Vízügyi Közlemények, 2004 (86. évfolyam)

3-4. füzet - Rövidebb tanulmányok, közlemények, beszámolók

624 Kovács József - Szabó Péter - Szalai József on the remaining data is shown on Figure 5. The hypothesis analysis made on the basis of the magnitude of the performance density (at 99.9% significance level in our case) gave positive results of the following periods: 4, 9,10.95 and 13.36 years. Table I was composed using the results obtained for those 143 observation wells that had longer than 30 year time series, showing the occurrences of the periods of various lengths. Table II shows the results of the statistical analysis of the periodicity of all observation wells, which were included in these investigations. Yearly periods were observed for 97% of the observation wells, as eventually expected. For longer periods no single well definable value could be determined. Therefore we defined given intervals around the various periods identified. For the 5 year periods we defined the interval 5±0.75 but this does not mean fixed upper or lower limits. The thus defined 5 year period was identified for 87.4% of the observation wells (Table II, Figure 6). We have defined an 11 year period as 10±1.5 years, on the basis of the similar approach and this corresponded to 68.8% of the observation wells (Table II, Figure 7). The 16 year period was defined for wells with 58 years long observation time series or longer ones and 27.5% of the wells fall into this category. The factors, which create these periods are results of random stochastic processes that also appear in the groundwater regime. The analysis of these changes can be carried out by the wavelet spectral analytic method. The precondition of applying this method is the equal-interval sampling and if it is not the case then re-sampling must be applied. In the case of the groundwater time series investigated a spline-function was fitted to the intervals between two sampling time points, then the repeated-sampling was made on this spline with 15 days intervals. Figure 8 shows, as an example, the normalised hydrograph and wavelet spectrum resolution of the observation well of Kunszentmiklós. In this wavelet spectrum the cold colours indicate the existence of the less likely periods, while the warm colours the more likely ones. The curve of the shape "U" marks the confidence interval. Analysing the time series of 80 observation wells with the wavelet spectrum resolution one finds that the ceasing of periods are bound to varying points of times, but this occurs in some periods in high number. For example, the annual period does not occur for 90% of the wells in the early 1970-ies (Figure 9). The results can be checked with a multivariate data-analytical method, with the hierarchic analysis of clusters. The dendogram obtained in this way (Figure 10) indicates, in harmony with the findings of the wavelet spectrum analysis, that the regime of the groundwater table was significantly different in the period of 1970-1975 from those of the groundwater level changes before and after this period.