Time-variant behavior of water chemistry is controlled by internal factors of a hydrologic system and by external stimuli. Periodic sampling of water provides data to compute this variation. The observed time variation can be modeled as a discontinuous time series, composed of time-systematic and stochastic components. The stochastic component is a residual attributed to random fluctuations in the parent water chemistry and in sampling analysis. Nonparametric tests confirm and amplify results of serial correlation. Tests are selected for their discrimination against specific time-variant alternatives to constant behavior. Several examples demonstrate the interplay between serial correlation and nonparametric techniques and some of the interpretative difficulties met in practice. (WRSIC-abstract)