Part 1: Making sense of sensors in lakes
Sensors can be stupid too.
This is the first blogpost in the Lake Sensors series where we trace the journey of CSEI’s Urban Lakes Initiative. If you would like to collaborate with us on lakes, please email us at csei.comms@atree.org.
A recent article in the Hindu talked about installing sensors in stormwater drains (SWDs). With the decreasing cost of sensing and rising popularity of ‘Internet of Things (IoT)’ solutions, cities around India are pursuing various versions of creating ‘digital twins’ — a virtual representation of physical assets or systems — under the Smart Cities Mission. The idea is that better data will lead to better decision-making.
Indeed, real-time dashboards have helped private, commercial, and industrial actors, especially in regions where water is scarce, use water more efficiently and recycle/reuse water more effectively. Companies like Fluxgen and WeGot are emerging players in this space. Articulating the value of open data, however, where there is no single stakeholder, is much more complicated.
Several questions that need to be answered are often not. Is this data even the right kind of data? Who will use this data? How will they use it? How will they ‘discover’ this dataset? And last but not least, ‘who will pay to collect, curate, and archive this dataset?’.
The purpose of this blog is to reflect on our own journey of the Bangalore Citizen’s Lake Dashboard to understand what worked and what did not work.
We will narrate our journey in the three phases it evolved: our attempts at IoT solutions for Bengaluru’s lakes, then a shift to citizen science, and finally, our move to the idea of lakes as living and learning spaces.
Phase 1: An IoT dashboard for citizens
Our original theory of change was to better inform the dozens of citizen groups on the health of their lakes, so they could be more informed and empowered to approach agencies.
The first set of decisions involved deciding which parameters should be measured and understanding what their levels would be. We had limited money and IoT sensors are expensive. So we needed to deploy the funds carefully. The ideal candidates for IoT are parameters that a) are consequential and b) change value frequently. Dissolved oxygen (DO) met these criteria.
Most Bengaluru lakes have sewage flowing into them either directly or through SWDs connected to the lake inlets. This could be raw sewage or secondary treated sewage from the centralised sewage treatment plants (STPs). Both of these sources are high in nutrients (nitrates and phosphates), which act as food for the algae, causing them to proliferate in lakes. The presence of excess algae causes large fluctuations in dissolved oxygen (DO) as they photosynthesise and decompose. The figure below shows the typical variation of DO throughout the day.
Adequate DO levels (minimum of 4 mg/l) are critical for the survival of aquatic life. Besides algal growth, DO in the lake also gets affected by other factors such as temperature, salinity, etc. Because IoT sensors collect data around the clock, we hoped to capture the early morning DO crashes that cause the fish kills.
So we bought a bunch of research-grade DO sensors and deployed them in two lakes — Jakkur and Kaikondrahalli. These lakes were fenced and had homeguards, who would be able to prevent theft or vandalism of sensors. Along with the DO sensors, we also installed sensors to measure lake levels and in the inlets and outlets to track the flow in and out of lakes.
Sensor Troubles
The first lesson we learned was that sensors are like babies. They have to be fed (batteries/solar energy), cleaned (to remove the muck), and educated (calibrated).
In measurement technology, sensor calibration refers to the process of defining the relationship between the value of the output signal by a measurement instrument and the parameter of interest. For example, say a sensor detects temperature by generating an electric pulse. Calibration is the process of specifying what size of the electric pulse corresponds to what temperature. The challenge is that as sensors are subject to the elements in real field conditions, they age and this relationship changes. So the calibration process has to be repeated periodically to ensure accuracy.
We were prepared for this and had factored this into our project planning. What we were not prepared for was that our lakes were so nutrient-rich that they fouled up and stopped recording data within a day or two. Simply put, when placed in the hot, wet, nutrient-rich environments of Bengaluru’s lakes, the sensors acted as a rich substrate for algae and other microorganisms to grow. Once the sensor was covered with a biofilm, it was no longer able to detect conditions in the lake.
This wasn’t the only problem: theft, vandalism and the sensor boxes attracting wildlife as ideal safe havens to build nests (and their young to chew on cables) led us to abandon the IoT enterprise. (In response to the vandalism, some local manufacturers like Krisnam Technologies have started building innovative sensor boxes that are camouflaged.)
In contrast, the water-level sensors were quite stable but here too there were major challenges. We chose to deploy Odyssey capacitance sensors because these were quite robust and the battery only needed to be changed every six months or so. But because they measured water level, they had to be installed vertically on a pole secured in the lake bed. So we had to wait till the lake bed dried up in the summer, to install the sensor. Sometimes the lake only dried up partially. But after we installed the sensor, thinking we had reached the lowest possible point, the lake shrunk further “stranding” the sensor on dry land. In many of the inlets and outlets, the sensor ended up getting stranded in a local stagnant pool of water, created by accumulated debris. So the sensor ‘thought’ it was measuring flowing water but it wasn’t really. In other cases, a builder had begun to dump debris into the drain upstream, blocking the flow altogether.
These challenges notwithstanding, we did get useful data that allowed us to draw useful insights into the social and ecological aspects of lakes.
What we learned from the IoT phase
First, sensor deployments have proved very useful to test specific hypotheses and deepen our understanding of lakes. But maintaining sensors over the long term requires a substantial commitment of time and resources. And this, in turn, requires a stronger ‘use case’ for the data than the one we had. This is equally a problem with the many digital twins created in an open data context without a clear user or use case.
Second, drawing insights in a highly dynamic environment like Bangalore is hard, because so many things changed simultaneously that despite the continuous monitoring, we still could not explain why nutrient levels, for instance, were changing the way they were.
Third, we were surprised about the extent to which the science was still unsettled, such that basic questions like “is the lake recharging groundwater?” and “what is a good nitrate level?” could not be answered clearly.
In Part 2 of this series we will discuss the next phase of our journey where we pivoted to citizen science. You can follow Mira community on Instagram for more on nurturing Bengaluru lakes. Our journey was funded by a CSR grant from Oracle.
