4.2 Interaction effect between 3D landscape metrics
Although numerous studies conduct multivariate analysis in urban flooding by using traditional statistical methods (i.e., pearson correlation, logistic regression, partial least squares, structural equation model) or machine learning method (i.e., random forest), fewer studies explore the interaction effect among explanatory factors (Ma et al., 2022). Urban morphology was complex and high heterogeneous, especially in the vertical direction. Existing 3D landscape metrics was still insufficient to analysis the urban canopy. Thus, it was necessary to explore the interaction effect among 3D landscape metrics. For example, more and aggregation impervious surfaces tend to rise the land surface temperature in urban heat island field. However, high-rise building can mitigate heat island effect through improve surface roughness and cast more shadows (Sun et al., 2020). How to simultaneous adjust the impervious surfaces and building height became vital to alleviate heat island. In our study, we revealed that enhancement effect exist between 3D building metrics. In particularly, when the 3D fractal as an given value, it is important to reduce 3D shape index to avoid urban flooding risks.