Effects of local and global meteorological data on predicting blast noise levels

Abstract

Predicting noise levels due to military blasts is important near training installations, since surrounding communities may be adversely affected. Prediction models, however, rely on atmospheric or weather data that is not readily available or easily attainable. In this study, blast noise level predictions 1-8 km from the source are made using two sources of meteorological data: publicly available data derived from weather stations that are tens of kilometers from a detonation site, and near-ground weather stations that are 2-6 km from the source. To predict the blast noise levels, statistical learning regression models are trained on each set of meteorological data and over 1000 measured blast events. Predictions from both meteorological data sources are compared to actual blast noise levels recorded during a field experiment by Valente et al. [2012], which were captured under a wide variety of atmospheric conditions. The root-mean-square errors for each model are compared, and the most important input pa...