Abstract Hydrologic models often exhibit inaccuracies in representing key hydrological fluxes due to uncertainties arising from the necessary simplification of complex processes and input data. Soil databases, commonly used in hydrological models, vary in format, resolution, and parameter range, leading to diverse approaches for generating soil inputs in process-based models. This study employs both linear (FOSM) and non-linear (iES) methods to quantify parameter and prediction uncertainty. A comparative perspective on how these approaches reflect uncertainty when using different soil databases is provided. The study area is the Mohaka catchment with an area of 2,428 km 2 , situated within the Hawke’s Bay Region of New Zealand. Four different soil databases were used in this study (FSL, S-map, HWSD, and ISRIC) with different spatial resolutions and the number of soil units covering the catchment. Although similar model evaluation metrics were obtained for streamflow simulation using the different soil databases, flow prediction uncertainty varied significantly for average, low, and high flows. For example, low and high flow predictions showed particularly high uncertainties for the global, low-resolution ISRIC database. Conversely, the local soil database S-map produced the lowest uncertainty range for low and high flow conditions. These findings highlight that while different soil databases may yield similar performance statistics during calibration, selecting those that minimise variance in key predictions can improve the reliability of model predictions. The findings emphasise the importance of selecting an appropriate soil database to enhance model reliability for the purpose under consideration.

Abstract This article investigates the temporal dependence structure and co-occurrence probability of compound climate events, specifically concurrent high temperatures and low precipitation, using a 55-year monthly time series (January 1965–December 2019) from the Tulcea meteorological station in Romania. Traditional univariate analyses often fail to capture the interdependent nature of climate variables, potentially leading to a significant underestimation of climate risk, especially in compound extreme events such as warm-dry periods. To address this limitation, the study employs a rigorous copula-based framework. This methodology, unlike correlation-based or single-variable approaches, rigorously separates the modeling of individual variable distributions (margins) from their joint dependence structure. The analysis is structured around two core directions: (Q1) Investigate if the dependence structure between temperature and precipitation changes over time, and (Q2) determine the exceedance of Maximum Temperature and Non-Exceedance of Minimum Precipitation at various thresholds. To answer (Q1), the study applied four statistical tests—including two-sample tests for Kendall and Spearman, a whole-surface Cramér–von Mises (CvM) copula test, and a direct test on finite-quantile probabilities—to compare "early" (1965–1991) and "late" (1992–2019) sub-periods. These tests were extended through seasonal pooling, rolling Mann–Kendall tests, and a changepoint scan to enhance statistical power and temporal localization. For the first time for the Dobrogea region, the findings show a general stability in the dependence structure across most months, with only mixed, non-coherent changes detected, suggesting that the underlying warm-dry coupling has not undergone a significant, sustained shift that coincides with the early/late split. To answer (Q2), the research quantified the empirical probability of a month being simultaneously unusually warm and unusually dry for various thresholds. This empirical probability was then compared against the independence baseline and against predictions from fitted parametric copulas (Frank and a data-selected family). This comparison revealed the degree to which the observed joint probability exceeds the probability expected under independence. This approach provides a practical, threshold-dependent assessment of compound risk.

Abstract Compositional regression models with a real-valued response variable can generally be specified as log-contrast models subject to a zero-sum constraint on the model coefficients. This formulation emphasises the relative information conveyed in the composition, while the overall total is regarded irrelevant. In this work, such a setting is extended to account not only for total effects, formally defined in a so-called $$\mathcal {T}$$ -space, but also for moderation or interaction effects. This is applied in the context of complex spatiotemporal data modelling, through an adaptation of the integrated nested Laplace approximation (INLA) method within a Bayesian estimation framework. Particular emphasis is placed on the interpretation of model coefficients and results, both on the original scale of the response variable and in terms of elasticities. The methodology is demonstrated through a detailed case study investigating the relationship between all-cause mortality and the interaction between extreme temperatures, air pollution composition, and total air pollution in Catalonia, Spain, during the summer of 2022. The results indicate that extreme temperatures are associated with an increased risk of mortality four days after exposure. Additionally, exposure to total air pollution, especially to NO 2 , is linked to elevated mortality risk regardless of temperature. In contrast, particulate matter is associated to increased mortality only when exposure occurs on days of extreme heat.