Network topologies are considered to be the fundamental factors that govern the network’s macroscopic fundamental diagram (MFD) and performance. Despite its significance, this domain remains underexplored. This study employed empirical data as evidence to examine the explicit connections between network topologies and the parameters of a specific MFD, in the form of a macroscopic Underwood’s model. Sixty-three 1 × 1 km dense urban networks were sampled in Hong Kong. The macroscopic Underwood’s models of the sampled networks were estimated based on taxi Global Positioning System (GPS) data. The results reveal that the network free-flow speed decreased with the average number of junctions per unit distance, which captures the frequency of network-induced stop-and-go vehicle motion. The network optimal density decreased with the degree density normalized by the trafficable area, which represents the intensity of the conflicts between traffic streams. A spatially variable macroscopic Underwood’s model was subsequently established.