Our design focuses on two goals. The first goal is to measure BOD, which is the most important factor in determining overall water quality. The second goal is to maximize efficiency and cost-effectiveness by reducing the amount of contact sensors required.
Biochemical oxygen demand (BOD), the industry standard water quality indicator, generalizes the degree of organic pollutants in a body of water. It is the most important value in measuring overall water quality compared to other values such as acidity or turbidity. A higher BOD value indicates greater amounts of organic matter demanded by oxygen-consuming microorganisms. However, BOD is not an easy parameter to measure. Traditionally, it is calculated by measuring the difference in dissolved oxygen over a period of five days at 20 degrees Celsius without light. This is not feasible to run autonomously, especially with the typical method which requires adding chemicals over time.
Recently, researchers have shown that tryptophan-like fluorescence has a strong correlation with BOD . This is measured by exciting the water at tryptophan’s maximum absorption wavelength at around 275 nm, and then observing the intensity of the emission around 350 nm, resulting from the fluorescence of tryptophan, an amino acid. Although a multiple regression has to be performed using more parameters, we believe it is the most viable option for cost-effective yet accurate water quality sensing. It also has the advantage of extremely low power consumption, as the light source only has to be switched on when taking measurements.
Current commercial optical BOD sensors are very expensive. We propose a cheap and easily fabricated design which would work as well as commercial offerings. The sensor setup is similar to a fluorometer, where an LED shines onto the water, and a photodiode detects the emitted light from the fluorescence caused by the LED. This value will then be regressed along with turbidity and other parameters. We intend to use acrylic (PMMA) with UV-transmissive properties as our windows around these components in order to allow the light to pass through.
Our second design goal is to minimize the use of contact sensors, in order to improve durability and reduce maintenance. Currently, most inexpensive water quality sensors, which would likely be found in IoT systems, are contact sensors, which are not conducive to long-term, low-maintenance deployment in the water due to corrosion and biofouling. However, as shown in BOD sensing, there have been several advances in measuring parameters optically, rather than through contact methods.
A similar setup to the BOD sensor can be used for turbidity. Although there are existing turbidity sensors for use in washing machines and other appliances, we found that they did not suit our needs due to their high current draw (~30 mA at 5V) and bulkiness. We propose another optical setup to determine turbidity, where an IR LED would shine across a small region of water and the resulting intensity would be measured via another photodiode.
For a pH sensor, a non-contact sensor is relatively difficult to produce. One method we will attempt is optical pH indicator sensing. This operates by observing the color of chemical pH indicators (e.g. phenol red) when exposed to water. The indicator should be attached to some permeable substrate, such as silicone or mylar, and measured with a color sensor. Since this is a complicated setup, we intend to use a regular pH probe as a backup.
Our full sensor node design consists of the aforementioned BOD, turbidity, and pH sensors, as well as a temperature sensor. The microcontroller used will be an STM32L011, which has a very low current draw. For communications, a LoRa module, such as the SX1231, will be employed to communicate with LoRa gateways connected to networks (such as The Things Network). We chose the LoRa protocol because of its long range and low power, allowing for fewer gateways and more remote placements. An application server will then handle further analysis and visualizations.
The cost of the electronics for a single sensor node should be less than $20 when mass manufactured. The network of sensors will be strategically placed in target areas of interest—the number of devices and placements to be determined based on the individual body of water.
With the use of the low-power optical sensors, as well as LoRa, we have calculated that the nodes can last well over a year on a 2300 mAh lithium ion battery when sending data every 30 minutes, without the need of an external rechargeable source such as a solar panel. This design should require almost no human interference, except battery replacement at the time of the annual maintenance checkup.
In developing regions of the world, networks of these sensors will drastically improve the speed and cost of water monitoring. It will also allow people to measure contaminants such as glyphosate, the most widely sprayed pesticide in the world, indirectly through BOD. This is significant because pesticides and water pollution have been leading causes of ecosystem deterioration for humans and wildlife. Furthermore, measuring BOD, which typically requires lab testing, can be measured instantly.