Comparison of sound propagation codes: Milstøy, BNoise and a PE-method

Ranges for military training must ensure that noise reaching neighbors is kept at an acceptable level. In Norway, strict noise limits are set by the authorities. To comply with these regulations, the Norwegian Defence and the Norwegian Defence Estates Agency needs accurate tools for calculating the noise propagation from noisy activity out to the neighbors of the training eld. The goal of the study documented in this report was to compare four different prediction methods to an experimental data set, and see which method provided the most accurate results. The prediction methods, in order of complexity, were: 1. Industry Noise (IN) model in Milstøy (current Norwegian standard method) 2. Nord2000road (N2R), a "quick" ray-tracing kernel included in Milstøy and currently under investigation as a possible upgrade 3. Fast Field Program (FFP) as implemented in the US noise assessment software, BNoise 4. Generalized Terrain Parabolic Equation (GT-PE), a research model utilizing the parabolic equation approximation to the acoustic wave equation. The predictions were compared to selected data from the Nortrial data set, with the speci c data physically recorded at Finnskogen, Norway. In the US, noise maps for large weapons are made with BNoise. BNoise is a tool with the same functionality as Milstøy. With regards to the basic assumptions, the propagation model in BNoise seems to be more advanced than both IN and N2R (in Milstøy). One difference between MS and BNoise is that BNoise pre-calculates noise propagation and stores it in lookup tables, while MS performs instantaneous calculations. The PE method can handle horizontal range-dependence, allowing the inclusion of terrain and horizontally-varying meteorology and ground type. Re nements of the PE are frequently addressed at professional conferences and the PE is generally considered state of the art. This report describes experiences with a recent implementation of the PE method to predict military shooting noise. Choosing between FFP and PE it seems that the advantage with the FFP is that it is proven to work, and is in use. Typically more advanced models seem to have a tendency to become unstable for realistic conditions. The PE could have computational advantages, and be more oriented towards the future due to its higher degree of exibility. To be able to make a decision about which direction to take, we here try to evaluate some properties of the different computational methods. The results of the study are not conclusive, but much was learned. One of the propagation conditions in BNoise consistently gave the best results, but other propagation conditions in that method were no better than any other method. The GTPE gave reasonable results in many cases, but the choice of which propagation condition parameters provide the best results, i.e. included terrain and type of meteorological pro le used, varied from case to case. Both the IN and N2R methods did not perform as well and had a strong tendency to overpredict the received noise levels. This report provides details on the data set used, overviews of each of the computational methods, and results of all trials. Further analysis is needed to determine which method best serves the needs of the Norwegian Defence.