Statistical independence of a series of ambient noise measurements

Abstract

The autocorrelation of a series of noise parameters (Leq, L10, L50, or L90) was examined. The noise parameters were derived from 15-min records of ambient noise collected at the same hour (0000, 0600, or 1800 h) on each of 100 (predominantly consecutive) days at a rural site. The same analysis was also performed for a suburban and a quiet, urban, residential site. The first zero crossing (indicating statistical independence) occurs at 9–13 days for the rural site. A run test for randomness (0.05 level of significance) was also used. The random hypothesis consistently failed for the rural site, showing that consecutive days do not produce a random string of data. It is shown that for an efficient ambient survey, samples should be separated by at least three days for rural sites, but that consecutive day samples are normally satisfactory for urban and suburban sites.