There has been much effort in the US and worldwide to measure, understand and manage natural soundscapes which are often complex due to a multitude of biological, geophysical, and anthropogenic influences. The sound pressure level is a time and space varying quantity that represents the aggregate of present sources. This work presents a predictive model relating seasonal sound pressure levels to geospatial features such as topography, climate, hydrology and anthropogenic activity. The model utilizes random forest, a tree based machine learning algorithm, which does not explicitly incorporate any apriori knowledge of acoustic propagation mechanics. The response data encompasses 271,979 hours of acoustical measurements from 192 unique sites located in National Parks across the contiguous United States. Cross validation procedures were used to evaluate model performance and identify GIS explanatory variables with predictive power. Using the model, the effect of individual explanatory variables on sound pressure level can be isolated and quantified revealing trends across environmental gradients. An example application of projecting predicted sound pressure levels across the Olympic peninsula is discussed. Because many wildlife habitats, geological processes, and anthropogenic impacts occur on a regional scale, the extent of acoustical analyses must be on similar scales.

/pdf/a-geospatial-model-of-ambient-sound-pressure-levels-in-the-56l4cf3acz.pdf

A geospatial model of ambient sound pressure levels in the contiguous United States.

Assessment of the environmental impacts of audible noise produced by human activities -- such as construction and operation of industrial facilities, vehicular traffic, and aircraft operations -- requires the use of sound-propagation computer programs capable of predicting the sound-pressure spectra that will be produced by these noise sources at distances as great as 10 km. Under acoustically unfavorable meteorological conditions, such as ground-based temperature inversions and noise-sensitive receptors being located downwind from intrusive sound sources, audible noise levels can be increased, even on the order of 20 dB. This paper presents a brief overview of computer programs currently available (``operational``) for predicting the effects of such meteorological conditions on audible noise levels in a format that allows gross comparison of the relative capabilities of the programs from the standpoint of engineering application to environmental impact assessment.

R.E. Liebich

Papers

Comparative review of available computer programs for modeling meteorological and terrain effects on long-range propagation of audible noise