2020 was a record-breaking year, but some of the reasons went unnoticed because of preoccupation with the pandemic. 2020 had the sunniest April and May on record and the third-hottest day ever recorded in August. It also had the wettest February on record with the highest number of flood warnings ever issued, as well as the wettest single day ever in October. It was warm, wet and windy.
The Met Office predict ever hotter drier summers and wetter warmer winters. With sea-levels also rising we need proactive solutions quickly.
Surprisingly, the UK has no single body responsible for flood control. The Department for the Environment and Rural Affairs (DEFRA) is regarded as “the policy lead” in England. However, actual decisions depend on district and borough councils, the Cabinet Office, the Department for Communities, the Highways Authority, water and sewerage companies, independent Internal Drainage Boards, coastal protection agencies and a variety of other bodies. Even though water flows without regard to political boundaries, areas that fall inside Wales and Scotland have been delegated to their respective regional assemblies, further impeding a unified response.
Although DEFRA specifically advocates a “free flow of information” between all these agencies, it is inevitably fragmented. Knowing what is happening at any given moment is further complicated by pumps and sluices under the control of landowners, and construction activities that change the flow of surface and subterranean water. Widespread housing developments on flood plains also change the risks. What is missing is the ability to gather instant, highly detailed information on complete water systems.
Monitoring and planning
Measuring water levels is not new: you can see old-style gauge boards beside many big rivers. DEFRA and other agencies rely on electronic measuring devices which are able to communicate with remote measuring stations. They provide an API which allows any interested party to access their data (http://environment.data.gov.uk/flood-monitoring/doc/reference) but that data is limited.
The imminent water level in one location is often less determined by height readings in that location than by flow-rates upstream, plus other factors. Water momentum, wind direction, the saturation of flood plains and the quantity of rain falling in any particular location are all hard to measure. Air humidity, temperature, soil condition, the water table, plant growth on riverbanks and debris in sewers all contribute to creating a complex picture.
Current systems don’t capture enough information, but they are also limited by how the data is used. Most water monitoring systems do little more than trigger an alert if the river level is high. This isn’t much better than a weather report telling you it is raining when you can look out of the window.
Alerts are often too inaccurate, imprecise or late for those affected to act. By using more sensors and better analysis we might achieve two things, earlier and more precise predictions and proactive flood prevention.
Coupling the IoT with AI
Some good news is the explosion in sensor devices that communicate via the Internet of Things. This means that more types of sensor can be deployed over larger areas cost-effectively. They require no independent wired or wireless networks and there is no reason for agencies to hold information silos. Local stakeholders can deploy and extract the information they require and easily share it with regional and national interests.
Many different kinds of sensor are available. For water level monitoring alone there are radar devices, electrical switches, gas bubbler pressure gauges, flowmeters, submersible pressure transducers. Especially interesting are ultrasonic devices such as the Wilsen sonic level .
IoT devices of this type use little power so they can function for years unattended, yet still connect with internet gateways 15km away, for example the LoRaWAN® system. This means it is feasible to deploy them over larger areas, such as tributaries, drainage ditches, fields, forests and culverts.
The other potential advance depends on AI. Not only can an AI-powered system extract better flood predictions from larger data sources, but it could also be applied to prevent flooding. By truly centralising all water management related information, AI can identify the systemic patterns that lead to loss of control over water. Solutions that could minimise flood disasters may be deployed far from the locations they devastate, replacing flood alerts with avoidance policies.
AI and IoT approaches to flood prevention are already being explored by Fujitsu in Japan, Google in Patna India, the Lassonde School of Engineering in Canada and in the town of Cary in North Carolina using the SAS Viya AI analytics platform and Microsoft Azure.