NPP-VIIRS夜间灯光数据的县级电力消费空间化研究

Spatialization Research of Electricity Consumptions at County Level Using NPP-VIIRS Nighttime Lights

以NPP-VIIRS夜间灯光数据和Landsat-8影像为数据源,通过CART算法将Landsat-8影像分为人工表面地类和非人工表面地类,为非人工表面地类对应的夜间灯光区域赋值0,将人工表面地类对应到夜间灯光区域,并求取各个县级行政区域内的灯光值总和,结合廊坊市的电力消费统计数据建立夜间灯光总量与电力消费量之间的二次多项式回归模型。根据模型生成廊坊市500 m格网大小的电力消费空间分布图。研究结果表明:模拟的各县电力消费量的绝对误差和相对误差较小,相对误差最大为0.03%,这说明夜间灯光结合Landsat-8土地利用分类方法模拟的县级电力消费量的精度较高。全局Moran's I指数为0.63,表明廊坊市电力消费在空间分布上呈现较显著的聚集性;局部空间自相关分析表明廊坊市的电力消费具有一定的差异性。

Taking NPP-VIIRS nighttime lights and Lansat 8 images as data sources,Landsat 8 images were classified into artificial surface and no-artificiality surface classes using CART algorithm. The values of nighttime lights corresponding to no-artificiality surfaces classes were assigned as zero while total nighttime lights in each county were counted coincided with artificial surfaces. Then total nighttime lights combining with electricity statistical data in each county were conducted for making correlation analysis,and then optimal quadratic polynomial mode were used as regression model. Based on electricity consumption value calculated with this regression model in each grid,a spatial distribution map of electricity consumption in Langfang was generated with resolution of 500 m. The results showed that absolute errors and relative errors were very small where max relative error was merely 0.03%,which demonstrated that the approach of nighttime light data fusing Landsat 8 images to simulate electricity consumption values had higher accuracy. Global Moran＇s I Index with 0.63 indicated that spatial distribution of electricity consumption presented relative significant aggregation features when local spatial autocorrelation results indicated certain divergences of electricity consumption in Langfang.