Stormwater runoff in green roofs (GRs) is represented by the runoff coefficient, which is fundamental to assess their hydraulic performance and to design the drainage systems downstream. Runoff coefficient values in newly installed GR systems should be estimated by models that must be feasible and reproduce the retention behavior as realistically as possible, being thus adjusted to each season and climate region. In this study, the suitability of a previously developed model for runoff coefficient determination is assessed using experimental data, and registered over a 1 year period. Results showed that the previously developed model does not quite fit the experimental data obtained in the present study, which was developed in a distinct year with different climate conditions, revealing the need to develop a new model with a better adjustment, and taking into consideration other variables besides temperature and precipitation (e.g., early-stage moisture conditions of the GR matrix and climate of the study area). Runoff coefficient values were also determined with different time periods (monthly, weekly, and per rain event) to assess the most adequate approach, considering the practical uses of this coefficient. The monthly determination approach resulted in lower runoff coefficient values (0–0.46) than the weekly or per rain event (0.017–0.764) determination. When applied to a long-term performance analysis, this study showed no significant differences when using the monthly, weekly, or per rain event runoff, resulting on a variation of only 0.9 m3 of annual runoff. This indicates that the use of monthly values for runoff coefficient, although not suitable for sizing drainage systems, might be used to estimate their long-term performance. Overall, this pilot extensive GR of 0.4 m2 presented an annual retention volume of 469.3 L, corresponding to a retention rate of 89.6%, in a year with a total precipitation of 1089 mm. The assessment of different time scales for runoff coefficient determination is a major contribution for future GR performance assessments, and a fundamental decision support tool.