I have had the chance to play with a prototype of a high precision Particulate Matter sensor that measures in real time PM1, PM2.5 and PM10.
It provides and updated information every second and thanks to the app each of these measurements are geolocated and activities may be tagged so you are able to track your exposure over time.
Particulate Matter is probably the most “famous” air pollutant, and air pollution is more and more pointed out for its impacts on our health and economy. Still, it seems that we don’t really know what to do about it as it feels that we have little impact as individuals about what’s in the air we breathe.
This is the purpose of this empirical study. It is not just about measuring air quality, it is also about documenting activities and locations to see if some trends can be unveiled and see if we could have more control than we thought on the air we breathe .
Set up and protocol
You can skip this section but I always feel like it is important to understand where the data is coming from and how it was collected.
The device used (PMSCAN from Groupe Tera) measures PM1, PM2.5 and PM10 every second and the data is transmitted via Bluetooth to a smartphone with a dedicated app. On the phone you can see in real time PM levels as values, charts or maps. An interesting feature is the possibility to program alarms such as : “if PM1 over 15µg/m3 for more the 30 seconds -> alarm” and this can be done for each PM size. This proved to be very interesting as you don’t spend all your time looking at the app. Each time an alarm was triggered I documented the situation (picture / comments) in order to be able to process that information in the end.
I also documented as well as I could the nature of my activities : cooking, walking, cleaning, working… in order to compile the data accordingly.
I kept the sensor with me for 48 hours and this revealed to be a bit challenging to find a place to attach without having it in the middle but also to make sure it doesn’t create false measurements which can happen if the air inlet scratches your clothes for example. I had the issue and it triggered an alarm because my clothes generated particules because the way I fixed it to my belt. So be careful about that, the device is really sensitive and accurate.
The data and results
You can find below the results. I down sampled the data to one value every 30 seconds for the graph but still makes a lot..
As you may have guessed the highlighted value ranges identify the different activities / locations : green is for home, red for walk, blue for drive and yellow for shops. This is the print for Saturday, same pattern goes for Sunday.
1 : Getting inside the car / closed windows. A bit obvious sitting down in the car puts in suspensions dusts that were laying on the seat.
2 : Handy work, looking for material/tools in the basement
3 : Stopping the car in an underground parking (-3)
4 : Back in the same parking
5 : Drilling holes in a cement wall at my son’s place
6 : Drilling / Cutting wood and cleaning up with a broom
7 : Back to the parking, putting the equipment in the car
8 : Cooking (oignons and meat for lasagna)
1 first learning : location is not enough
Easy to point out for indoor measurements, same conclusions can be drawn for outdoor activity (walking), even if statistically it would have been better to have more data. Unlike gases (like NO2, O3…) that rapidly expand in every directions, particulate matter is very sedentary in a way that from a side of a street to the other you may have totally different values.
Exception : the parking seems to be a constent hotspot, more about that below.
2 second learning : activity and ventilation play a key role
Cooking, handy work and cleaning generate very high levels of PM in the air.
Cooking for example, it reached more than 108 µg/m3 of PM10 in an open kitchen and it took almost 35 minutes before getting back to normal (opened windows at start + 12 min). Then got another peak when opening the oven.
Compared to outdoor exposure, the worst I got was 28 µg/m3 at a busy intersection and 56 µg/m3 after a gust of wind.
3 third learning : PM concentrations may vary very rapidly in time
In matter of seconds you can go from a very low to a very high concentration.
Over that week-end I spent 96 % of my time indoor (scary..), 2% walking outdoor, 2 % walking in public indoors.
Statistically, it is hard to draw significant analysis with such difference between indoor and outdoor time but with these basic elements we still get sound results. Considering PM10 exposure, 97% comes from indoor time, 2% from walking outdoor and 1% from public indoor.
Based on alarms and activity tracking I managed to break things further down with the results that can be found below.
I summed the amount of PM10 in µg/m3 for each subcategory (tab 1), then calculated the amount of time for each (tab 2) in order to get a ratio (tab 3) of the amount of PM10 per time unit.
And again handy work, cooking and parking stand out, by far.
Looking at the raw figures I was surprised that parking was not even higher than this and here is what I found out : in some cases I get high values (but not that high) and in other cases I have extremely high values so on average it is not that bad. So I did some more tests and this is my conclusion :
I have lower values when I am getting back to my car but very high values when I arrive with the car, even if I arrive to my spot freewheeling, engine off.
My conclusion is that the high level of PM doesn’t com from my exhaust but it comes from the PM sitting on the parking floor and put in suspension when the vehicle arrives. This would be no surprise as, because there is no rain to wash it off, PM just accumulates on the floor waiting.
Some specific events, situations I did not really expect led to high concentrations of PM :
– cat litter : renewing cat litter generated an impressive peak at 288 µg/m3
– car : what is better, closed windows, opened windows or air recycling. I always felt that fresh air from an open window was best and….no. Air recycling (Golf-Volkswagen) not only limited but dropped the level of PM to very low levels (around 1 µg/m3). With just closed windows, after PM peak when seating , levels are around 10 µg/m3. With opened windows it changes all the time from 4µg/m3 to 37µg/m3 on this experiment (driving in town). I will do more testing on that specific topic trying to remove all other variables (moving in the car, trip length..)
– broom/vacuum : I always wondered how compared broom VS vacuum. No debate, I reached my highest scores cleaning drilling dusts, reaching 1230 µg/m3 in seconds. Same spot/situation with the vacuum (basic one) and the sensor lightly reacted, not triggering any alarm.
There is still a lot to extract from that dataset but these first results have given me ideas on how to improve my data collection process and what will be the next things I will test : more on driving, parking but also spend more time outside in order to have more reliable data. I will also improve my activity classification/tagging in order to make the analysis process easier.
From my personal example, on that specific week-end I was responsible for 97% of the PM I have been exposed to. It doesn’t mean that we can change everything but it means that we have much more control on the quality of the air we breathe than I thought. It also means that traditional air quality maps we can find on the web are important but far from enough to accurately assess the quality of the air we breathe in real life, I mean taking consideration outdoor but also indoor air quality. And in most cases even if outside air quality may not always be good, chances are that this air is better than the one in your house / office… so ventilate your house!
Second conclusion, even if there is no doubt on geographical injustice — where you live is what you breathe — , the nature of your profession might be an even bigger issue : working on construction sites, kitchen (restaurants..), closed spaces and so may highly impact your health. Even though this is no surprise, it is amazing to see how much effort is put on creating world wide maps but in the end we still know so little about what we really breathe.
Do not hesitate to get in touch if you want the full dataset, if you have test ideas or if you want to have a try at this prototype.