Low-power Internet of Things (IoT) sensing nodes can be deployed on the surface of different water bodies for various purposes, including water quality monitoring and pollution detection. Two of the most formidable challenges toward such goals are: 1) making the nodes resilient against rough water and extreme weather conditions and 2) enabling the nodes to establish reliable wireless links. In this article, we share our experience in deploying low-power and resilient IoT nodes on the surface of different water bodies—on a small lake, North Biscayne Bay, Crandon Beach, and South Beach, in Miami, Florida. Furthermore, the article closely examines how link quality was affected by predeployment configurations as well as the characteristics and the motion of the waters. Based on the analyses of a vast amount of statistics, the article establishes a theoretical (mathematical) and generalized model to characterize and predict link quality fluctuations. We shall show that the realization of the model using the Kalman Filter enables link quality prediction with accuracy exceeding 90%.