Makara: Guardian of the Chennai Rivers

Background

I grew up in the city of Chennai in Tamil Nadu in Southern India. Three rivers flow through Chennai before draining into the Bay of Bengal in the east - The Adyar, Cooum, and Kosasthalaiyar Rivers. These rivers have historically provided critical benefits, serving as sources of drinking water, irrigation, habitat for diverse species as well as water management for floods and the monsoons. The Cooum River, in particular, was vital to the city's development. Early buildings and gardens were constructed to face the river, which was a hub of activity.

The Adyar River estuary and mangrove forests are also important ecosystems for a wide variety of organisms from aquatic life to birds, both resident and migratory.

The Buckingham Canal connects all three rivers, and has its own interesting history. It was constructed between 1806 and 1882 as an artificial waterway that is nearly 800 kilometers, used for transporting timber and bulk goods. After its importance declined due to railways and highways, the canal's use in Chennai diminished after the 1960s, though it remains in good condition elsewhere. It acts as a flood buffer zone and was largely helpful in reducing the impact of the 2004 Tsunami in Chennai.

See https://www.google.com/maps/d/u/0/viewer?mid=14VJso-AdAaShFQPaLgH16aQeM7zp6ZY&ll=13.136464798281425%2C80.04329680399482&z=11

Problem

While Chennai's rivers have historically provided immense benefits, centuries of neglect and urbanization have caused severe problems. Unfortunately, all three rivers are extremely polluted, due to human encroachment and the discharge of effluents, and damaged ecology due to dams, embankments and redirected fresh water flow.

The Buckingham Canal is the most polluted of the water ways, with the discharge of untreated sewage into it directly. Interconnecting the rivers, it therefore has a direct significant impact on the water quality of the rivers themselves.

The mere mention of Cooum conjures up images of a dirty, smelly water body. The pollution of the Cooum river has been an issue that takes centre stage in all discussions around improving life in Chennai. The fate of Chennai’s other water bodies, the Adyar river and Kosasthalaiyar river and the Buckingham canal also mirror that of the Cooum.

The Cooum River originates from the surplus course of Cooum tank in Tiruvallur District. It runs east for a distance of about 65 kilometres and confluences with the Bay of Bengal downstream of Napier Bridge, traversing a distance of 20 kilometres within Chennai city limits

According to scientists, around 1950, Cooum had 49 species of fish, and by the late 1970s, this was reduced to 21 species. However, to date, there are no fish in the river in the urban areas, owing to highly toxic pollutants found in the river water. Fortunately, the good news is that about a 40 km early rural stretch of this river is still not as polluted. Originating in the Kesavaram Tank, it is relatively unpolluted upto Thandidurai and Avadi.

The Adyar River originates at Adhanur Lake near Guduvancheri. Surplus water from Chembarambakkam joins near Thiruneermalai giving it shape as a river and flows 42 km through the districts of Kancheepuram, Chengalpet and Chennai. It confluences with the Bay of Bengal near Adyar.

The Kosasthalaiyar River originates near Pallipattu in the Thiruvallur district. It runs for approximately 136 Km. The northern tributary, the Nagari River, originates in the Chittoor district of Andhra Pradesh and joins the main river near the backwaters of the Poondi reservoir. It flows into the Bay of Bengal at the Ennore Creek in the northern part of Chennai.

The rivers are important resources (that are currently not utilized) in a bustling metropolitan area. Chennai is dependent on monsoon rains and ground water for its water supply, and mismanagement of the water table has resulted in catastrophic flooding. During the Tsunami of 2004, the rivers, along with the canal helped prevent extensive damage.

The Adyar River estuary has a diverse ecosystem of birds, both migratory and resident.

The Kosasthalaiyar river is considered largely clean in its upper reaches, closer to its origin near Pallipattu, and before it enters the highly urbanized and industrial areas of North Chennai. Specifically, the stretch upstream of the Poondi Reservoir and leading up to areas like Red Hills is generally referenced as being less polluted. This water is vital for filling the Poondi Reservoir, a key source of drinking water for Chennai.

Hope

There have been several attempts at cleaning up the rivers and canal over the years. The Government of Tamil Nadu formed the Adyar Poonga Trust (APT) in 2006 to develop an Eco Park in 58 acres near Adyar creek called Tholkappia Poonga. This trust was later renamed as Chennai Rivers Restoration Trust (CRRT) in 2010. Owing to the severity of the problem and continued problems of encroachment and pollution, progress has been slow.

During the pandemic of Covid-19 and its resultant lockdown, it was observed that the water quality of the Chennai rivers increased substantially (See https://www.frontiersin.org/journals/marine-science/articles/10.3389/fmars.2021.659686/full)

If the rivers are restored back to health, we can see that it will have an immense impact on the ecology, not only for humans but also aquatic life and birds, and other wildlife too. For humans, the rivers will help provide another source of drinking water, transportation, and ecological connection. For aquatic life, other wildlife and birds, it will be a boon ecologically and can have a transformational effect on the environment.

As we have seen, there are early stretches (outside of the Chennai metropolitan area) of the rivers that are not as heavily polluted.

Just as Delhi’s Yamuna Revival or Ahmedabad’s Sabarmati project showed results with focused intervention, the Chennai rivers too can bounce back — if citizens, planners, and policymakers act together.

The National Water Quality Monitoring Programme (NWMP) is an Indian initiative, coordinated by the Central Pollution Control Board (CPCB), to assess and monitor the quality of surface and groundwater across the country. Here is the data for Adyar and Cooum Rivers for 2023:

The data is collected through periodic manual monitoring. There is no continuous real-time monitoring being done however.

Having constant visibility into water quality, aquatic life and ecosystem along the entire length of the river(s) will help pinpoint problem areas to focus on. More importantly, it will also celebrate and highlight positive stories of thriving ecosystem areas and enable human connection.

Let us consider an analogy. A Continuous Glucose Monitor (CGM) provides far more useful information than point-in-time Blood Sugar prick-test measurements (i.e. fasting/post-prandial). The CGM data can help individualize blood sugar responses to foods and lifestyle factors and blood-glucose management efforts. It can thereby help change food and lifestyle behaviors for positive measurable impact and improvement in blood glucose management.

Similarly, real-time continuous monitoring having a constant stream of data – of water quality measurements, ecological factors such as aquatic life and other wildlife concentrations – at points along the length of the river can help focus not only on problem areas but also highlight what is working and thereby

• 1) help change human behavior
• 2) help informcleanupefforts for positive impact.
• 3) educatethepublic on the river ecosystem (both aquatic life as well as above water wildlife near the rivers)
• 4) celebrate the successes

Initiatives focusing on changing public perception of the river, reconnecting it with the city, and creating a more positive relationship with the waterway will go a long way towards successfully rehabilitating these rivers and transforming the landscape.

The Department of Environment and Climate Change has announced the Tamil Nadu Lake Quality Monitoring Project to be implemented in water bodies through real-time monitoring and artificial intelligence-based data analysis. It is yet to be implemented, however. This is the perfect time for a citizen science project such as Makara to fill the gap and provide a solution.

Why Makara?

The Makara is a legendary sea creature in Hindu and Buddhist mythology. It is associated with water and with protection, and is considered the vehicle (Vahana) of Varuna, the God of water, and Ganga, the river Goddess.

Source: https://en.wikipedia.org/wiki/Makara

The Makara project aims to highlight the good, the bad and the ugly along the stretches of the 3 rivers to better understand and highlight the unseen (thriving ecosystem vs. increasing pollution, improvement in quality with cleanup efforts vs. effluent dumps) and use the data to inform as well as change behaviors. This system would of course be equally applicable to any other river ecosystem in any other part of the world.

Source : https://www.past-india.com/photos-items/buckingham-canal-in-madras-or-chennai-old-print-1922/

Source: https://www.ajlajournal.org/articles/re-imagining-open-spaces-along-the-buckingham-canal-chennai-india

Makara Project

The goal of the Makara Project is to get quantitative as well as qualitative data regarding the wellness of the river and its ecosystem. Makara will not only measure water quality parameters but also measure aquatic and wildlife type and concentration.

A set of floating, anchored sensor pods will be deployed periodically along the length of the river.

Data such as Polluted River Stretches For Restoration of Water Quality will be utilized to identify the worst polluted stretches and optimize the placement of Makara sensors. The Table below shows which stretches of the Adyar and Cooum Rivers are the most polluted.

Each sensor pod will measure water quality parameters such as temperature, pH, conductivity, dissolved oxygen, turbidity, and specific conductance. Additionally, each sensor pod will include two AI cameras, one for underwater and one for over water. The underwater cameras can be used to catalog both the presence (or absence) of aquatic life as well as of trash and other pollutants. The presence and concentration of different species can be monitored and logged. The above water AI camera can similarly be used to identify above water conditions and also wildlife.

The units will be powered by rechargeable batteries along with solar panels. One interesting stretch goal will be attempt to harvest kinetic energy from the water flowing where possible. The sensor pods will be low power devices with optimization of power consumption by reducing usage and using sleep modes.

The sensors will communicate via a Meshtastic network to send the data to central stations which will consolidate the data and run further analyses and generate reports with the data using cloud services.

These reports and analyses can

• 1) Inform the cleanup efforts around dynamic changing conditions of water quality - correlating with environmental and weather conditions, diurnal variations, effluent discharges from industry and power plants, sewage discharge from storm water drains, etc.
• 2) Identify the aquatic life (species, concentration, time of day, etc)
• 3) Identify above water wildlife (species, concentration, time of day, etc)
• 4) GIS mapping of river water quality over the length of the river
• 5) prediction of poor water quality and fish kill-off events before they happen for intervention

There are two forms of information dissemination:

• Through public display monitors at central station locations
• Through website access

There will be two types of data:

• Planning and control data for the city
• Public data for the general populace

Eco Parks will play a vital role in Makara's goal of helping change public perception for positive transformation. They will host central stations and also provide public display monitors with public water quality, aquatic life and wildlife data that can be integrated into the narrative of the Eco Park as an attraction.

The Tholkappia Poonga or Adyar Eco Park (also known as Adyar Poonga) is an ecological park set up by the Government of Tamil Nadu in the Adyar estuary area of Chennai, India. In fact the Adyar Poonga Trust is the body that was later renamed to the Chennai Rivers Restoration Trust (CRRT).

EFI Kanagam Eco Park is another effort by the Non-Profit group Environmentalist Foundation of India (EFI) partnering with the Greater Chennai Corporation to create an Urban Eco Park in a piece of abandoned land along the Cooum River, transforming into an urban oasis and sports park.

It is the goal of the Makara Project to place base stations in these Eco Parks, which will be used to disseminate public-interest related data analyses, reports and imagery to showcase positive stories of river health and also display photographs of cataloged aquatic life and wildlife around the rivers that have been collected from the Makara Sensor Pods.

The reinforcement of these Eco Parks as guardians and disseminators of knowledge of the river and ecosystem will spur the development of more Eco Parks along the river's edges, to further help change human perception and habits, transform urban planning and identifiation and remediation of water polluters and pollutants and revitalize Chennai's rivers!

Architecture:

The high level architecture is shown here. Makara sensor nodes collect water quality data and also perform image classification of aquatic life, above water wildlife and trash/detritus. The data gets transmitted over a Meshtastic network to a base station that is located at a suitable place (like the Eco Parks). The base station will have access to power and also cloud connectivity and send the data to a cloud service for additional data consolidation and analytics and reports. The analytics and reports are of two types - 1) for public data with an emphasis on storytelling and information dissemination and 2) that would be used by the city/CRRT for monitoring and planning - identifying polluters and pollution events and acting to prevent deterioration, showcasing cleanup efforts and providing real-time continuous monitoring data to bodies like National Water Quality Monitoring Programme (NWMP) and acting as stewards and evangelists for adoption to other rivers and water bodies.

In order to break down the solution in manageable chunks as we develop the project from idea to prototype to production, we first focused on the individual Makara Sensor Pods and how they would function, while still laying the framework for a broader system.

For this project, the Makara Sensor Pod was designed with the various subsystems in mind, and utilizing off-the-shelf components as much as possible to derisk the initial prototype and to enable fast iteration and testing. Subsequent iterations of the prototype would further integrate different subsystems into the main design to optimize and bring down the cost of production.

The diagram below shows the various subsystems of the Makara Sensor Pod:

The diagram below shows the Makara V1 Prototype. There is a central main board that integrates the various subsystems, which have been carefully designed to operation within the constraints of a remote aquatic environment (see Design Considerations below).

The diagram below shows the main Makara Sensor V1 prototype board. It has been given for fabrication and assembly and we are awaiting it to integrate and test and calibrate the systems.

Here is an artist's (ChatGPT) rendering of the floating buoy Makara Sensor Pod!

Design Considerations

Water Quality:

Water quality is often described by different indicators such as

• temperature - Measuring temperature helps to understand the magnitude and variability of temperature fluctuations and anticipate the consequences for water quality and ecosystem health
• dissolved oxygen (DO) - Less than 5 mg/L: Stresses most fish. Less than 3 mg/L: Too low for many aquatic organisms to survive. Less than 1 mg/L: Hypoxic, often devoid of life.
• pH - The pH in most natural water environments is between 6.0 and 8.5. pH values below 4.5 and higher than 9.5 are considered lethal to aquatic organisms, and pH values less extreme can interfere with reproduction and other essential biological processes.
• total dissolved solids (TDS) - High levels of TDS diminish water quality, making it unsuitable for drinking and irrigation. Generally, freshwater has a TDS level between 0 and 1, 000 mg/L. This is dependent on regional geology, climate and weathering processes, and other geographical characteristics that affect sources of DO and transport to water systems.
• conductivity : Drinking water - > 500 µS/cm: Levels higher than 500 µS/cm may indicate contamination and reduced water quality. Irrigation - > 3.0 dS/m: Unsafe for irrigation, as it indicates strong saline conditions that will reduce most crop yields.
• suspended sediment - An increase in turbidity or suspended solids can negatively affect aquatic health
• nutrients
• metals
• hydrocarbons
• industrial chemicals

For measuring some of these parameters, there are 3 types of liquid sensors that are available for the Maker, at differing levels of accuracy, cost and continuous usage. Atlas Scientific, DF Robot (Gravity sensors) and Seeed Studio (Grove sensors) offer relatively low cost liquid sensors. Of these, Atlas Scientific are the most pricey but the most robust and accurate. While DF Robot's and Seeed Studio's sensors are somewhat similar, DF Robot offers more extensive sensors with better support (such as analog isolation power boards). I chose DF Robot's sensors as a middle-of-the-road approach.

When utilizing multiple liquid sensors, as they are all immersed in liquid, they can potentially interfere with each other. Providing isolated power for each liquid sensor is crucial towards eliminating/reducing interference and improving measurement quality. These add additional costs to the project but are required. I'm currently using off-the-shelf analog isolation power boards but hope to integrate them into the main board to reduce the overall cost of each sensor.

Another factor to consider is lab grade vs. industrial grade liquid sensors. Since the sensors will be immersed in liquid continuously, industrial grade sensors are preferred. One factor that propelled DF Robot sensors over the Grove sensors was the availability of more industrial grade sensors.

From a cost and proof-of-concept perspective, I chose pH, TDS, EC and Turbidity sensors, with industrial grade where practical (cost) and available. Dissolved Oxygen sensors were too pricey and so I omitted them in version 1, though they (and other sensors) can always be integrated as required after running tests on the prototype and reducing costs through consolidation.

Another consideration is the use of external ADC for measurement. The ADCs on the ESP32S3, while being 12 bit, suffer from non-linearity especially at the extreme voltage ranges. Considering that we'd likely be measuring high levels of pollutants, it seemed prudent to rely on an external ADC (I chose the ADS1115 (which is also 16 bit and has 4 channels), and decided to integrate with an off-the-shelf board in the interest of time. Subsequent versions of the prototype can have this ADC directly integrated on the main board).

Power:

Since these devices will be deployed in relatively remote areas without direct access to power, they will be run using Solar energy and batteries.

The devices themselves also have to be low power devices with optimized power consumption, sleep modes, etc.

With Solar charge controllers, you have MPPT, or PWM charge controllers and also linear charge controllers that offer "near-MPPT" performance but with less complexity and cost.

The CN3791 is an example of True MPPT, while CN3165 and the BQ25185 are linear charge controllers offering near-MPPT performance.

As a remotely deployed IoT device, we'd also like to monitor the battery levels that get reported along with other sensor data.

Many simple circuits use a simple voltage divider along with an ADC pin on the microcontroller to read an interpret the voltage to battery level. However, a chip like the MAX17048 on a breakout like the Adafruit MAX17048 LiPoly / LiIon Fuel Gauge and Battery Monitor makes it easier to hook everything up and read the values via I2C directly. It would be great is solar charge management boards offer an integrated solution with a battery monitor, but unfortunately, there is none. I felt it better and easier to integrate with the MAX17048 breakout vs. trying to get battery voltage out to the main board and read it with an ADC pin and voltage divider, etc.

Ultimately, I would try to integrate the solar charge manager and battery monitor directly into the main board.

An additional point of consideration for power is utilizing low power modes and optimizing and reducing the power consumption. It is not sufficient to use low modes on the MCU if the connected circuitry continues to draw power even when the MCU is sleeping. I therefore added high-side switching controlled by the microcontroller so it can power down the sensors etc when it is done with measurements and goes to sleep, and power up the sensors when it is ready next to read data. There may need to be some conditioning time factored in for reading some of the sensors.

Microcontroller and Networking:

Since the devices will be deployed in relatively remote areas along the river (about 20-40 miles), it was decided to utilize Lora radio with Meshtastic, a decentralized peer-to-peer mesh network.

A lot of Meshtastic supported hardware devices (https://meshtastic.org/docs/hardware/devices/) utilize the Semtech SX1262 as the Lora transceiver. SeeedStudio's Wio-SX1262 is one such module that is plug-n-play with their Xiao ecosystem.

There were a few considerations for the microcontroller:

• Meshtastic support
• Low power modes
• Castellated design for direct integration on main board
• Enough horse power for AI models

SeeedStudio's Xiao ESP32s3 ticked off all the boxes and so is utilized in this design. The Xiao ESP32s3 Sense includes an integrated camera and so appeared very attractive. However, our design utilizes two cameras, each running different AI models and so we instead opted to utilize Grove AI Vision 2 Camera modules which can run a model directly (thus leaving the ESP32s3 for other models and functionality as desired) - they also integrate with the base MCU via I2C and so we could integrate two cameras as desired. This choice also allowed for the Xiao ESP32s3 + Wio-SX1262 (directly plugged into the top connector of the Xiao) which optimized placement.

Edge AI and Image Classification

I decided to utilize Gove AI Vision 2 Camera modules. The image classification models can be off-loaded to the individual camera modules, leaving the base MCU free to handle its workload and incorporate other models and functionality as desired.

The underwater camera will run classifications of fish and trash. We will be using existing open source models and fine-tuning them. We will use Edge Impulse to help deploy the models to the Grove AI Vision 2 Camera modules.

The above water camera will be used to identify and ideally classify wildlife, and objects like floating trash, etc.

In order to optimize solar/battery usage, we will periodically wake up and check and record any positive classification into the available SD cards.

Future Work:

For this iteration, the problem and solution have been identified, and the design for the Sensor Pod subsystems has been thought through and the initial prototype developed. I'm still waiting for the design fabrication and assembly (getting the problem and solution, and careful design of the subsystems to satisfy robust remote deployment with clean data measurement took awhile, so that delayed the PCB fabrication and assembly).

The next steps of the project are:

• 1) Assemble the Makara Sensor Pod circuitry
• 2) Write the basic firmware and test out the functionality
• 3) Deploy existing image classification models to the AI cameras to test the systems
• 4) Perform calibration of the liquid sensors
• 5) Test meshtastic and receiving the data at the base station
• 6) Put the electronics in a waterproof enclosure
• 7) Test out the integrated Sensor Pod in controlled environments (e.g nearby pond or pool)

If everything goes well, the next steps would be:

• 1) Develop the base station firmware
• 2) Have a rudimentary backend for data collection and for the creation of analytics and reports
• 3) Display visual storytelling report on base station

Of course, while all of this is progressing, also connect with entities like the Chennai River Restoration Trust, Environmentalist Foundation of India (EFI) and others to find other like-minded organizations and individuals to make Makara a reality!

Thank you and Happy Making! 😊