[Guest Post] Citizen Science for Watershed Action: Big Data Meets Fusion Tables

Illah Nourbakhsh is Professor of Robotics at Carnegie Mellon University (CMU) and head of the Robotics Masters Program. He is director of the Community Robotics, Education and Technology Lab (CREATE) at CMU and is a Kavli Fellow of the National Academy of Sciences.


Background

Industrial action, such as the recent Marcellus Shale natural gas drilling in Pennsylvania, West Virginia and Ohio, is endangering local populations’ drinking water supplies while simultaneously revolutionizing local economies in some of the most depressed regions of the United States. Millions of gallons of fresh water are being used, tainted, then dumped as part of a hydraulicfracturing process that is turning the eastern United States into one of the largest natural gas reservoirs in the world.  However, this complicated process by which natural gas is extracted and its impact on watersheds is difficult to monitor especially given that  the industrial chemicals are often hidden away as intellectual property. Today, local and state governments are torn between the desire to protect citizens' access to drinkable water and the need to generate income by leasing land to industry. Awareness is growing rapidly about the fact that industrial action endangers water quality in complex ways. The eastern U.S. has become a highly charged battleground with farmers, ranchers, environmentalists, politicians and watershed organizations on both sides of the issue.

While some websites, such as www.marcellusmedia.org, are doing an outstanding job of informing the public about the issues and inaccuracies swirling around this natural gas drilling debate, the lack of reliable data has been a major roadblock for informed decision making. Without reliable data, there is no pathway for hydrogeological models to improve, for theories of causal links between industrial action and water quality to be assessed, and for communities to have the peace of mind that comes with knowing that reliable data protect them against sudden changes in water quality (as demonstrated in documentaries such as Josh Fox’s Gasland.)

Current water monitoring systems are expensive, brittle, difficult to install, and time-consuming to maintain.  Devices can cost at least $1,500 (the most popular is the “Solinst” brand sensor) and users must wade into the water stream to install and periodically retrieve data manually. At this rate, water-monitoring organizations are spending most of their resources on the expensive systems and are limited in their ability to map entire watersheds on an efficient scale that can tell the entire story. The systems also fall victim to the elements and are destroyed by ice flows in the winter. As experts have explained to us, the ability to map water quality along a stream, densely in time and space, would be game-changing in the scientific understanding of how streams, watersheds and aquifer intrusions relate to pollution. Furthermore, the lack of transparent open access to water quality measurements greatly reduces the inherent value of the data.

While there is an increase in demand for accessible water quality monitoring tools, so are the needs for cheap, easy to use sensors that seamlessly add data to an on-line resource that is open to all.

Meet WaterBot: An Affordable Water Quality Sensor

Figure 1 - WaterBot (6x4x2 in^3).  It costs roughly $50 to make

Luckily the price of custom electronics continues to fall rapidly, while the quality is increasing. Today, you may find affordable wireless devices that run on a set of batteries for an entire year!  We have designed and begun pilot production of WaterBot (figure 1), a small standalone water quality sensor that measures water conductivity and temperature, stores values recorded every five minutes, then wirelessly reports the values opportunistically using the Zigbee 802.15.4 protocol to any host computer that approaches the river. The host computer, in turn, automatically redirects all data to Google Fusion Tables, which are publicly available on our website, waterbot.org.

The website is able to show not only WaterBot data but any water quality data run through the WaterBot software. We have demonstrated comparable results to existing sensors by producing side-by-side visualizations of both Solinst and Waterbot data collected from the same river collection point and stored in exactly the same way using Fusion Tables (see figure2). It also includes simple and familiar viewing options for comparing the data at different points in time. 

Figure 2: A side-by-side comparison between WaterBot and the Solinst measurements of water temperature (blue curve) and conductivity (red curve). The measurements between the two instruments are nearly identical, yet the WaterBot is considerably more affordable. This is great news for projects looking to track water quality at state or national scales.

How Fusion Tables Encourage Citizen Science 

The Fusion Tables functionality is critical to our project because it directly supports the democratization of data, and this is core to the mission of empowering local citizens to make the most informed decisions. Furthermore, Fusion Tables provides immediate visualization and feedback in a way that scales to arbitrarily large data sets, and this will be ever more important as citizens upload and share massively large data sets in the coming years, from air quality and water quality data to contextual environmental information such as wind direction, barometric pressure, etc. We have populated FusionTable’s zoomable viewer (see below), which is able to show arbitrarily large datasets interactively, using pilot WaterBot data collected at Nine Mile Run, which is one of the largest urban watershed cleanup programs in the United States.  Measuring temperature and conductivity in Nine Mile Run is of great value because, together, these values define Total Dissolved Solids, which we can track to demonstrate how rainfall and weather changes influence water quality in our watershed over the next several years. Using Fusion Table’s zoomable viewer, we encourage you to explore how zooming in reveals diurnal cycles, and then further zooming reveals the digitization limit of the WaterBot sensor electronics.  It is this ability to seamlessly move between time scales that makes the data particularly powerful in environmental applications, where many time periods can provide important scientific and community clues.

The WaterBot project is in its early days. We have successfully piloted six deployments, and have demonstrated that the resulting data has great value- by examining the conductivity spikes in the Nine Mile Run watershed in Pittsburgh, Pennsylvania, we were able to track sewage overflow events due to heavy rainfall, a problem that is commonplace in Eastern cities where the combined sewer-rain gutter systems were never designed for the population density that we now have.

Six counties are preparing to take on WaterBot on a larger scale in Pennsylvania and West Virginia, epicenters of the Marcellus Shale drilling controversy. As schools and citizen groups take this on, adopting local waterways, measuring them and sharing the data using Fusion Tables, we believe citizens’ abilities to directly impact policy will be greatly amplified, armed as they will be with real data that is easy to visualize and communicate in powerful ways.

Big data is an ever-strengthening trend. If we can put the tools in the hands of the public to capture big data, to store it, examine it, find patterns and tell the resulting stories, then we have a chance at bringing greater public good and rationality to the resource extraction processes that are today running roughshod because of a lack of information and accountability.

Illah Nourbakhsh

CREATE Lab, Carnegie Mellon Robotics Institute

Editor's Note:
In addition to thanking Prof. Nourbakhsh, I would like to thank his students Jessica Pachuta and Max Buevich. As well my colleagues from Google: Hector Gonzalez and Heidi Lam.