Deep convective clouds (DCC) play an important role in heavy rainfall and severe weather. However, limited availability of all-day cloud physical property data has hindered understanding of diurnal variations of DCC physical properties and their relationships with rainfall over South China and the adjacent seas. This study investigates diurnal variations of DCC physical properties and their relationships with rainfall using all-day cloud products from the DaYu Cloud Analysis System (DaYu–CLAS). The results show pronounced regional heterogeneity in DCC diurnal variations. Inland and eastern coastal DCC exhibit afternoon peaks, with the eastern coastal region showing larger diurnal amplitudes. Oceanic DCC are active from nighttime to early morning, whereas DCC on the southern flank of the Yungui Plateau exhibit bimodal patterns with early morning and afternoon peaks. Relationships between cloud physical properties and rainfall rate (RR) show similar patterns across subregions. Cloud optical thickness and cloud water path (CWP) are positively correlated with RR, cloud top height shows a nonlinear positive relationship, and cloud effective radius exhibits a threshold-like behavior near 29 μm. Based on these relationships, a machine learning model is established to estimate RR using cloud physical properties, meteorological variables, and geographic location. The model shows reasonable performance across all subregions (<em>R</em> > 0.64) and reconstructs general RR diurnal features, with CWP identified as the dominant predictor. These findings improve the understanding of DCC diurnal variations and demonstrate the potential of cloud physical properties for convective rainfall estimation and monitoring.