The hydrological regime is the main factor governing the functioning of floodplain rivers. An understanding of its dynamic leads to a better idea of the system behaviour for which proper methods must be used. We analysed the daily water level of the Paraná River during the last century at three-gauge stations using linear and non-linear tools to characterise the hydrological dynamic and to analyse to what extent chaotic behaviour prevails. The three water level time series were characterised as non-linear and non-stationary by power spectrum, autocorrelation function, and surrogate test analyses. A strange attractor was developed when the phase space was reconstructed, having a low dimensional chaos, supported by the correlation dimension, the positive maximum Lyapunov exponents, and the recurrence quantification analysis. In line with this, the system resulted unpredictable with a threshold by sample entropy, and with an intermediate hydrological complexity, while Hurst exponent characterised the system as persistent and with sensitive dependence on initial conditions. In a general overview, all the evidence obtained indicated that the Paraná River behaviour is at the edge of chaos. A latitudinal gradient of decreasing chaoticity was observed as the floodplain extent increases, whereas complexity was highest at the intermediate river station due to the inflow of tributaries with different hydrology. This paper attempts to offer some additional insights into the understanding of hydrological behaviour of floodplain rivers and proper methods to catch their complexity.