Assessment of PM2.5 exposure and respiratory health risks for vulnerable populations in urban underpass: a predictive MLR approach in Chiang Mai, Thailand.

Chiang Mai, Thailand, experiences severe air pollution episodes annually, with particulate matter (PM2.5) posing significant health risks-particularly for road tunnel users who face prolonged exposure within semi-enclosed, poorly ventilated environments. Despite the province having the highest number of road underpasses in Thailand, PM2.5 dynamics and associated health risks within these structures remain understudied. This study aimed to develop a multiple linear regression (MLR) model for forecasting PM2.5 concentrations at the Don Kaew Road Tunnel in Chiang Mai and to quantify the health exposure risk for vulnerable population groups using a comparative risk assessment framework. Air quality and meteorological data were collected over approximately 1 year (August 2022-July 2023), yielding 385,622 records from five low-cost DustBoy sensor stations installed along the tunnel. Variables included CO, CO2, NO2, SO2, O3, PM10, PM2.5, TVOC, CH₂O, temperature, relative humidity, air pressure, wind speed, wind direction, and rainfall. The MLR model was developed using a stepwise method in R (v4.3.0), with analyses stratified by weekday/weekend and day/nighttime periods. Health risk was assessed using inhalation rate, daily intake, and relative risk calculations based on Thai and WHO exposure factor standards. The MLR model achieved a high goodness-of-fit (R2 = 0.9995, p < 0.001), with PM10, temperature, relative humidity, and monthly and daily time variables as significant predictors. Strong positive correlations were observed among PM2.5/PM10 (r = 0.9995), CH2O/TVOC (r = 0.9999), and CH2O/CO2 (r = 0.9999), while negative correlations were found for temperature/relative humidity (r = - 0.7967) and relative humidity/O₃ (r = - 0.5939). PM2.5 concentrations were significantly higher on weekdays than weekends, peaking on Tuesdays, and were elevated during daytime and dry season months. Mean PM2.5 inside the underpass (65.42 µg/m3) was significantly higher than the aboveground control site (42.18 µg/m3; t = 8.432, p < 0.001), representing a 1.55 times higher intake per trip. When comparing the most vulnerable population (infants in the underpass) against the reference group (healthy adults at the surface), the relative risk reached approximately 3.5 times. The MLR model effectively captured PM2.5 variability within an urban road tunnel and demonstrated that even brief transit exposures (approximately 2 min) in unventilated underpasses generate disproportionate health burdens for vulnerable populations. These findings underscore the need for improved mechanical ventilation and health-informed urban planning in Chiang Mai's expanding underpass infrastructure.