Urban agglomeration has become the most salient feature of global urbanization in recent decades, while spatiotemporal patterns and evolution remain poorly understood in urban agglomerations, which limit the decision-makers to make more informed decisions to improve the regional environment. Here we selected one of the most rapidly urbanized regions in the world – Pearl River Delta Metropolitan Region (PRDR), located in southern China, as the case. Landsat images spanning from 1995 to 2015 were used to retrieve land surface temperature (LST). Four types of regional heat island (RHI) degree were defined for further analysis. Then multi-scale spatiotemporal patterns and characteristics of RHI were identified with the help of cloud-based computing, spatial and landscape analysis. We found that (1) traditional urban heat island (UHI) appears as an RHI on an urban agglomeration scale. In PRDR, we found RHI expended with increasing connectivity, especially in the estuary areas where isolated UHI gradually merged during the rapid urbanization. (2) The contribution of main cities in PRDR to RHI and the evolutionary trends and pattern, which is changed from a west-east to a southwest-northeast gradient, have been revealed. (3) Considering the scale effect and different RHI categories, we revealed that during the urbanization, the aggregation of the RHI is significant on a larger-scale, and the area of 4 °C ≤ Relative LST ≤ 8 °C is the stable and high-risk area, which provide scientific bases for the governance of the thermal environment on the regional scale. (4) The study also indicates the cooling effect of forests and water is better than that of grassland, while the cooling effect of grassland is uncertain. The methods and results of this study not only have implications on environmental planning and management in the PRDR but also provide useful insights into the thermal environment research and practice in other urban agglomerations.