Abstract
The factors which control the crest settlements of the Kremasta Dam (Central Greece), one of the highest earthfill dams in Europe, were analyzed on the basis of the threshold correlation. The latter is a new stochastic technique which permits identification of parts of time series which are highly correlated on the basis of the optimization of the correlation coefficient between two parameters in combination with a high-pass filter of gradually increasing width.
Based on a unique, >35 years long monitoring record consisting of leveling data, reservoir levels and precipitation levels we formed three time series describing the parameters which may affect the dam geometry (fluctuations of the reservoir level, of the rate of change of the reservoir level and of the rate of rainfall) and a fourth time series describing the dam settlements (settlement index fluctuations). Threshold correlation analysis permitted to identify critical levels for each of the above three parameters influencing the dam settlements. It was found that none of the three parameters alone can explain excessive dam settlements, but if the critical values of the three parameters are at the same time exceeded (i.e. under a conditional Boolean probability), the rate of the settlements exceeds its critical value. This approach may prove useful for other dams, but also in the analysis of data in other fields of science and engineering.
© de Gruyter 2009
