Integrated multi-scale assessment of particulate matter in coastal environments: long-term analysis, mobile monitoring, and spatial prediction techniques

Abstract

This study presents a multi-scale analysis of Particulate-Matter (PM) concentrations across the coastal region in India using long-term air quality datasets, mobile monitoring, and advanced spatial-interpolation techniques. Long-term observations (2016–2025) were analysed to examine interannual and seasonal variations in PM across the coastal region. Fine-scale spatial variability in urban microenvironments (February 2026) was investigated using an instrumented two-wheeler platform equipped with real-time aerosol and meteorological instruments, including DustTrak, an Optical-Particle-Sizer (OPS3330), and low-cost sensors deployed in different orientations. The spatial distribution and prediction of PM at both regional and microenvironmental scales were further evaluated using Inverse-Distance-Weighting (IDW), Ordinary-Kriging (OK), Gaussian-Process-Regression (GPR), and Tabular-Prior-data-Fitted-Network (TabPFN-V2) interpolation methods. Long-term analysis showed strong seasonal PM variability, with winter PM 2.5 peaks (~ 60 µg/m 3 ) and $$\:{\text{P}\text{M}}_{10}$$ maxima (~ 65 µg/m 3 ) associated with southwesterly winds. At the regional scale, OK and GPR showed superior PM 2.5 prediction performance, with the lowest RMSE values (2.5–2.6 µg/m 3 ) across the Indian coastal state of Goa. Mobile monitoring revealed PM 2.5 hotspots (> 200–400 µg/m 3 ) and elevated total number concentration near major traffic corridors, while microenvironmental interpolation showed that TabPFN generated smoother and more spatially coherent PM patterns with the lowest RMSE (~ 51 µg/m 3 ) compared to IDW, OK, and GPR. Sensor orientation analysis showed improved particle sampling and lower-bias for front-and right-facing sensors due to enhanced airflow exposure.