The Varaždin aquifer represents the main source of water for public supply, agricultural, and industrial purposes in the Varaždin County in NW Croatia. In the last decades, this area has experienced contamination of groundwater with nitrates. This study describes the conceptualization of the Varaždin aquifer for the purpose of developing numerical model of groundwater flow and nitrate transport. Within the study, three important elements are defined: aquifer geometry, recharge from precipitation, and other boundary conditions. 3D aquifer model revealed that Varaždin aquifer consist of three layers: upper aquifer, semipermeable interlayer, and lower aquifer. The Wetspass-M model was used for the assessment of spatial and temporal distribution of water balance components for the period 2008–2017. Results of the model indicate that the average annual precipitation is distributed as 34% groundwater recharge, 21% surface runoff, and 45% actual evapotranspiration. The maps of equipotential lines show the behavior of the aquifer system and define boundary conditions, i.e., recharge and discharge areas of the aquifer: an inflow boundary from Drava River and accumulation lake Varaždin on the northwest and north, no flow boundary on the west and south, and an outflow boundary on the east.

Exploring the interaction between precipitation, surface water, and groundwater has been a key subject of many studies dealing with water quality management. The Varaždin aquifer is an example of an area where high nitrate content in groundwater raised public concern, so it is important to understand the aquifer recharge for proper management and preservation of groundwater quality. The NW part of the Varaždin aquifer has been selected for study area, as precipitation, Drava River, accumulation lake, and groundwater interact in this area. In this study, groundwater and surface water levels, water temperature, water isotopes (²H and ¹⁸O), and chloride (Cl⁻) were monitored in precipitation, surface water, and groundwater during the four-year period to estimate groundwater recharge. Head contour maps were constructed based on the groundwater and surface water levels. The results show that aquifer is recharged from both Drava River and accumulation lake for all hydrological conditions–low, mean, and high groundwater levels. The monitoring results of water temperature, chloride content, and stable water isotopes were used as tracers, i.e. as an input to the mixing model for estimation of the contribution ratio from each recharge source. The calculation of mixing proportions showed that surface water is a key mechanism of groundwater recharge in the study area, with a contribution ratio ranging from 55% to 100% depending on the proximity of the observation well to surface water.

Floods are defined by maximum water levels or flow of high-water waves. Here, we defined the deterministic method for the calculation of the probability of a high discharge event, named as the Probability Of Success (POS). The POS method previously developed for petroleum subsurface systems has been modified for the surface hydrological system with the purpose of flood prediction. The case study of this research is the small basin of Kašina Stream on Medvednica Mt. (NW Croatia). The data are obtained upstream from the hydrological station Gornja Kašina. The POS model is defined by four categories. Each geological category is described with accompanied events and probabilities. Floods are defined by four categories: total precipitation, total water flow, basement, and maximal water capacity in soil. The categories total precipitation and basement were divided into two sub-categories each: quantity and duration; porosity and soil depth. Data are collected for a hydrometeorological event, namely an intensive convective storm on 24–25 July 2020, when Zagreb was locally hit by heavy urban floods. The presented probability method yielded a probability of 1.76% that such an event could happen to the station. However, the flooding was not recorded. A comparison of the real event and the predicted probability supported the adequacy and applicability of the method, showing it has high reliability. The presented probability model could be easily applied, with small modifications, to the entire area of Northern Croatia for the prediction of small basin flooding events.