3.4 Interactive effect between 3D building metrics
Among the building metrics, the GeoD factor detector demonstrated that three-dimensional fractal was the most influential factors, followed by building density (q = 0.186) and building congestion degree (0.176), then building height (0.153) (Table 3). The results were inconsistent with BRT analysis results, which also suggested occlusion effects in BRT model.
The GeoD interaction detector indicated that enhancement effect exist between building metrics. The interaction effect of building congestion degree with building density (BCD\(\ \cap\ \)BD) was greater than that from the two sub-factors individually. Here, ‘\(\cap\)’ represents interactions. These finding indicated that the occurrence of urban flooding would be increased when the building congestion degree included building density. For building coverage ratio and building height, the interaction effect was double-factor enhancement, which suggested urban flooding risks would be enhanced when building coverage ratio included building height. The interaction effect was nonlinear between building height and three-dimensional fractal. It means that the explanatory power of the GD model would not increase when adding both into the model.