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

Accueil du site → Doctorat → Grèce → Time series forecasting under distribution shift in environmental applications

University of the Aegena (2022)

Time series forecasting under distribution shift in environmental applications

Mammas, Konstantinos of Anastasios

Titre : Time series forecasting under distribution shift in environmental applications

Auteur : Mammas, Konstantinos of Anastasios

Etablissement de soutenance : University of the Aegena

Grade : Doctor of Philosophy (PhD) 2022

Climate change affects all regions of the world. Ice is melting and the sea is rising. In some regions extreme weather events and rainfall are becoming more frequent while others are experiencing higher temperature and droughts. Heavy rain and other extreme weather events are becoming more frequent as well. This can lead to floods and decreasing water quality, but also decreasing availability of water resources in some regions. The Mediterranean region has been considered as one of the most responsive regions to climate change, based on the results from global climate change projection scenarios. Over the next years it is projected that temperature will increase more in Europe, compared to other regions. According to the IPPC report (AR6 Climate Change 2021), a precipitation decrease is projected during summer in the Mediterranean extending to northward regions. The regions in the northern part of Europe will become wetter while in the southern part it is expected that global warming will result in dryer climate conditions. The increasing water availability in the north and decreasing in the south will have a significant impact on the duration and intensity of water scarcity in already water scarce areas in southern Europe. At the same time, droughts will last longer and become more intense in the southern and western regions of Europe, while drought conditions will become less extreme in the northern and northeastern parts of Europe. The impacts of climate change result in severe economic challenges. Exposing the present economy to global warming of 3°C would result in annual financial losses of 175 billion EUR (2°C : 83 billion EUR and 1.5°C : 42 billion EUR). In the present PhD dissertation, we forecasted meteorological droughts using data-driven methods. Initially, we explored the variability of rainfall in drought prone areas and fitted statistical models to describe meteorological droughts. Furthermore, we built a drought forecasting framework that reduces the bias introduced to the data and consequently to the fitted models under climate change scenarios. To support our experiments, we employed popular modelling techniques, including, statistics, machine learning and deep learning.

Mots clés : (Prediction) time series ; Forecasting ; Environment ; Droughts


Page publiée le 26 novembre 2022