Ryan: Welcome back, everyone, to the third and final live session of our boot camp. Today, we're going to cover an exciting topic: how to design a network of air quality sensors. Traditionally, there has been limited guidance on this, but we're fortunate to have Dr. Maggie, my colleague, with us today. She recently completed her PhD dissertation on this topic and will share her insights.
Maggie: Thank you, Ryan, for the introduction. My name is Maggie, and I'm excited to present to all of you today. I recently completed my doctorate at UCLA in environmental science and engineering. Working at Clarity for the past year has provided me with valuable research insights. A portion of my dissertation focused on sensor network design, which is an emerging topic. I'm thrilled to share the importance of this topic with all of you.
To set the framework, air pollution is one of the greatest public health crises of our time. The World Health Organization estimates that 7 million premature deaths can be attributed to air pollution exposure. The image on the slide illustrates the wide range of health impacts caused by air pollution, including respiratory illness, cardiovascular disease, bodily inflammation, and premature mortality. Given the prevalence of these health impacts, it is crucial to have effective air monitoring solutions that can help us understand exposure and localized air quality impacts.
In previous bootcamps, you discussed the difference between regulatory and hybrid monitoring networks. Bridging the importance of health with managing air quality, many countries, including the United States, require the establishment of regional air monitoring networks. These networks determine compliance with national ambient air quality standards designed to protect public health and the environment. However, regional monitoring networks have limitations, such as limited spatial coverage and the need for extensive resources and expertise.
On the other hand, low-cost sensors like Clarity's Node S have gained popularity for increasing spatial coverage of air monitoring. They are being used by various organizations, including schools, to understand air quality impacts and make decisions related to student health. The US EPA and other government agencies have recognized the potential of low-cost sensors and funded projects that utilize them for research, education, and measuring local air quality.
With the emergence of this new technology, designing a low-cost sensor network raises important questions: where should sensors be placed, how many should be purchased, what sources should be measured, and what reasoning goes into these decisions?
However, limited research and diverse project scopes make it challenging to provide comprehensive guidance that fits all projects. Recognizing this industry need, I took on the task of studying sensor network design for my dissertation.
I conducted a comprehensive literature review using scientific databases and gray literature sources, in addition to surveying professionals in the low-cost sensor field. Through this research, I identified key variables that should be considered in sensor network design. These variables include project goals, budget, available air quality knowledge, pollution sources, community expertise, and optimal sensor placement. If you're interested in learning more about my research, my dissertation is linked for reference.
If you would like to see more about my research, my dissertation is actually linked on this slide that I think will be shared later. But this is really just a very high level of the work that I did. So I used the literature that I found to develop what I am calling the Sensor Network Design Decision Framework, which is based on elements that seem to be consistent for all projects.
So in this, these three boxes that you see here are the core items that are needed to set up a low-cost sensor network. In the previous boot camp presentations, we really talked about the importance of establishing project goals, so I'm just emphasizing here on how important it is to have those project goals so that you can better understand how to move forward when you're designing a sensor network.
Now, these arrows here are drawn in both directions and are pointing to and from each of these steps to indicate that this is a very iterative process. You may decide when you study your project goals that you're interested in an entire city; however, you may decide based on your budget and other factors that you really only have enough funding or capacity for maybe a small community. Thus, your study area might change, and your project goals might change.
And then around these core, I drew this circle, and I'm going to show you the boxes that I drew around it. The reason that it's a circle and not kind of like a process where you see arrows going ABC is that, again, this is a very iterative process. You may or may not review different variables as you move throughout the process.
So, starting from the top, it is really important that you establish early on what the community input or professional expertise is for your project. So, understanding what the impact might be on the ground, understanding kind of who's at the table and whose input is going to be valuable when moving forward in your project.
Moving clockwise, there are some quantitative methods that were identified in the literature that primarily use a lot of computer modeling, machine learning simulations, and statistical estimates. A lot of them use what's called a gridded approach, where they essentially put a grid over a spatial area and they use different variables such as historic air quality data or land use data to essentially weigh what areas would benefit from sensors the most. These are pretty in-depth processes that definitely require a lot of expertise, and again, all of this literature is listed in the literature review if you're interested in reviewing it.
Moving on to some other variables, budget is often the most limiting factor when determining what you can or how you can set up your low-cost sensor network, primarily because it's dependent on how many sensors you want to buy and for what pollutants, which will also change the price. So that could definitely be very limiting.
Another variable to consider is location feasibility, so understanding the characteristics about the location where the low-cost sensor will ultimately need to be placed. For example, if the low-cost sensor needs power and connectivity, are those available in the location that you want to deploy it? And you also have to consider if the project has permission to place the low-cost sensors where they want to. So if they want to place it on a city streetlight or some type of location that's owned by another entity, they're going to have to get permission to do that. Otherwise, they'll have to solicit volunteers, maybe from schools or from residents who are interested in hosting the sensor. I think it's also important to note, for location feasibility, a lot of air quality experts recommend doing a co-location or locating the low-cost sensor next to the reference monitor to ensure that the data being collected is accurate. So that's another question to ask: Do you have access to that low-cost sensor?
Ryan: Let's see, um, so we got a question, I guess, you know, getting a little more into the specifics of citing locations. So, it's from an anonymous attendee, but the question is, "I think that sensors are placed at a representative location with well-mixed air and at a particular height. So, how do you approach determining these? And yeah, for example, in one picture in the PowerPoint, there was a sensor shown to be fixed on a building wall. In other cases, sometimes they're attached to a pole or, you know, other locations. So, yeah, I guess, you know, getting into the details of actually choosing the right location to deploy the sensor for airflow and whatnot, is that something that you want to weigh in on, Maggie?
Maggie: Yeah, I could probably give, again, a high level, but I know that we have, for our customers and our knowledge base, there are citing requirements and guidelines that we have. The ideal place that you would want to put it is just above the breathing height to make sure that you can capture the proper pollutant mix. And if you're looking at exposure, making sure that you're capturing what people are actually breathing. The goal is to place it somewhere in an unobstructed area. However, with low-cost sensors, we know that people are working with different landscapes all the time. So, being aware of the most optimal placement is probably important. I've seen it for certain communities where it was placed under and in, like, under a bay, a school bus bay, which was giving them really high readings. And so, we had to then recommend that they move it to a more open area. Also, when you're working with low-cost sensors, it's important that you don't have it too low because you don't want it to get stolen. So, really working within all of these variables is important. But I know that we have more specific information available, so I'll just leave it at that as kind of that high level.
Ryan: Yeah, thanks, thanks Maggie. And yeah, there are some great resources on our knowledge base with respect to sensor deployment, tips and tricks. We also have a white paper with some guidance on that. And we can make sure to add links to both of our white papers to the materials hub that we have for the Air Sensor Boot Camp. And yeah, I guess, on that note, we do have a question from another anonymous person. What is the calibration process for clarity sensors look like? I will mention, not sure if everyone was able to attend all three of our boot camp sessions, but we did dive into that in a good degree of detail on the first session with Jack, our air quality data scientist, talking a bit about how Clarity approaches co-locating and calibrating sensors and some of the results from our calibration approach. But yeah, I don't know, Maggie, if there's anything else you would want to weigh in on that side of things, on the calibration process and how it relates to network design.
Maggie: I would encourage you to check out Jack's session from June 7th and the white paper on performance targets for air quality sensors. The EPA and others have been grappling with determining what should be considered adequate or useful data from an air sensor. While it may not reach the gold standard of FRM or FEM data, there are guidelines published by the EPA that provide some goals for air quality sensor data performance and data quality objectives.
Mosinkan: I've installed 15 sensor platforms in Nashik City, India for my PhD project. It becomes tedious to visit the host when something goes wrong with the sensor platform, and missing data is a significant loss. Is there any method to predict lost data with the help of the nearest sensor platforms?Ryan: That's a good question. I didn't focus on that aspect in my research, but it is possible to develop a model to compensate for lost data. However, the best approach is to have a data completeness expectation in place with the project and ensure the sensors are working properly. Service level agreements and data quality objectives can help with this.
Maggie: I'd like to add that EPA also has a data completeness requirement, which is typically 75% of data points. Monitoring the health of your sensor nodes through a dashboard can help identify issues without the need for immediate troubleshooting. It's essential to have notifications for any problems that arise and follow best practices for managing the network.
Ryan: Unfortunately, we're almost out of time for today, but we'll compile written answers to the remaining questions and provide them in the materials hub. Don't forget to submit your homework by next week for the certificate of completion. Keep an eye out for future boot camps, and feel free to reach out to Clarity for any assistance. Thank you all for joining, and special thanks to Dr. Maggie for her excellent presentation.
