This is the second blog post in a series written by the 2026 recipients of the Duke University Libraries Summer Research Fellowship for LIFE Students. You can read the first post here. Yubraj Bhandari is a senior majoring in Physics & Mathematics.
I grew up thinking severe air pollution meant nothing but an unexpected holiday. That was the reality growing up in the mountain valley of Nepal, where the government closed schools whenever pollution levels turned severe. On the worst days, the country’s capital, Kathmandu, topped global pollution rankings. Headlines made the crisis sound urgent, yet somehow temporary. A few days later, school would reopen and life would move on. I rarely stopped to ask what that air was doing to us, or what was happening beyond the capital once the headlines faded.
A layer of haze covers Kathmandu. The capital often receives the most attention when Nepal’s air pollution reaches dangerous levels.
Years later, a conversation with a Duke graduate student changed the meaning of those headlines. He described the crisis as a two-layer fold: (1) Nepal’s air was dangerously polluted but (2) the country lacked the monitoring needed to understand who was exposed and when. Kathmandu’s smog was photographed, ranked, and discussed. With even higher levels of pollution transcending national borders, my hometown of Terai received far less attention. For the first time, the gap in the data felt both scientific and personal.
The World Bank’s 2025 report Towards Clean Air in Nepal put the scale of that problem into numbers. The report estimates the population-weighted annual PM2.5 exposure at about 37 micrograms per cubic meter in the Kathmandu Valley and 39 micrograms per cubic meter in the Terai. Both levels were more than seven times the World Health Organization guideline of 5 micrograms per cubic meter. Air pollution now ranks as Nepal’s leading risk factor for life expectancy, exceeding both tobacco and malnutrition, and shortens the average Nepali’s life by an estimated 3.3 years.[3]
Wanting to do something, I applied for the Duke University Libraries LIFE Summer Research Grant to conduct preliminary research. Our Bass Connections team began preparing low-cost PM2.5 sensors for new deployments in Nepal, while I focused on a basic question: Does outdoor air-quality data reflect what people actually breathe indoors?
At first glance, Nepal’s national monitoring network seems to offer a strong foundation. It has grown from six PM10 sampling stations in 2002 to 30 air-quality stations today[1, 4]. Yet coverage on a map does not always become reliable data in practice. In 2024, the Department of Environment reported data from only 16 of those 30 stations.[4] Even a working outdoor station answers only part of the exposure question. It can describe the air around the station, but not necessarily the air inside a nearby home, classroom, or office.
That missing piece matters in Nepal, where household conditions can strongly shape indoor air. An analysis of the 2016 Nepal Demographic and Health Survey found that 66 percent of households used unclean cooking fuel and 45 percent lacked a separate kitchen.[2] Another study of 824 household kitchens in Bhaktapur found that mean PM2.5 ranged from 80 micrograms per cubic meter in homes using electric stoves to 656 in homes using biomass.[5] Cooking smoke does not remain neatly indoors. The World Bank identifies cooking, along with industry and mobility, as one of the Kathmandu Valley’s three largest sectoral sources of PM2.5.[1] Outdoor pollution enters buildings, while emissions produced inside return to the shared air. The boundary between the two is far more porous than our monitoring systems suggest.
Before the new sensors could travel to schools and homes, they first had to spend time alongside a reference-grade monitor. This process is formally referred to as colocation. During collocation, each device samples the same air under the same conditions, allowing us to identify and statistically correct for differences in measurements in and across individual sensors before deployment. However, the process required more time and technical expertise than I had expected. Fortunately, I received valuable support from VayuDrishti, a Kathmandu-based startup developing low-cost air-quality sensors, and researchers at ICIMOD. Collocation may not be the most visible part of the project, but it is what makes every later comparison credible.
Four low-cost PM2.5 sensors record the same air during collocation. The comparison helps us identify differences among devices before deployment.
While the new sensors underwent collocation, I turned to an existing indoor and outdoor sensor pair in a household in Belauri, Kanchanpur. I matched the two datasets by hour and obtained 2,717 paired observations from December 2025 through April 2026. The two lines often moved together, which showed how strongly outdoor conditions shaped the air inside. But they did not move together perfectly. Indoor PM2.5 was higher in 76 percent of the matched hours. Across the study period, the mean reached 122 micrograms per cubic meter indoors and 104 outdoors, with the largest gaps during morning and evening household activity.
Indoor and outdoor PM2.5 followed similar daily patterns in Belauri. Indoor levels often rose higher during morning and evening household activity.
What caught my attention was not simply that one number was higher than the other. It was the relationship between the two. Outdoor pollution set much of the background, but household activity could push indoor exposure even higher at the times when people were cooking, eating, and beginning or ending their day. An outdoor monitor may describe the neighborhood accurately while still missing the air that a person actually breathes in a kitchen.
This was a pilot, not a portrait of Nepal. The measurements came from one household, and covered only part of the year. As such, much remains to be explored. For carrying this project on into the future, my primary interest lies in collecting data across locations where vulnerable groups live: hospitals and schools. An added layer of complexity arises when the ventilation architecture and build material of Nepali houses are taken into consideration, opening pathways to answer questions about the impact of air pollution on vulnerable groups, and the preparedness of Nepali infrastructure for episodes of pollution spikes.
I presented these preliminary findings at the Kathmandu Geo Lab Student Research Symposium. There, the poster became a starting point for conversations with students and researchers from Kathmandu Geo Lab and the Institute of Engineering at Pulchowk Campus, as well as people developing air-quality sensing technology in Nepal. Their questions pushed me beyond the graph itself.
Presenting preliminary findings at the Kathmandu Geo Lab Student Research Symposium. Photos courtesy of Kathmandu Geo Lab.
The project also exposed a more basic obstacle. Finding environmental data can itself become a research task. Existing measurements sit across different websites, documentation varies, and missing periods are not always obvious. One of my next steps is therefore to fully launch the Nepal air-quality dashboard and its source database. I hope it will give students and researchers a single place to find, compare, and use existing datasets alongside measurements from our own sensors.
This summer served as a prelude to the larger study ahead. During the semester, the sensors will support a Bass Connections project led by Professor Henri Gavin. Alongside the existing household pair, we plan to deploy four newly collocated indoor-outdoor pairs in Belauri, expanding the network across two schools, a government office, a concrete house, and an Indigenous Rana Tharu mud-and-straw house.
Deployment is only part of the next stage. We also want to better understand the physics behind how low-cost sensors estimate PM2.5 and why their readings sometimes differ.
Temperature and humidity can affect sensor performance, so we plan to combine pollution measurements with local weather data and study whether these factors can help us correct the readings. Together, these steps will move us from asking whether indoor air differs from outdoor air to understanding why, when, and for whom it differs.
More than half of Nepal’s population lives in the region, and monthly average PM2.5 can approach 100 micrograms per cubic meter during parts of the dry season. The World Bank estimates that 68 percent of the Terai’s exposure comes from outside Nepal. Nepal has also endorsed a “35 by 35” goal, which aims for an annual average of 35 micrograms per cubic meter by 2035.[1] Local measurements cannot stop pollution at the border, but they can help Nepal distinguish what it can reduce at home from what it must address through regional cooperation.
I would like to thank Duke University Libraries for their assistance and support with this research.
References
[1] World Bank. (2025). Towards Clean Air in Nepal: Benefits, Pollution Sources, and Solutions. World Bank Open Knowledge Repository
[2] Ghimire, S., Mishra, S. R., Sharma, A., Siweya, A., Shrestha, N., & Adhikari, B. (2019). Geographic and socio-economic variation in markers of indoor air pollution in Nepal: evidence from nationally-representative data. BMC Public Health, 19, 195. https://doi.org/10.1186/s12889-019-6512-z
[3] Air Quality Life Index. (2025). Warning: Breathing is hazardous to health. Energy Policy Institute at the University of Chicago. https://aqli.epic.uchicago.edu/post/warning-breathing-is-hazardous-to-health-2025-en
[4] Department of Environment. (2025). Status of Air Quality in Nepal in 2024. https://lib.icimod.org/records/ztbs9-fwd88
[5] Pokhrel, A. K., Bates, M. N., Acharya, J., Valentiner-Branth, P., Chandyo, R. K., Shrestha, P. S., Raut, A. K., & Smith, K. R. (2015). PM2.5 in household kitchens of Bhaktapur, Nepal, using four different cooking fuels. Atmospheric Environment, 113, 159–168. https://doi.org/10.1016/j.atmosenv.2015.04.060
The post LIFE Summer Research Grant Reflections: What People Actually Breathe: Indoor vs Outdoor Air Pollution in Nepal appeared first on Duke University Libraries Blogs.
Discover more from DrWeb's Domain
Subscribe to get the latest posts sent to your email.
