【作者】 刘劲松；

【导师】 许清海；

【作者基本信息】 河北师范大学 ， 生态学， 2009， 博士

【摘要】 地理学和景观生态学研究领域,尺度问题倍受关注,相关工作主要集中在尺度效应、尺度转换和尺度选择等方面。人口密度,作为研究人口分布的常用测度指标,是开展多尺度人地关系研究的理想视角。地理学界大致有三方面人群关注人口密度研究。(1)人口地理学领域：重视从自然环境、社会经济和人口政策方面探讨人口密度的成因,同时从历史的角度探讨影响人口密度的因素,进而揭示人口地理分布的区域差异；人口密度统计多以县级以上行政区为单元。(2)地图学领域：新近工作主要集中于编制基于经纬格网的人口密度图。(3)地理信息系统领域：侧重构建人口密度模型。纵观国内外人口密度研究,人口地理学较少关注人口密度的尺度问题,尽管地理信息系统已开始关注多尺度人口密度,建模过程中也考虑了地理要素的影响,但常忽视制约不同尺度人口分布的地学机理。地图学多编制单一尺度的人口密度图,多尺度人口密度图较少,也较少关注人口密度与地理环境的关系。这些都束缚了多尺度人口密度真实内涵的理解和运用。尺度推绎中的可塑性面积单元问题(MAUP),学术界存在两种看法：(1)地球信息科学领域认为MAUP是统计学谬误或生态学谬误,并认为MAUP是空间分析的八大障碍之一。(2)生态学家为了准确、全面地认识地理格局和生态过程,认为MAUP不容忽视,并认为MAUP包括尺度效应和划区效应。Openshaw和Taylor认为应重视MAUP的地理学意义,而不应把其视为纯粹的统计学或数学问题,并认为MAUP是研究空间格局的重要突破点。但受专业背景限制,生态学家未能找到开展MAUP研究的量化指标,制约了MAUP研究进展。国内学者在讨论MAUP的时候,不约而同地选择了人口密度这一指标,说明人口密度是研究MAUP较为可行的量化指标,但较少关注其背后隐含的地学问题。本研究以石家庄地区为例,基于人口密度MAUP的尺度效应和划区效应,获取多尺度人口密度；利用多元统计分析,明确多尺度人口密度间的协同变化关系,遴选人口密度特征尺度,确定制约特征尺度人口密度的主要因子,刻画人口密度等级结构体系,寻找特征尺度人口密度影响因子与人类行为的对应关系,揭示隐藏在人口密度MAUP背后的地理学规律,增强人地关系研究中对“人类行为与地理环境关系”的认识和理解。运用主导数据库、多元统计分析等方法,开展多尺度人口密度的定量研究。人口密度主导数据库包括两个技术环节：(1)编制最小粒度人口密度图,(2)编制人口密度尺度推绎模型。编制最小粒度人口密度图过程中,依次编制了县、乡、村、街区人口密度图,并借助街区人口密度获得了最小粒度人口密度图(R=0),该图网格大小为100米×100米,图中人口密度大于200000人/平方公里的格子仅有12个,说明空间统计误差比重甚微,而且此类误差在人口密度尺度推绎中未得到传播,说明最小粒度人口密度图为多尺度人口密度研究提供了可信的数据基础。多尺度人口密度的形式化表达借助Focalmean函数,分别利用圆形、方形滤波算子编制尺度上推模型,获得了R=0-99和1×1-199×199两个系列的多尺度人口密度图。在多尺度人口密度基础上,开展特征尺度遴选,编制最佳因子法、Sheffield指数法和标准Sheffield指数法等3种特征尺度遴选程序,尝试定量化的特征尺度遴选。CI指数可以作为特征尺度遴选的定量化模型,其中参数P是特征尺度遴选的关键控制参数。以CI指数的计算结果为线索,结合尺度相关分析,辅以目视判读,遴选出5个特征尺度,即R=0、R=3、R-12、R=29、R=99,初步分析了特征尺度人口密度的制约因素。多尺度人口密度计算比较发现,基于圆形滤波因子的尺度上推模型优于方形滤波因子的尺度上推模型。上述两尺度上推模型均能保证多尺度人口密度的数据一致性,且随着尺度增加,两序列的人口密度标准差均按幂函数方式急剧降低,说明随着尺度增大,人口密度的空间变异逐步降低、空间异质性逐步降低。相关分析显示,R≥30的各尺度人口密度相关性明显增强,尺度R=99的人口密度表现出区域共轭特征。特征尺度人口密度影响因子分析表明,(1)人口密度与格局、过程存在尺度依赖关系。其中,人文因素在小尺度(R=0)上对人口分布发挥主导作用,如：新兴商业居住区人口密度均大于80000人／平方公里,这不仅满足了开发商的利润追逐,也考虑了消费人群的生活成本。(2)人口聚居与党政机关、企事业单位的空间分布存在较大关联,省、市、区政府周边地区人口密度较高,介于5-8万人／平方公里,企业生活区的人口密度差异明显,华北制药厂周边地区人口密度达5-6万人／平方公里,棉三、棉四等传统企业周边人口介于3-5万人／平方公里。这些均属于计划经济体制下的人口分布特征。京广铁路沿线周边地区(中华大街与平安大街之间)是石家庄的主商业中心,但这里人口密度最低,只有1-2万人/平方公里。该区白天人口多、流动强度大,与夜间人口密度形成明显反差,说明户籍人口密度仅反映人口的居住格局,不反映人口的工作格局。现阶段城市人口居住地与工作地有分离之趋势,人口工作地倾向于物质流便捷、信息通畅的区域,人口居住地倾向于物业好、成本低、环境质量美的区域。(3)自然和区位因素在各个尺度上对人口格局具重要控制作用,如：河流影响着中小尺度(R=3、11)的人口密度,地貌格局控制着较大尺度(R=29、99)的人口密度。(4)河流、地貌等影响因子对人口分布的影响在各尺度有不同程度体现。(5)河北省历史人口密度分析表明：元代是现今人口发展的起点,通过现代人口密度,很难看到元代以前的人地关系痕迹,现代人口密度图中人地关系痕迹的时间跨度当不超过700年。(6)特征尺度人口密度是人地关系尺度效应的集中体现,是人地关系矛盾制衡的历史结果。矛盾制衡是人口密度与格局、过程产生尺度依赖的根本原因,是形成人口密度MAUP的根本原因。石家庄地区人口密度影响因子分析表明,人口密度MAUP中不仅包含自然行为,也体现人类意志。与其消除看似谬误的MAUP,不如借助MAUP开展多尺度人口密度的定量研究,认清不同尺度人地关系的主要矛盾,协调不同尺度人地矛盾冲突,优化区域配置。本文是对人口密度MAUP地理学解释的初步探讨,仍有许多问题值得进一步商榷。(1)尺度推绎模型的数学基础有待完善。目前的尺度推绎是假设在欧氏空间中各点、各向同性,且尺度均衡。该假设在小区域、低精度下可近似成立,但在国家、洲际尺度尚显不妥。由于欧氏空间的任何对象都对应一个整数维,这就意味着欧氏空间的尺度变换是简单可逆的。但现实世界是分数维,用欧氏空间描述地学空间实体的尺度变换,无法实现简单可逆。地图代数、球面离散网格、分形几何、划区效应等领域的研究进展将改善人口密度尺度推绎模型的数学基础。(2)人口密度尺度上推属于相邻尺度人口密度推绎模型。是否可以用于人口密度跨尺度推绎?是一个悬而未决的问题,模型的尺度适用范围值得进一步推敲。(3)尺度上推是由复杂到简单的过程,故人口密度尺度上推模型较易构建。利用贫乏的大尺度信息反推出复杂的小尺度信息并不容易,尺度下推模型是尚待解决的难题。人口密度尺度上推实现了对人口格局的变焦观察,但人口密度尺度推绎还无法实现自动变焦功能,深入开展尺度下推的约束机制研究,对揭示尺度转换重要节点的变异机制至关重要。努力尝试将环境要素格局及过程等约束条件融入尺度下推模型,对现代地理科学的创新性研究及其重大应用产生积极的推动作用。(4)能否以特征尺度为基点,明确给出人地系统各相关要素统一的大尺度、中尺度和小尺度概念,为跨学科研究提供层次、标准统一的尺度模板?有待不同学科特征尺度研究进展。(5)有待开展更大尺度(全省或华北地区)人口密度尺度上推工作,探讨季风气候背景下,人口密度是否与大气流场、温度、降水等宏观因素存在尺度依赖关系。更多还原

【Abstract】 Scale was paid much attention in studies of geography and landscapes. Most works are focused on scale effects, scale transform and scale selections. Population density, as a common-used index of population distribution, was an ideal view point in study of multi-scale relations between human and environment.There are three groups in geography paying attentions to the population density:(1) is in the group of population geography. They focuse on natural environment, social economy, population policy and history to discuss the cause and factor of population density, and to reveal the regional difference of population distribution. The statistic of population density was usually based on the unit of county or upper district. (2) is in the group of cartology. They fasten on the population density based on longitude-latitude grid in recent works. (3) is in the group of Geographical Information System. They think much of making population density models.Reviewing all population density studies of native or foreigners, the scale of population density was not paid much attention in population geography. Despite multi-scale population density was started to be considered in research works of Geographical Information System, and geographical factors were also thought of in modeling. But, the geographical mechanism of controlling multi-scale population distribution was always ignored. Single-scale population density maps were more than multi-scale ones and the relationship between population density and geographical environments were disregarded in cartology. These limited the understanding and application of the real meaning of multi-scale population density.There were two opinions for the Modifiable Areal Unit Problem(MAUP) in academic field. (1) Geographical information system thought MAUP is a statistical or ecological error and being one of the eighth obstacles in space analysis. (2) Ecologists thought MAUP was unignorable in order to realize geographical pattern and ecological process more accurately and comprehensively. They suggested that MAUP should include scale effect and zoning effect. Openshaw and Taylor thought MAUP shouldn’t be considered pure statistical or mathematical problem, its geographical meaning should be thought much of and MAUP were the breakthrough point in studying spatial pattern. But for the constraint of speciality background, ecologists couldn’t find the quantized index and restricted the development of MAUP study. Native population geographical researchers all selected the index of population density in discussing the MAUP, which meant that population density was feasible quantized index, but the geographical problem implied behind was concerned little.In this paper, a case in Shijiazhuang district, multi-scales population densities were obtained basing on the scale effect and zoning effect of population MAUP. The cooperation relationship between multi-scale population densities were defined using multiple statistic analysis, the major factors controlling population densities were determined by selecting characteristic scales, and the impacting factors of characteristic scale population density and its relationship to the human actions were shown in grading framework system of population density. Finally, the geographical rule behind in the MAUP of population density was found out.Dominant database and multiple statistic analysis were used to study multi-scale population density quantitatively. Population density dominant database included two main technique steps:(1)compilating minimum grain map of population density (2) compilating calculation model of population density. During the course of compilating minimum grain map of population density, population density maps of county, township, village and district were compiled successively and minimum grain population density map was gained by means of district population density (R=0). There were only 12 grids whose population density were more than 200 000/km2 in the grid of 100m×100m map, which means that spatial statistic error was slight and the error wasn’t spread in scaling of population density. Minimum grain population density map supplied credible data basis to study of multi-scale population density. Focalmean Function was applied to formal description of population density. Circular and rectangle filtering operator were used to compile up-scaling model and two series of multi-scale population density maps from R=0 to R=99 and from 1 X 1 to 199X199. Basing on the multi-scale population density, the characteristic scales were selected quatitatively by optimal coefficient, Sheffield index and standard Sheffield index programme. CI index could be used as quantitative model of characteristic scale selection, in which parameter P was the key in characteristic scale selection. Basing on CI index, combining scale correlation analysis and with the aid of visual interpretation, five characteristic scales were selected (R=0, R=3, R=12, R=29,R=99) and impact factors of characteristic scale population density were analyzed initially.Comparison of multi-scale population density showed that the calculation model basing on circular filtering wave was better than that on rectangle filtering wave. These two up-scaling models all could ensure the consistency of multi-scale population density and as the scale expanding, the standard error of both series of population density declined sharply by the means of exponential function, which meant that as the scale enlarging, the spatial heterogeneity of population density decreased gradually. Correlation analysis showed that the relationship among all scales of R≥30 enhanced obviously, while the population density of R=99 appeared regional conjugation.Analysis of impact factors of characteristic scale population density shown that (1) there was scale-dependent correlation between population density, population pattern and population process, in which artificial factors played dominant roles in population distribution on small scale(R=0), for example, the population density of new business-inhabitant district was greater than 80000 person/km2, developers were fulfiled with gain and consumers were acceptable life cost; (2) there was close correlation between human habitant and official, enterprise and institution distributions, population densities were high in adjacent areas of province, city and district officials, where population densities were 50 000 to 80 000 persons per square kilometer. Population densities were different evidently in habitant districts of factory, there are 50 000 to 60 000 persons per square kilometer around the areas of Huabei pharmaceutical factory, while 30 000 to 50 000 persons per square kilometer around the areas of the 3th and the 4th cotton mills. All these population distribution were the characteristics in the planning economy system. The area along the Beijing-Guangzhou Railway (between Zhonghua and Pingan street) was the main commercial center, but the population density was only 10 000 to 20 000 persons per square kilometer. Population in this area was higher as massive flowed population in daytime, and was evident contrast to low population in nighttime, which shown that registered population only reflected habitant pattern, didn’t reflect employ pattern. At present, resident areas are forward to separate with working places. Working places are forward to select in the areas of convenient matters and informations, while residential area prefered to choose good property, low cost and graceful environment. (3)Each population pattern was controlled by natural and location factors, e. g. the population of middle and small scale(R=3、11) were influenced by rivers, and the population of large scale(R=29、99) was controlled landforms. (4)The impacts of river and landform on population distribution embodied in different scales. (5) The historical data of population density in Hebei province showed that the Yuan Dynasty was a start of present population pattern. Human marks to the environment couldn’t be found in modern population patterns before the Yuan Dynasty. (6) Characteristic scale population density embodied the scale effect of human and environment, and was the result of historical development. Contradiction balance was the basic reason causing scale-dependence of population density, pattern and process, which was also the basic reason forming MAUP of population density. The impacting factors analysis of population density in Shijiazhuang city revealed that MAUP included not only the natural characters, but also included the human behavior.This paper discussed the population MAUP preliminaryly and many problems remained to study further.(1) Mathematics base in scaling model need improvement. Hypothesis of present scaling was based on the points and the directions homogeneous in Euclid space and on the scales equilibrium. This hypothesis could be approximately tenable in small region and low precision, but it was incorrect in nation and continent scale. Any object corresponded to an integer dimension in Euclid space, which meant that the scaling was reversible in Euclid space. But the real world was fractional dimensions; scaling description of geographic space with euclid space couldn’t realize the simple reversion. The progress in map algebra, sphere disperse grid, fractal geometry and zoning effect would improve the mathematics base of scaling.(2) Up-scaling of population density belonged to adjacent-scale scaling, whether it could be used to cross-scale scaling was a problem in suspense, and the application scope need to be discussed.(3) Up-scaling was the process from complex to simple, so the up-scaling model of population density was easy to construct. But it was difficult to infer complex small-scale information from simple large-scale information, so down-scaling was a difficult problem to be solved. Up-scaling of population density realized the variable-focus observation of population distribution, but population density scaling couldn’t realize automatic focusing. Further research of controlling mechanism of down-scaling was important to reveal key-point variation mechanism of scaling. Trying to combine the pattern and process of environment into down-scaling model would promote modern geography innovation and its application greatly.(4) Basing on characteristic scale, if large, medium and small scale could be endowed special scale meaning in different factors of human-environment correlation system, providing uniform level and standard scale template? This need research progress of different subject.(5) Up-scaling of population density in lager-scale (Hebei province of Huabei region) need to be carried out, probing into the scale-dependent correlation of population density and macroscopic factors such as atmospheric flow field, temperature and precipitation in the monsoon background.更多还原