Evolution of Gradual and Abrupt Trends in Nighttime Lights and Responses to Land Drivers via BFAST01 and Geographically Weighted Regression

Abstract

Nighttime lights (NTLs) provide an unparalleled view for understanding urban environments. The high-frequency NTL intensity changes (NTLICs) and their land drivers in megaregions are less investigated. Here, monthly Visible Infrared Imaging Radiometer Suite-Day Night Band (VIIRS-DNB) NTL images from 2014 to 2020 were regarded as time series modeling input data. Furthermore, the 7-year mean values of five types of Points of Interest Density (POID), Road Density (RD), Normalized Difference Vegetation Index (NDVI), and Normalized Difference Built-up Index (NDBI) were used to extract features as land drivers of NTLIC. A simplified version of the Breaks For Additive Season and Trend (BFAST01) algorithm was adopted to detect major changes in monthly NTL intensity at the pixel level in the Beijing–Tianjin–Hebei (BTH) megaregion, China. Geographically weighted regression (GWR) was used to address the spatial nonstationarity of the land drivers. The results showed that (1) increasing trends (63.35%) and decreasing trends (4.57%) were mainly observed in central and coastal cities, respectively. This indicates that the spatial characteristics of the NLTIC were unbalanced; (2) reversal change type 8 (37.29% of abrupt changes) was the dominant type, mainly occurring around April 2017, which may be related to the non-capital function redistribution policy; (3) NDVI and RD were the best one of two physical drivers and six human activity drivers, respectively; and (4) the R2 of the GWR reached 0.60, a 20% improvement over ordinary least squares regression. This study provides an insightful understanding of the dynamics of NTL intensity and its response mechanisms in megaregions.

Keywords

• BFAST01
• geographically weighted regression (GWR)
• land drivers
• nighttime lights (NTLs)
• urbanization