Date of Award
Winter 2018
Project Type
Dissertation
Program or Major
Statistics
Degree Name
Doctor of Philosophy
First Advisor
Ernst Linder
Second Advisor
Jennifer M Jacobs
Third Advisor
Philip J Ramsey
Abstract
Future projections of extreme precipitation can help engineers and scientists with infrastructure design projects and risk assessment studies. Extreme events are usually represented as return levels which are equivalent to upper percentiles of an extreme value distribution, such as the Generalized Pareto distribution, which is used for exceedances above a certain threshold.
My dissertation focus is on uncertainty quantification related to estimation of future return levels for precipitation at the local (weather station) to regional level. Variance reduction is achieved through spatial modeling and optimally combining suites of climate model outputs. The main contribution is a unified statistical model that combines the variance reduction methods with a latent model statistical downscaling technique. The dissertation is presented in three chapters: (I) Single-Location Bayesian Estimation of Generalized Pareto Distribution (GPD); (II) Multiple-Location Bayesian Estimation of GPD with a Spatial Latent Process. (III) Spatial Combining of Multiple Climate Model Outputs and Downscaling for Projections of Future Extreme Precipitation.
Recommended Citation
Zhao, Meng, "A combined model of statistical downscaling and latent process multivariable spatial modeling of precipitation extremes" (2018). Doctoral Dissertations. 2430.
https://scholars.unh.edu/dissertation/2430