Uncertainties of band sound levels when estimated from monochromatic outdoor sound propagation calculations

Octave band and third octave bands sound levels are widely used in environmental acoustics, and most numerical models for outdoor sound propagation estimate them through the combination of several monochromatic calculations. As some of these models can be very time consuming when long range distance calculations are needed, the number of monochromatic calculations performed for each band is often necessarily very limited. That induces uncertainties on the estimate of the noise level of the band that can be large when the spectrum amplitude varies widely inside a band. This is the case for example when interferences due to ground effects occur. The paper presents estimates of these uncertainties depending on frequency band and on source-receiver locations. The uncertainties are evaluated thanks to a quasi Monte-Carlo technique from many calculations using a Rudnick propagation model and a Latin Hypercube Sampling numerical experimental design for input parameters (source-receiver locations). The results provide practical indications on the minimum number of monochromatic calculations to do per band, in order not to exceed a given uncertainty level.