Extreme rainfall analysis is crucial for flood hazard assessment. A common approach is to probabilistically analyze observed annual maximum rainfall values, which are then fitted to an extreme value distribution to estimate the probability of maximum rainfall. However, when continuous time series data are available, it is more effective to use methods that incorporate all rainfall data. Such a method is the point process (PP) method, which analyzes both the frequency of exceedances above a given threshold and the values of those exceedances. This study applies the PP method to uninterrupted daily rainfall records from 1901 to 2023 from the National Observatory of Athens meteorological station in Thiseion. Preliminary analysis using annual maxima did not reveal any significant trend but showed monthly seasonality in precipitation. A threshold rainfall of 10 mm was firstly selected for the PP model by examining the stability of parameter estimates across a range of values. Maximum likelihood estimation was then employed on 1509 rainfall data points over the 10 mm threshold using two different models: (a) a stationary model and (b) one that incorporates the observed seasonality into the shape and location parameters of the PP model. A likelihood ratio test for the two nested models confirmed that incorporating seasonality is statistically significant (p-value < 2.2 ∙ 10^-16). Finally, the results from the PP method were compared to those from the National Flood Risk Management Plans (NFRMPs) derived using the annual maximum approach method. The stationary model’s estimates for various return periods were consistent with the NFRMP values. However, the non-stationary model, which accounted for seasonality, produced estimates up to 16% higher for the upper 95% confidence interval. Therefore, the results of this study indicate that incorporating seasonality (non-stationarity) in the probabilistic extreme analysis of rainfall makes a significant difference in the rainfall design values.