基于DMSP/OLS数据的城市中心城区提取新方法

近年来,DMSP/OLS夜间灯光数据已被广泛应用于城市空间格局和城市化过程、人类活动及效应、经济水平、生态环境影响、电力能源消耗量和灾害分析等领域。本文利用DMSP/OLS的稳定灯光数据产品,首先,以辅助资料的空间对比法确定城市中心城区范围最小阈值,然后,提取DMSP数据中的每个极大值及一定范围内的临近值,作为城市中心城区范围,并将提取的范围与作为辅助资料的MODIS土地覆盖数据中相应的城市中心城区范围作线性回归分析,进而获得估算城市中心城区范围的线性模型。经验证其结果预测吻合度达到82.06%,平均相对误差为8.47%,说明利用该方法提取城市中心城区范围具有一定的可行性,可较准确地提取发展水平不同的城市的中心城区范围。

A New Method of Extracing Areas of Center City Regions Based on DMSP/ OLS Data

In recent years, the DMSP/OLS nighttime light data have been widely applied to various fields such as urban space and urbanizations, human activity and effects, economic level, ecological effects, electric energy consumption and hazards analysis. In order to get the research results of high-precision, accurate center city regions should be generally obtained before researching. This study utilized DMSP/OLS stable lights imagery to estimate the areas of center city regions. Firstly, we set the minimum threshold of center city regions according to the spatial contrast method based on auxiliary data. Subsequently we extracted every maximum of urban nighttime light and the values which closed to each maximum in a certain range from the DMSP data. These regions should be corresponding to center city regions. However, because of the problem of“lighting overflow”and other influences, the results were not very accurate. Trying to make a correlation analysis about center city regions's areas between DMSP data and MODIS land cover data, the results showed that they were significantly related. Then an center city regions model could be established between them. Considering the different levels of urban development, there may be two different relationships between center city regions' areas which respectively extracted from DMSP data and MODIS land cover data. Therefore, cities were divided into two classes. The maximum of urban light in one class was equal to 63 and it was less than 63 in the other one. As a result, two different models were established between MODIS land cover data and DMSP data based on the maximum value of urban light, thus, it is feasible to accurately estimate the center city regions' areas by DMSP data. At last, the results of verification showed that there was a high percent of prediction fitness (82.06%) and the mean relative error was 8.47%, which showed the reliability of the linear models. So, this paper introduced a new method which could simultaneously and accurately extract the center city regions with different levels of urban development. This work is an important preparation for further studies in urban development and so on.