地理学报 ›› 2005, Vol. 60 ›› Issue (5): 817-826.doi: 10.11821/xb200505013

• 区域发展 • 上一篇    下一篇

基于空间马尔可夫链的江苏区域趋同时空演变

蒲英霞1, 马荣华2, 葛莹1, 黄杏元1   

  1. 1. 南京大学城市与资源学系,南京210093;
    2. 中科院南京地理与湖泊研究所,南京210008
  • 收稿日期:2004-12-21 修回日期:2005-06-27 出版日期:2005-09-25 发布日期:2010-09-09
  • 作者简介:蒲英霞 (1972-), 女, 汉, 山东莒县, 讲师, 主要从事空间数据挖掘、GIS空间分析和区域分析等研究。 E-mail: puyingxia@yahoo.com
  • 基金资助:

    国家自然科学基金项目(40301038)

Spatial-Temporal Dynamics of Jiangsu Regional Convergence with Spatial Markov Chains Approach

PU Yingxia1, MA Ronghua2, GE Ying1, HUANG Xingyuan1   

  1. 1. Department of Urban and Resources Sciences, Nanjing University, Nanjing 210093, China;
    2. Nanjing Institute of Geography and Limnology, CAS, Nanjing 210008, China
  • Received:2004-12-21 Revised:2005-06-27 Online:2005-09-25 Published:2010-09-09
  • Supported by:

    National Natural Science Foundation of China, No. 40301038

摘要:

以江苏省77个县域1978~2000年人均GDP数据为资料,基于空间马尔可夫链方法,研究江苏省区域趋同的时空动态演变特征。首先按照全省人均GDP平均水平,将所有县域划分为低、中低、中高和高4种类型,计算其马尔可夫转移概率矩阵;然后将区域类型转移的空间分布格局加以可视化;最后,以每个区域在初始年份的空间滞后类型为条件,构造空间马尔可夫转移概率矩阵。结果表明:(1) 自改革开放以来江苏省一直存在"俱乐部趋同"现象,在1990~2000年期间更为显著。(2) 那些区域自身及其邻居同时向上转移的地区全部集中在苏南,而区域自身或邻居有一方或双方均向下转移的区域绝大多数位于苏北。(3) 江苏省区域人均GDP类型转移显著受到地理背景的制约。在1978~1990年和1990~2000年期间,一个落后区域以落后地区为邻时,其向上转移的概率分别为0.148和0.025,低于平均概率0.2和0.042;一个富裕区域以富裕区域为邻时,其向上转移的概率分别为1.0和0.991,高于平均概率0.987和0.984。这表明,发达的区域背景对区域转移起到了正面影响,而欠发达的区域背景则产生了负面影响,进一步为江苏省"俱乐部趋同"现象的存在提供了空间上的解释。

关键词: 区域趋同, 时空演变, 空间自相关, 空间马尔可夫链, 江苏省

Abstract:

Based on the per capita GDP dataset at the county level in Jiangsu province from 1978 to 2000, this paper attempts to apply spatial Markov chains to investigate the spatial and temporal characteristics of regional convergence in Jiangsu. Firstly, all the per capita GDP data in Jiangsu are classified into 4 different classes by annual provincial average. Due to the changes in the regional development strategies over time, the whole period is then divided into two sub-periods (1978-1990 and 1990-2000), and two Markov transition probability matrices for these two periods are estimated respectively for comparison. Secondly, two kinds of maps are accordingly made in order to visualize spatial patterns of class transitions, one for region, and the other for region and its neighbors. Finally, conditioning on each region's spatial lag at the beginning of each year, spatial Markov matrices for the two different periods are constructed. The conclusions are drawn as follows: (1) The process of regional convergence in Jiangsu has been globally characterized by "convergence clubs" since 1978, but this trend in the 1990s is sharpened and statistically different from that of the period from 1978 to 1990. (2) Those regions and their neighbors that both experience upward mobility are located in southern Jiangsu, while the regions or their neighbors that move downwardly are mostly found in northern Jiangsu. (3) Regional per capita GDP class transitions in Jiangsu are highly constrained by their geographical neighbors. If a poor region is surrounded by poor regions, the probabilities of moving upward for the periods from 1978 to 1990 and 1990 to 2000 decrease to 0.148 and 0.025 respectively, while they average 0.2 and 0.042 in traditional Markov matrices. It suggests that poor regions are negatively affected when surrounded by other poor regions. Conversely, if a rich region is surrounded by rich neighbors, the probabilities of moving upward for those two sub-periods increase to 1 and 0.991, while they are 0.987 and 0.984 on average in traditional Markov matrices. It suggests rich regions are positively influenced by other rich regions being surrounded. These empirical analyses provide a spatial explanation to "convergence clubs" detected in traditional Markov method.

Key words: regional convergence, spatial-temporal dynamics, spatial autocorrelation, spatial Markov chains, Jiangsu province