The Pantanal wetland is one of the least explored regions of South America. It is characterized by an outstanding flora and fauna adapted to a seasonal flood pulse controlled by a dry and a wet season within each year. The resulting inundation covers in average an area of approximately 150 000 km2 and is seen as the most important driver for ecological integrity. Evaporation from the large floodplain is supposed to influence the climate of the whole continent. The regional groundwater is connected to the surface water and plays an important role for the characteristic flooding regime by regulating the wetland's water table. The water balance assessment of the wetland and the internal water exchange between surface and groundwater is therefore of high relevance for the conservation of the Pantanal biodiversity. Despite of its importance, water balance studies including groundwater–surface water interactions based on field data are rarely undertaken. This is mainly due to the remoteness and difficulty in accessing this area, which results in lack of data. In our study, we developed a new tracer-based model to simulate the spatio–temporal surface and subsurface fluxes for a range of water bodies. The model was able to simulate these fluxes considering a dynamic simulation of inflow and outflow using a newly collected 2-year dataset of water levels, stable water isotopes and chloride collected from several water bodies in the northern Pantanal region. Quantitative differences between water bodies according to their location in the floodplain were determined by the flooding regime and connectivity as well as site-specific characteristics, such as hydraulic conductivity and water depth. Our model simulated water balance fluxes with a Nash–Sutcliffe efficiency of 0.61, whereas it simulated stable water isotopic compositions better than chloride. We present the first study based on field data for the Pantanal, which is able to quantify water balances fluxes. Because their representation in global climate and land cover products is insufficient, our simulation results are valuable for validating large-scale models.