constrained variables)进行排序,概率分布较为均匀或较为重要的约束变量通常放在前面进行模拟赋值,例如性别、年龄等,其他社会经济属性特征如种族、职业状况等则通常放在后面进行模拟研究[24]

运用条件概率技术在合成虚拟人口数据集的过程中由于涉及随机分布和随机选择原理,因此每次运行结果都会具有差异性。使用该技术需要多次运算,一般至少需要5000次迭代过程,然后计算平均值可以作为最终模拟结果。条件概率技术具有一定的运算优势,例如容易建模,运算速度快,不需要微观个体调查样本,不同属性特征的条件概率可以从其他公开的政府数据或汇总层面的统计表中获取,通过整合不同类型的数据源合成虚拟人口空间数据集等。然而该技术也存在一定缺陷,例如对约束变量的构成以及排序等非常敏感。此外,如果选取较多的约束变量去模拟人口属性概率空间分布特征时可能会产生较大误差[25]

2.2 加权

2.2.1 确定性加权 与条件概率明显不同,确定性加权技术并不涉及随机分布原理,空间微观数据不是虚拟合成的,而是依赖于对已有的微观调查样本进行权重赋值。具体而言,确定性加权技术主要基于一个简单方程式对微观个体数据不断赋予新的权重,使其与不同地域单元的人口属性空间分布特征相匹配[24]。例如,利用微观非汇总层面的家庭调查数据和宏观汇总层面的人口普查统计表,通过确定性加权技术对家庭调查样本赋予权重,并随着约束条件的增加对初始权重进行不断调整,直到加权后的家庭调查样本的属性特征与真实的人口普查统计表之间达到一定的拟合优度,否则需要对约束条件进行重新选择或结构调整。该技术运算简单,被广泛应用于空间微观模拟研究中,具体的权重赋值过程可以被定义为[26]

ni=wi×sijmij(2)

式中: ni表示对调查样本i赋予新的权重;j表示该样本的某一社会经济属性; wi表示调查样本i的初始权重; sij表示个体i及属性j在汇总层面的人口普查统计表中对应的元素; mij表示个体i及属性j在微观调查样本数据表中对应的元素。

一般而言,每个调查样本的初始权重赋值为1,最终赋予的权重通常为小数。运用该技术首先需要把选择的约束条件或约束变量,如性别、年龄、受教育程度等进行排序,依据其先后顺序逐步进行权重赋值过程,不断用新的权重替代初始权重,直至最后一个约束变量按照公式(2)完成权重赋值过程。由于最后一个约束变量的拟合优度通常较好,因此在进行排序时将最为重要的约束变量放置最后进行加权[24]。依次完成所有约束变量的加权过程即为一次迭代或循环,将该循环的新权重作为初始权重按照公式(2)进入第二次循环,重新对所有约束变量依次进行权重赋值过程。一般而言,完成10次以上的迭代或循环之后得到的最终权重相对比较稳健,通常可以作为最终结果进行模拟研究[26]

确定性加权技术操作简单,运算速度较快,生成的最终权重相对比较稳健。但缺点也较为明显,对约束变量的构成及排序比较敏感。同时,由于其假定空间地域单元具有社会经济属性的同质性特征,因此在模拟不同地域空间单元人口分布异质性特征较为明显的时候产生较大误差。已有一些学者对该技术进行改进。例如,Smith等[27] 首先将不同的空间地域单元按照一定的社会经济属性特征进行聚类分析,对具有相似人口属性空间分布特征的地域单元进行确定性加权,然后再分别对不同类型的空间地域单元进行权重赋值过程,生成具有区域典型性特征的空间人口数据集进行模拟研究,这样可以减少基于同质性假设所产生的误差。

2.2.2 概率加权 区别于确定性加权通常选择全部调查样本作为数据基础,概率加权主要是从调查样本中随机选择子样本进行权重赋值过程,运算原理不同于确定性加权[23]。其中,模拟退火技术作为一种组合最优化方法,可能是最为常用的概率加权方法,该方法是对爬山算法(hill climbing algorithm)的提炼和改进[24]。在爬山算法运行过程中,如果随机选择的替代样本或元素提升模拟人口选择集的整体拟合优度,即减小模型误差,该替代元素则被选入目标人口数据集;反之,则被剔除。爬山算法的缺点在于随机选择潜在的替代元素过程中不能后退,生成的人口数据集可能是次优组合而非最优组合[28]

模拟退火技术在爬山算法基础上进行改进,放松该算法的基本假设以克服其缺点。对于模拟退火技术而言,为了寻求最优组合,一些随机选择的替代元素即使让模型拟合优度暂时变差,但只要满足一定条件,同样被选入目标人口数据集。换言之,是否接收一个潜在的替代元素主要取决于以下方程式[28]

pδE=exp(-δET)(3)

式中: δE表示模拟人口数据集的拟合优度可能产生的变化;T表示在选择集随机替换一个元素使得结果变差所能接受的最大程度。由于替代元素是随机选择和实时评估的,那些使得模型拟合优度提升的替代元素自动被接收,而那些使得模型拟合优度暂时变差的替代元素如果其 pδE大于一个介于0到1之间的随机数时则被接收,否则被删除。如公式(3)所示, δE(即模型误差的增加值)越小,替代元素被接收的可能性则越大;相反,设置的临界值T越小,替代元素被接收的可能性则越小[28]

由此可见,模拟退火算法在寻求最优组合过程中同时允许前进和后退,包含Metropolis算法,这是确定性加权和条件概率等技术所不具备的[24]。此外,该算法对选取的约束条件或约束变量的整体拟合优度同时进行评估,因此不受约束变量结构或顺序的影响。模拟退火算法从调查样本中每次随机选择或替换一个样本,实时计算模型拟合优度,按照公式(3)决定替代元素即微观个体调查样本是否被接收或删除,不断重复此过程,直至找到人口最优化组合。该算法优势明显,但缺点是不容易建模,运算速度较慢,需要实际调查的家庭或个人样本数据。

总体而言,条件概率、确定性加权、模拟退火三种技术优缺点各异,可以根据不同的研究目的、数据特性等进行选择。已有相关研究证明加权技术要显著优于条件概率,尤其是模拟退火技术在选择约束变量等方面更具灵活性,在寻求最优组合过程中更具有效性,是较为流行的一种空间微观模拟技术[23]。例如,Harland等[24]在英国利兹地区不同空间尺度上对比这三种技术的模拟结果,发现模拟退火算法生成的虚拟人口数据集与真实的人口普查数据更为匹配,模拟误差最小。此外,相比条件概率,模拟退火技术生成的人口数据集不仅能够更好地代表现实世界的人口属性空间分布特征,同时可以最大程度的保留实际调查的家庭样本人口的丰富属性特征,解决由于数据限制所产生的相关问题,更好地应用于城市系统微观模拟研究中。

3 评价指标

使用空间微观模拟技术生成的虚拟人口数据集通常需要通过一些拟合指标评估模拟结果的优劣。由于模拟结果通常是对未知的大样本微观个体数据进行的预测和估算,需要将微观模拟结果汇总至一个合适的空间尺度,然后对选择的约束变量和其他非约束变量的人口属性空间分布特征进行评估[29]。其中,最为常用的空间微观模拟评价指标是Total Absolute Error (TAE),可以被定义为:

TAE=ij|Tij-Eij|(4)

式中: TijEij分别表示人口普查统计表中针对单元格 ij中真实的观测到的数据以及通过模拟技术生成的数值。

TAE统计指标易于操作和理解,主要计算不同地域空间单元格中模拟生成的结果对比真实世界的数值之间的绝对误差,也被称为分类误差[24]。该统计量属于测量误差的绝对指标,通常受到目标人群或样本总量的影响。因此,在评估模拟结果优劣的时候也会用到一些相对指标,例如Cell Percentage Error(CPE)和Standardised Root Mean Square Error (SRMSE)。其中CPE可以被定义为:

CPE=TAEN×100(5)

式中:N表示不同地域空间单元格或属性对应的人口数值。此外,相对统计量SRMSE可以被定义为[30]

SRMSE=ij(Tij-Eij)2m×nijTijm×n(6)

式中:mn分别表示相应矩阵的维度。

绝对指标和相对指标分别对模拟结果的不同方面进行评估,可以在空间微观模拟研究中进行综合评价。使用这些统计指标对约束变量进行的评估通常被称为内部验证。此外,还可以对模拟结果进行外部验证,即通过利用其他政府部门或研究团体公开的相关数据,以及通过其他文字资料比如政府报告、发表的文献等,对模拟结果进行评估 [27]。总而言之,对模拟结果进行验证是非常重要的一个步骤,可以采用多种指标或方法检验模拟结果的优劣。

4 空间微观模拟方法在城市研究中的应用

4.1 国际研究进展

在传统城市研究中,由于受到个体数据保密性等原因限制,通常情况下难以获取大样本微观个体数据,或是某些重要变量,比如收入、健康状况、通勤距离、交通出行碳排放等在政府公开的普查数据中不存在或微观空间尺度上难以获取,很大程度上限制了相关研究的开展。空间微观模拟方法可以在不同空间尺度,尤其是精细化空间尺度上通过整合不同数据源合成大样本微观个体数据集,模拟微观活动主体或其某些重要变量的空间分布特征及演变趋势,进而对城市相关政策的空间效应进行评估[2,18,20]。例如,Ballas等[29]基于空间微观模拟方法在精细化空间尺度上合成利兹地区劳动力人口的微观数据集,进而探讨不同地区人口就业可达性的空间差异,并对劳动力市场政策变化的空间影响进行评估;运用IPF加权技术,Anderson[26]对威尔士2003—2005年的家庭贫困率进行微观空间尺度的动态模拟,以及Tanton[31]对澳大利亚不同地区居民的收入水平以及贫困程度等进行空间微观模拟研究。

该方法在微观空间分析方面具有较大优势,近些年来部分学者将其应用于城市居民健康研究中,用来模拟微观空间尺度不同人群的健康行为以及健康效应[8,32]。例如,基于2001年英国人口普查数据和老龄化人口调查面板数据,Clarke等[32]通过模拟退火技术对英国老年人口多种疾病的患病率进行空间微观模拟,并用模拟的患病率结果对不同地域空间单元的健康医疗设施分布情况进行政策评估;也有部分学者通过条件概率或加权技术等分别对城市居民的吸烟率[33]、肥胖率[8]、患病率[32]、残疾水平[34]等进行空间微观模拟研究。

此外,随着交通拥堵、空气污染和能源消耗等问题凸显,微观模拟技术也逐渐应用于交通出行、能源利用等研究中[35]。相比传统计量方法,微观模拟技术能够对居民日常活动-移动行为的时间、空间以及交通方式等的出行决策过程进行有效的、非汇总层面的过程模拟分析,能够较为完整地复制居民复杂的活动-移动系统,因此可以更为准确的估算和预测真实世界的出行行为[36]。与交通研究中传统四阶段方法相比,微观模拟技术具有三个优势: 在计算和存储大量的、多维度的出行概率矩阵方面具有优势; 可以更为直接的模拟居民复杂的出行决策过程,以及研究出行链及其时空制约机制; 能够模拟生成关于居民多种活动与出行统计量的详细信息,而非简单的汇总指标如平均值[37]

为了更好的理解和估算大样本微观个体居民的日常交通行为,微观模拟技术在西方发达国家得到较为广泛的应用。例如,相关学者开发MIDAS (Microanalytic Integrated Demographic Accounting System)模型,试图通过对出行行为的动态建模以及社会经济属性等的微观模拟,进而对出行需求进行预测[35]。该模型主要基于一系列条件概率,对家庭成员的社会经济属性进行动态模拟,进而预测家庭小汽车拥有量的情况以及家庭成员日常活动-移动行为。其他微观模拟系统包括RAMBLAS (Regional Planning Model Based on the Micro-Simulation of Daily Activity Patterns),用来预测荷兰埃因霍芬地区居民日常交通出行量的空间分布[38];以及美国德克萨斯大学奥斯汀分校研究团队开发的CEMDAP (Comprehensive Econometric Microsimulator for Daily Activity-Travel Patterns),主要基于活动分析法对居民日常活动与出行行为进行模拟和预测[36]

Hunt等[39]对早期微观模拟技术在城市或交通领域中的应用进行较为系统的综述,包括加拿大研究团队开发的ILUTE(Integrated Land Use, Transportation, Environment)以及美国的UrbanSim等模拟软件或系统[40]。这些模拟系统主要运用条件概率或蒙泰卡罗抽样合成大样本居民日常出行模式,进而对交通出行量、能源利用以及城市系统等进行模拟分析和动态预测。近些年,相关学者开始运用加权技术研究大样本微观个体居民通勤行为的空间分布特征,并对能源利用等进行模拟分析。例如,Lovelace等[41]运用IPF技术对英国约克郡地区受到石油价格影响最大的通勤者空间分布及其社会经济属性进行空间模拟,发现农村地区通勤者对石油价格的敏感性或脆弱性最大;即使是在某些富裕地区,这种脆弱性也会在通勤者中普遍存在。

4.2 国内研究现状

相比而言,尽管发展中国家目前面临多种城市问题,如交通拥堵、空气污染以及健康危害等,关于空间微观模拟技术的研究主要集中于发达国家,在发展中国家开展甚少[42],可能与以下三个因素有关: 空间微观模拟技术是一种相对比较复杂的地理计算方法,需要C++或JAVA编程,模型开发相对具有挑战性。 关于精细化空间尺度城市活动主体的模拟软件较为缺乏,尽管有少数几个软件面向公众开放使用,例如ILUTE 和CEMDAP,但是由于这些软件需要大样本详细的交通出行调查数据,同时对数据格式要求比较严格,因此在软件使用方面受到较大限制[43]缺乏微观空间尺度的相关数据[22]。空间微观模拟方法主要基于家庭、个人等微观分析单元进行模拟研究,需要精细化空间尺度的详细信息,但是这些数据在发展中国家,尤其中国,较为缺乏。例如,在中国,没有大样本详细的、微观空间尺度的交通出行调查数据;即使是每十年进行一次的全国人口普查数据,政府通常只在较大尺度,如城市、区县等层面公开人口社会经济属性统计表,而微观空间单元,如街道、居委会层面的数据通常难以获取。

随着多种数据源的获取性不断增强以及计算机技术的快速发展,相关学者通过开展活动日志调查或大样本的公交刷卡数据等,采用不同的微观模拟方法对精细化城市模拟研究进行探索。例如,龙瀛等[4,17]采用多智能体方法基于多种数据源初步合成北京市全样本的居民个体数据,并在微观空间尺度上建立对城市形态、交通能耗以及环境影响之间关系的定量分析。利用北京第五次和第六次人口普查数据以及居民活动日志调查,马静等[43, 44]首次采用模拟退火技术对2000年北京大约72万城市居民和2010年大约100万城市居民的日常出行行为以及交通碳排放进行街道层面的动态模拟,并对比分析2000—2010年北京城市居民出行行为以及交通碳排放的时空演变趋势,在此基础上进一步针对不同情境下2030年交通出行碳排放进行预测和评估。

5 讨论

空间微观模拟方法虽然在数据整合、过程模拟以及政策评估等方面具有一定优势,但也存在局限性,运用该方法时需要注意几个问题[23]

(1)加权技术是目前合成大样本空间微观数据集的主要方法,需要同时使用汇总数据,如人口普查统计表以及微观调查样本数据。其中,微观调查样本数据除了包含重要的目标变量(如交通出行行为)之外,还需要涵盖与人口普查数据相匹配的一系列属性特征,如性别、年龄、受教育程度、职业状况等。两套数据中共有的属性变量可以被选择作为空间微观模拟的约束条件或约束变量,但是需要注意约束变量与目标变量之间应该具有显著的相关关系,可以通过科学理论和数学模型进行选择。约束变量的选择优劣至关重要,直接影响模拟效果[27]。 (2)微观调查样本数据作为合成大样本空间数据集的重要基础,应该具有代表性,满足一定的样本量需求,较大的调查样本量可以更好的捕捉不同空间地域单元的人群异质性特征[23]

(3)使用空间微观模拟技术需要通过多种指标对模拟结果的有效性进行检验 [24]。一般而言,约束变量的拟合优度相对较好,误差较小,通常会在文章中汇报其TAE。而对于外部验证,需要将模拟结果与其他尚未在研究中使用的数据源进行对比,例如,Edward等[45]基于英国人口普查数据和全国健康调查数据,通过空间微观模拟技术对成年人肥胖率的空间分布进行模拟研究,并采用与肥胖率相关的其他来源的癌症数据进行外部验证。但是,通常情况下由于相关外部数据难以获取,对模拟结果进行外部验证较为困难。

6 结论

空间微观模拟方法主要基于家庭、个人等微观分析单元,通过整合不同层面的数据源,可以合成大样本空间微观数据集,同时保留小样本传统调查数据的丰富属性以及宏观数据的大样本特征,在一定程度上实现小样本调查数据与“大数据”的有效结合,克服单一数据缺陷,提供一种更为灵活、有效的研究方法和技术手段。但是,空间微观模拟方法作为计算机模拟研究中的一个分支,其本身也存在一定的问题和局限性,需要与其他方法,如计量模型、人工智能体模拟、以及定性方法等相结合,才能更好的发挥作用[46]。同时,计算机模拟研究需要在科学理论思想的指导与规范下进行,努力“将地理学的经验建模、理论解析和模拟实验有机结合起来”[47],才能更为准确的理解和分析各种城市问题产生的原因、过程和影响,深入研究宏观层面的城市系统与微观层面的个体行为之间的相互作用机理,提出更具有针对性的空间政策或规划措施,以改善城市环境,提高居民生活质量。

The authors have declared that no competing interests exist.


参考文献

[1] Orcutt G H.

A new type of socio-economic system

. The Review of Economics and Statistics, 1957, 39: 116-123.

[本文引用: 1]     

[2] Merz J.

Microsimulation: A survey of principles, developments and applications

. International Journal of Forecasting, 1991, 7: 77-104.

https://doi.org/10.1016/0169-2070(91)90035-T      URL      [本文引用: 3]      摘要

This paper surveys microsimulation as an instrument to analyze and forecast the individual impacts of economic and social policies. The basic principles of microsimulation, which is a tool for practical policy advising as well as for research and teaching, are pointed out and the static and dynamic (cross-section and longitudinal) approaches are compared to one another. Present and past developments of microsimulation models — with emphasis on the static version — and their areas of application are reviewed, focusing on the us and Europe as leaders in this field. The requirement profile and components of a static microsimulation model are discussed. To illustrate the mechanisms involved in a microsimulation model, the paper concludes with a concrete example: the Static Sfb 3 Microsimulation Model based on a relational database system.
[3] Clarke G P.

Microsimulation for Urban and Regional Oolicy Analysis

. London: Pion, 1996.

[本文引用: 2]     

[4] 龙瀛, 沈振江, 毛其智.

城市系统微观模拟中的个体数据获取新方法

. 地理学报, 2011, 66(3): 416-426.

https://doi.org/10.11821/xb201103014      URL      [本文引用: 2]      摘要

目前自上而下的城市系统宏观模拟并不能解决城市这一复杂系统中出现的部分问题,城市系统微观模拟(如多主体系统MAS)已经成为城市系统模拟的新思路,其主要是基于个体数据(如个人、家庭、公司或建筑物)开展的。国际国内这方面的应用都受到个体样本稀缺的限制。微观模拟所需要的个体样本数据是原有的统计制度所不能适应的,尤其是在中国,个体样本在统计公报或年鉴中不公开,仅可通过典型调查来补充。本文旨在探索稀疏数据环境下构建城市系统微观模拟的个体样本数据的新方法。该方法基于已有的多源统计数据、典型调查数据以及个体的通用规则,反演出个体样本的属性信息和空间分布,进而可以以GIS图层的形式直接作为微观模拟的数据基础。通过本方法获取的样本,能够符合已有的统计资料,并遵照了样本的基本特征,可以作为现有数据条件下的微观模拟模型的数据输入。同时该方法的应用简单,统计意义上的准确度高,适合我国统计制度下的微观模拟模型的构建。

[Long Ying, Shen Zhenjiang, Mao Qizhi.

Retrieving individual attributes from aggregate dataset for urban micro-simulation: A preliminary exploration

. Acta Geographica Sinica, 2011, 66(3): 416-426.]

https://doi.org/10.11821/xb201103014      URL      [本文引用: 2]      摘要

目前自上而下的城市系统宏观模拟并不能解决城市这一复杂系统中出现的部分问题,城市系统微观模拟(如多主体系统MAS)已经成为城市系统模拟的新思路,其主要是基于个体数据(如个人、家庭、公司或建筑物)开展的。国际国内这方面的应用都受到个体样本稀缺的限制。微观模拟所需要的个体样本数据是原有的统计制度所不能适应的,尤其是在中国,个体样本在统计公报或年鉴中不公开,仅可通过典型调查来补充。本文旨在探索稀疏数据环境下构建城市系统微观模拟的个体样本数据的新方法。该方法基于已有的多源统计数据、典型调查数据以及个体的通用规则,反演出个体样本的属性信息和空间分布,进而可以以GIS图层的形式直接作为微观模拟的数据基础。通过本方法获取的样本,能够符合已有的统计资料,并遵照了样本的基本特征,可以作为现有数据条件下的微观模拟模型的数据输入。同时该方法的应用简单,统计意义上的准确度高,适合我国统计制度下的微观模拟模型的构建。
[5] Orcutt G, Caldwell S, Wertheimer R.

Policy exploration through microanalytic simulation

.The Urban Institute, Washington D. C., 1976.

[本文引用: 1]     

[6] Orcutt G, Merz J,

Quinke H, eds. Microanalytic Simulation Models to Support Social und Financral Policy. Amsterdam:

North-Holland, 1986.

[本文引用: 1]     

[7] Ballas D, Clarke G, Dorling D, et al.

Using geographical information systems and spatial microsimulation for the analysis of health inequalities

. Health Informatics Journal, 2006, 12: 65-79.

https://doi.org/10.1177/1460458206061217      URL      PMID: 17023399      [本文引用: 2]      摘要

The paper presents a spatial microsimulation approach to the analysis of health inequalities. A dynamic spatial microsimulation model of Britain, under development at the Universities of Leeds and Sheffield, uses data from the censuses of 1971, 1981 and 1991 and the British Household Panel Survey to simulate urban and regional populations in Britain. Geographical information systems and spatial microsimulation are used for the analysis of health inequalities in British regions in a 30 year simulation. The interdependencies between socio-economic characteristics and health variables such as limiting long-term illness are discussed. One of the innovative features of the model is the estimation of variables such as household income at the small area level, which can then be used to classify individuals. The health situation of different simulated individuals in different areas is investigated and the role of socio-economic characteristics in determining health is evaluated.
[8] Edwards K, Clarke G.

The design and validation of a spatial microsimulation model of obesogenic environments for children in Leeds, UK: SimObesity

. Social Science & Medicine, 2009, 69: 1127-1134.

https://doi.org/10.1016/j.socscimed.2009.07.037      URL      PMID: 19692160      [本文引用: 3]      摘要

Obesogenic environments are a major explanation for the rapidly increasing prevalence in obesity. Investigating the relationship between obesity and obesogenic variables at the micro-level will increase our understanding about local differences in risk factors for obesity. SimObesity is a spatial microsimulation model designed to create micro-level estimates of obesogenic environment variables in the city of Leeds in the UK: consisting of a plethora of health, environment, and socio-economic variables. It combines individual micro-data from two national surveys with a coarse geography, with geographically finer scaled data from the 2001 UK Census, using a reweighting deterministic algorithm. This creates a synthetic population of individuals/households in Leeds with attributes from both the survey and census datasets. Logistic regression analyses identify suitable constraint variables to use. The model is validated using linear regression and equal variance t -tests. Height, weight, age, gender, and residential postcode data were collected on children aged 3-13 years in the Leeds metropolitan area, and obesity described as above the 98th centile for the British reference dataset. Geographically weighted regression is used to investigate the relationship between different obesogenic environments and childhood obesity. Validation shows that the small-area estimates were robust. The different obesogenic environments, as well as the parameter estimates from the corresponding local regression analyses, are mapped, all of which demonstrate non-stationary relationships. These results show that social capital and poverty are strongly associated with childhood obesity. This paper demonstrates a methodology to estimate health variables at the small-area level. The key to this technique is the choice of the model's input variables, which must be predictors for the output variables; this factor has not been stressed in other spatial microsimulation work. It also provides further evidence for the existence of obesogenic environments for children.
[9] Tanton R.

A review of spatial microsimulation methods

. International Journal of Microsimulation, 2014, 7(1): 4-25.

[本文引用: 3]     

[10] Birkin M, Clarke M.

SYNTHESIS: A synthetic spatial information system for urban and regional analysis: Methods and examples

. Environment and Planning A, 1988, 20: 1645-1671.

[本文引用: 3]     

[11] Birkin M, Clarke M.

The generation of individual and household incomes at the small area level using synthesis

. Regional Studies, 1989, 23: 535-548.

https://doi.org/10.1080/00343408912331345702      URL      [本文引用: 1]      摘要

ABSTRACT BIRKIN M. and CLARKE M. (1989) The generation ofindividual and household incomes at the small area level using synthesis, Reg. Studies 23, 535--548. This paper describes a method for generating estimates of small area income distributions at the micro-level. The methodology is based on the combination of iterative proportional fitting to estimate missing data and micro-simulation. This method generates synthetic individuals and households together with their associated attributes for enumeration districts in the study area. Data sets that are used include the 1981 Census of Population together with the New Earnings Survey. Results from the modelling exercise are used to illustrate both the types of information that are produced and the flexible nature of aggregation when a micro-data approach is adopted. The study area used is the Leeds Metropolitan District but the results should prove of more general interest. BIRKIN M. et CLARKE M. (1989) La génération des revenus individuels et ménagers par petites zones à partir d'une synthèse, Reg. Studies 23, 535--548. Cet article présente une méthode de générer des estimations de la distribution des revenus des petites zones par micro-analyse. La méthodologie est fondée sur une combinaison de l'ajustement proportionnel itératif pour évaluer les données manquantes, et de la micro-simulation. Cette méthode génère des individus et des ménages artificiels dotés de leurs caractéristiques et habitant les zones de recensement à l'intérieur de la zone d'étude. Les ensembles de données utilisés comprennent le recensement de la population (1981) conjointement avec l'enquête sur les rémunérations courantes. On se sert des résultats provenant de la modélisation afin de démontrer à la fois la gamme des renseignements qui en découlent et la flexibilité de l'agrégation quand on utilise une micro-analyse. La zone d'étude en question est le métropole de Leeds, mais les résultats sont d'intérêt général. BIRKIN M. und CLARKE M. (1989) Eine Beschreibung der Erzeugung von Einzel- und Haushaltseinkommen auf der Ebene des Kleinraums mit Hilfe der Synthese, Reg. Studies 23, 535--548. Dieser Aufsatz beschreibt auf Mikroebene eine Methode zur Beschaffung von Sch01tzungen der Einkommensverteilung innerhalb eines Kleinraums. Die Methodik beruht auf einer Kombination einer sich wiederholender proportionalen Ausrüstung zur Sch01tzung von fehlenden Angaben, und Mikrosimulation. Diese Methode gestattet es, synthetische Einzelpersonen und Haushalte sowie dazugeh02rige Attribute für Z01hlbezirke in dem zur Untersuchung stehenden Gebiet zu schaffen. Die Angaben, die dazu herangezogen wurden, schlossen die Volksz01hlung vom Jahre 1981 und den New Earnings Survey ein. Ergebnisse von den Modellübungen werden dazu benutzt, sowohl die Arten von Angaben, die gemacht werden, wie auch die flexible Natur der Aggregation beim Mikrodatenverfahren zu erl01utern. Als Untersuchungsgebiet dient der Grosstadtraum von Leeds, doch die Ergebnisse dürften sich allgemeineren Interesses erfreuen.
[12] Ballas D, Clarke G.

GIS and microsimulation for local labour market analysis. Computers,

Environment and Urban Systems, 2000, 24: 305-330.

https://doi.org/10.1016/S0198-9715(99)00051-4      URL      [本文引用: 1]      摘要

This paper outlines the findings of ongoing research on ways of employing and combining geographical information systems (GIS) and microsimulation methodologies for the evaluation and analysis of local labour market problems and policies. First, it is shown how data sets from different sources are integrated to create an urban labour market GIS. In the context of this information system, ways of mapping thematically the local labour market demand and supply are presented. In addition, travel-to-work flows are mapped and it is demonstrated how this data can be used to build local labour market Spatial Interaction Models. Further, it is shown how GIS can be combined with microsimulation techniques to highlight urban problems and enhance the analysis and evaluation of potential social and employment policies. In particular, it is shown how GIS and microsimulation can be used as tools in order to analyse a region's economy and labour market and to estimate the degree of labour market segmentation and socio-economic dualism within an urban system. Also, it is outlined how what-if spatial policy analysis of local labour markets can be performed (i.e. simulating new policy initiatives, firm closures, changes in benefit policies and performing regional multiplier analysis). Finally, the paper presents outputs from SimLeeds, which is a spatial microsimulation model for the Leeds labour market, and explores the potential of GIS combined with microsimulation modelling to create a new framework for the formulation, analysis and evaluation of socio-economic policies at the individual or household level.
[13] Brown L, Harding A.

Social modelling and public policy: Application of microsimulation modelling in Australia

. Journal of Artificial Societies and Social Simulation, 2002, 5(4): 6.

[本文引用: 1]     

[14] Vovsha P, Petersen E, Donnelly R.

Microsimulation in travel demand modeling: Lessons learned from the New York best practice model

. Transportation Research Record, 2002, 1805: 68-77.

[本文引用: 1]     

[15] 黎夏, 叶嘉安.

基于神经网络的元胞自动机及模拟复杂土地利用系统

. 地理研究, 2005, 24(1): 19-27.

https://doi.org/10.3321/j.issn:1000-0585.2005.01.003      URL      [本文引用: 1]      摘要

本文提出了基于神经网络的元胞自动机(Cellular Automata),并将其用来模拟复杂的土地利用系统及其演变.国际上已经有许多利用元胞自动机进行城市模拟的研究,但这些模型往往局限于模拟从非城市用地到城市用地的转变.模拟多种土地利用的动态系统比一般模拟城市演化要复杂得多,需要使用许多空间变量和参数,而确定模型的参数值和模型结构有很大困难.本文通过神经网络、元胞自动机和GIS相结合来进行土地利用的动态模拟,并利用多时相的遥感分类图像来训练神经网络,能十分方便地确定模型参数和模型结构,消除常规模拟方法所带来的弊端.

[Li Xia, Yeh Gar-On.

Cellular automata for simulating complex land use systems using nueral networks

. Geographical Research, 2005, 24(1): 19-27.]

https://doi.org/10.3321/j.issn:1000-0585.2005.01.003      URL      [本文引用: 1]      摘要

本文提出了基于神经网络的元胞自动机(Cellular Automata),并将其用来模拟复杂的土地利用系统及其演变.国际上已经有许多利用元胞自动机进行城市模拟的研究,但这些模型往往局限于模拟从非城市用地到城市用地的转变.模拟多种土地利用的动态系统比一般模拟城市演化要复杂得多,需要使用许多空间变量和参数,而确定模型的参数值和模型结构有很大困难.本文通过神经网络、元胞自动机和GIS相结合来进行土地利用的动态模拟,并利用多时相的遥感分类图像来训练神经网络,能十分方便地确定模型参数和模型结构,消除常规模拟方法所带来的弊端.
[16] 陈彦光, 周一星.

细胞自动机与城市系统的空间复杂性模拟: 历史、现状与前景

. 经济地理, 2000, 20(3): 35-39.

URL      [本文引用: 2]      摘要

细胞自动机(CA)现已忧为城市地理研究的重要模拟实验工具,有人甚至认识它是21世纪的新的范型。本文对“粗中有细”的方式,全面地介绍了细胞地理学的起源与沿革,论述了CA的基本思想、组成要素、模型性质及其在城市地理学中的应用领域,指出了存在问题,并预言了其发展前景.

[Chen Yanguang, Zhou Yixing.

Cellular automata and simulation of spatial complexity of urban systems: History, present situation and future

. Economic Geography, 2000, 20(3): 35-39.]

URL      [本文引用: 2]      摘要

细胞自动机(CA)现已忧为城市地理研究的重要模拟实验工具,有人甚至认识它是21世纪的新的范型。本文对“粗中有细”的方式,全面地介绍了细胞地理学的起源与沿革,论述了CA的基本思想、组成要素、模型性质及其在城市地理学中的应用领域,指出了存在问题,并预言了其发展前景.
[17] 龙瀛, 毛其智, 杨东峰, .

城市形态、交通能耗和环境影响集成的多智能体模型

. 地理学报, 2011, 66(8): 1033-1044.

https://doi.org/10.11821/xb201108003      URL      [本文引用: 3]      摘要

城市能耗占全球能耗的比重随着城市化率的不断提高而增大,交通能耗作为城市能耗的重要构成部分,已有较多研究证明城市形态对其具有显著影响,这些研究多属于城市间层次,而少有城市内的研究对城市形态与交通能耗、环境影响的关系进行定量识别。本文拟建立城市形态、交通能耗和环境的集成模型,对单一城市内的不同空间组织(即城市形态),如土地使用方式、开发密度、就业中心的数量和分布等,对潜在的通勤交通能耗和环境影响的关系进行定量识别。该模型采用多智能体(multi-agent)方法,一方面针对同一假想空间采用蒙特卡洛方法根据约束条件生成多个城市形态,并采用就业地斑块数目、平均斑块分形指数、香农多样性和平均近邻距离等14个指标表征城市形态。另一方面,固定数量的居民agent在所生成的每个城市形态内,选择居住区位和就业区位,根据通勤距离和社会经济特征选择交通方式,进而计算通勤交通能耗和环境影响,在城市层面统计通勤交通能耗和环境影响总和。最后分析城市形态与通勤交通能耗和环境影响的定量关系,主要得到以下结论,①对于不同的城市空间布局和密度分布,通勤交通能耗的弹性范围约为3倍;②城市形态评价指数中,就业中心斑块的数量是对通勤交通能耗影响最大的变量;③多种城市形态所对应的通勤交通能耗基本呈正态分布。此外,还对城市形状对通勤交通能耗的影响进行了识别,并针对假想空间的多个典型城市形态(如紧凑与分散、单中心与多中心、TOD政策、绿隔政策),进行了通勤交通总量的计算,进而对典型规划理念进行了定量对比。本模型不仅可以用于识别城市形态与通勤交通能耗和环境影响的定量关系,定量对比典型的规划理念,还可以用于空间规划方案的能耗和环境影响评价。

[Long Ying, Mao Qizhi, Yang Dongfeng, et al.

A multi-agent model for urban form, transportation energy consumption and environmental impact integrated simulation

. Acta Geographica Sinica, 2011, 66(8): 1033-1044.]

https://doi.org/10.11821/xb201108003      URL      [本文引用: 3]      摘要

城市能耗占全球能耗的比重随着城市化率的不断提高而增大,交通能耗作为城市能耗的重要构成部分,已有较多研究证明城市形态对其具有显著影响,这些研究多属于城市间层次,而少有城市内的研究对城市形态与交通能耗、环境影响的关系进行定量识别。本文拟建立城市形态、交通能耗和环境的集成模型,对单一城市内的不同空间组织(即城市形态),如土地使用方式、开发密度、就业中心的数量和分布等,对潜在的通勤交通能耗和环境影响的关系进行定量识别。该模型采用多智能体(multi-agent)方法,一方面针对同一假想空间采用蒙特卡洛方法根据约束条件生成多个城市形态,并采用就业地斑块数目、平均斑块分形指数、香农多样性和平均近邻距离等14个指标表征城市形态。另一方面,固定数量的居民agent在所生成的每个城市形态内,选择居住区位和就业区位,根据通勤距离和社会经济特征选择交通方式,进而计算通勤交通能耗和环境影响,在城市层面统计通勤交通能耗和环境影响总和。最后分析城市形态与通勤交通能耗和环境影响的定量关系,主要得到以下结论,①对于不同的城市空间布局和密度分布,通勤交通能耗的弹性范围约为3倍;②城市形态评价指数中,就业中心斑块的数量是对通勤交通能耗影响最大的变量;③多种城市形态所对应的通勤交通能耗基本呈正态分布。此外,还对城市形状对通勤交通能耗的影响进行了识别,并针对假想空间的多个典型城市形态(如紧凑与分散、单中心与多中心、TOD政策、绿隔政策),进行了通勤交通总量的计算,进而对典型规划理念进行了定量对比。本模型不仅可以用于识别城市形态与通勤交通能耗和环境影响的定量关系,定量对比典型的规划理念,还可以用于空间规划方案的能耗和环境影响评价。
[18] Ballas D, Clarke G, Dorling D, et al.

SimBritain: A spatial microsimulation approach to population dynamics. Population,

Space and Place, 2005, 11: 13-34.

https://doi.org/10.1002/psp.351      URL      [本文引用: 3]      摘要

In this paper we present an account of a 3-year research project that is aimed at dynamically simulating urban and regional populations in Britain. In the context of this project we are using data from the 1991 UK Census Small Area Statistics (SAS) and the British Household Panel Survey (BHPS), in order to dynamically simulate the entire population of Britain into 2021 at the small area level. This paper discusses the structure, aims and objectives of SimBritain and presents some preliminary results. Firstly, alternative spatial microsimulation strategies are discussed and their advantages and drawbacks are outlined. Next, the difficulties in calibrating and validating dynamic microsimulation models such as SimBritain are highlighted and ways to tackle these difficulties are explored. The paper then presents some model outputs that highlight the geographical variation of a wide range of socio-economic variables through the 1990s. Moreover, in light of these outputs, the paper discusses the potential of SimBritain for policy analysis. Copyright ? 2005 John Wiley & Sons, Ltd.
[19] Miller E J, Douglas J, Abraham J E, et al.

Microsimulating urban systems. Computers,

Environment and Urban Systems, 2004, 28: 9-44.

[本文引用: 1]     

[20] Mannion O, Lay-Yee R, Wrapson W, et al.

JAMSIM: A microsimulation modelling policy tool

. Journal of Artificial Societies and Social Simulation, 2012, 15(1): 8.

[本文引用: 2]     

[21] Birkin M, Clarke M.

Spatial microsimulation models: A review and a glimpse into the future. Population Dynamics and Projection Methods

. Springer, 2011.

[本文引用: 2]     

[22] 龙瀛, 茅明睿, 毛其智, .

大数据时代的精细化城市模拟: 方法、数据和案例

. 人文地理, 2014, 29(3): 7-13.

URL      [本文引用: 2]      摘要

以地块作为基本空间单元并以城市活动主体作为模拟对象的精细化模拟是未来城市模型研究的重要方向,大数据(big data)时代的到来也为其提供了重要发展机遇。本文重点对精细化城市模型的主流建模方法进行了介绍,包括元胞自动机(Cellular Automata,CA)、基于主体建模(Agent-based Modelling,ABM)和传统的微观模拟(Microsimulation,MSM)这三种自下而上的微观模拟方法。之后结合精细化城市模型的高标准数据需求问题,对国际上通行的用于精细化模拟数据合成(population synthesis)的方法进行了综述,并给出笔者近年来在精细化城市模拟方面的多项实践案例,最后提出了以GIS为平台,结合CA/ABM/MSM方法,构建我国精细化城市模型的框架体系和关键技术,以期支持我国大城市地区空间政策的制定和评估。

[Long Ying, Mao Mingrui, Mao Qizhi, et al.

Fine-scale urban modeling and its opportunities in the “big data” era: Methods, data and empirical studies

. Human Geography, 2014, 29(3): 7-13.]

URL      [本文引用: 2]      摘要

以地块作为基本空间单元并以城市活动主体作为模拟对象的精细化模拟是未来城市模型研究的重要方向,大数据(big data)时代的到来也为其提供了重要发展机遇。本文重点对精细化城市模型的主流建模方法进行了介绍,包括元胞自动机(Cellular Automata,CA)、基于主体建模(Agent-based Modelling,ABM)和传统的微观模拟(Microsimulation,MSM)这三种自下而上的微观模拟方法。之后结合精细化城市模型的高标准数据需求问题,对国际上通行的用于精细化模拟数据合成(population synthesis)的方法进行了综述,并给出笔者近年来在精细化城市模拟方面的多项实践案例,最后提出了以GIS为平台,结合CA/ABM/MSM方法,构建我国精细化城市模型的框架体系和关键技术,以期支持我国大城市地区空间政策的制定和评估。
[23] Hermes K, Poulsen M.

A review of current methods to generate synthetic spatial microdata using reweighting and future directions. Computers,

Environment and Urban Systems, 2012, 36: 281-290.

https://doi.org/10.1016/j.compenvurbsys.2012.03.005      URL      [本文引用: 6]      摘要

Synthetic spatial microdata enable analyses of artificial populations in the form of individual unit record files at a small area level. They allow analyses of estimates of variables that are otherwise not available at this small area level, while preserving the confidentiality of personal data. This type of data has mainly been used to provide more detailed census data and for spatial microsimulation modelling: for example to analyse social policy and population changes, transportation, marketing strategies or health outcomes. We argue that many potential applications for synthetic spatial microdata remain to be developed. One reason for this is the lack of information about and confidence in this type of data. Introductory literature about creating synthetic spatial microdata and discussions on the decisions that need to be taken during the data generation process are rare. In this paper, we therefore review currently existing methods to generate synthetic spatial microdata in a manner which will support most readers who are considering this approach, and we address the main issues of the data generation process with regards to analyses of neighbourhood level data. We discuss further possible applications of these data and the importance of synthetic spatial microdata.
[24] Harland K, Heppenstall A, Smith D, et al.

Creating realistic synthetic populations at varying spatial scales: A comparative critique of population synthesis techniques

. Journal of Artificial Societies and Social Simulation, 2012, 15: 1-24.

[本文引用: 9]     

[25] Voas D, Williamson P.

An evaluation of the combinatorial optimisation approach to the creation of synthetic microdata

. International Journal of Population Geography, 2000, 6: 349-366.

[本文引用: 1]     

[26] Anderson B.

Estimating Small Area Income Deprivation: An Iterative Proportional Fitting Approach

. Centre for Research in Economic Sociology and Innovation (CRESI) Working Paper 2011-02, University of Essex: Colchester, 2011.

https://doi.org/10.1007/978-94-007-4623-7_4      URL      [本文引用: 3]      摘要

Small-area estimation, in particular the estimation of small-area income deprivation, has potential value in the development of new or alternative components of multiple deprivation indices. These new approaches enable the development of income distribution threshold-based measures of income deprivation as opposed to benefit count-based measures of income deprivation and so enable the alignment of regional and national measures such as the Households Below Average Income with small-area measures. This chapter briefly reviews a number of approaches to small-area estimation before describing in some detail an iterative proportional fitting based spatial microsimulation approach. This approach is then applied to the estimation of small-area HBAI rates at the small-area level in Wales in 2003–2005. This chapter discusses the results of this approach, contrasts them with contemporary ‘official’ income deprivation measures for the same areas and describes a range of ways to assess the robustness of the results.
[27] Smith D M, Clarke G P, Harland K.

Improving the synthetic data generation process in spatial microsimulation models

. Environment and Planning A, 2009, 41: 1251-1268.

[本文引用: 3]     

[28] Williamson P, Birkin M, Rees P.

The estimation of population microdata by using data from small area statistics and samples of anonymised records

. Environment and Planning A, 1998, 30: 785-816.

https://doi.org/10.1068/a300785      URL      PMID: 12293871      [本文引用: 3]      摘要

"Traditionally, estimates of the number of people in small areas (the smallest geographical units for which data are available) have been disaggregated only by age and sex. More recently, much research effort has been directed towards developing some form of enhanced small-area population estimation, in which the population in a small area is disaggregated not only by age and sex, but also by a wide range of additional economic and social characteristics. Solutions to this problem currently include account-based demographic models, often used by local authorities."
[29] Ballas D, Clarke G P.

The local implications of major job transformations in the city: A spatial microsimulation approach

. Geographical Analysis, 2001, 33: 291-311.

[本文引用: 2]     

[30] Knudsen D C, Fotheringham A S.

Matrix comparison, goodness-of-fit, and spatial interaction modeling

. International Regional Science Review, 1986, 10: 127-147.

[本文引用: 1]     

[31] Tanton R.

Spatial microsimulation as a method for estimating different poverty rates in Australia. Population,

Space and Place, 2011, 17: 222-235.

https://doi.org/10.1002/psp.601      URL      [本文引用: 1]      摘要

This paper uses spatial microsimulation to calculate small area poverty severity and headcount poverty in Australia using the Foster-Greer-Thorbeck method. The estimates of poverty severity across Australia, and for certain subgroups of the population, are then compared with headcount poverty rates. Areas of difference are highlighted, and the reasons for some of these differences are investigated further. Because the spatial microsimulation gives unit record data on incomes in each small area, we can compare the microsimulated income distributions for the small areas to see what drives the difference between headcount poverty and poverty severity.We find that the headcount poverty rate is in the same quintile as poverty severity for 63% of people in Australia. The headcount poverty rate tends to be lower than poverty severity in areas where there is extreme poverty, and this is confirmed by looking at the microsimulated income distribution of areas where the differences were greater. This may be because poverty severity is giving greater weight to the extreme poor compared with headcount poverty. There were very few areas where the headcount poverty was higher than poverty severity.Looking at older single people, the picture was very different. Only 36% of aged single people were in the same quintile. In many areas, the headcount poverty was much higher than poverty severity. Areas where the headcount poverty rate was higher than the poverty severity tended to be areas where there were a number of people very close to the poverty line, which is common for aged single people on a pension. Copyright 2009 John Wiley & Sons, Ltd.
[32] Clark S, Birkin M, Heppenstall A.

Sub regional estimates of morbidities in the English elderly population

. Health & Place, 2014, 27: 176-185.

https://doi.org/10.1016/j.healthplace.2014.02.010      URL      PMID: 24631924      [本文引用: 3]      摘要

61Recognises that an ageing population will bring challenges to society, particularly in the domain of health care.61Use of spatial microsimulation to represent a spatially relevant local population.61Exploit the longitudinal nature of the sampling population to produce spatially relevant future estimates of morbidities in the elderly population.61Discusses how this approach allows for a more interrogative exploration of the model outcomes.61Suggests how this work can be taken forward to produce future forecasts.
[33] Tomintz M, Clarke G, Rigby J.

The geography of smoking in Leeds: Estimating individual smoking rates and the implications for the location of stop smoking services

. Area, 2008,40(3): 341-353.

https://doi.org/10.1111/j.1475-4762.2008.00837.x      URL      [本文引用: 1]      摘要

Smoking is an important and topical health issue in the UK. In Leeds, the target is to reduce smoking prevalence by 9 per cent by 2010. However, a key unknown is the likely distribution of smokers across the city. This paper aims to estimate individual smoking rates using microsimulation and hence evaluate the performance of established stop smoking services, e. g. how well do they serve local areas and to what extent does attendance vary between service point locations? Location-allocation modelling is employed to test various location scenarios and provide insight into where to optimally place centres.
[34] Lymer S, Brown L, Yap M, et al.

2001 Regional disability estimates for New South Wales, Australia, using spatial microsimulation

. Applied Spatial Analasis and Policy, 2008, 1(2): 99-116

https://doi.org/10.1007/s12061-008-9006-4      URL      [本文引用: 1]      摘要

Estimating disability levels in older Australians and their demographic and socio-economic profiles is essential for identifying the need for aged care services and for the development and implementation of effective social policy on ageing. Small area estimates are produced from the spatial microsimulation model ‘CareMod’ which is based on the 1998 ABS Survey of Disability, Ageing and Carers (SDAC; ABS (1999). Disability, Ageing and Carers: User Guide, Australia. Canberra: ABS), and up-rated to 2001. Estimates are generated by reweighting the SDAC confidentialised unit record file to create ‘synthetic’ datasets for each Statistical Local Area in New South Wales (NSW), Australia. Disability levels and the need for aged care in NSW was determined across a range of age groups. The results show that there is significant variation across NSW in disability levels and the need for aged care services by older persons. With increasing age, the presence of severe or profound disability increases and, as a consequence, this subgroup of aged persons has a greater potential need for high level care. The coastal areas of NSW have a greater requirement of care provision with respect to overall numbers, as a consequence of the greater overall population. However, many inland towns have greater rates of disability amongst the elderly. The research findings will assist in the strategic planning and improved targeting of aged care services, especially in identifying areas of unmet need at the small area level.
[35] Goulias K G.

Forecasting the impact of sociodemographic changes on travel demand: Experiments with a dynamic microsimulation model system

. University of California Transportation Center, 1992.

[本文引用: 2]     

[36] Bhat C, Guo J, Srinivasan S, et al.

Comprehensive econometric microsimulator for daily activity-travel patterns

. Transportation Research Record, 2004, 1894: 57-66.

[本文引用: 2]     

[37] Vovsha P, Petersen E, Donnelly R.

Microsimulation in travel demand modeling: Lessons learned from the New York best practice model

. Transportation Research Record, 2002, 1805: 68-77.

[本文引用: 1]     

[38] Veldhuisen J, Kapoen L, Timmermans H.

RAMBLAS: A regional planning model based on the microsimulation of daily activity travel patterns

. Environment and Planning A, 2000, 32: 427-444.

[本文引用: 1]     

[39] Hunt J D, Kriger D S, Miller E.

Current operational urban land-use-transport modelling frameworks: A review

. Transport Reviews, 2005, 25: 329-376.

https://doi.org/10.1080/0144164052000336470      URL      [本文引用: 1]      摘要

Various alternative frameworks are available for modelling urban land‐use–transport interaction. This paper provides a detailed review of six of these frameworks that have been or are currently being used to develop operational models. The intention is to indicate what is the general nature of the current state of practice and what is now available for practical modelling work in the area. The intention is also to compare the current state of practice with what might be the ideal in various respects. The six frameworks reviewed (ITLUP, MEPLAN, TRANUS, MUSSA, NYMTC‐LUM and UrbanSim) are considered in terms of their representations of physical systems, decision‐makers and processes, along with various more general modelling and implementation issues. None matches the ideal as envisaged here in all respects. However, a wide range of policy considerations can be handled explicitly with what is available, and more recent developments show an encouraging trend towards expansion in the scope of what can be considered. Further strengthening of the behavioural basis and relaxation of some of the more restrictive assumptions would appear to be both appropriate and likely in the future.
[40] Waddell P.

Modeling urban development for land use, transportation, and environmental planning

. Journal of the American Planning Association, 2002, 68(3): 297-314.

https://doi.org/10.1080/01944360208976274      URL      [本文引用: 1]      摘要

Large urban areas have come under intense pressure to respond to federal mandates to link planning of land use, transportation, and environmental quality; and from citizen concerns about managing the side effects of growth including sprawl, congestion, housing affordability, and loss of open space. Planning models used by metropolitan planning organizations were generally not designed to address these issues, creating a gap in the ability of planners to systematically assess them. UrbanSim is a new model system developed to respond to these emerging requirements and has now been applied in 3 metropolitan areas. This paper describes the model system and its application to the area of Eugene-Springfield, Oregon.
[41] Lovelace R, Philips I.

The ‘oil vulnerability’ of commuter patterns: A case study from Yorkshire and the Humber, UK

. Goeforum, 2014, 51: 169-182.

https://doi.org/10.1016/j.geoforum.2013.11.005      URL      [本文引用: 1]      摘要

Motorised transport infrastructure and increasingly specialised labour markets have resulted in energy intensive commuter patterns in many parts of the world. This is cause for concern due to the possibility of oil price shocks and the need to restrict the combustion of fossil fuels to minimise the chances of runaway climate change. The paper investigates methods to identify the spatial distribution and socio-demographic profiles of those who are vulnerable to high oil prices. It does this by use of four metrics of oil vulnerability which were developed using a spatial microsimulation model and applied to the case study region of Yorkshire and the Humber, UK. The metrics capture different aspects of vulnerability and highlight the importance of translating conceptual definitions into practical metrics. The geographically aggregated results coincide with the literature: rural areas are associated with the highest levels of vulnerability. The individual level results indicate that vulnerability can be prevalent even in seemingly resilient areas. Ultimately, we conclude that the social and spatial distribution of oil vulnerability depends on how an energy-constrained future is envisioned. Creating localised metrics about the future is a challenging task fraught with danger but could, provided that these metrics are interpreted with sufficient humility, aid the development of equitable policies to encourage resilience, as part of a worldwide transition away from fossil fuels.
[42] Yagi S, Mohammadian A K.

An activity-based microsimulation model of travel demand in the Jakarta Metropolitan Area

. Journal of Choice Modelling, 2010, 3: 32-57.

https://doi.org/10.1016/S1755-5345(13)70028-9      URL      [本文引用: 1]      摘要

The goal of the study reported in this paper was to develop a comprehensive activity-based modeling system in the context of developing countries, providing accurate estimates which are expected to serve as better inputs for evaluation of different transportation policy scenarios. The case study is Jakarta, Indonesia as one of the largest metropolitan areas in Asia. The modelling system primarily adopts a tour-based structure in which the tour is used as the unit of modeling travel instead of the trip, preserving a consistency in destination, mode, and time of day choices across trips.
[43] Ma J, Heppenstall A, Harland K, et al.

Synthesising carbon emission for mega-cities: A static spatial microsimulation of transport CO2 from urban travel in Beijing. Computers,

Environment and Urban Systems, 2014, 45: 78-88.

https://doi.org/10.1016/j.compenvurbsys.2014.02.006      URL      [本文引用: 2]      摘要

Developing low carbon cities is a key goal of 21st century planning, and one that can be supported by a better understanding of the factors that shape travel behaviour, and resulting carbon emissions. Understanding travel based carbon emissions in mega-cities is vital, but city size and often a lack of required data, limits the ability to apply linked land use, transport and tactical transport models to investigate the impact of policy and planning interventions on travel and emissions. Here, we adopt an alternative approach, through the development of a static spatial microsimulation of people’s daily travel behaviour. Using Beijing as a case study, we first derive complete activity-travel records for 1026 residents from an activity diary survey. Then, using the 2000 population census data at the sub-district level, we apply a simulated annealing algorithm to create a synthetic population at fine spatial scale for Beijing and spatially simulate the population’s daily travel, including trip distance and mode choice at the sub-district scale. Finally, we estimate transport CO2 emission from daily urban travel at the disaggregate level in urban Beijing.
[44] Ma J, Mitchell G, Heppenstall A.

Exploring transport carbon futures using population microsimulation and travel diaries: Beijing to 2030

. Transportation Research Part D, 2015, 37: 108-122.

https://doi.org/10.1016/j.trd.2015.04.020      URL      [本文引用: 1]      摘要

Evaluating transport policy for cities in developing countries is often constrained by data availability that limits the use of conventional appraisal models. Here, we present a new ‘bottom-up’ methodology to estimate transport CO2 emission from daily urban passenger travel for Beijing, a megacity with relatively sparse data on travel behaviour. A spatial microsimulation, based on an activity diary survey and two sample population censuses, is used to simulate, for Beijing’s urban districts, a realistic synthetic population, and their daily travel and CO2 emission over 2000–2010. This approach provides greater insight into the spatial variability of transport CO2 emission than has previously been possible for Beijing, and further, enables an examination of the role of socio-demographics, urban form and transport developments in contributing to emissions over the modelled period. Using the 2000–2010 CO2 emission estimates as a baseline, CO2 emissions from passenger travel are then modelled to 2030 under scenarios exploring politically plausible strategies on transport (public transport infrastructure investment, and vehicle constraint), urban development (compaction) and vehicle technology (faster adoption of clean vehicle technology). The results showed that, compared to the trend scenario, employing both transport and urban development policies could reduce total passenger CO2 emission to 2030 by 24%, and by 43% if all strategies were applied together. The study reveals the potential of microsimulation in emission estimation for large cities in developing countries where data availability may constrain more traditional approaches.
[45] Edwards K L, Clarke G, Thomas J, et al.

Internal and external validation of spatial microsimulation models: Small area estimates of adult obesity

. Applied Spatial Analasis and Policy, 2011, 4(4): 281-300.

https://doi.org/10.1007/s12061-010-9056-2      URL      [本文引用: 1]      摘要

Spatial microsimulation models can be used to estimate previously unknown data at the micro-level, although validation of these models can be challenging. This paper seeks to describe an approach to validation of these models. Obesity data in adults were estimated at the small area level using a static, deterministic, spatial microsimulation model called SimObesity. This model utilised both Census 2001 data and the Health Survey for England for 2004-2006. Regression analysis was used to identify the covariates that were the strongest predictors of obesity and these were used as the model input variables. The model was calibrated using regression and equal variance t-tests. Two methods of external validation were undertaken; aggregating obesity data to a coarser geographical level at which obesity data was available, and secondly using small area level cancer data for tumour sites known to be correlated to obesity. The output obesity data were mapped and statistically significant hot (cold) spots of high (low) prevalence of obesity identified. Both internal and external validation showed low errors, suggesting this was a satisfactory simulation. Statistically significant hot and cold spots of (simulated) obesity prevalence existed, even after adjusting for age. This paper emphasises three steps to validation of spatial microsimulation models: 1. Accurate simulations require strong correlations between the input and output variables; 2. It is essential to internally validate the models; 3. Use all means possible to externally validate the model.
[46] 柴彦威, 赵莹, 刘云刚.

城市地理学研究方法的进展与展望

. 中国科学院院刊, 2011, 26(4): 430-435.

https://doi.org/10.3969/j.issn.1000-3045.2011.04.010      URL      [本文引用: 1]      摘要

城市地理学研究方法受哲学思潮、相邻学科、技术进步等因素的多重影响。通过积极消化西方已有 的理论与方法,并分层次吸收与中国本土密切结合的内容.城市地理学研究方法建立起了多种路径,强调方法研究与应用案例结合,传统方法与前沿方法兼顾。城市 地理学思想与方法的系统研究,从学科体系上弥补了中国城市地理学偏重于经验与实证研究的缺陷,引入了行为主义、结构主义、后现代主义等新的方法论;从知识 共享与人才培养上,组织了方法论的主题研讨会,营造出积极的理论争辩氛围,形成以中青年学者为主的方法研究团队。未来中国的城市地理学研究方法,当以科学 主义与实用主义相结合,外生地理学与内生地理学相统一,积极推进地理资源共享平台与研究者网络建设。

[Chai Yanwei, Zhao Ying, Liu Yungang.

Research Progress and prospect of urban geography methodologies and methods

. Bulletin of Chinese Academy of Sciences, 2011, 26(4): 430-435.]

https://doi.org/10.3969/j.issn.1000-3045.2011.04.010      URL      [本文引用: 1]      摘要

城市地理学研究方法受哲学思潮、相邻学科、技术进步等因素的多重影响。通过积极消化西方已有 的理论与方法,并分层次吸收与中国本土密切结合的内容.城市地理学研究方法建立起了多种路径,强调方法研究与应用案例结合,传统方法与前沿方法兼顾。城市 地理学思想与方法的系统研究,从学科体系上弥补了中国城市地理学偏重于经验与实证研究的缺陷,引入了行为主义、结构主义、后现代主义等新的方法论;从知识 共享与人才培养上,组织了方法论的主题研讨会,营造出积极的理论争辩氛围,形成以中青年学者为主的方法研究团队。未来中国的城市地理学研究方法,当以科学 主义与实用主义相结合,外生地理学与内生地理学相统一,积极推进地理资源共享平台与研究者网络建设。
[47] 陈彦光.

地理学理论研究和科学分析的一般方法探讨

. 地理科学, 2009, 29(3): 316-322.

https://doi.org/10.3969/j.issn.1000-0690.2009.03.002      URL      Magsci      [本文引用: 1]      摘要

根据近年来地理学家和物理学家对城市发展和演化分析的有关成果,总结出地理理论研究方法的一般模式:第一步,根据观测数据建立经验模型;第二步,构造假设、建立理论方程并且求解,求解的结果要与经验模型及其参数一致;第三步,基于第二步的假设条件进行计算机模拟实验,模拟结果要与观测的结果一致.第一步用到的方法从地理学"计量运动"时期开始发展,至今已经比较成熟;第二步用到的方法发展多年,但进展较慢;第三步用到的方法目前正在快速发展,但没有与第二步的方法有机结合.强调的方法在于,将地理学的经验建模、理论解析和模拟实验有机结合起来,形成一个完整的、不同步骤相辅相成的研究程序.这一套方法在一定条件下可以推广到地理应用研究领域,推广之后的第二步可以大为简化,但第三步则会更加复杂.

[Chen Yanguang.

Exploring general research method of theoretical geography with three steps

. Scientia Geographica Sinica, 2009, 29(3): 316-322.]

https://doi.org/10.3969/j.issn.1000-0690.2009.03.002      URL      Magsci      [本文引用: 1]      摘要

根据近年来地理学家和物理学家对城市发展和演化分析的有关成果,总结出地理理论研究方法的一般模式:第一步,根据观测数据建立经验模型;第二步,构造假设、建立理论方程并且求解,求解的结果要与经验模型及其参数一致;第三步,基于第二步的假设条件进行计算机模拟实验,模拟结果要与观测的结果一致.第一步用到的方法从地理学"计量运动"时期开始发展,至今已经比较成熟;第二步用到的方法发展多年,但进展较慢;第三步用到的方法目前正在快速发展,但没有与第二步的方法有机结合.强调的方法在于,将地理学的经验建模、理论解析和模拟实验有机结合起来,形成一个完整的、不同步骤相辅相成的研究程序.这一套方法在一定条件下可以推广到地理应用研究领域,推广之后的第二步可以大为简化,但第三步则会更加复杂.

/