The doctoral dissertation analyses the water regime of coastal karst Vransko Lake in Dalmatia (Croatia). This is a crypto-depression in the immediate vicinity of the sea with which it communicates through the karst underground and through a canal dug back in the 18th century. For that reason, the lake water becomes saline in extremely dry periods, when the salinity increases several times in relation to the standard level, endangering the lake ecosystem protected as a nature park. The paper analyses the lake water balance and precipitation trends, air temperature, inflows to the lake and oscillations in the levels of the lake, the sea and the chloride content. It has been identified that Vransko Lake is a water resource where the impact of the climate change/variations is already felt and that their continuation might jeopardize the existence of the lake as a freshwater/brackish system. Due to karst characteristics of the basin and the lake area, the inflows of water to the lake and losses of water from the lake are only partially monitored in hydrological terms, which is why a methodology for their overall evaluation has been developed. The model investigations included the definition of a conceptual model of the action of the lake system, as well as of the model of inflow into the lake system. Simulation modelling of hydrologic processes related to the investigated area of Vransko Lake as well as of the wider region was applied. A model was developed for the evaluation of overall annual inflows into the lake system in the conditions of the expected continuation of the already registered climate change/variations until the year 2100, using the results of estimation of the distribution of annual precipitation and air temperatures obtained from the RegCM3 and Aladin models. Using the machine learning models, annual changes of hydrological conditions in the lake and its quality (chloride content) were modelled. The obtained results show that the application of the models of this generation facilitates both the discovery of interrelations in systems as complex as aquatic systems in the karst and the simulation of the behaviour of such systems in the modified circumstances, with system status forecasts on the basis of characteristic hydrologic indicators - water level, scope of salinization, etc.