Assessing the influence of the sampling strategy on the uncertainty of environmental noise measurements through the bootstrap method

Abstract

The present study deals with the influence of the sampling parameters on the uncertainty of the equivalent continuous noise level in environmental noise measurements. It is also intended to assess the possibility to define, as much as possible, a general methodology which enables the definition of the parameters once the precision level is defined. The results were obtained through the resampling of a sequential set of equivalent noise levels, acquired at 5min intervals throughout a continuous measurement period of 17days, thus including the three reference periods, day, evening and night. From these data, the written resampling software allows one to recreate several sampling strategies which consider the duration periods and their frequency, as they relate to simulating different measurement strategies. Since the software tool simulates a large number of repetitions of the data acquisition campaign, generating randomly the time location of the sampling periods, the evaluation parameter is the standard deviation of the equivalent noise level, for the considered reference period.For short term measurements, not only the integration time but also the number of samples obtained in order to characterize a given time period have influence on the uncertainty of the result. This influence is higher for the time periods where sound levels suffer a greater variation, such as during the night. If both parameters, time and number of measurements, are not carefully selected, it may result in high uncertainty which contributes to the loss of precision of the measurements.Another interesting outcome of this work is that it has been possible, for the three distinctive temporal variation patterns, to derive mathematical expressions representing the effect of the sampling parameters on the uncertainty of the equivalent noise level. Despite the differences in the time patterns of the equivalent noise level along the three reference periods, the three expressions obey to the same structure, presenting only variations in their numerical coefficients. This fact indicates the possibility of the existence of a general model to define the effect of the sampling parameters on the uncertainty of the equivalent noise level.