The comparison of road traffic noise levels from prediction models and actual noise measurement

This research focuses on the comparison of the traffic noise levels that are calculated from prediction models, and the noise levels from actual traffic noise measurement. The research was conducted on Regent Street, Sandy Bay Road and the Tasman Highway, Hobart. At each site, the traffic noise measurement was operated seven days continuously, with site topography and at least 10 hours of traffic volume studies. Repeat measurements were taken at each site. Three prediction models: T-Noise, STAMSON4.1 and ITFNS1.1 were used for predictng traffic noise levels at each study site. The study showed that in day-time, all recorded noise levels under normal conditions exceeded both NSW road traffic noise criteria and Tasmanian road noise guideline (Leq 15 hr = 60 dB(A)) and (L10 18 hr = 63 dB(A), respectively), and most of the night-time noise levels exceeded both night-time criteria of NSW and Tasmania (Leq 9hr = 55 dB(A) and Leq 1 hr = 55 dB(A), respectively). Traffic volume studies showed that there were approximately 850 veh/hr with 11 % trucks on Regent Street, 1600 veh/hr with 13% trucks on Sandy Bay Road and 3200 veh/hr with 6% trucks on the Tasman Highway. The comparison studies showed that, within the absolute noise levels comparison, on Regent Street and Sandy Bay Road, T-Noise showed the greatest accuracy with the smallest variation, an average of 0.76 dB over prediction on Regent Street and 1.12 dB under prediction on Sandy Bay Road. STAMSON4.1 provided less accuracy than T-Noise, with an average under prediction of 1.86 dB on Regent Street and 2.51 dB under prediction on Sandy Bay Road. ITFNS1.1 exhibited unreliable predictions for all study sites. None of the models presented reliable results for the Tasman Highway study site. The results from correlation tests on predicted and actual noise levels showed that T-Noise and STAMSON4.1 provided no significant difference in the correlation coefficient values. At the urban road sites T-Noise provided the greatest correlation of R2 = 0.57, while STAMSON4.1 provided R2 = 0.43, with only R2 = 0.006 gained from ITFNS1.1. For highway conditions, T-Noise provided the greatest correlation of R2 = 0.64, while R2 = 0.43 was offered by STAMSON4.1. However, in correlation tests of urban and highway conditions combined, both models showed poor results with R2 = 0.08 and R2 = 0.27 from T-Noise and STAMSON4.1 respectively. These results provide an indication of the accuracy of these models for this application. In the case of correlation between traffic volumes and the accuracy of noise level predictions; only T-Noise shows a correlation between these two factors indicating a propensity for improvement in its accuracy for this application.