地理研究 ›› 2008, Vol. 27 ›› Issue (6): 1243-1252.doi: 10.11821/yj2008060003

• 土地资源与利用 • 上一篇    下一篇

城市植被覆盖变化可预测性及其空间格局的定量递归分析——以深圳市为例

李双成1, 赵志强1,2, 高 阳1, 王仰麟1,2   

  1. 1. 北京大学城市与环境学院资源与环境地理系 地表过程分析与模拟教育部重点实验室,北京 100871;
    2. 北京大学深圳研究生院,城市人居环境科学与技术重点实验室,深圳 518055
  • 收稿日期:2008-03-31 修回日期:2008-08-18 出版日期:2008-11-25 发布日期:2008-11-25
  • 作者简介:李双成(1961-),副教授,河北平山人。主要从事于地表格局与过程复杂性计算与模拟。 E-mail:scli@urban.pku.edu.cn
  • 基金资助:

    国家自然科学基金重点项目(40635028);国家自然科学基金项目(40771001)

Determining the predictability and the spatial pattern of urban vegetation using recurrence quantification analysis: A case study of Shenzhen City

LI Shuang-cheng1, ZHAO Zhi-qiang1,2, GAO Yang1, WANG Yang-lin1,2   

  1. 1. College of Urban and Environmental Sciences, Peking University, Beijing 100871,China;
    2. The Key Laboratory for Environmental and Urban Sciences, Shenzhen Graduate School, Peking University, Shenzhen 518055, China
  • Received:2008-03-31 Revised:2008-08-18 Online:2008-11-25 Published:2008-11-25
  • Supported by:

    国家自然科学基金重点项目(40635028);国家自然科学基金项目(40771001)

摘要:

由于人口的快速增加和经济要素的大量聚集,城市是下垫面性质改变最为显著的区域。本文应用定量递归分析技术对1999~2006年逐旬的1km SPOT-VEGETATION NDVI数据进行分析,以期揭示NDVI时间系列的非线性动态特征及其空间格局。研究结果表明,深圳市的NDVI时间系列的RQA特征值居于随机系列和确定性系列之间,属于含有随机成分的确定性系列。不同覆被类型下NDVI系列的特性不同,林地区的NDVI系列规则性最高,农地区次之,建设用地区最差,表明人类活动作为一种噪声对于NDVI系列性质的扰动。用二阶Rényi熵作为NDVI系列的可预测性指标,其空间格局表现为海拔高、坡度大、林地分布集中的西北和东南部可预测性高,而海拔相对低、建设用地相对集中的中南部地区可预测性低。

关键词: NDVI, 可预测性, 定量递归分析, 深圳市

Abstract:

The property of underlaying surface in urban area is being significantly changed due to rapid increase of population and accumulation of large amount of economic elements. It is important to analyze the variation of urban vegetation and its predictability for scientific planning of the ecological landuse. In this paper recurrence quantification analysis (RQA), an extension of recurrence plots (RPs), was used to measure the predictability of Normalized Difference Vegetation Index (NDVI) series and the spatial patterns by using the SPOT-VEGETATION NDVI data with 10-day temporal resolution and 1 km spatial resolution from 1999 to 2006. The results indicate that all indices of RQA of NDVI series in the study area lie between the stochastic series and deterministic series, suggesting that the property of NDVI series in Shenzhen City belongs to deterministic chaotic time series. The nonlinear properties of NDVI series differ in different landuse and landcover types, and are generally characterized by the highest regularity in woodland, higher regularity in cropland, and lower regularity in the built-up area, inferring impacts of human activities being as a gauss white noise on the NDVI series. Moreover, the second order Rényi entropy, K2, was proposed to indicate the long-term predictability of NDVI time series. The statistical values of K2 in the whole study area are: maximum value 0.76, minimum value 0.32, mean value 0.60, and standard deviation 0.06. The analysis of spatial autocorrelation in ArcGIS 9.2 suggests that K2 shows a better performance for the discerning of the spatial differentiation of K2, resulting from different landuse and landcover. On the whole, spatial distribution of K2 exhibits significant regional differentiation, characterized by lower k2 values, i.e., higher predictability in the northwest and southeast region with higher elevation and dense vegetation cover, and higher k2 values, i.e., lower predictability in the middle and southern residential area, suggesting that human disturbance may be the major driving factor for causing non-stationary dynamic characteristics of NDVI series. The combination of RPs and Geographical Information System is a useful approach not only in displaying the spatial patterns of RPs computing results but also in analyzing relationships among the influencing factors.

Key words: NDVI, predictability, recurrence quantification analysis, Shenzhen