Speaker
Description
Strategic noise maps (SNMs) represent one of the few available data sources on acoustical information at high-spatial resolution. However, SNMs provide traffic noise predictions, not accounting for other acoustic properties. Recently, studies demonstrated promising results in modelling additional acoustic properties based on land-use (LU) data.Here, we assess the predictive performance of a LU-based model trained and externally validated to the Ruhr region, Germany. The model generates high-spatial predictions for five acoustic properties: Total Noise, Intelligibility, Sharpness, Biophony, and Acoustic Dominance. Acceptable predictive performance on external test data was achieved only for Total Noise and Acoustic Dominance.Building upon these results, we compare LU-based Total Noise predictions to Traffic Noise predictions from SNMs. Predictions were made for 737 participants from a population-based cohort-study. LU-based predictions were on average 6.8 dB higher than SNMs estimates (range: 49.8–69.7 vs. 35.2–72.2 dB resp.), and the two measures correlated only moderately (Pearson's r = 0.64). This suggests that Total Noise is only partially related to Traffic Noise, likely because LU-based models capture a broader composition of sound sources. Overall, these results indicate that LU-based models may represent a complementary approach to conventional SNMs, but currently only for a limited set of acoustic properties.