Hello, my name is Ufuoma and I am interning with Public Lab this summer from the Providence office. I wanted to share my summer project on air quality with the hope of getting feedback and/or collaborators and also to learn about related work!
Literature Review:
Past contributors have written great posts discussing the pressing environmental and health impacts of deteriorating air quality as well as the financial barrier (upwards of $14,000) to producing widespread data compliant with Federal Reference Methods (FRM). see @Schroyer's post introducing the Dustduino). Because of this expense, cities may not have enough FRM sensors to create an adequate representation of the air quality in a location. For example, Dallas, TX, population of 1.3 million and land mass of 385 square miles has just 5 of these sensors (source). In comparison, Providence, RI, population of only 180,000, land mass 21 square miles, has 4 of these sensors (source). As you can see, the deployment of these sensors are neither proportional nor standardized across the United States.
Schroyer discusses the ever more popular method of using low cost (less than $200) sensors to increase the spatial density of air quality data. There has also been great work analyzing the quality and limitations of low cost methods built around the Shinyei PPD42NS particulate matter (PM) sensor that the DustDuino uses (see @Willie 's post that summarizes academic studies performed in Berkeley and Xi'an). The results of these studies were that the sensor shows promise when calibrated and at longer intervals of at least 1 hour or more.
Objective:
My objectives for this project are two-fold. In the Berkeley paper's 'conclusion and future work' section, they discuss how the work needs to be extended in terms of ease of use so that this method of calibrating low cost sensors can be more readily adopted by community scientists with varying levels of technical expertise. So my first objective is to develop readable and repeatable protocols for gathering, organizing, calibrating and interpreting air quality measurements from low cost sensors.
My second goal was a method inspired by a Columbia researcher, Franziska Landes, who was studying soil quality in New York City for the presence of lead. She showed two sets of data. The first was generated independently of the neighborhood community members and found just 14% lead presence exceeding federal standards in their samples. The second data showed samples at locations where parents in the community said their kids play - these samples returned 85% lead presence exceeding federal standards. My second objective is to use a similar framework of community inclusion in my approach to identify neglected sample areas. Federal entities choose the locations of these monitors in what seems to be a randomized way but it is important to gather data that can be put into context for different communities. I plan to work with organizations and communities here in Providence to gather data that is meaningful and interpretable to them.
Deliverables:
Projects like PurpleAir are helpful for this idea of putting communities in charge of their science and data collection. However, there have been multiple reports of PurpleAir drastically reporting 36-48% higher pollution levels than FRM sensors (source). I want to work to reduce the cloud of doubt that we have about our data when using low cost sensors like this.
By the end of this project I will have
(1) Developed relationships with local communities and organizations
(2) Published an open source and interpretable method for deploying and calibrating a low cost sensor (such as DustDuino or PurpleAir).
(3) Published an open source method for data visualization
(4) Contributed to the body of knowledge about the viability and limitations of low cost sensors
4 Comments
Some hopefully helpful comments: particulate matter in air is very different than lead concentration in soil. PM2.5 has a regional aspect. To see this compare the four sensors over time for Providence that you mention (I'm assuming they are for pm2.5 and not pm10?) Use the shortest time periods available. The four sensors will track quite well. (Why is that the case?) I'd also suggest talking to the gov't agencies that placed the four sensors to learn about their placement criteria. I doubt that randomness is a criteria unless there is a unique problem in Providence. Most often they are placed far from any known local sources because most people do not live near a known local source. I have worked with issues of sand mining in Wisconsin. People want to understand what is coming from the mine not what is regional. How do you differentiate that from the regional aspect? Not easy.
Low cost sensors are low cost for reasons. Usually they are not factory calibrated.(The DYLOS is factory calibrated.) Their components are not of the highest quality so their specs will show high variability. Therefore whatever model you want to examine, you should work with at least two or three examples; you will need to show how to calibrate and get an idea of how the calibration will change over time and under different conditions. Public lab has been trying to deal with low cost sensors for a number of years. The difficulties are why you are now being tasked to try again. It is also important to take into consideration the reasons why someone wants to monitor. Different reasons call for different criteria for a "good" monitor for the purpose.
You've set yourself quite a task. Good luck.
Is this a question? Click here to post it to the Questions page.
Reply to this comment...
Log in to comment
Hi, @sagarpreet -- do you think the sensor data @jiteovien referenced in http://www.dem.ri.gov/programs/air/air-monitoring.php would be possible to display on the new leaflet environmental layers map?
@jeffalk - thanks for your input. We should also have a Dylos somewhere that can be part of this work if it's helpful.
> you will need to show how to calibrate and get an idea of how the calibration will change over time and under different conditions
I especially like the interest in the strategy via @willie of extending the resolution of high-end sensors, which should directly address the calibration question. I wonder how fast sensors "drift" and what they're sensitive (as false positives, say) to that the FRM sensors are not.
I also agree that making easy-to-follow guides on some of these steps may help more people do this kind of work. Let's start with what's simple but useful, and build out from there!
On Wed, Jul 18, 2018 at 5:05 PM \<notifications@publiclab.org> wrote:
Is this a question? Click here to post it to the Questions page.
Reply to this comment...
Log in to comment
Hi, I just would like to share a bit of work I did here: https://publiclab.org/notes/mprof9/11-07-2017/low-cost-sensor-for-air-quality-monitoring It is a home made VOC measuring system based on Arduino and a Figaro gas sensor. There are a few comparisons with parallel official measurements made by the Regional Environmental Agency in a different location a few km away. The main use is to monitor the ambient air especially during the winter when lots of wood stoves are running in my area for domestic heating. An automatic system for the opening of windows is linked to this VOC monitoring device, so that when the air is cleaner during the day windows open to refresh the indoor air. Calibration of the sensor is still an open question of course and I am willing to get to some solution hopefully also by sharing with this community. Thanks for reading. Mauro
Reply to this comment...
Log in to comment
Just trying to collect up a list of resources related to co-location approaches as we discussed on #opencall --
https://publiclab.org/wiki/pm-dev lists some
SCAQMD or South Coast Air Quality Management District has been connected with Public Lab folks a lot and has resources on this: https://en.wikipedia.org/wiki/South_Coast_Air_Quality_Management_District
Their website: http://www.aqmd.gov/
Reply to this comment...
Log in to comment
Login to comment.