The aim of my thesis is to study different aspects of hydrological modelling. The study have been performed with the “GR” hydrological models developed by the hydrological research team of IRSTEA, institute where I completed my thesis work. Three basins with different locations and hydro-climates have been analysed. Before running the “GR” models, the hydro-climatic features of the watersheds have been analysed in order to distinguish their nival or pluvial character. Secondly a set of simulations has been performed in order to understand how the different calibrations could affect the simulated discharges: in particular, the components of the Kling-Gupta efficiency goodness-of-fit criteria provided information on the ability to reproduce different aspects and parts of a simulated hydrograph, while the parameters of the models have been studied to understand their sensitivity to the different calibration inputs. Furthermore the role of the inclusion of a snow accounting model has been analysed comparing the streamflows simulation with and without the “CemaNeige” module: the snow melting resulted to be an essential process for the nival basins and neglecting it resulted in poor simulations. Moreover, to demonstrate that the simulation of parts of the hydrograph can be improved through objective functions based on transformations of the output, some tests have been performed and the results showed that high flows were well reproduced without transformation, while calibrating on inverse and square root transformation allowed the best reproduction of low flows and of a wider range of discharges respectively. The last subject analysed was the effect that the presence of a dam can have on the catchment behaviour: few years of observed discharges and of reservoir water levels measured in a fourth basin have been compared. The analysis confirmed that the dam played a role of mitigation of the extreme discharges, reducing the high flows and enhancing the low ones.