Modelling the urban acoustic environment using land use-based gradient boosting.

Area of Science:

• Urban planning and environmental science
• Acoustic ecology and public health

Background:

• Urban noise negatively impacts human health, but beneficial acoustic properties are understudied.
• Lack of high-resolution methods beyond noise metrics limits research on health-promoting urban acoustic environments.

Purpose of the Study:

• To evaluate land-use type (LUT) models for predicting urban acoustic properties.
• Focus on four acoustic indices (Articulation Index, Bioacoustic Index, Link Density, Sharpness) and LAeq.
• Compare model performance against strategic noise maps and literature.

Main Methods:

• Trained gradient boosting models using 2,746 acoustic measurements from Bochum and 90 from Essen.
• Models predicted acoustic indices based on 77 LUTs within 50m and 300m buffers.
• Evaluated models on independent test sets from both cities.

Main Results:

• Link Density predictions showed low deviation (RMSE 0.17-0.21) with R² of 0.27-0.3.
• LAeq predictions achieved moderate performance (RMSE 4.4-4.8 dB(A), R² 0.46-0.52).
• LUT models outperformed strategic noise maps for LAeq prediction in Bochum.

Conclusions:

• LUT-based models offer a scalable, high-resolution method for characterizing urban acoustic environments.
• These models show potential for predicting Link Density and LAeq, advancing health research.
• Findings support designing healthier urban spaces with improved acoustic quality.