Low-power Internet of Things (IoT) sensing nodes can be embedded into various physical environments to monitor vital parameters. Some of these environments impose rough and extreme operation conditions, severely limiting the performance of these nodes. Modeling these environments is vital to make the nodes adaptive. In this letter, we propose a model to estimate the complex motion of nodes deployed on the surface of different water bodies. The model relies on received power statistics only. Experimental results confirm that the model is reliable, achieving an estimation accuracy of 93%.