Floodaware - Near Realtime Granular Flood Forecast and Modelling using IoT

International audience Flash flooding is one of the most frequent natural disasters of the modern world, forming a major threat to life and property. Floodaware is a system designed to give early warnings in flood prone areas, giving people the extra few minutes needed to save lives. The Floodaware system connects live sensors and rainfall data to provide a near realtime prediction forecast of flooding to residents, as well as enhanced modelling capabilities using the GAMA platform. Floodaware uses network of ultrasound based water level sensors connected to The Things Network (TTN) to record levels at sections of the streams in the catchment. The water levels are used to calibrate the model of the catchment to best reflect catchment dynamics and provide large amounts of empirical data for further modelling. Live rainfall data is then used within the model and to provide early warnings and forecasts of floods to local authorities. The model of the catchment features a number of interconnected layers which interact with each other. The layered approach allows to exploit certain features of geometry to accelerate computation and give an understandable conceptual model. The first layer is the rainfall or cloud layer which is comprised of agents given particular rainfall payloads and geometries. At the time of each agent’s precipitation the intersection of the agent and underlying sub catchments is calculated to convert the level of rain (in mm) to a volume. Once calculated the payload is transmitted to the underlying sub catchment to allow for flow within the sub catchment, and overall flow through the whole network. The flow within the network is based upon the Watershed Bounded Network Model (WBNM). WBNM is a commonly used hydrological model for modelling floods in catchments. In Floodaware the WBNM equations have been used to govern the flow within sub catchments, and the flow downstream towards the outlet. Floodaware uses GAMA to create a more accessible instance of the WBNM logic, and connect it with richer data. The modelling harnesses GAMA’s GIS capabilities by accessing a PostGIS enabled PostgreSQL database. The database allows for simple execution and setup of different modelling scenarios, and an excellent way to store, maintain and access relevant data. Initialisation, runtime, and final results are all stored within the database to log results and keep a record of experiments performed outside Floodaware’s live operation. GAMA’s GIS features and database access have made it a useful and accessible platform for the purpose. More development and support for newer data storage methods will help enrich GAMA’s future in the modelling world. Object based data formats would be especially useful as they closely parallel the descriptions of agents in GAMA. The integration of data storage and presentation within Floodaware are greatly enabled by GAMA and allow for greater interactivity and understanding of the workings of the model.