An Extended Model to Predict Reflections from Forests

A model to calculate the sound exposure level due to forest reflections is presented. The model is based on the work published earlier in this journal in which a single tree was solely represented by a vertical cylinder. Comparisons with measurements revealed two weaknesses of the chosen approach. On the one hand it showed that levels could not be predicted correctly at shorter propagation distances. And on the other hand the vertical directivity, i.e., the reflection characteristic with elevation, which was observed in the measurements, could not be reproduced. It turned out that the modeling approach had been chosen too simplistic and that the contributions from the branches cannot be neglected entirely. Therefore the model has been extended to take reflections at the canopy, represented by a number of spheres, into account as well. The model has been implemented in the new Swiss simulation model for railway noise sonRAIL. Validations of this implementation with measurements generally show a good agreement with a standard deviation of 2.9 dB(A) up to propagation distances of 500 m. For greater distances the model tends to overestimate the resulting sound exposure level. However the corresponding attenuations at such distances are in the range of 80 dB(A) and therefore the resulting sound exposure is unlikely to yield a relevant contribution to any noise exposure. The deviations up to 500 m can primarily be explained by the distinct properties of the forests involved. In order to further increase the accuracy of predictions specific information about the forest properties would be needed. Mainly the type of trees (hardwood or softwood), the average tree diameter and height as well as the forest density have an influence. However, such information is generally not available and therefore one has to rely on statistical averages, as it was done for this publication. Additionally the accuracy could be increased by applying a more sophisticated propagation calculation with the focus on ground attenuation and meteorological effects. The associated additional computational effort, however, will question an application for noise mapping.