Long-term A-weighted day-evening-night sound levels (Lden) was modelled using LUR (Helbich et al., 2025). The LUR model was based on noise measurements from 232 monitoring sites of the Regional Health Inspectorates in the five biggest cities in Bulgaria (75 in Sofia, 45 in Plovdiv, 45 in Varna, 37 in Burgas, and 30 in Ruse) between 2018 and 2022. The fixed measurement stations were adjacent to traffic lanes, industrial sites, and residential and recreational areas. Field workers from Regional Health Inspectorates took measurements with calibrated sound level meters three times a day (at least two measurements were during peak traffic hours) over two daytime periods between 07:00 and 19:00 h following ISO 1996–2 protocol. Each measurement was taken approximately 1.5 m above the ground for 15–20 min. For each sampling point, we had sound level data for each year from 2018 to 2022. Due to the high correlation and stability of the measured sound levels over time, we used the average values from 2018 to 2022 as the dependent variable in our LUR models.

Noise levels were predicted using the extreme gradient boosting algorithm and transport- and land-use-related predictors derived from OpenStreetMap and the 2018 pan-European Urban Atlas. The model has demonstrated a ten-fold cross-validated R² of 0.68, RMSE of 4.739 dB(A) and mean absolute error of 3.595 dB(A). The model’s residual Moran’s I was −0.039 (p = 0.186).

We mapped the traffic noise estimates by superimposing a 50 m grid on each city to obtain predicted noise levels in dB(A).

Downloads

The noise data are available as part of this publication.