| 13 Aug 2025
Spatiotemporal variation in rainfall predictability in Serbia under a changing climate
Abstract. This study examines whether the predictability of precipitation dynamics in Serbia has been influenced by climate change. We apply Generalized Weighted Permutation Entropy (GWPE) to evaluate the temporal structure of daily precipitation series using the parameter q, which filters subsets of small (q < 0) and large (q > 0) fluctuations. The analysis covers data from 14 weather stations between 1961 and 2020.
Entropy values for q = 0 and q = 2, corresponding to Permutation Entropy and Weighted Permutation Entropy respectively, remained stable spatially and temporally. In contrast, GWPE values for q = -10 and q = 10, representing the predictability of small and large fluctuations, exhibited significant spatial and temporal variation between two 30-year subperiods. Entropy values for q = -10 were consistently lower, indicating that small precipitation fluctuations are more predictable than large ones. In several locations, significant changes in entropy occurred despite relatively stable annual precipitation amounts. In others, annual totals varied while entropy remained constant. These findings suggest that climate change has influenced the predictability of precipitation in Serbia. By filtering fluctuations across scales, GWPE effectively reveals underlying changes that may be masked by standard statistical measures.
I find this manuscript and its findings both interesting and valuable for advancing our understanding of the complexity and predictability of rainfall records, particularly in light of the current uncertainty surrounding general rainfall predictions. The methods used for data analysis are appropriate and reproducible, and I especially commend the application of the relatively new entropy-based calculation method. Overall, I recommend this manuscript for publication in EGUsphere.
My questions and comments are as follows:
The use of two climatic time periods for separate analyses is clear. However, I am also interested in whether analyzing the entire available period (1961–2020) would reveal visible changes in the GWPE plots (e.g., crossovers) that might indicate shifts in predictability between the two climatic periods. Could you provide results or insights for the whole period?
The GWPE values are presented with three significant digits. Does this imply that the error of calculation lies in the third digit? More generally, is there a way to estimate or quantify the error associated with this method?
In Figures 3, 8, and 9, would it be possible to display the differences on a color scale distinct from those used for rainfall amount or GWPEC in the two periods? Similarly, in Tables 2 and 3, could you add columns showing the differences in values between the two periods?
For q>0, GWPE values all appear higher than 0.5, except for Sremska Mitrovica during the 1991–2020 period. This suggests a possible pattern—has a similar behavior been observed in other types of real-world data?
Thank you for your responses and for the valuable contribution to this special issue.