XGBoost-based calibration of the portable PAQMON 1.0 PM monitor in outdoor conditions

This study examines the applicability of the XG calibration procedure for the PAQMON 1.0 portable platform in field conditions. Although PAQMON 1.0 is designed to accommodate additional sensor modules, this study utilized a version adapted explicitly for monitoring respirable particulate matter concentrations and meteorological parameters. Measurements were conducted in Bor, Serbia, during both the heating (GP) and non-heating (NGP) periods to account for seasonal climate and air quality variations. The platform demonstrated stable performance without significant signal drift or hardware failures. The correlation between PM2.5 concentrations and the reference Grimm EDM180 monitor was R = 0.65 (p ≤ 0.001) during GP and R = 0.40 (p ≤ 0.001) during NGP. The application of the XGBoost (XG) model significantly improved the accuracy of PM2.5 measurements compared to linear regression models. The adjusted R² values were 0.68 for GP and 0.44 for NGP, while NRMSE values ranged from 0.43 to 0.83. Although the XG model exhibited slightly lower performance during NGP, its calibration enabled the PAQMON 1.0 platform to achieve a Data Quality Objective (DQO) of 50% in both measurement campaigns. These findings confirm the practical applicability of XG calibration for enhancing the accuracy of low-cost sensor systems in air quality monitoring.