19,000 will die today not from high-blood pressure, smoking, or lack of access to food and water, but from simply breathing.
While often overshadowed by more immediate health risks, such as vehicle collisions or heart attacks, exposure to poor polluted air is quietly killing 1/8 of the world.
Per the World Health Organization, ambient air pollution now accounts for about 7 million deaths per year, making it the fifth leading cause of premature death across all ages and sexes.[1] Furthermore, more than 90% of these deaths occur in low and middle-income regions (mainly in Asia and Africa) which do not have the resources to combat the global health epidemic.[2] Outside of human suffering, air pollution is also massive burden on the global economy as well. When accounting for disease-adjusted life years (DALYs) and productivity stints (i.e. worker illness, damaged crops, general reduced productivity, traffic accidents, acid rain damaging buildings and crops, etc), the World Bank estimates that air pollution costs the world economy about $225 billion year to year.[3]
Recognizing this, scientists, researchers and policymakers alike have begun to address this epidemic through implementing measurement systems across the world. Conducted primarily for legislation surveillance and scientific research purposes, air pollution monitoring networks in major cities such as Beijing, Delhi, Paris, and many others have given environmental officials and researchers the insights necessary to fully understand and combat air pollution.[4] With greater understanding of air pollution trends, states have begun to inform and protect the public through continuously updated monitoring mediums and creating new laws, such as Delhi’s even/odd driving protocol or China’s War on Pollution policies, to lessen pollution and ultimately citizen exposure to harmful pollutants.
Nonetheless, monitoring networks are still limited by their size and fail to provide individual families direct insights on their personal exposure to harmful pollutants. A new wave of systems, low-cost air quality devices, aims to bridge these gaps. While limited in the accuracy of their measurements — studies and programs have shown promising potential the proliferation of low-cost monitoring devices has. As a recent Columbia University and Volvo-funded research project in the Mathare slum outside of Nairobi showed, regardless of the accuracy or usability of data provided by low-cost monitoring devices, exposure to such systems increase civilian consciousness of air pollution risks and invigorates local communities to lobby their officials or even address the issue through grassroots programs (i.e. planting trees, walking to work) as well.[5] In addition, these systems allow for a new-age process of research and policy-making called “Citizen Science”, in which civilians collect and provide expansive data to officials which in turn, create policy to address insights seen in the vast sums of information.[6] Despite these promises, however, certain communities continue to question the benefits of mass low-cost monitoring deployments due to validity concerns in their data.[7] Regardless, these monitoring devices are more economically accessible, user-friendly, mobile, and enable individuals to track their local exposures to various pollutants and relevant data points, among them carbon monoxide (CO), particulate matter (PM2.5), nitrogen oxides, temperature and humidity.[8] All in all, enhancing global awareness and combat against air pollution.
In the following, a detailed description to create a low-cost personal air monitor and an example of its use is provided in an effort to allow individuals to create their own monitors and in turn, begin adjusting their lifestyles to address health risks brought on by air
Sensors, Microcontroller and Wiring
Our device consists of 12 individual pieces ranging from sensors to wires and casing. Beginning with sensors. The device uses two sensors wired to a breadboard and micro-controller which all together give the monitor the ability to conduct and collect data on three pertinent measurements, temperature, humidity and PM2.5.
Beginner with the simpler device — the DHT-11 Integrated Sensor measures local temperature and humidity levels. In terms of the engineering behind the device, the temperature sensor uses thermistor technology in that a small resistor is located within the device whose electrical resistance changes as the temperature around the thermistor fluctuates. In contrast, the humidity sensor uses a moisture-holding material between two electrodes, thus as moisture changes as does electrical resistance allowing for humidity readings.
Following this, the PPD42NS sensor provides PM2.5 readings, which as mentioned in the introduction, are vital to understanding pollution in an individual’s locality. The PPD42NS uses a nephelometer to measure PM2.5. The process begins with a small heater drawing air through a tunnel in the device in which an LED light shines on the air. In turn, the device analyzes the amount of scattered light due to particles moving through the device and provides a measurement of PM2.5 levels in pcs/m3.
To collect and send these measurements, the sensors are connected to an Adafruit Feather HUZZAH ESP8266 Arduino micro-controller. Arduino micro-controllers, first designed in Italy and now a major cornerstone of the global “maker movement”, essentially act as a very small, easily programmable computer that allow users to control sensors (or any connected device for that matter), read data, and send it for further storage to a third-party. Our micro-controller, the Feather HUZZAH, provides us the ability to interact with our sensors and through integrated wifi technology and battery charger capability, collect data portably and send it to servers for further analysis. The micro-controller is also equipped with an Adafruit SSD1306 (OLED) display screen that allows direct view of the micro-controller’s connection process and real-time measurements.
All of the devices are connected through an Adafruit half-size breadboard and breadboard jumper wires, which like any breadboard, allows for solder-less wiring to integrate the components of the monitor together. The following picture provides a view of the main components of the monitor and their connections throughout the breadboard.
3D-Printed Parts and Waterproof Enclosure
In addition to the electrical components of the monitor described above, the project also relies on a set of 3D printed parts and a waterproof enclosure to assist in proper placement, final assembly, and protection of the monitor. These parts are a 3D-printed monitor case and lid, a 3D-printed waterproof box adapter, and finally, the waterproof enclosure itself.
Beginning with the 3D-printed components, the monitor case and lid protect the device from movement or damage, essentially securing all the solderless wires, micro-controller, display, and DHT-11 sensor in place. In addition, we also rely on a 3D-printed box adapter piece which connects the monitor to the inside of the waterproof case. Links to the relevant Thingiverse page (with corresponding STL) files are provided below. Finally, it is worth nothing that a “slicer” program may be needed to convert .stl files to .gcode files to be compatible with 3D printers, links to two recommended slicer software programs are included below.
Thingiverse: https://makerware.thingiverse.com/thing:2857819
Slicer Programs: http://slic3r.org/ AND https://grid.space/kiri
Finally, a waterproof enclosure is also recommended to protect the device when used outside. This Adafruit Large Plastic Project Enclosure with a clear top allows us to make the monitor both weatherproof while still allowing for a complete view of the device, a photo of the monitor within the enclosure is shown below. Nonetheless, some saw work is necessary to open a hole in the bottom of the case to allow for air to enter the enclosure, and subsequently the PPD42NS monitor enabling PM2.5 readings.
Battery and Solar Panel
To allow for mobile monitoring, our monitors rely on self-powering system consisting of a solar PV panel, charge controller, and lithium polymer battery. Beginning with the latter, our device contains a LiPo (lithium polymer) 3.7V, 2,500 milliamp-hour battery which can roughly provide enough energy for the monitor to run for two weeks on one full charge. To supplement and charge this battery, the system uses a Medium 6V 2W solar panel. This panel is waterproof, scratch and UV resistant and extremely lightweight and durable. Finally, we must also use a charge controller to control power supply to the monitor (switching between PV and battery as necessary), block the solar PV panel from charging the battery when it is full, and handle the appropriate voltage provided to the monitor in cases where the solar PV panel is overstimulated by sunlight. In short, balancing the entire transference and use of power through the monitor, battery, and solar PV panel. Furthermore, we also have the ability to charge our battery directly through a mini-USB port on the charge controller should the PV panel not be sufficient (photo below). Finally, it is worth mentioning that some soldering is needed in the assembly of the charge controller.
Overall Costs and Cost Per Device
Links to each of the devices needed to create this monitor:
Adafruit Feather Huzzah with ESP8266
DHT-11 Basic Temperature-Humidity Sensor
Adafruit SSD1306 FeatherWing OLED Display
Adafruit Lithium Ion Polymer Battery
Data Connectivity and Hosting
As mentioned earlier, the device has the ability to connect to the internet. Through the Wifi integrated Feather HUZZAH ESP8266 micro-controller and associated Arduino code (described in greater detail below), the device can connect to any wifi provider given that the wifi SSID and password are known. In turn, we have connected are devices to ThingSpeak servers to which our data is uploaded and stored. While in theory these methods are soundproof, actual application has shown that issues with wifi interruptions will distort data collection. Furthermore, should there be any issues with ThingSpeak servers, our data hosting is also vulnerable to distortion or loss.
Description of the Code
Our micro-controller is controlled through an original Arduino set of code that instructs and conducts all steps from activation to ultimate data dispersion for storage. To begin, we declare sets of global constants and include all relevant libraries. Processes included in this section are specifying the Wifi settings, ThingSpeak settings (for data collection and storage), DHT-11 settings (indicating the variables to measure and pin locations on the micro-controller/breadboard), the PPD42NS (same process as DHT11), and the SSD1306 OLED Display.
Following this, we enter the main setup function of our code. We begin by declaring the “pinMode” (i.e. locations) for data input of the two sensors and continue by setting up the display (i.e. text size, text color, rotating screen, declaring button A as pause and button C as a resume function). Next, the code connects the micro-controller to your specified Wifi and begins a “stopwatch” which indicates time between measurements. Once that stopwatch reaches the specified sample time, the code reads (editing if measurements are above a maximum value) the data and displays it on the OLED display. Subsequently, the micro-controller connects to ThingsSpeak’s servers and uploads the data to the appropriate library (using unique API keys that have already been specified in the global constants). Finally, the code resets the stopwatch, effectively sending the code back to the beginning of the measurement loop. This process continues until an intentional or unintentional interruption. The code can be seen and accessed through GitHub below. A link to download the Arduino integrated development environment is also included.
Air Quality Monitor Code: https://github.com/colinmccormick/Georgetown_STIA315_AQ_monitor
Arduino IDE: https://www.arduino.cc/en/Main/Software
Description of Testing
In order to prove the usability of the device, a 24-hour monitoring test was devised and conducted on the Georgetown University campus in Washington, DC. For background, the primary student gym Yates Field House and outdoor turf field Kehoe Field are located on top of one another in the west corner of campus directly next to the university’s heating and cooling plant. To measure the impacts this plant has on air pollution in the area and provide a recommendation on best times to exercise per air quality levels — the device was placed directly on Kehoe field, shown below.
Placement of the device, however, was imperative. First of all, being that the device relies on a solar energy for power, the PV panel had to be placed in a position that would expose it to sunlight throughout the day. In addition, as mentioned above, our PPD42NS PM2.5 sensor requires that air be pulled in (being that it is engineered as a nephelometer). Thus, the monitor has to be located on an elevation to allow for the transference of air. The positioning of the device in shown in the picture below.
In terms of effectiveness, the monitor was extremely successful in measuring temperature, humidity, and PM2.5, and no issues were seen in the sensors in the first 14 hours of testing. The final 10 hours however brought many issues. Specifically, continuous provision of power, data hosting and transference, and sensor issues all proved to both be major dilemmas. Beginning with power, having charged the battery for 2.5 hours prior to testing, the monitor died about 14 hours into the test (around 6:00 AM). This can be attributed to the panel not being exposed to sunlight throughout the evening and thus relying solely on reserve battery power which unfortunately was not full and in turn, died quickly. In addition, once the monitor’s battery was replaced and measurements began around 12:00 PM the following day, although the monitor was conducting measurements and displaying them on the OLED screen, the data failed to transfer over to ThingSpeak. Thus, intermittent pauses in data collection can be seen around the 12:00 PM to 1:30 PM time range on May 9th. Finally, around 1:30 PM on May 9th, our DHT-11 temperature and humidity sensor stopped conducting measurements and began providing null values. After rewiring the monitor, the DHT-11 began conducting measurements again around 2:30 PM. Finally, the new battery was also not fully charged and died about 30 minutes prior to end up the test, around 3:30 PM on May 9th.
The following graphs show the change in temperature, humidity, and PM2.5 throughout the 24-hour test conducted from 16:17 May 8th to 4:00 PM May 9th . As mentioned above, it is worth noting the 6-hour gap between 6:03 May 9th to 12:00 May 9th due to battery issues, intermittent pauses from 12:30 PM to 1:30 due to data hosting issues, and failure on the DHT-11 sensor from 1:30 PM to 2:30 PM on May 9th.
Despite interruptions, the graphs clearly shows a trend in PM2.5 levels as it relates to temperature in that as temperature drops as do PM2.5 levels. Something else worth noting, however, is the amount by which PM2.5 levels jumped (and varied) when the cooling and heating plant is running, primarily during normal work hours. Finally, weather was clear and ranged from 57 degrees fahrenheit at a low to 77 degrees fahrenheit at a high thus there is no extraneous weather pattern (i.e. rain, high winds, etc) that may have skewed the data. Thus, in terms of recommendations, the best time to exercise on Kehoe Field if trying to avoid high levels of PM2.5 is during the early morning or late evening hours.
Suggestions for Improvement
With regards to improvement, my primary concerns lie with power provision and data transference/hosting. To create a viable low-cost air quality monitor, we must be sure of its continuous ability to track and report data. Thus, any improvements to battery size or solar power acquisition will assist in making a truly legitimate air quality monitor. In addition, we must refine the data hosting and transference abilities of the monitor by either amplifying its wifi connection capabilities or carefully choosing locations that will not be prone to wifi shortages. Next, some long-term analysis may be in order to check if it would be worth looking into a more reliable temperature/humidity monitor. Although the sensor was entirely checked and rewired to the monitor after it failed, it continued to only produce null values. While the cause of this could very well still be due to user error (since the sensor acted effectively throughout the 5-month course leading to this test), some research on possible substitutes would not hurt. Finally, with some coding, the monitor can also begin to track many other data points such as averages, percentile rankings, daily high/low values, percentage change from previous days, and more throughout measurement cycle, which would offer even more insights on to air quality levels.
Closing Thoughts
Building and programming the device were fairly simple processes given the time, focus, direction and detailed instructions that were provided in our course. Likewise, testing the device was also uncomplicated, only requiring the press of a button and proper placement to allow all sensors to act accordingly. With this specific test, we have provided the over 7500 students in the Georgetown student body and insight on best times to work out. More importantly, however, we have proved the legitimacy of a monitor of this type.
Nonetheless, solar power did prove to be an issue and thus room for improvement exists within this regard. While charging the battery to full capacity will certainly increase longevity (up to two weeks perhaps when coupled with daily solar charges) — any additions to power provision will only allow for longer tests and less interruptions. In addition, as mentioned in the suggestions for improvement, issues with data transference and sensor reliability also will need to be addressed in the final production of a monitor of this type and cost.
All in all, though, the device performed as expected and provided a direct view into a vital component of air quality, PM2.5 levels, as well as insights on how temperature and humidity influence particulate matter levels. In conclusion, this monitor (or any low-cost monitor of the type) holds potential in two routes — widespread data collection and more importantly, in increased awareness of air quality concerns. While researchers, scientists and environmental officials will continue to rely on traditional, high-cost monitors — value will still be found in the large trends across wide testing areas that low-cost monitors will provide. Finally, as shown through the Mathare experiments and our class interest itself, giving a person (even a very general level) view of their exposure to air pollution will undoubtedly influence their decision-making process in both their own exposure to harmful pollutants and their lifestyle choice that may be attributing to global air pollution.