IoT-based, municipal flood warnings

Extreme weather events are leading to flooding more and more frequently. Floods primarily occur when there is precipitation over a longer period of time (continuous rainfall), as water bodies are gradually fed by several tributaries and flooding occurs as a result. The way in which flood warnings are to be issued is governed by federal state regulations. The rather general definitions of the federal state regulations leave room for manoeuvre in the design of the warning. In many cases, floods are reported based on water levels. One focus here is on larger bodies of water (1st order bodies of water, e.g. Rhine, Weser, Ems). Smaller to medium-sized bodies of water (2nd and 3rd order) are not or rarely taken into account in nationwide flood reporting systems. 

A message is generally based on an assessment of the current water level situation. If a defined threshold value is exceeded, a flood alert is issued. Municipalities are increasingly expressing the desire to consider flood risks on a small scale in order to take local conditions more into account and to monitor 2nd and 3rd order watercourses in particular, as these rise particularly quickly during heavy and continuous rainfall. For this reason, many local authorities are increasingly endeavouring to establish their own information and warning systems. So far, however, there have been many smaller isolated solutions in the form of different approaches and no holistic, standardised solutions that actually issue reliable warnings.

In order to avoid each municipality developing its own solution, standardised procedures for setting up a small-scale flood warning system should be established. This was the reason for developing a concept for a standardised procedure for piloting a small-scale, IoT-based flood warning system. One advantage of a small-scale sensor network is that causes and effects can be better linked within the municipality. The concept envisages a 3-stage approach: 

Stage 1: Development of a flood information system, 

Stage 2: Analysing the causes and effects of flooding, 

Stage 3: Expansion into a complete flood warning system. 

In the first stage, warnings based on threshold value exceedances are possible. In the long term, solutions are to be developed that enable flood warnings based on forecasts. The central research question here is the extent to which previous forecasting models can be improved using small-scale data (dynamic and static). Advances in deep learning methods in particular promise greater accuracy in flood forecasting.