Assessment of road noise at long distances from the source; a comparison between measurements and sound prediction models evaluating the effects of noise barriers

Abstract

Predicting how sound propagates from road traffic noise is essential for environmental noise assessments and for designing effective mitigation strategies. However, the accuracy of current theoretical models tends to become more uncertain as the distance from the source increases, especially when factors such as vehicle speed, traffic volume, and terrain complexity increase, or when noise barriers are introduced. Although there has been research on long-distance sound propagation, few studies have examined the differences between predicted and measured sound levels beyond 150 meters, particularly in cases where noise barriers are involved. This thesis investigates discrepancies between theoretical predictions and short-term field measurements of road noise at distances ranging from 180 to 350 meters, both with and without noise barriers. Sound levels predicted by two commonly used models, Nord2000 and CNOSSOS-EU, were compared against measurement data from four locations with varying terrain conditions. These were short-term measurements (1-3h) that followed the NT ACO 0039 standard. Additionally, a simpler prediction model was developed in MATLAB using standard calculation formulas to assess its performance against the more complex models. The results show consistent discrepancies between predicted and measured values, particularly in complex terrain. Nord2000 achieved the closest agreement, with RMSE values around 3.3–4.1 dBA, while CNOSSOS-EU showed slightly higher deviations of 3.7–4.5 dBA. These error levels are consistent with previous studies and may be interpreted as indicative of model prediction uncertainty. The inclusion of noise barriers appeared to increase prediction errors. However, further data collection is necessary before drawing statistically significant conclusions. Understanding how noise propagates over long distances is key in order to evaluate the reliability of prediction models. Continued research is needed to strengthen confidence in noise assessments at extended ranges