Using the internet of things to predict the future of flooding in Dublin
18 July 2017
Jamie Cudden (Smart City Programme Manager, Dublin City Council), Professor Linda Doyle (Director of CONNECT, Trinity College Dublin) and Dr Frank Smyth (Executive Director, CONNECT, Trinity College Dublin) at the deployment of rainfall and flooding sensors around Dublin city.
According to estimates by Dublin City Council, flood damage to the capital’s infrastructure can range from €2 million to €100 million per annum, with an average annual cost of around €8 million. This figure is set to increase due to rising sea levels and more intense rainfall events.
Over recent months, CONNECT (the Science Foundation Ireland Research Centre for Future Networks at Trinity College Dublin) has been working with Smart Dublin – a collaboration of the four Dublin city councils – to use Internet of Things (IoT) technology to help predict flooding events and better manage the situation once flooding has occurred.
We have now deployed networked sensors in several locations around the city to gather flood-related data. The goal is to use information from these sensors to better understand the nature of city flooding and, as a result, make informed decisions to mitigate its effects.
Pervasive Nation – IoT testbed
At the heart of our project with Smart Dublin is Pervasive Nation, which is CONNECT’s nationwide IoT testbed. Funded by Science Foundation Ireland, Pervasive Nation is a Low Power Wide Area Network (LPWAN). This type of network has been attracting a growing amount of interest in recent years. LPWANs are designed for the specific purpose of gathering small amounts of data at regular intervals from large numbers of sensors. It is ideal for use with sensors that must remain in-situ for long periods of time and which need batteries with long lifetimes.
There are several flavours of LPWAN and Pervasive Nation uses LoRa/LoRaWAN, which offers the benefit of secure bi-directional communication between the sensors and the network, while also offering mobility and localisation services.
The physical layer uses a form of spread-spectrum modulation based on wide-band linear frequency modulated pulses. The level of frequency increase or decrease over time is used to encode the data to be transmitted, i.e. it is a form of chirp modulation. This form of modulation enables LoRa systems to demodulate signals that are 20dB below the noise floor when the demodulation is combined with forward error correction (FEC).
LoRa/LoRaWAN is also a permissionless network and therefore, just like a WiFi network, anyone can deploy one. While LoRa networks can, in principle, operate at different frequencies, they generally tend to be operated in the ISM bands as these are licensed-exempt. Pervasive Nation follows this trend, operating at 868 MHz, a frequency which also supports long ranges.
The relatively low cost of LoRa basestations, the permissionless nature of the network and the fact that licensed-exempt spectrum can be used, means it is possible to deploy a network without the need for support from large-scale network operators and without the barrier of high costs associated with spectrum purchase. Deploying a LoRa network can be more onerous than a WiFi network, however, since good coverage requires basestations to be placed at heights. LoRa networks offer the same kind of potential though for open innovation already been seen in the WiFi world.
A significant portion of Ireland is now covered by the Pervasive Nation network with full coverage achieved in Dublin. The Pervasive Nation basestations (or ‘access points’, as they are called in the IoT world) are distributed around the country and, from these, data is backhauled to the network backend in the cloud. The backend of the network includes the network server (the network controller and monitor) which direct the data to where it should be processed and stored, and to the application enabling platform (AEP).
The network server and AEP for Pervasive Nation is hosted on Waterford Institute of Technology’s datacentre. The AEP supports the registration of sensors on the network and also interprets the data collected from those sensors. It also provides a framework on which applications can be built.
Two types of rainfall sensors
We are currently using two types of rainfall sensors on the Pervasive Nation network. One is manufactured by Casella and sends a signal each time 0.2mm of rainfall occurs. The other is a low-cost unit – more than a factor of 10 times cheaper than the Casella option – and is capable of measuring 0.3mm of rainfall. The sensors are integrated with LoRa radio chipsets to allow data to be transmitted over the Pervasive Nation network.
Both sensor units have a similar operating technique based on a tipping bucket mechanism that tilts each time the specified rainfall amount is reached. A message is then communicated via the LoRa radio (RN2483) to signify that the tip has occurred and allows a calculation of the rainfall amount.
There are currently 17 gauges deployed in eight locations around the capital identified by Dublin City Council. By the end of July 2017, that number will rise to 24 and will be supplemented by four lower cost weather stations.
The current phase of experimentation involves testing the sensors on the Pervasive Nation network, evaluating their accuracy, understanding their robustness, reliability and longevity, and carrying out basic analysis of the collected data.
In many IoT applications the aim is to deploy many low-cost sensors rather than a few highly expensive sensors. Early experimentation is often focused on whether this trade-off is possible as well as ensuring that basic functionality is delivered.
The Pervasive Nation team in CONNECT is also working on several other features including the design of new sensors that might be used on the network, antennas suitable for the system, low cost radios for LoRa, future LoRa access points, geolocation techniques using LoRa, network server features, enhanced security features, mechanisms for easy bulk deployment of sensors, new supports for application design and new business models for LPWAN initiatives.
We will see the real value of a flood monitoring system when data will usefully inform decisions by city authorities. This will require a critical mass of sensors placed in relevant locations with the development of more analytics to mine the data.
To be a success, it will also require an acceptance of this kind of practice. The data from these sensors will be very useful to the local authority. It will, for instance, also inform many day-to-day activities such as gully clearance. The data could also be used in the planning processes for new builds.
Analyse data to reveal patterns
Fundamentally, IoT offers the possibility to gather and analyse data at a scale that reveals patterns which are not observable by a human expert, but which can help to enhance that expertise.
Our Pervasive Nation network is allowing the testing and trialling at scale and under real conditions. It is open to industry, academics, and agencies who are interested in understanding and exploring how LPWAN IoT applications might help their businesses, research and stakeholders.
Our work with Smart Dublin to battle city flooding is off to a promising start.
CONNECT is hosting ‘Get Connected with Pervasive Nation’ in Croke Park on Wednesday, 13 September 2017. It is a free event that will provide an introduction to the Pervasive Nation network. Full details here. Contact: email@example.com.
Prof Linda Doyle is director of CONNECT, the Science Foundation Ireland Research Centre for Future Networks at Trinity College Dublin