Could the Internet of Things (IoT) protect developing areas of the world from catastrophic flooding? A group of researchers in India thinks so.

Each year, floods cause devastation around the globe because villages cannot afford to take preventative measures. Current flood detection systems can cost over $1 million, which is far too expensive for many communities.

In an effort to find more cost-efficient flood detection methods, researchers in West Bengal, India,  proposed the installation of a modified mesh flood forecast system in a river basin. This system would use artificial neural networks and a wireless sensor network connected with the IoT to predict the probability of flooding. Most importantly, their model would cost less than $1,000 per unit, making it reasonable for any town to better protect the homes and lives of its people from flooding.

“Flooding is a major problem in our region and as far as we know, the current mechanisms used for flood prediction are not as simple and lightweight as our proposed solution,” said Prachatos Mitra, researcher from the Institute of Engineering and Management in Kolkata, India. “We wanted to develop a model to help towns that cannot afford or utilize expensive prediction methods.”

Similar to the image below, the researchers designed a modified mesh system for the wireless sensors that would be more economical. This mesh network topology is more energy efficient than other layouts because the sensors relay data along other sensors to connect to the central node. This requires less power than having each sensor transmit to the central node directly.

Flood forecasting sensor mesh layout

 

Additionally, the mesh design provides alternate routes between sensors, creating a backup route if one fails, so fewer sensors are required compared to other IoT solutions. While this design isn’t effective for large-scale projects, it works well for smaller ones.

Rather than implementing a typical statistical model, the researchers used an artificial neural network for processing data. In existing flood detection, statistical methods need various data to be calculated with high accuracy over a long period of time. As a result, they aren’t as adaptable as artificial neural networks. Furthermore, due to their computer algorithm structure and the availability of open source code, artificial neural networks are also easier to implement.

Structured to mimic human neuron connections, this network is a computer program where artificial neurons receive data and continually evaluate parameters to match given outputs. In flood forecasting, the network would receive water flow, rainfall and humidity data to estimate water levels.

An artificial neural network becomes more accurate as it receives more data. Therefore, when the network receives past flood levels, it adjusts the value it places on each parameter until it matches the water level of previous floods. Eventually the network recognizes the pattern by weighing the various parameters, and then is able to correctly predict a flood’s water level.

A simulation of the Ganges River from 2006 to 2014 revealed the artificial neural network showed a significant correlation between the parameters and flood results (see image below). This means when the parameters were set into the network, the prediction for the water level was closely associated to the actual water level recorded during the Ganges River floods.  

Table of artificial neural network's flood forecast results

 

The next step in the flood forecasting model is to implement the sensor network into a real case and use the data from the test to further refine the number of sensors and parameters needed for the network.

Apart from flood forecasts, the model can also be applied to appraising the viability of development projects located near rivers. The model is easily installed, and doesn’t require an expensive central device or high processing power for prediction.

The model is still in development. The researchers’ model was created with the help of Prof. Dr. Himadri Nath Saha, Asst. HoD Department of Computer Science Engineering, Institute of Engineering and Management, Kolkata.

The researchers hope all towns and villages will soon be able to afford and implement a flood forecasting system that can better prepare them to limit any and all damage as a result of floods.

Learn more about flood forecasting in IEEE Xplore.