Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Doctorat → Italie → Compound weather and climate events : dependence and causality issues

Politecnico di Milano (2021)

Compound weather and climate events : dependence and causality issues

Rahimi, Leila

Titre : Compound weather and climate events : dependence and causality issues

Auteur : Rahimi, Leila

Université de soutenance : Politecnico di Milano

Grade : Dottorato 2021

Résumé partiel
Composition of multiple climate drivers or/and hazards characterizes compound weather and climate events. Understanding this kind of events needs to analyze the complex causal chain, which could cause extreme impacts. Estimation of the dependence between characteristics/drivers (random variables) of these kind of events, is in its infant step and of significant importance in the field of hydrology, meteorology and risk assessment. To do this estimation, the combination of multiple climate drivers have to be considered, because the composition of them could push an event to extreme levels by a factor of up to some points, compared with variables being independent. Specially spatial and temporal dependencies as main classes of dependencies are respectively considered to estimate the dependencies of random variables at the same time, and inter-temporal situation. In addition, complex interacted physical processes, cause weather/climate related extreme events in multiple temporal and spatial scales. When drivers combine the impacts of events intensify, especially when they occur in succession / simultaneous (such as drought and heat wave, heavy rainfall and saturated soils or global or regional synchronized floods or heatwaves). Local and short timescales of compound events are felt in different spatial and temporal scales, in addition climate change and local-scale changes are significant issues to how deal with and model non-stationarity of compound events. To model and analyze compound events, understanding the components of compound events is an essential critical issue. Compound events may have modulators, drivers, hazards and impacts. Hazards do not need to be extreme in statistical senses to make an extreme impact. Generally, compound events are distinguished in four classes : 1) Preconditioned, 2) Multivariate, 3) Spatially Compounding and 4) Temporally Compounding, but it is not always easy to identify a strict bound to fit an compound event in a specific class. Separation of each type of compound events is a challenging issue. In this study for each class of compound events a case study research has been done to detect and analyze multiple drivers. To consider a “Preconditioned Events”, the effect of the rainfall intensity on the unsaturated (pre-existing condition) 3D corner slope under historical (1961-2005) and future (2006 and 2100) conditions (under climate change) has been investigated. To do this investigation two types of soil (clayey and sandy) have been considered. Rainfall caused an increment of pore water pressure u in the unsaturated corner slope which led to a considerable reduction in the soil suction and the safety factor. By increasing the daily rainfall intensity, the slopes with high friction angle and low cohesion are more at risk, especially after increasing rainfall intensity under climate change, and safety factor of slopes reduced remarkable. This work is very useful to manage the infrastructure in downstream of slopes, also when climate change also in view. In the “Multivariate class of compound events”, the statistical dependence between flood peak Q, flood volume V and flood duration D has been investigated using a worldwide database of daily discharge. Results of this chapter shed light on the compound nature of flood events rainfall-driven, where the dependence between flood event characteristics (Q, V, D) emerges as a consequence of the relation of such characteristics to the rainfall input variables (I, W) that control the hydrographs. In addition, this result puts light also on the multivariate modeling of flood event characteristics (Q, V, D) stating that there is not a causal priority among these variables to be used in conditional analysis and modeling.


Page publiée le 20 mars 2022