The hydrological impact of large-scale forest fires in a large basin is investigated on both a daily and an hourly basis. A basin of 877 km2 was chosen, with 37% of its area having been burnt in the summer of 2007. Five models are employed, namely SWAT (semi-distributed), GR4J, GR5J, and GR6J (lumped) for the daily time step, and HEC-HMS (semi-distributed) for the hourly time step. As SWAT and HEC-HMS implement the SCS-CN method, the change in the Curve Number (CN) from pre-fire to post-fire conditions is estimated along with the post-fire trend of CN for both time steps. Regarding the daily time step, a 20% post-fire increase in CN proved necessary for the accurate streamflow prediction, whereas ignoring this led to an underestimation of 22% on average. On an hourly time basis, CN was 95 for burnt areas after the fire, with a mildly decreasing trend after the third year and still above 90 until the fifth year. When neglecting this, peak flow is seriously underestimated (35–70%). The post-fire trend lines of CN for the two-time steps showed statistically equal slopes. Finally, GR models accurately predicted runoff while constraining one model parameter, which proved useful for the realistic prediction of other variables.