Are urban landscape patterns associated with human activity intensity? Evidence from nighttime light data

Human activity is a key indicator of urban vitality. However, the spatially dependent and nonlinear relationships between urban landscape patterns and human activity intensity remains an underexplored area of research. Using the Luojia1-01 nighttime light data as a proxy for human activity intensity and fine-scale 3D building information to quantify urban landscape patterns, this study applied both the multiscale geographically weighted model (MGWR) and the extreme gradient boosting regression model (XGBoost) to investigating such relationships. Wuhan, the largest metropolis in Central China, was selected as the study area for the application of our methodology. The prediction results identified road area density and floor area ratio as the dominant factors affecting Nighttime light (NTL) intensity, while building height, building plan area ratio, and building volume ratio had minimal impacts. The same landscape metric in different built environments might have distinct impacts on NTL intensity, especially for the spatial dominance metrics like the 2D and 3D largest patch index. However, discrepancies were found between the two models, concerning the importance level and effect directionality of road length density and facade area ratio. This was due to the linear assumption adopted by MGWR, which might underestimate feature importance and obscure threshold effects when employing large bandwidths. Our findings highlighted significant impacts of urban landscape patterns on human activity intensity. The relationships tended to follow a simple nonlinear pattern, which were more effectively captured by the XGBoost model.