The white skeleton of the Maeslantkering, the massive floating water barrier protecting Rotterdam from high seas, dazzles in the sunshine. This engineering marvel is one of the largest moving structures in the world. Each arm is as large as the Eiffel Tower and weighs twice as much. If a storm surge above three meters is anticipated in the port of Rotterdam and its hinterland of 1.5 million people, the Maeslantkering automatically starts to close, flooding the two arms which move the barrier itself into place and dropping it into the waterway to form a watertight seal. The yearly test closing of the barrier, which seals the entrance to Europe’s busiest port, costs up to 30 million euros ($40 million).
The construction of the Maeslantkering was one of the high points in the Netherlands’ traditional water management strategy—build your way out of the waves—but rising water levels and spiraling costs have prompted the Dutch to try something new: Analytics.
According to IBM 66 percent of the population of the Netherlands lives in an area prone to flooding either by sea or rivers. The country spends 7 billion euros a year on managing its water systems: A complex web of sea gates and levees, canals and locks, drainage ditches and pumping stations, as well as the common or garden sewage and drinking water systems. That cost will rise by 1-2 billion euros by 2020.
Digital Delta is a 12-month research program investigating how to integrate and analyze water data from a wide range of existing and new data sources in order to reduce the cost of future water projects by 20-30%. The project involves IBM, the Rijkswaterstaat, researchers from the University of Delft, and several other partners who filled me in over a lunch of Dutch Tosti at a meeting spot in the shadow of the Maeslantkering.
There are three ways in which leveraging data can cut costs, according to Djeevan Schiferli, a water management executive at IBM. “You can better design your future infrastructure. You get longer use of the existing infrastructure, postponing investing in new infrastructure. The third element is on the maintenance of the infrastructure, not putting 10 kilometers of a levee into maintenance but only 100 meters.” All the big infrastructure investment decisions made by Rijkswaterstaat are aided by models developed by Deltares, a Dutch water research institute. “They are sensitive to the starting conditions,” says Feron. “With better starting conditions, better quality models, up to date and more data sources, the predictions will get more accurate. The decisions about which infrastructure we need where and where we can skip an investment.”
The Dutch water system is already one of the most highly monitored in the world. “We do biological, chemical, and hydrographical information and we also do a lot of measuring from ships on the rivers and canals, “ explains Feron.”We have a combination of current measurements—the water going through rivers, river discharge. We operate the major locks. During high water and low water it's critical not to leak too much water from one system to the other. Wave heights. That's important for predicting storm surges. For wind and waves we cooperate with the meteorological service. We have hydrographic and meteorological sensors on the North Sea and on the shore. We monitor shipping.”
That’s just the Rijkswaterstaat, but most of the Netherland’s levees and pumping systems are managed by one of the 27 local water boards. Water boards were first established in the 13th century by farmers who needed to maintain local water systems. They operate independently of other government bodies and levy their own taxes, which I grudgingly pay myself in Amsterdam. While the “Johnny-come-lately” Rijkswaterstaat, which was only established in 1798, shares data with local water boards and they with each other, it’s often done informally and offline. “So if there is an organizational change,” says Feron, “the system doesn't work anymore. It depends on people and if they know each other.” Neighboring water boards may end up working against each other when they both, for example, start preventative pumping of water out of their systems when rain is predicted.
Digital Delta aims to change all that by establishing a central registry of data sources all available in a standardized data format. The registry will be built by IBM, but will become public property at the end of the project. The first step is connecting existing and new data sources to the central system. These include precipitation measurements, water level and water quality monitors, levee sensors, radar data, and model predictions as well current and maintenance data from sluices, pumping stations, locks, and dams. One business case Digital Delta will address is connecting the data from water level gauges, which measure the water level in rivers and canals, with the gauges from local water boards. They all have different and incompatible interfaces.
For this and other reasons, adding new sensors is usually a hassle. Nick van de Giesen, professor of water management at Delft University, wants to experiment with new types of sensors. “I want to be able to install a sensor, scan a QR code and that's the end of it. Small companies like Alert Solutions have nice sensors but they spend 90% of their time and effort and swearing on getting the rest of the chain set up.” The rest of the chain means getting easy, automated access to the sensor and other data sources.
“Rijkswaterstaat monitors the rivers and they go up and down the rivers twice a year with their ships and they measure depth to see if there needs to be dredging, “ says IBM’s Schiferli, “One small company said 'Listen, we can take in the depth readings of commercial shipping and give you a more accurate, more timely 3-D overview and we think we can do it cheaper.’ When we asked companies like this why are you not implementing this solution they say it costs us one third to two thirds of our budget to find the data, get access to the data, and validate the data.”
Once the data registry is established, it will become easy to connect new data sources, access them in a standard format via an API, and combine them. One of Van De Giesen’s colleagues at the University of Delft, for example, has been using call center data on water complaints to predict sewer flooding. “How do you know whether a flood was due to rainfall or lack of maintenance?” asks Van De Giesen. “If you start with the complaints and add inflow, you can do much more. The complaints come in transcripts. We have someone who reads the transcripts and says 'It's overflow' or one of about five categories. We have 20,000 transcripts but that a subsample of everything we have in Rotterdam. With these 20,000 classifications you should be able to automate it and use the data from the whole of Rotterdam. And why would it not work in Utrecht?”
Once the data is available, it can be used by IBM’s Intelligent Operations for Water (IOW) software to build a systems view of the entire water system based on the data sources available and a semantic model. At a later stage in the project, IBM plans to utilize IOW’s analytics to enable completely new use cases. “If you look at the analytics there's descriptive—12 times a year a drain floods—and predictive analytics using Machine Learning — what was the combination of data streams which led to the flooding?“ says Schiferli. “The third level is prescriptive: ‘If the drain is clogged and the weather forecast predicts heavy rain, we need to do the following.’ Eventually the data registry built by Digital Delta may contain not only data, but applications and models as well. “If his colleague has a simulation of the sewer system using his rain gauge data, she can decide to have that model as part of the registry.”
IOW has an application development model which includes an SDK third parties can use to build water applications such as leak detection, flood management, or water quality applications. So small companies and researchers can participate by building their own applications on top of the registry. Feron foresees one rather more unusual way in which the data could be used: to inform Dutch citizens about the risks they face. “We have such a high risk level, but there is such a safe feeling," he says. “They are really surprised that a flood risk of once in 100 years means that it can be tomorrow. We need to accept certain unsafety. Accept that once in 100 years, for example, one of these polders will flood and the people who live there know that can happen. One University of Delft project has algorithms on what happens when some river or dyke goes, and you can use that to show people more or less real-time what happens. ‘This part of Delft stays dry. This sector we evacuate.’ Visualizing that will help people to accept it. They decide for themselves what they should do. This open data and visualization and apps will be the technology that will help communicate with the public.”
The rising sea levels caused by climate change means that the flooding risk the Netherlands has faced for hundreds of years will become an issue in other parts of the world. “This country lives below water but this is happening many places in the world, “ concludes Schiferli, “New Orleans, Jakarta, Adelaide, which is dry but also has heavy rain. All the same questions. If the Dutch have got the situation under control, why are they going to the next phase?”
[Image: Flickr user Goldsardine]