Spatial interpolation of noise monitor levels

Abstract

Continuously recording noise monitoring stations provide feedback of the noise environment at monitor locations. While this feedback is useful, it only provides information at a few point locations, and in many cases it is of interest to know the noise level(s) at the locations between and beyond noise monitoring locations. In this study, we test the accuracy of several spatial interpolation models with experimental data collected during the Strategic Environmental Research and Development Program (SERDP) Community Attitudes Towards Military Blast Noise study. These datasets include 9 months of blast noise events captured at two different study locations. In both cases, a small number of monitors (e.g., 3–9) were located over a large region of interest (e.g., 1–8 km2), thus providing realistic operational conditions. The utility of deterministic (e.g., nearest neighbor, Delaunay triangulation, thin plate splines, etc.) and stochastic (e.g., geostistical or kriging) interpolation models for estimating single-event and cumulative noise levels is examined using leave-one-out cross validation. The accuracy of each approach is assessed with the root-mean-square-error (RMSE), and we discuss the practical implications of implementing such approaches in real-time systems.