Extremely hot summers and cold winters can drastically reduce agricultural production, increase energy consumption, and lead to hazardous health conditions. Quantification of precipitation extremes is important for flood planning purposes. Thus, understanding and predicting the spatial and temporal variability and trends of extreme weather events is crucial for the protection of socio-economic well-being, and also for understanding global warming and mitigating its regional impact.
Tools for statistical modeling of univariate extremes are well-developed. However, extending these tools to model spatial extreme data is an active area of research. One of the challenging issues in spatial extreme value modeling is the need for spatial extreme value techniques in high dimensions, since most of the multivariate extreme value theories only work well for low dimensional extreme values. In this proposal novel general (multivariate, nonseparable, nonstationary, non-Gaussian) spatial-temporal statistical methods for modelling of extreme events are proposed, to produce maps of precipitation and temperature return levels, to estimate trends and variability of extreme weather events, and to provide uncertainty measures.
The results of this research have the potential to become a sound foundation for both studying the impact and planning for mitigation of global warming regional effects. The multivariate spatial tools will enable calculation of probabilities of serious weather related events, such as the combination of temperatures near freezing and heavy precipitation, which can have extremely serious effects on forests, or simultaneous modeling of minimum and maximum temperatures which is the form of information about heat waves needed for realistic impact assessment.