Detecting precipitation patterns remains a central challenge in hydrological sciences due to the non-linear nature of atmospheric dynamics and the growing influence of climatic variability. This study investigates the evolution of a 64-year daily precipitation series (1961–2024) at the Tulcea meteorological station (Dobrogea, Romania) and introduces a novel adaptation of the Most Probable Precipitation Method (AMPPM), shifting its application from a regional spatial framework to a temporal one. Shannon Entropy is used as a measure of “climatic disorder. ” Model evaluation incorporates Mean Error (ME), Mean Absolute Error (MAE), and Root Mean Squared Error (RMSE), which here measure structural divergence rather than predictive accuracy. Results demonstrate that the Synthetic Representative Series (SRS) isolates the stable climatic signal, reducing the global coefficient of variation (cv (%) ) to 70. 96% and mitigating extreme skewness typical of coastal convective activity. Seasonal entropy analysis reveals divergence: winter entropy decreases through signal stabilization (minimum 2. 00 bits in March), whereas July–October entropy increases, highlighting previously hidden high-frequency daily oscillations. The aggregated Tot₆4 series achieves a final entropy of 2. 75 bits, confirming a complex, multi-state daily precipitation process. MAE and RMSE values for the SRS (e. g. , October: MAE = 1. 20, RMSE = 4. 53; Tot₆4: MAE = 1. 40, RMSE = 4. 58) indicate that the SRS captures dominant precipitation patterns with minimal deviation, comparable to or better than the moving average approaches.
Alina Bărbulescu (2026) studied this question.