Characteristic and design soil parameters: use of statistics

This paper emphasises that the geotechnical parameter governing the occurrence of a limit state is not the point value as measured by a laboratory or field test, and this has important implications for the determination of soil parameters with statistical methods. Two confusing concepts in statistical soil mechanics are discussed: the modelling of innate variability by a random field model, and the distinction between population mean, mean value of a failure domain and sample mean. These two concepts are intrinsically related. This paper demonstrates that most of the confusions or dilemmas associated with the characteristic value and the use of partial material factors may be removed by considering appropriate probabilistic or statistical concepts. It needs to be emphasised that ‘considering’ does not equate to ‘routinely using’. For a sandy soil, it is argued that the methodology of taking a prudently assessed critical state friction angle as the design value is statistically consistent with the definition of the design value in Eurocode 7.