Mixed accuracy of nighttime lights (NTL)-based urban land identification using thresholds: Evidence from a hierarchical analysis in Wuhan Metropolis, China

Identifying and monitoring urban land is essential for sprawl management. The use of nighttime lights (NTL) data has been reported as a suitable approach for identifying urban land across large regions, but the accuracy of urban land classification using these data is seldom discussed, particularly in small- and mid-sized cities. This paper provides a hierarchical framework for analyzing the accuracy of several DMSP/OLS- and NPP VIIRS-based NTL metrics at three nested levels (the overall Wuhan metropolis [WHM], nine cities comprising WHM, and 36 counties comprising the nine cities) using threshold approaches. Comparative analyses show mixed results, ranging [59.72%, 99.79%] and [0%, 83.96%] for map- and class-level accuracies, respectively, at the three nested levels. Moreover, NPP VIIRS is generally superior to DMSP/OLS for classifying urban land across the entire WHM and most cities/counties. Findings suggest map-level accuracy (over 95%) may be overinflated for certain NTL-based metrics, as the metrics produced relatively low class-level accuracies, around 60%. Pass time, spatial resolution of the data product, and certain situations (toll and railway stations, construction sites, less developed urban areas, and reflective surfaces near urban areas) are demonstrated as notable factors impacting NTL-based urban land identification. The findings from this study contribute to a better understanding of the appropriateness of using these metrics for urban land identification in different cities/scenarios and the development of more formalized frameworks for assessing applications in large-scale regions.