Drone-Based Hyperlocal Particulate Matter Monitoring Dataset - Busan New Port Construction Site

This dataset contains second-resolution particulate matter (PM₁, PM₂.₅, PM₁₀) measurements collected by a UAV-mounted Sniffer4D sensor at the Busan New Port North Container Phase 2 Hinterland Development Project (Mt. Yokmang, Busan, South Korea). Data were collected over three field campaign days (January 16, 17, and 23, 2025) using a DJI Matrice 350 RTK drone flying at altitudes of 30–105 m above ground level. Flights covered two primary PM emission sources: blasted rock loading by excavator (AT#1) and dump truck unloading (AT#2). Morning sessions involved both activities simultaneously; lunch sessions involved AT#2 only. The dataset comprises 22 Excel (.xlsx) files and a README.md, totaling 2.22 MB. Each file contains 30 variables including GPS position, 3D spatial distances to emission sources, PM concentrations, and meteorological parameters (temperature, humidity, pressure, wind speed and direction) at 1-second intervals. This dataset was used to train tree-based machine learning models (XGBoost, CatBoost, GBR, LightGBM, Random Forest) for hyperlocal 3D PM prediction, incorporating a novel Distance-Weighted Aggregation strategy for spatial feature engineering. Funding: National Research Foundation of Korea (NRF), Grant No. 2022R1I1A3073717.