Question: What are some activities for using the Simple Air Sensor as part of a broader air quality project?

Could this demonstration be linked back to the particle size (2.5 microns) that the air sensor is set to detect? Maybe start a discussion on which of the 3 settings that the air sensor can use is best for detecting which common irritants and the applications of each?

Here's an example of teaching with similar air sensors from the EPA: https://www3.epa.gov/airnow/teachers/gh_pmsensorkit_handoutandinstructions.pdf

Hi @nshapiro - these are all good questions, let me try to go through them one by one, thanks!

the documentation needs to be vastly improved

I'm hoping that we can get into a lot the "what's this good for" through this question, and to build out documentation as we go.

What do you mean by "high precision" in describing the sensor? In my understanding Plantower doesn't use that language when describing the product. How might that language be misleading to non-experts?

Sorry, I don't know where the "high precision" text is, can you point it out?

Oh, i see - on the store page -- i believe that was just copied from the page where we bought the sensor, but I agree we should change that. I'll edit that page now, thanks!

How does a single PMS 1003 sensor differ from the Purple Air (so that people don't think they are buying just a cheaper version of the purple air w/out quant and logging features)?

The Purple Air is not open source, so it's hard to know exactly what's going on in software, but it is based around two Plantower PMS5003 sensors, and part of our interest in DIY/open source devices is to provide transparency and accountability around what exactly closed-source devices are doing. I think there are people who've spoken directly with the Plantower and Purple Air teams and have more information on what's going on inside, so this is one way to get at this question. The Purple Air uses 2 sensors for redundancy, but I'm not sure if it averages the data, or just throws it out if they don't match.

What does it mean that these sensors are factory calibrated?

Since they're not open source, it's hard to say, but I know that folks organizing big Purple Air monitoring projects are re-calibrating them after purchase. There is a lot of information about this in the context of Purple Air's use of them in:

• https://publiclab.org/questions/warren/01-04-2018/how-good-is-the-purpleair-sensor-and-what-does-it-detect
• https://publiclab.org/notes/guolivar/01-08-2018/thoughts-on-low-cost-air-quality-sensors
• https://publiclab.org/purpleair

What conditions are these sensors good for and what conditions make them wonky?

I'm interested in developing not only materials around this, but also trainings and event models to help people plan study designs, and since the Simple Air Sensor uses the same sensor as the Purple Air, I hope it will make for a good platform to empirically determine how conditions affect readings and build knowledge around this in a group context.

What kind of particulate matter do these sensors work best in (ie they tend to preform worse in traffic-dominant pm than wood smoke-pm)? How fast does factory calibration deteriorate?

These are all great questions for any device that uses Plantower sensors, and because the Purple Air does, there are some threads about this connected to the Purple Air topic:

• https://publiclab.org/questions/Cbarnes9/11-09-2018/what-type-of-aerosol-was-used-for-the-calibration-of-the-purpleair-lasers (esp. this comment)
• https://publiclab.org/questions/Cbarnes9/11-09-2018/does-particle-composition-that-the-purple-air-monitors-measure-greatly-skew-the-data-output

What are the cut off points for the different colored lights and why?

The intent of the light isn't to tell you what is safe or not, but to show relatively what the sensor is detecting. Because of the cost and complexity of the Purple Air, it's harder to understand the effects of, say, a small breeze, or the position (1ft off the ground vs. 4ft) so we are hoping people can try these kinds of things with the Simple Air Sensor to get a better understanding of what the Plantower sensor is "good at". But you can look in the source code on lines 70-75 to see how we map the pm25_standard value to a color wheel. And, if you connect it to the Arduino IDE, you can read the numeric data on the serial port.

Which would also lead to a question of error (which remains unanswered here: https://publiclab.org/questions/wu_ming2/02-02-2019/how-do-i-calculate-error-for-averages) as the user might want to know what the lights mean and if it changes to a scarier color how accurate + precise that change is.

I agree some people will want to know this, but perhaps they would need to read the data over serial, or work with a different device. This certainly isn't the best device out there, and what we're trying to offer is a kit which can help people with questions or use cases that are quite different from that of another device. But we're open to ideas on how some of these things could be done with a Simple Air Sensor. Maybe it could blink, or make a noise, or something?

As to the error, I guess this is a good question for all Plantower-based devices, and I'll see if we can connect that question you linked to up with some of the other threads regarding Purple Air devices, and also see if we can get some folks who have offered input on this in the past to chime in.

Why should one be interested in just the 2.5 data if the sensor counts (estimates) three sizes?

Well, the way it's set up now, we can only output one type of data. I'm interested in how to change this, so if you have any ideas, I'd love to hear them. We could maybe use 3 lights, although it'd be harder to read, or we could try adding a switch. Or we could think about some kind of blinking pattern.

Could you specify on the store page who this sensor is for? I sincerely don't understand how this sensor can be "not intended for data collection" while here you are asking for how it can be used for an air quality project.

It's for some of the use cases described on this page, and some listed at https://publiclab.org/simple-air-sensor#Uses -- and we are hoping to build these out into well-documented activities that people can do. Broadly, it's intended to help people who do not want to program a sensor or build electronics, but want to try using one of the recent generation of Plantower-like sensors in scenarios where immediate feedback is helpful, and non-quantitative data is still useful. This might be alongside more expensive devices, in an event to build knowledge about the effects of conditions like wind, temperature, humidity on sensor readings, or in a classroom setting.

Nick, I'm going to reach out because I'm hearing some frustration on your part. We're grateful for your input and thank you for asking some great questions.

Ah, and regarding What kind of particulate matter do these sensors work best in (ie they tend to preform worse in traffic-dominant pm than wood smoke-pm)? How fast does factory calibration deteriorate? -- maybe you already read this, but @guolivar discussed the calibration in relation to those two types of particulate matter in this post: https://publiclab.org/notes/guolivar/01-08-2018/thoughts-on-low-cost-air-quality-sensors#regulated+air+quality+measurements

It depends, I have found that the Plantower factory calibration does a pretty good job for wood smoke dominated aerosols but not that great for traffic dominated aerosols. At the moment I'm testing their response to "natural" sources (sea salt, dust/sand, etc) as I'm preparing an article reporting on what we've learned about these units.

Maybe @guolivar has published this report by now?

Just relying quickly as its late in Newfoundland. Thanks for your response! It would be great if the limitations of the device and what is still unknown could be listed up front and very clear language was used on the store page to describe what applications this device was good for like "Try using this kit for educational and experimental use, but please try a more expensive and validated sensor like x if you are trying to understand your air quality." I think @mimiss 's example, with a modified handout, is a great one! The examples that you link to now of speculative uses doesn't make that clear to me. In my experience, intuitive color coding (from green is good to red is dead) can make people even more freaked out than quantitative readouts that have little meaning for them.

Are there questions that PL is specifically hoping to answer w/ this kit? I see these questions https://publiclab.org/questions/OrionAllgaier/03-13-2019/questions-from-the-university-of-wisconsin-eau-claire but they don't seem like they would be answerable through building this or similar kits.

Thanks @cfastie for your comment. Some of it however is not clear to me so I'd appreciate some clarification. You write: "To be reminded that the precision and/or accuracy of the Plantower sensor should be scrutinized, examine Figure 10 here. Note that the relationship among PM1, PM2.5, and PM10 is constant (there is always more PM10 and less PM1). Over the course of two days in my house it is unlikely that the relationship among the three size classes of particles would be that constant." To my understanding the descriptions of each in terms of aerodynamic diameter have an upper limit but no lower limit. So, particles that fit PM10 are also PM2.5 as are PM1 and there will always be at least as much PM10 as PM2.5, ie PM10 will never be less that PM2.5, etc. And when I examine the figure you refer to, I do not see the relationship as "constant", but though PM10 is always at least as great as PM2.5, for example, the amount by which it is greater is not constant. Perhaps I am misunderstanding what you are trying to say? Thanks for a clarification.

Hi jeffalk,

Thanks for pointing out that the PM10 index includes all particles smaller than 10µm. If that is true, you are correct that PM10 should never be smaller than PM2.5 or PM1.0.

You are also correct that the difference between PM10 and PM2.5 is not constant. However, the proportional relationship among the three indices, although not perfectly constant, does not vary much and varies in a predictable way.

Figure 1. Upper: An eight hour period of data collection showing three particulate indices (part of "Figure 10" linked above). Lower: the relationship (ratios) between the indices in the upper graph. The value of PM10 is about 1.1 times larger than the value of PM2.5, and the value of PM10 is about 1.7 time larger than the value of PM1.0. The ratios vary as a function of the magnitude of the PM indices (when the indices are larger, the ratios are larger).

Figure 2. Scatterplot showing the relationship between PM2.5 and PM10 for the data in Figure 1. Linear regression indicates a strong association between these two indices. The correlation coefficient for these indices is also high (r=0.93).

These three indices estimated by the Plantower sensor are highly correlated. This is consistent with the idea that all of the parameters output by the sensor are derived from a single set of laser scattering measurements. The equation used to derive each parameter apparently starts with the same set of measurements.

The actual concentrations of 10µm and 2.5µm particles in air might be correlated in many air samples. The high correlation presented here (Figure 2) might be unrealistic for samples of natural air and more likely to be an artifact of the math used to derive the estimates.

The correlation shown above is not perfect possibly because additional information is used (there are parameters in the equations other than the scattering measurements) or because some scattering measurements are weighted differently (e.g., depending on the magnitude of other scattering measurements).

The message is that the Plantower sensor apparently uses the same set of equations to estimate 12 different indices of particulate matter regardless of the characteristics of that particulate matter (shape, composition, density, size distribution, color, temperature, humidity).

Meaningful interpretations of the output of the sensor benefit from some appreciation for how the sensor works and how the software derives the output. For example, for a range of laser scattering measurements the equations probably produce non-zero values for all of the indices. So for a sample of air with only PM1.0 particles, an erroneous value might be output for PM10 (e.g., about 1.7 times larger than PM1.0) even though no particles larger than 1.0µm were present.

Chris

Building on this in the thought of a classroom setting, students could wear simple air sensors, manually log their data, then compile it on a map to identify spots in their school where they'd like to install other, more permanent sensors. manually logging the data when the simple air sensor goes off can also add info you wouldn't get from a purple air, like what it smells like, what's going on at the time, what is the weather like, etc. It could help determine what qualities of the air you'd like to investigate more thoroughly.