Testing meteorological classifications for the prediction of long-term average sound levels

Abstract

The paper describes the results of a preparatory study for the planned introduction of a harmonised method to assess long-term sound levels in Europe. The purpose of this study is to quantify the error in the prediction of long-term sound levels when propagation calculations refer to a limited number of between 1 and 121 representative meteorological situations. This error is determined with respect to the average sound level resulting from parabolic-equation simulations for 1452 meteorological situations of consecutive 6-h intervals during one year. The meteorological situations were taken from meteorological tower measurements at Garching near Munich, Germany. The various meteorological conditions enter the model through two-parameter logarithmic–linear vertical profiles of the effective speed of sound which were fitted to the tower measurements. While source power and ground impedance were held constant, the model results elucidate the sound level variations due to the meteorological variability. A spread of 18 dB was found in the calculated instantaneous levels at 200 m range for rigid ground. The spread was 42 dB at 1000 m. Because of the asymmetric frequency distribution of wind directions, a spread of still up to 5.5 dB was found in the calculated annual average levels at 1000 m range for various directions of propagation. As a consequence of the specific meteorological conditions the annual average night-time level exceeds the annual average day-time level by up to 5 dB. Only 25 logarithmic–linear profile classes are necessary to determine the long-term average sound level at 1000 m range with an accuracy better than 2 dB. The use of purely linear profile classes slightly impairs the results, whereas the use of purely logarithmic profile classes cannot be recommended.