The spatiotemporal characteristics of mean and maximum dry spells (DSs) in Croatia were analysed through general climatology and the return values associated with different return periods. An analysis was performed across seven regions with different precipitation regimes comprising data from 131 meteorological stations for 1961-2015. Daily precipitation limits of 1, 5, and 10 mm were used to define dry days, and DSs were defined as consecutive series of dry days. Generalized extreme value (GEV) and generalized Pareto (GP) distributions were applied to the DS annual maxima (AM) and the peak over threshold (POT) series, respectively. The respective distribution parameter methodology and interpretation combined both frequentist and Bayesian approaches. The regional mean DS lengths clearly indicated a dominance of DS durations in easternmost Croatia during the cold season (mainly winter) and in the southern Adriatic during the warm season (mainly summer). Elsewhere, combined climatic influences were reflected, with the shortest DSs found in the highlands. This spatial pattern was more pronounced for DS5 and DS10, whereas differences in DS1 durations were not evident. Spatial DS distribution return values associated with various return periods obtained by the GEV and GP distributions followed the general DS climatology. The visual inspection of an appropriate threshold for GP-POT modelling, using the GP parameter stability graphs and L-moment ratio diagrams, indicated that the 75th percentile value was best for all three DS categories. Although the estimates obtained by the two methods could not be directly compared, the GP model generally found longer DS return values than those obtained by the GEV, particularly in the regions with long DS. This study provides a climatological basis for drought parameters and drought risk assessment in Croatia. Moreover, the results are relevant for adjacent regions, complementing and enriching the current DS characteristics knowledge in southeastern Europe.
