The contribution presents an implementation of Kalman filtering for estimation of unknown parameters of a well-known hydrostatic-seasonal-time model which is used in dam safety procedures. The unknown parameters present a system state of the KF, which is iteratively improved by new information entering with new observations–in our case measured relative displacements. The analyses showed that the KF can detect statistically significant changes in dam behaviour by testing measurement innovations after a successful initialization phase and stabilisation of the parameters. Since the KF is implemented on the HST-model the algorithm enables detrending of reversible and irreversible deformations, and analysis of a long-term trend for each measurement epoch. The whole process is strongly influenced by the process noise intensity scalar which defines weighting between the process and the measurement noise. Defining an appropriate value of the process noise intensity scalar is the main challenge of the algorithm, where the convergence of a posteriori error covariance matrix is used as the main criterion in the tuning process for the case study. In the case study, time series of measured displacements and water level in impounding reservoir of an embankment dam for a period of 21 years are used to test the proposed algorithm. Measurements were captured with a fully automatized monitoring system based on tachymetric observations.