[Zurück]

L. Crocetti:
"Impacts of Climatic Oscillations on Precipitation in an extended Mediterranean area";
Betreuer/in(nen): W. Dorigo; Department für Geodäsie und Geoinformation, 2019; Abschlussprüfung: 06.05.2019.

Oceanic-atmospheric oscillation patterns, described by so-called climate modes, have a strong impact on the variability in the terrestrial water cycle. However, the relation between climatic oscillations and hydrology is not yet fully understood due to uncertainties in the observations and the co-varying behavior of multiple oscillations. A better knowledge about these connections is needed to provide better predictions about climate and hydrology. In this study, the impact of 17 major climate modes on monthly precipitation anomalies in an extended Mediterranean area between 28.5N - 56.5N and 10W - 46E is analyzed. The climate modes are expressed through their corresponding Climate Oscillation Index (COI), used to describe the state of the atmosphericoceanic circulations. A supervised learning approach, called least absolute shrinkage and selection operator (LASSO) regression is used to quantify the in uence of these teleconnection patterns (e.g., North Atlantic Oscillation, East Atlantic West Russia Pattern) on precipitation anomalies. Precipitation is an important component of the hydrological cycle and one of the most dominant climatic drivers for water availability besides potential evaporation. The LASSO regression is a data-driven method that uses automatic feature selection and regularization, which in this study is used, to identify oceanic-atmospheric controls on precipitation anomalies and to disentangle the impact of individual climate modes. The methodology considers cross-correlations in the features, i.e. Climate Oscillation Indices. Time lags ranging between zero and five months are introduced in every feature to account for potential lagged response of precipitation anomalies to ocean-atmospheric oscillations. The LASSO model is fitted for each grid point in two ways. Once, by only using the time series of the grid point and additionally by adding the information of the eight neighboring grid points. Besides using all months of the year to build the model, the analysis is also performed for each season separately. Both of these steps increase the coefficient of determination R2 derived from the LASSO regression and therefore improve the predictive performance of the LASSO model. For validation of the regression models two cross-validations and a significance test using the Benjamini-Hochberg procedure are applied. The results gained by the LASSO regression show that in specific hot spot regions up to 70% of the precipitation anomalies can be explained by the modes of climate variability. Adding the information of the neighborhood into the model increases the explained variance R2 significantly. Analyzing the influence of each COI shows that the signal of the East Atlantic Pattern (EA), East Atlantic West Russia Pattern (EAWR), Northern Annular Mode (NAM), and North Atlantic Oscillation (NAO) have a significant impact in the western parts of the investigated area during wintertime. These results help to improve the general understanding of how the individual climate modes affect different parts of the extended Mediterranean area.

https://repositum.tuwien.ac.at/urn:nbn:at:at-ubtuw:1-122829