Topological data analysis is an offshoot of the classical area of topology where here data is studied qualitatively by looking at topological features present in the dataset. Features such as components, holes and voids are the focus of persistent homology which is the main branch of topological data analysis. In this talk, we present our recent work on the application of persistent homology in the development of a flood early warning detection system. Water level data from Kelantan River is used to investigate variations in topological features present prior to a flood event. These variations are then classified using the theory of classical slowing down to indicate flood and non-flood events. The efficacy of this approach is reported in the conclusion of this work.