Sampling frequency is one of the most crucial factors in the design of groundwater quality monitoring systems. Monitoring systems in general have two major objectives: (1) to describe natural processes and long-term changes and (2) to serve as alarm-systems and detect single pollution events. A comparison between two data sequences of different sampling frequency - weekly and monthly - is made through an example of the groundwater quality monitoring system in the karstic region of the Transdanubian Mountains in Hungary. Hydrogeochemical time series were first decomposed into their components: trend, periodicity, autocorrelation, and rough in succession. In order to identify outliers within the rough, Exploratory Data Analysis (EDA) was applied. Optimal sampling frequency was determined based on the analysis of the above components. Results have shown that: (1) seasons shorter than two months do exist in the studied time series which cannot be captured by monthly sampling; (2) for monitoring seasonal processes samples should be collected at the Nyquist frequency (at least two samples per period); for pollution detection autocorrelation lag-time (or semi-variogram range in time) should determine the sampling distance; in the lack of autocorrelation property the analysis of outliers should guide the sampling design; (3) cross-correlation analysis between precipitation and the observed parameters indicative of pollutant travel time yields valuable additional information on the pollution sensitivity of the hydrogeological system.