地球信息科学学报 ›› 2018, Vol. 20 ›› Issue (1): 68-78.doi: 10.12082/dqxxkx.2018.170224

• 地理空间分析综合应用 • 上一篇    下一篇

基于夜间灯光与LUC数据的川渝地区人口空间化研究

胡云锋1,*(), 赵冠华1,2, 张千力1,2   

  1. 1. 中国科学院地理科学与资源研究所,北京 100101
    2. 中国科学院大学,北京 100049
  • 收稿日期:2017-04-30 修回日期:2017-08-25 出版日期:2018-01-20 发布日期:2018-01-20
  • 通讯作者: 胡云锋 E-mail:huyf@lreis.ac.cn
  • 作者简介:

    *作者简介:胡云锋(1974-),男,江西赣州人,博士,副研究员,主要从事遥感监测与区域可持续发展评价研究。E-mail: huyf@lreis.ac.cn

  • 基金资助:
    国家重点研发计划项目(2016YFB0501502、2016YFC0503701);高分专项(00-Y30B14-9001-14/16)

Spatial Distribution of Population Data Based on Nighttime Light and LUC Data in the Sichuan-Chongqing Region

HU Yunfeng1,*(), ZHAO Guanhua1,2, Zhang QianLi1,2   

  1. 1. Institute of Geographic Sciences and Natural Resources, Chinese Academy of Sciences, Beijing 100101, China
    2. University of Chinese Academy of Sciences, Beijing 100049, China
  • Received:2017-04-30 Revised:2017-08-25 Online:2018-01-20 Published:2018-01-20
  • Contact: HU Yunfeng E-mail:huyf@lreis.ac.cn
  • Supported by:
    National Key Research and Development Program of China, No.2016YFB0501502, 2016YFC0503701;Key Project of High Resolution Earth Observation System, No.00-Y30B14-9001-14/16.

摘要:

高精度的人口空间分布数据是开展小尺度人口活动变化规律研究的关键数据。夜间遥感影像对于反映人类社会活动具有独特的能力,因而被广泛的应用于社会经济领域的空间数据挖掘。本研究以DMSP/OLS夜间灯光数据、NPP/VIIRS夜间灯光数据、常住人口统计数据、土地利用数据为数据源,在县级尺度上建立逐步回归模型,构建川渝地区人口空间分布数据;并随机抽取研究区内500个乡镇,以常住人口统计数据为真实数据,对人口空间化结果进行精度检验。结果表明:① 2种夜间灯光数据与人口均有较高的相关性,相关系数均在0.76以上,NPP/VIIRS夜间灯光数据与人口的相关性要高于DMSP/OLS,且受拟合模型的影响不大。② 与人口相关性较高的土地利用类型有多种,耕地、林地对人口空间分布也有影响,在建模时不应只考虑建成区。③ 在2种夜间灯光数据分别与土地利用与土地覆被数据(Land Use/ Land Cover, LUC)结合时,使用DMSP/OLS夜间灯光数据和NPP/VIIRS夜间灯光数据得到的逐步回归模型的复相关系数R2分别为0.796、0.817,模型拟合率较高,而基于NPP/VIIRS夜间灯光数据得到的人口空间化结果分辨率(500 m)相比DMSP/OLS提高了一倍(1 km),中心城区与周边城区人口密度变化更加自然,更符合实际人口分布情况。④ 与LUC数据结合时,使用NPP/VIIRS夜间灯光数据得到的人口空间化结果精度要高于DMSP/OLS夜间灯光数据,表明NPP/VIIRS夜间灯光数据相比DMSP/OLS更适用于人口数据空间化研究。

关键词: DMSP/OLS, NPP/VIIRS, 人口模拟, 逐步回归, LUC

Abstract:

Spatial distribution data of the population with high-precision is the important data to study the law of the variation of population activity at small-scale. Remote sensing images of nighttime light have the unique ability of reflecting human social activities. Thus, they were widely used in spatial data mining of the socio-economic field. In this study, DMSP/OLS nighttime light data, NPP/VIIRS nighttime light data, resident population data and land use data were used as data sources. Then, we used these data to build the stepwise regression at county scale. The spatial distribution data of population in Sichuan and Chongqing area were established based on the stepwise regression model. Finally, we took the resident demographic data of the randomly selected 500 townships as the practical data to assess the accuracy of spatial distribution data of the population. The analysis shows that: (1) both of the two nighttime light data have high correlation with the population. The correlation coefficients are both above 0.76. The correlation of NPP / VIIRS night light data and population is higher than DMSP / OLS. The fitting model does not change the results. (2) There are many types of land use that are highly relevant to population. Farmland and woodland can also affect the spatial distribution of population. Thus, built area should not be considered as the only type of land use for building the population distribution model. (3) When the two nighttime lights were combined with LUC( Land Use/ Land Cover), the complex correlation coefficient (R2) of the stepwise regression model using DMSP / OLS nighttime light data and NPP / VIIRS night light data is 0.796 and 0.817 respectively, and the model fitting rate is higher. Compared with the results based on DMSP/OLS (1 km), the spatial resolution of population based on NPP/VIIRS nighttime light data increases to 500 m. The change of population density is more natural from the central city to the surrounding urban area, and the population distribution is more real. (4) When combined with LUC data, the results obtained with NPP / VIIRS nighttime light data were more accurate than DMSP / OLS nighttime light data, indicating that NPP / VIIRS nighttime light data is more suitable for the research of spatial distribution of population than DMSP / OLS.

Key words: DMSP/OLS, NPP/VIIRS, population simulation, stepwise Regression, LUC