Monsoon rains are an important fresh water supply for agricultural activity, while extreme rainfalls during a monsoon season frequently cause flash floods. In this study, a nonlinear causation measure of event synchronization is used to set complex networks of extreme rainfall during the Australian summer monsoon (ASM) development between 1st November and 1st March. We adopted Tropical Rainfall Measuring Mission-based satellite rain rate estimates from 1998 to 2015. Examining several standard network centrality measures, such as degree and local clustering, we revealed the multiscale nature of ASM development, which previously was only studied by weather analysis methods. The land-sea contrast in surface heating critical for ASM is depicted clearly by the degree centrality. In addition, both the clustering coefficient and the community structure show critical change in spatial pattern matching with the climatological average onset time of the ASM during late December. The former is likely related to the interaction between synoptic forcing and mesoscale convection during monsoon onset, resulting in characteristic changes in the rainfall field. One of the network communities also extends spatially during the onset, revealing critical information from the near-equatorial region to ASM and would be applicable to monitor monsoon development. Results from this study further support that network measures as defined by a single parameter of rainfall have enormous potential for monsoon onset prediction.