The nighttime light (NTL) data can capture light signals from urban buildings, road facilities, and vehicles. The NTL images in urban built-up areas have continuous spatial distribution and brightness value significantly higher than that in surrounding areas. Therefore, many studies utilize it for the urban built-up area extraction [1
]. Currently, the typical NTL images are provided by Defense Meteorological Satellite Program/Operational Linescan System (DMSP/OLS) and National Polar-Orbiting Partnership’s Visible Infrared Imaging Radiometer Suite (NPP/VIIRS) [7
]. Their spatial resolutions are 1 km and 500 m, respectively. However, the study of extracting urban built-up areas using NTL data only is limited to a large scale due to the low resolution, which makes it hard to obtain high-precision urban built-up area results on small and medium levels [9
]. Meanwhile, the area of the urban built-up area extracted from the NTL image is larger than the actual range due to the blooming effect [10
], and the extraction accuracy is low. Elvidge et al. [2
] also believe that the blooming effect is one of the main causes of overestimating the urban built-up areas.
In recent years, many studies [11
] combined multi-source data with NTL data to build urban indices for the urban built-up area extraction to improve its accuracy on small and medium levels. Several studies have demonstrated that abundant information on urban built-up areas can be obtained by utilizing multi-source data with different characteristics [13
]. The human settlement index (HIS), vegetation adjusted NTL urban index (VANUI), and enhanced vegetation index (EVI) adjusted NTL index (EANTLI) are the most broadly utilized urban indices. They assume an inverse relationship between vegetation and urban built-up area to mitigate NTL blooming effect and characterize urban built-up areas [15
]. Lu et al. [15
] proposed the HIS by combining NTL images with the normalized difference vegetation index (NDVI) images. HIS overcorrects the light signals in peri-urban areas. Besides, NTL blooming effect is still obvious in bare soil areas, where NDVI values are zero. Zhang et al. [16
] proposed VANUI combining NDVI and NTL, which enriches urban fringe information and is broadly utilized in urban built-up area extraction. However, VANUI has a limitation in the peri-urban areas where both vegetation values and NTL values are high. In addition, it is not suitable for some desert cities in North America regarding the unobvious relationship, between urban built-up areas and vegetation. According to the same principle, Zhuo et al. [17
] built the EANTLI by combining EVI with NTL. Compared with NDVI, EVI can weaken the effects of atmosphere and soil background on vegetation index. Therefore, it can promote the accuracy of the urban built-up areas extracted without the shortcomings of NDVI. However, the EANTLI value may be abnormally high especially for the mixed pixels in the water land boundary region, which increases the misclassification error.
Previous articles ignored combining NTL and point of interest (POI) data to build an index for urban built-up area extraction, while actually POI is positively correlated with urban built-up areas. POI is a kind of social sensing data produced by human activities and contains a wealth of location and attribute information. The abrupt changes of its density at the boundaries between urban and surrounding suburbs and rural regions make it easier to extract urban built-up areas [18
]. At present, the main method using POI is to set a threshold value for the kernel density of POI to obtain accurate urban built-up area results. Some studies have demonstrated that there is a good coupling relationship between NTL and POI, which has high consistency and broad applicability in the study of urban spatial structure [20
]. POI brings convenience in obtaining the boundary of urban built-up areas accurately and helps to make up the shortages of the low resolution and the blooming effect of NTL data. Therefore, this paper combined NTL and POI data to build a new index for urban built-up area extraction. It is urgent to clarify the effects of the combination on the extraction improvement. In addition to POI, this paper also introduces a natural remote sensing data, land surface temperature (LST). Many studies have shown that the land surface temperature is positively correlated with the distribution of urban land cover [22
]. The characteristics of urban built-up areas can be enhanced by LST from a natural perspective, which is different from the humanistic and social perspective of POI. Some [14
] chose the combination of NTL and LST to extract urban built-up areas. Among them, He et al. [24
] combined the NTL with LST to extract dynamic information of global urban expansion by the fully convolutional network. Further, Zhang et al. [25
] proposed the temperature and vegetation adjusted NTL urban index (TVANUI) for the purposes of characterizing urban built-up areas and reducing the blooming effect. Their satisfactory results proved that the combination of LST and NTL has great potential in improving the extraction accuracy of urban built-up areas. Therefore, this paper combined NTL with POI and LST, to establish the POI and LST Adjusted NTL Urban Index (PLANUI) for the study on the extraction of urban built-up areas.
The widely used NPP/VIIRS images with 500-m resolution were introduced into this study to explore the effects of PLANUI. Moreover, the Luojia 1-01 satellite was successfully launched in June 2018 and began to provide NTL images with 130 m resolution and 250 km width. This data greatly enhances the spatial resolution of NTL and was proved to have a great ability to extract urban areas [26
]. Therefore, this paper also introduced Luojia 1-01 images into the urban built-up area extraction experiments to verify whether PLANUI is suitable for NTL data with increasingly high spatial resolution.
The primary purpose of this paper is to propose the PLANUI that combines NTL images with POI and LST data to reduce the blooming effect of NTL and enhance urban built-up areas features for promoting the extraction accuracy. The secondary objective is to explore whether PLANUI is suitable for NTL with a high resolution like Luojia 1-01.