The chapter proposes the use of physics in AI (machine learning and neural networks) for spatiotemporal analysis of hydro-climatic events. Hydro-climatic events stem from climatic variability and the dynamics of the hydrological cycle. AI can efficiently learn the pattern of climatic study data. In remote sensing and GIScience, hidden patterns are commonly found through computer vision techniques, such as in satellite image classification. Similarly, atmospheric and ocean circulations have classically been studied using numerical models built on physical laws. However, numerical models used in this way are limited by the availability of datasets, implicit patterns within the data etc. On the other hand, AI requires large resources to optimize or generalize, which sometimes give no result. Thus, to get the best of both worlds, we propose to combine physical laws with AI in hydro-climatic studies. The chapter includes various techniques which have development potential. We propose some theoretical foundations, for example the development of loss functions. We also explore some interesting applications where AI and synthetic data can be helpful additions to simulations, for example the study of relatively new climatic phenomena such as sudden stratospheric warming. Finally, we demonstrate by an example how physics-informed neural networks (PINNs) can be used to analyze precursors of hydro-climatic events, such as storm events caused by low air pressure due to atmospheric heating.