Replication Data: Traffic noise assessment in urban Bulgaria using explainable machine learning

This data package contains the input data and analytical results of a machine learning-based traffic noise prediction model for the five largest Bulgarian cities. The model was trained based on measurement data from 232 fixed-site monitors. Noise measurements (A-weighted, in dB) were taken from 2018 to 2022 through calibrated sound level meters three times a day over two daytime periods between 07.00 and 19.00 h following ISO 1996-2 protocol norms. The noise predictions obtained through the extreme gradient boosting model and transport- and land-use-related predictors are shared as GeoTIFF files for each city on a 50 m raster. The raster layers are shipped in the projected coordinate system for Bulgaria (EPSG code: 7801). The data processing and statistical analyses are carried out in the R environment. The R scripts, the input data, and the predictions are provided in this data repository. Detailed information about the file naming and data is given in the read-me file. This dataset is accompanied by a research article (Helbich et al. 2025, Traffic noise assessment in urban Bulgaria using explainable machine learning, Sustainable Cities and Society, in press). Contact persons are Marco Helbich (m.helbich@uu.nl) and Angel Dzhambov (angel.dzhambov@mu-plovdiv.bg).