Thursday, September 5, 2013

Understanding air pollution with a simple dust sensor

Outdoor air pollution, in the most extreme cases, can be immediately identified even without any special training. It casts a haze over cities, collects on streets and buildings, and provides dramatic fodder for the news. Even when the air pollution isn't actually visible, we can smell when something isn't quite right.

I previously wrote about how difficult it can be to obtain basic environmental data, and how government budget cuts are threatening air monitoring networks in several states. It now appears that other countries are making hard decisions about which monitors to keep, and which monitors to shut down. The Guardian reported recently that up to 600 air quality monitors, including monitors for nitrogen dioxide and particulate matter (PM), could be shut down across the United Kingdom.

Yet for all the attention the media pays to outdoor pollution, people spend only about 1 to 2 hours outdoors (and that's only in the pleasant summer months) according to one University of Newcastle study. According to the EPA, we spend about 90 percent of our time indoors. We spend the vast majority of our time indoors, so it makes sense monitor pollution in the home.

Indoor Air Pollution (IAP) is an especially big problem in developing countries, where 60 to 90 percent of households still rely on coal and wood for heat and food preparation. About 36 percent of acute lower respiratory infections and about 22 percent of chronic obstructive pulmonary disease (COPD) in the developing world are caused by IAP (pdf). In one study of women in China, researchers found that a 10 μg/m3 (microgram per cubic meter) increase in PM1 (ultrafine particles smaller than one micrometer) was associated with 45 percent increased risk of lung cancer.

IAP isn't just a concern in developing and BRIC nations, though. Similar problems exist for the rural poor in the US and Canada, where indoor pollution exceeds the World Health Organization air quality guidelines in up to 80% of homes. As in BRIC nations, these homes rely greatly on burning organic fuels.

Air quality at home can be an issue even for homes that don't burn wood or coal. Indoor air pollution can come from "molds, bacteria, viruses, pollen, animal dander and particles from dust mites and cockroaches," according to the American Lung Association.

Indoor air pollution ranks among the top five environmental risks to public health, the EPA says. Indoor pollution levels may be two to five, and sometimes 100, times higher than outdoor pollution.

All that makes the indoors a great place to put a dust sensor.

More about the sensor

The DustDuino is a project I first announced on this website in May. Its goal is to provide a cheap means for citizens to crowd-source accurate environmental data. The DustDuino is not actually a sensor, but rather is a sensor node, which is a device that gathers, processes, and sends sensor data.

The first version of DustDuino uses the Sharp GP2Y1010 infrared LED dust sensor to obtain particulate matter (PM) readings. It's a fairly handy bit of electronics for the price, considering its output is easily converted into EPA's standard unit of measurement for particulate matter (μg/m3 or micrograms per cubic meter) and can generate 100 dust readings every second.

But most importantly, it can theoretically sense changes in dust as small as 1μg/m3 with the Arduino Uno's ATmega 328P microcontroller. For reference, the maximum National Ambient Air Quality standard for PM10 pollution (coarse particulate matter 10 micrometers in diameter and smaller) is 150 μg/m3 , averaged over a 24-hour period.

It's important to note this sensor, like any cheap sensor, has its limits. Sharp notes in documentation that the sensor should not "be used for or in connection with equipment that requires an extremely high level of reliability and safety," such as nuclear power control applications, or in medical or life support systems. A study at the Karlsruhe Institute of Technology noted that these sensor tended to not be calibrated very well from the manufacturer, with a maximum deviation between sensors almost reaching 100 μg/m3 .

In tests, the GP2y101 tended to be especially noisy, but averaging the readings over any time period longer than five minutes yielded meaningful measurements. Yet researchers concluded the device was well-suited for "Participatory Sensing applications, such as creating maps pointing out pollution hot spots or the erection of inexpensive measurement grids, as well as personal appliances, e.g. exposure logs or warning systems."

That's exactly what the DustDuino is intended for.

First 24-hour trials of the dust sensor

Over the course of three days, the dust sensor node was deployed in three separate locations: in the living room (5/16), in the basement (5/21), and in the back yard (6/4). Every minute, the node would take a reading from the sensor, interpret the reading in terms of mass concentration, and send that data via a home WiFi network to a Xively server.

After all the data was posted, I queried the Xively server to send back the data in CSV format (Comma Separated Value). I placed the CSV data into Microsoft Excel, which I used to graph and compare the data. Above is a graph which compares sensor readings from all three locations.

Three things stand out in the particulate matter data (circled in the graph above):
  1. Living room dust levels jumped first thing in the morning. The blip in the living room data occurred between 14:00 and 15:00 UTC, which translates to 9 am and 10 am CST. On this, a Saturday, that corresponded to the time I woke up and began making breakfast (a delicious breakfast of bacon). Cooking is associated with an increase in airborne particulates in this household, which is consistent with the literature on indoor pollution. Also note blips at 11:45 am and 5:41 pm; both times when food is prepared in this household.
  2. Outdoor dust levels rise in the morning, and peak in the afternoon. This is generally consistent with other research that monitored outdoor particulate matter over the course of a day. PM levels in the environment increase as human activity increases during the day. However one possible exception to this rule occurs during the winter, when cold temperatures lead to an increase in the burning of fossil and organic fuels for heat.

  3. Basement levels seemed to increase slowly over the day. There apparently aren't many studies that compare pollution between multiple floors of a house. One study did find that "one person walking in the basement resulted in a much higher human exposure but a lower source strength than one person walking on the first floor." (Ferro, Kopperud, & Hildemann, 2004) However this study took place in a finished basement, which had a much smaller "mixing volume" than did the first floor, which allowed "concentrations to build up." This isn't exactly the situation in this house, as basement is approximately the same area as the first floor, is unfinished (concrete floor), and has no walls.

    What we do know is that things like radon gas can collect in basements over time (pdf). We also know from research (and personal experience) that wet basements do a very good job at growing mold, which contributes to PM10 pollution. This basement can be particularly troublesome when it comes to water infiltration, and the humidity can exceed 50% at times. Unfortunately I did not record the actual humidity of the basement at the time, nor was the sensor able to differentiate between mold and other particulate matter, so I could not determine the culprit. Yet I'm inclined to believe this slow upward trend in basement PM could be linked to an increase in mold growth.

Comparing indoor and outdoor air quality

So each location seemed to exhibit a unique pattern of particulate matter concentration over time, but what can we say about air quality in general? Was the air better or worse inside, in terms of particulate matter?

Above is a chart called a "high-low-close" chart, or a "stock" chart. As its name suggests, it's used to compare how high or how low stocks may trade in the market. It also can be used to illustrate scientific data, which is exactly what this one is doing.

The three bold, vertical lines represent the three locations where the dust sensor was placed. The high point of each line (which terminates in a diamond shape) corresponds to the maximum dust concentration that the sensor registered for any given location. Conversely, the low point of each line represents the lowest concentration of particulate matter that was measured for that location. You can probably guess that the red triangle corresponds to the average reading for any one location.

What can we interpret from the stock chart of the dust readings?
  1. The basement had the smallest range of dust readings, but the highest average over 24 hours. Human activity kicks up dust, and there was limited human activity in the basement for this period. Having said that, this is an enclosed, damp environment that is conducive to mold growth, which likely is the reason for the high average.
  2. The living room had the highest peaks of any tested area. Again, human activity kicks up particulates, which the dust sensor reads as short spikes of high dust concentration.
  3. The back yard had the lowest average dust readings over 24 hours. Lo and behold, over a 24 hour period, the air outside had fewer airborne particulates on average. This is consistent with the research above that shows homes can have higher PM levels than the outdoors.

Comparing readings to air quality standards

While it's nice to know whether you're experiencing higher-than-average dust levels at home or in the back yard, the data doesn't become that useful in the broad scheme of things until it's actually compared to air quality standards, and accurate measurements from elsewhere. After all, what does it mean to have elevated PM10 levels? Or low PM10 levels?

I've done some research and collected air quality standards from the United States and abroad, and also found some real-world measurements from EPA PM10 monitors. I've collected those in the above table, and included my own data samples for reference.

Interestingly, OSHA has very high limits on indoor air quality for general dust (more than 33 times the federal limit for outdoor air quality). An aforementioned study in China found indoor particulate matter concentrations that exceeded 1.5 times the American outdoor limit.

The Illinois Environmental Protection Agency (IEPA) hasn't published an annual air quality report on its website since 2011, so I used that report to find the single highest 24-hour period sampled in the state. It turns out there are only four PM10 monitors in the state (three in Cook county, and one in Madison county), and one of those monitors is set to be discontinued according to the 2014 Ambient Air Monitoring Network Plan.

The EPA has an online Air Quality Index calculator to help citizens find out what it kind of air pollution it takes to turn "good" air into "moderate" or "unhealthy" air. As it turns out, at 55 micrograms per cubic meter, the AQI changes from "good" to "moderate."

The European Union has much more strict standards for air quality. The 24-hour limit for PM10 pollution is 50 μg/m3, which is three times the American limit.

The measurements taken by the Sharp sensor, if accurate, show the outdoor and indoor air quality is much better than all of these benchmarks.

Summary: It's "good enough"

We can't really be sure whether the tiny, inexpensive Sharp dust sensor took accurate readings of particulate matter concentrations in the environment. That would require calibration with equipment costing several times more than this humble infrared device. But from laboratory tests and this simple study, we can say this device produces fairly precise readings which are consistent with the research and literature on air quality in the home and outdoors. It can measure trends in air quality, both short-term and long term, and is suitable for crowd-source efforts to identify point-source pollution in the home or outdoors.

If I had to pick on one shortcoming of the dust sensor, it wouldn't actually be its accuracy. If the device is precise, accuracy likely can be dialed-in with the correct testing equipment and procedures. Rather, its biggest drawback is an inability to determine exactly what kind of particulate matter is wafting in the air.

Is the dust PM2.5, or PM10? Or particulates greater than 10 micrometers? The answer matters, because the smaller the dust particles are, the greater risk they pose to health. It also matters because to compare this data with any of the air quality standards or air quality research, we must be able to determine the particle size. We don't know the range of particle sizes from this sensor, because Sharp has not provided that information.

For a $10 sensor, the Sharp GP2Y1010 is "good enough" for many applications. People wanting to compare the air quality of their homes or outdoor spaces can do so easily and cheaply with this sensor. This sensor also could be calibrated to mitigate some of the accuracy issues.

However, there may be other sensors on the market at about the same price point, that use similar methods to measure airborne particulate matter, that might provide better accuracy and size discrimination of airborne particles. I'm currently testing one such sensor, and will have results in time.

1 comment:

  1. Hi, very interesting and detailed post, all this extra information you found seems very useful for what im doing, Thanks!

    Anyway, i have a question that maybe you could answer me:

    I have a sharp sensor that i tried to use with the chris nafis [] code, but i get outdor messurements with negative values very often, since my sensor it wasn't calibrated exacltly the way him sensor was. and i commented a line that hardcoded any value below zero to zero because i thinked it was ugly.

    now my question is, ¿how can i find the zero value in volts for my sensor (when pollution is zero)?