atmospheric research

May 2024 - present

updated 2025

Inventing a rover-AI system to make clean air accessible for 2.4K+ Dallas students

Cost-effective, mobile rover-AI system engineered to evaluate impacts of microclimatic interactions on pediatric asthma

A collage of various electronics and robotics parts featuring Arduino boards, sensors, wires, a motor assembly, and a small outdoor rover with a pinwheel mounted on top.

30 second synopsis

I invented a cost-effective, mobile rover-AI system with Arduino-based aerial sensors (patent-pending) to make outdoor air quality monitoring more economically feasible for urban schools susceptible to pediatric asthma, and used it to clarify the extent to which fine and coarse particulate matter; carbon dioxide; and carbon monoxide concentrations affect pediatric asthma vulnerability (or PAV) in Dallas ISD schools with low vs. moderate vegetation. The system is currently deployed in 4 elementary schools impacting 2,400 students, resulting in an average 22% reduction in >PM2.5 levels, and its insights are being used to lobby school district officials to implement 2 atmospheric safety protocols in Dallas ISD elementary schools for the 25-26 school year.

2.4K+

Dallas ISD students impacted

elementary schools trialing rover

~22%

average reduction in >PM2.5

atmospheric safety protocols

Background

Analyzing the problem

To understand current gaps in research and methodology on pediatric asthma vulnerability (PAV), I reviewed 40+ peer-reviewed studies on the impact of microclimatic interactions on pediatric asthma in urban schools.

Full lit review

My (messy) research doc

A tilted view of an academic paper titled “Evaluating the impact of urban greenness on pediatric asthma rates: Implications for school policy,” surrounded by related charts, tables, and reference pages.

A U.S. map shows asthma prevalence among children aged 0 to 17 by state, shaded in four blue gradients ranging from 4.4% to over 9.8%, with white indicating unavailable data.

Distilling gaps in literature

I distilled my research findings into 4 key insights Despite the prevalence of pediatric asthma among urban children, many studies obscured how specific pollutants or particle sizes interact with vegetation to impact PAV by grouping them into broader categories and using aerial sensors that are expensive and unreliable because they rely on pollutants reaching fixed collection points such that they fail to capture spatiotemporal variation. Thus, there is a need among schools to have the personalized data, cost-effective technology, and specificity necessary to respond to PAV threats in their unique microclimates.

Unclear how microclimates impact PAV

The effects of specific microclimatic parameters on pediatric asthma vulnerability (PAV) and how varying levels of vegetation influence pollutant exposure in urban school environments are poorly understood at the census tract level.

Insufficient analysis of individual pollutants

Many studies have aggregated pollutant data (e.g., into particulate matter, nitrogen oxides, or volatile organic compounds), which obscures the distinct effects of individual pollutants or pollutant sizes on PAV.

Limited exploration of microclimatic interactions

Existing research lacks analysis of how specific meteorological factors (e.g., temperature, humidity, wind speed) influence the dispersion and concentration of pollutants to affect PAV in urban school environments where vegetation varies. Although some studies correlated microclimatic conditions with pediatric asthma, none conclusively determined whether these factors contribute to PAV or if their effects are mediated through interactions with vegetation.

Absence of standardized metrics for PAV

The lack of a standardized metric for assessing PAV has produced conflicting findings and inconsistent proxies in literature that made it challenging to compare studies and draw causal conclusions about the impact of microclimates on PAV.

Limitations of traditional aerial sensors

Current sensors are expensive and inefficient because they are stationary (rely on pollutants reaching fixed collection points), rather than removing PM in areas of highest concentration. Furthermore, traditional sensors overlook how pollutants are diffused via meteorological factors—wind velocity, pressure, temperature variations (e.g., urban heat islands, thermal inversions).

The ends and the means

Experimental design

To address these gaps, the goal was to clarify to what extent fine (>PM0.3) and coarse (>PM2.5) PM and carbon oxides (CO and CO₂) correlate with pediatric asthma vulnerability (PAV) in urban schools with low (<0.2) vs. moderate [0.2-0.6] NDVI (vegetation) at the census tract level by testing a cost-effective, mobile AI rover equipped with Arduino-based aerial sensors for each microclimatic parameter.

Variables and conditions

Hypothesis

I hypothesized that schools with lower NDVI (vegetation) would experience a higher PAVI compared to those in more vegetated tracts due to increased exposure to microclimatic parameters—particulate matter (PM), carbon monoxide (CO), and carbon dioxide (CO₂).

Independent variables

>PM0.3

>PM2.5

CO₂

NDVI

Dependent variables

A high pediatric asthma vulnerability index (PAVI) is characterized by a high rate of emergency department visits and high chronic obstructive pulmonary disease frequency coupled with low inhaler prescriptions, where α, β, and γ represent weighting coefficients.

Macroscale PAVI = α(ED visits) + β(COPD) - γ(Inhaler prescription)

Experimental and control groups

I selected 5 schools in census tracts with low NDVI (<0.2) for the experimental group and 5 with moderate NDVI [0.2, 0.6] as controls, and engineered a rover to sense these pollutants at 3 sublocations per school across 2 dates.

Controls

To control for meteorological confounders, I defined site selection criteria to determine which schools to collect data from and recorded microclimatic meteorological parameters (temperature, humidity, and wind speed) to ensure homogenous conditions.

The method behind the madness

Methodology

To identify meaningful patterns in PAV, I selected 5 schools in census tracts with low NDVI for the experimental group and 5 with moderate NDVI as controls, and engineered a rover to sense these pollutants at 3 sublocations per school across 2 dates.

180

trials

research sites

sensors engineered

Selecting sites

I selected schools for the experimental group based on 6 exclusion criteria by using a random forest classifier to identify schools with moderate NDVI (greenness) and exclude them to isolate the effects of sparse vegetation (low NDVI).

Experimental inclusion criteria

Dallas ISD public schools in census tracts of NDVI <0.2 and ≥5 years of PAV data, enrolling ≥500 students, and spaced >3 miles apart.

A choropleth map displays a city divided by neighborhood, shaded by a scale from -1.0 to 0.6 representing NDVI, with black circles marking five experimental schools and black triangles marking five control schools.

Collecting data

I gathered data on the selected atmospheric parameters at 10 selected Dallas ISD schools, divided into 5 experimental and 5 control locations to test my hypothesis. To control for confounders, I conducted a total of 180 trials by collecting data at 3 designated sublocations, with 3 replicates per sublocation, on 2 dates (10 sec for 3 min/trial), for each school.

Data collection was conducted exclusively in public areas, adhering to posted signage at each site indicating publicly open hours and on non-operational days to comply with Dallas ISD policy.

A collage of three aerial satellite images shows different high school campuses with numbered labels identifying buildings, sports fields, and surrounding areas.

Rover

I engineered a novel, cost-effective, mobile rover with Arduino-based aerial sensors for each microclimatic parameter to maximize data coverage and overcome the mobility limitations of traditional aerial sensors. I self-assembled sensors using Arduino-based parts.

Engineering the rover

I engineered a novel, cost-effective, mobile rover with Arduino-based aerial sensors for each microclimatic parameter to maximize data coverage and overcome the mobility limitations of traditional aerial sensors. Since there wasn't an affordable way to microsense wind speed, I engineered a proximity sensor + pinwheel to count rotations over time.

A collage of three images shows a homemade rover with orange and blue wheels, various electronic components, and a mounted pinwheel, both in outdoor grassy terrain and an indoor workspace.

Final rover with sensors

A top-down image displays a small motor connected to a blue gear beside a larger black gear, placed on a flat metallic surface.

CAD 3D-printed wheels

Three close-up images show different configurations of Arduino-based electronic circuits connected with colorful jumper wires, breadboards, and sensors, including SD card modules and microcontrollers.

CO₂ and CO sensors

Three images show electronic components and microcontrollers, including an Arduino board and a breadboard circuit, with the final image featuring a hand holding a windmill-style sensor setup connected to wires in front of computer monitors.

Meteorological sensors

A close-up image shows a blue electronic sensor unit mounted on a blue bracket, with wires connected below and a blurred classroom or workshop environment in the background.

>PM2.5 and >PM0.3 sensor

The data pipeline

The rover transmitted data to a multi-step machine learning algorithm visualized on a mobile app that modeled pollutant dispersion relative to temperature, humidity, and wind speed to identify high-exposure microclimates, predict changes in pollutant levels, and provide school-specific recommendations to reduce PAV.

Processing data

After transporting the rover to each school (each with 3 sublocations and 3 replicates, 10 sec for 3 min/trial), the rover either autonomously or remotely followed a standardized path and transmitted sensor data to a mobile app.

1. Each data point is linked with specific coordinates where the rover is located.
2. Calculate spatial covariance between data points to assess how pollutant concentrations vary with distance.
3. A kriging model (covariance matrix) predicts pollutant concentrations at unmeasured locations.
4. Use DBSCAN clustering to identify pollutant hotspots by grouping proximal areas with high pollutants.
5. Gaussian pollutant dispersion model accounts for pollutant movement based on meteorological conditions.

Visualizing data

The model built a spatiotemporal map of pollutant concentrations by visualizing pollutant dispersion relative to temperature, humidity, and wind speed to identify high-exposure zones.

Two smartphone screens show a map of North Dallas High School displaying predicted average PM2.5 concentrations over time, with a time slider feature allowing users to forecast pollution levels based on meteorological data.

Interpreting macroscale PAVI

PAVI was highest in areas adjacent to major roads, parking lots, and building perimeters due to pollutant accumulation, reduced dispersion, and limited vegetation buffering. Lower PAVI was observed in open fields and downwind locations where pollutants disperse more freely.

A phone screen shows a pollutant heatmap of North Dallas High School with outlined zones and a color-coded microscale PAVI score, showing predicted air quality values based on meteorological conditions.

Recommending interventions

The model generates recommendations on optimal intervention (e.g., vegetation, HVACs, HEPA) to reduce PAV.

Spatiotemporal maps

The app uses a hybrid Kriging interpolation + DBSCAN clustering model to generate spatiotemporal PM2.5 dispersion maps by using Gaussian pollutant transport simulations with analysis of meteorological covariance.

Recommendations

The app applies a proximal policy optimization algorithm to recommend vegetation and air filter placement, density, and species selection by identifying pollutant hotspots, dominant wind flow paths, and areas with high microscale PAVI.

A phone screen shows a map of North Dallas High School highlighting areas with recommended vegetation interventions to reduce PM2.5, suggesting tall evergreen trees planted in two staggered rows at a minimum height of 20–30 feet.

The magical moment

Results

I found that experimental schools had significantly higher levels of CO₂ and coarse PM, which were strongly correlated with increased PAV, especially in downtown and southeast Dallas, and inversely related to NDVI, suggesting that areas with lower vegetation density experienced higher PAV.

p > 0.0001, *p ≤ 0.0001, and **p < 0.0001 represent significance values derived from unpaired two-sample t-tests for the average values of each parameter in the experimental group compared to the control group.

Experimental schools recorded 27% higher >PM0.3 concentrations than control schools on average

Experimental schools recorded 55% higher >PM2.5 concentrations than control schools on average

Experimental schools recorded 2% higher CO concentrations than control schools on average

Experimental schools recorded 19% higher CO₂ concentrations than control schools on average

Experimental schools recorded 13% higher air temperature than control schools on average

Experimental schools recorded 20% higher humidity than control schools on average

Experimental schools recorded 14% higher wind speed than control schools on average

Spatial trends

I found, by using spatial regression analysis on PCCI data, that census tracts of experimental schools had a higher PAV compared to control schools, meaning that Dallas ISD schools in areas with lower NDVI had significantly higher levels of coarse PM and CO. This corresponded to an increased PAVI, supporting my hypothesis.

Two maps of Dallas show pediatric asthma data: one shows emergency department visits per 90 days by neighborhood, and the other shows annual inhaler prescriptions, both shaded in blue gradients to indicate severity levels, with schools and Seagoville labeled.

Experimental group

Spatial regression analysis revealed that tracts in downtown Dallas showed a moderately high PAVI. Those in the urban periphery, despite proximity to downtown, showed a lower PAVI.

Control group

Tracts in Seagoville showed a high PAVI, indicating over-reliance on preventive medication due to localized environmental triggers rather than health emergencies. North Dallas showed low PAVI, likely due to consistent healthcare access preventing acute asthma incidents in higher-risk areas.

Discussing results

Insights

I distilled my research findings into 4 key insights.

Coarse PM concentrations are significantly greater in low-NDVI schools

Experimental schools with lower NDVI experienced significantly higher levels of PM (to a greater extent coarse PM [>PM2.5]), which were strongly correlated to increased PAV, especially in downtown and southeast Dallas.

Higher concentrations of fine, and to a greater extent, coarse PM in experimental locations suggest that areas with lower vegetation cover experience elevated air pollutant levels. Vegetation filters and traps PM through their stomata or absorbs them through leaf cell membranes, after which it can be released through a process called resuspension, which is often dictated by wind speed. In areas with dense vegetation, PM is more likely to be captured before it can contribute to harmful exposure levels, while in low-vegetation areas, reduced filtration can result in higher coarse PM levels. The data indicate that coarse particles (>PM2.5) were elevated to a greater extent than fine particles (>PMO.3), potentially because they are more easily trapped by vegetation due to their larger size. The lack of vegetation in experimental schools could allow these larger particles to remain suspended longer.

CO₂ concentrations are significantly greater in low-NDVI schools

Higher CO₂ levels in experimental locations with low NDVI suggest a reduced capacity for natural CO, absorption and sequestration capabilities found in more vegetated environments. Vegetation naturally absorbs CO, from the atmosphere through photosynthesis by converting it into oxygen, which suggests that the lack of greenery in these areas potentially inhibits this absorption.

Low vegetation density worsens PAV

The inverse relationship between NDVI and temperature, CO₂, PM, and humidity suggests that low vegetation density directly contributes to worse air quality and higher PAVI in experimental schools. Specifically, spatial regression analysis (r=0.72) showed a significant inverse relationship between PAV and NDVI, while fine PM and CO₂ were positively correlated with elevated PAV. This suggests that low greenery and high atmospheric pollutants interact to increase PAV in urban schools.

Meteorological factors mediate high-PAV microclimates

Experimental schools experienced significantly higher microclimatic temperatures, indicating a strong urban heat island effect caused by impervious surfaces replacing natural vegetation. Site-specific wind flow, temperature gradients, and vegetation density also created concentrated >PM2.5 exposure zones near traffic-adjacent school microclimates in low-NDVI areas.

Limitations of findings

Social determinants of health, such as higher poverty rates in experimental locations, is a confounding variable that correlates with increased rates of chronic diseases and could have increased PAV independent of environmental exposures. As such, future studies should include more locations while controlling for social determinants of respiratory health.

Sharing results

Mentored by Dr. Jennifer Honda at the University of Texas at Tyler, I published an independent research paper in a Harvard-affiliated journal, where I got some valuable feedback from PhD students and post-doctoral fellows.

DOI: 10.59720/24-269

Read the paper

A collection of journal pages from the Journal of Emerging Investigators features a student research article titled "Rover engineered to evaluate impacts of atmospheric parameters on pediatric asthma in Dallas," including text, data charts, and photos of the rover construction process.

Significance

So what?

The rover-AI system serves two purposes: For schools, it provides a cost-effective tool that helps them know exactly when, where, and how to most effectively use interventions to reduce pollutant concentrations in the most efficient time frame. For future studies, it offers a more accurate data collection method by using sensors to capture spatiotemporal variations across larger areas and time intervals.

This is the first study to assess the impact of microclimatic parameters on pediatric asthma prevalence and to integrate machine learning to spatially visualize pollutant dispersion.

PAV + ML

Three smartphone screenshots show an environmental data app with maps of North Dallas High School with air quality metrics, including PAVI scores and PM2.5 concentrations, data collection visuals, and hotspot analysis.

The rover-AI system’s portability, affordability, and intelligence informs policy decisions tailored to schools' unique microclimatic conditions rather than averaged across a ZIP code.

Personalized data

Three photos show a homemade mobile rover built from electronics and cardboard, equipped with sensors, wires, and a colorful spinning pinwheel, positioned outdoors on grass.

This study’s spatial insights are being used to lobby school officials to implement 2 atmospheric safety protocols in Dallas ISD elementary schools for the 25-26 school year.

Impacting policy

A set of visuals showing PM2.5 air pollution data, including a map of Dallas neighborhoods, a bar graph comparing PM2.5 concentrations between control and experimental sites, and a mobile app displaying air quality data for a school campus.

Currently deployed in 4 Dallas ISD elementary schools totaling 2.4K+ students, the rover-AI system overcomes the mobility limitations associated with traditional sensors.

Impacting people

A collage of satellite images shows three elementary school campuses—David G. Burnet, Seagoville North, and Adelle Turner.

What I'm up to now

Currently, I am augmenting the AI-rover system to correlate atmospheric data with PAV proxies in 4 regions in Dallas County to inform a local bill proposal and working remotely with UT Tyler CS students to produce a symposium publication by predictively modeling PAV using analyses of microclimatic variables through LSTMs and spatial autocorrelation. More details coming soon!

Taking it into competition

#sciencefair2025

In February 2025, I had the opportunity to showcase my project at the Dallas Regional Science and Engineering Fair, where I placed 3rd for the earth and environmental science category among 1.1K+ students with a $200 cash prize and advanced to the Texas Science and Engineering Fair.

State results

Regional results

A detailed research poster explaining a mobile rover-AI system designed to study how microclimatic factors affect pediatric asthma, featuring background, methodology, results, and policy implications.

Competing at TXSEF

Although I initially placed 3rd, I advanced to the state level because the 2nd place winner chose not to attend. At the state fair, I went on to earn 2nd place in the same category. Around this time, also placed 1st at the 2025 NASA GLOBE Program International Virtual Science Symposium!

Photos from 2025 DRSEF Beal Bank Awards Banquet

A collage of students presenting and receiving awards at a science fair, with one student standing beside a project display board.

GSTCA

I was also awarded a $2K residential scholarship to conduct research at the Governor's Science and Technology Champions' Academy at Southern Methodist University, where I researched the use of unmanned aerial systems (UAS) on the efficiency of radio signals passing between cell towers and the use of UAS for mapping and identifying urban food deserts and infrastructure deserts.

Presenting my project

Here are some of my favorite questions asked by judges who grilled my project (and my stressed-out responses to them).

How did you get the idea?

My dad is a middle school teacher in DFW, and he’s told me about how common inhalers are at school—how kids are constantly reaching for them, especially during recess. It made me wonder: what’s in the air that’s triggering so many asthma cases? At the same time, I realized that most air quality data comes from stationary sensors that don’t reflect what kids are actually breathing in on school grounds. That's when I thought, "what if I could bring the sensor to the pollution instead of waiting for the pollution to reach the sensor?" That’s how I came up with the idea to build a mobile rover that could track pollution in real-time where they spend the most time: at schools.

Could you describe the timeline for this project?

From June to July 2024, I focused on background research, writing a literature review paper, and identifying key gaps to shape the goal and experimental design of my study, including hypotheses and variables, while also selecting sites for data collection. In July, I constructed the mobile rover, organized logistics, and conducted the data collection. By August, I completed the database analysis of pediatric asthma vulnerability (PAV), formulated conclusions, discussed findings, and explored insights, limitations, future research directions, and policy implications. By September, I translated my work into a research paper, which was accepted by a Harvard-affiliated journal. Finally, from November to January, I had the opportunity to present my findings to DISD school officials and the school board during my computer science end-of-semester final symposium, which contributed to the implementation of new atmospheric safety and healthcare equity policies.

What obstacles or unexpected results did you encounter?

I think the significant obstacle I encountered was engineering my rover. Specifically, aligning the particulate matter sensors with the optical filters required a lot of precise adjustments to avoid interference from ambient light, which proved challenging due to the rover's compact design. Additionally, integrating the CO₂ and CO sensors demanded a lot of calibration to ensure accurate readings across varying environmental conditions, which was complicated by inconsistent power supply issues affecting sensor performance. These technical difficulties required iterative adjustments and extensive testing to ensure that all sensors provided reliable and consistent data throughout the study.

On what basis did you reach your conclusion?

I reached my conclusion by performing statistical analysis on the atmospheric data collected from the rover. I utilized ANOVA to compare the mean concentrations of coarse PM and CO₂ between schools with low and moderate NDVI. This was followed by post hoc Tukey HSD tests to identify specific differences between groups. The analysis revealed significant differences in pollutant levels, with lower NDVI areas showing higher concentrations. Additionally, I performed regression analysis to examine the correlation between pollutant levels and pediatric asthma vulnerability (PAV), and used correlation coefficients to validate the strength of these relationships. This approach confirmed that reduced vegetation was associated with increased pollutant levels and elevated PAV.

What could be done to strengthen your work?

To strengthen the work, I could better control the potential confounding impact of social determinants of health (SDOH), such as higher poverty rates in experimental locations, which could independently influence chronic disease rates and skew the results. Incorporating a larger sample size from additional DISD locations and implementing more robust controls for SDOH would enhance the validity and generalizability of the findings, ensuring that the observed effects on PAV are more accurately attributed to environmental exposures rather than socio-economic factors.

Paying it forward

What was unique about this research project was that it combined 2 of my favorite passions: niche data science research and urban sustainability. I decided to share this interest with others through 3 clubs / organizations.

Science fair + research club

It was through this club that I was first inspired to start my research project as a freshman — and along the way, I made some great friends and unforgettable memories. I'm so proud to have helped increase Frisco High School’s advancement rate at the Texas Science and Engineering Fair—from 0% in 2024 to 67% in 2025 among Frisco ISD competitors.

A group of nine people, including former Secretary of Education Miguel Cardona, are smiling during a virtual video call, with names displayed below each participant.

sustain + ability

Through my agtech business Agsight, I help to promote local environmental justice and sustainability efforts in Frisco ISD and beyond through cross-campus initiatives (e.g., sustainability research) and the Global Sustainability Scholars Program.

Reflections

What I learned

Don't be afraid to find + create new opportunities

Building my rover-AI system taught me that me that I don’t have to wait for opportunities to show up because I can create them myself. When I first realized how overlooked air quality monitoring was in schools vulnerable to pediatric asthma, I didn’t have thousands of dollars of resources or a lab—just an idea and a strong reason to pursue it. I taught myself how to engineer aerial sensors, built prototypes in my garage, and reached out to school administrators until someone said yes. That one “yes” turned into a system now helping thousands of students breathe cleaner air and sparking policy conversations I never imagined I’d be part of. More than anything, this experience taught me that real impact often starts with stepping into the unknown—and that the best opportunities are sometimes the ones you make for yourself.

Inspiration comes from unexpected places

Inspiration doesn’t always strike in a lab or classroom—sometimes, it starts with something as simple as having a conversation with a parent or watching a wheeze after recess on a bad air day. This made me start asking questions I hadn’t considered before: Why aren’t schools monitoring air quality? What if there were a low-cost way to help? But as I started tinkering with sensors, digging into asthma studies, and walking through schoolyards with different vegetation levels, I found myself completely obsessed with a challenge I hadn’t seen coming. It reminded me that real inspiration comes from paying attention to the world around you and caring enough to do something about it.

Engineering is not linear

Working on this project taught me that engineering is anything but linear—it’s messy, unpredictable, and full of moments that feel like you're taking two steps back for every one step forward. I started with a simple goal: make outdoor air quality monitoring more accessible. But what followed was months of trial and error—sensors that gave faulty readings, code that broke the night before testing, late nights staying up for science fair, and more versions of the AI pipeline than I can count. Each problem forced me to rethink my assumptions, dig deeper into research I hadn’t expected to need, and sometimes scrap whole parts of the design to start over. But through that process, I learned that real innovation happens in loops, dead ends, and breakthroughs that come when you least expect them.

Enjoy the process

What surprised me most wasn’t just how much I learned inasmuch as it was how much I loved learning it. I’ll never forget the late afternoons I spent in club meetings, whiteboarding ideas with friends, or that one four-hour conversation with my science teacher. Traveling to the state science fair, presenting to judges, sharing my work with other students, and researching on SMU’s campus, surrounded by people who were just as curious and obsessed with problem-solving as I was. What started as a personal project turned into a shared experience, full of moments that made all the debugging and late nights worth it. It taught me that it's about the community, the discovery, and finding joy in the process itself.

More work

About me

Helping 50+ specialty crop farmers grow more equitably

agsight

Jan 2024 - present