Statistical Methods for Flood Hazard Models and Crop Yield Projections

Restricted (Penn State Only)
- Author:
- Roth, Samantha
- Graduate Program:
- Statistics
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- May 23, 2024
- Committee Members:
- Bing Li, Professor in Charge/Director of Graduate Studies
Enrique Del Castillo, Major Field Member
Klaus Keller, Special Member
Murali Haran, Chair & Dissertation Advisor
Nicole Lazar, Major Field Member
Robert Nicholas, Outside Unit & Field Member - Keywords:
- multiresolution
multifidelity
Gaussian process
statistical downscaling
statistical bias-correction
computer model calibration
uncertainty quantification
environmental statistics - Abstract:
- My dissertation is motivated by two important problems in environmental risk, riverine flooding and corn yield projections. Adequate preparation for riverine flooding in a changing climate hinges on high spatial resolution flood projections from flood hazard models. Corn yields are also potentially impacted by climate change and therefore require accurate projections based on models of the climate and of the relationship between the climate and yields. My thesis consists of four projects that address statistical challenges in combining complex computer simulation models with observational data. In my first project I propose a new emulation-calibration approach capable of handling multiple spatial resolutions of high-dimensional computer model output. My second contribution is a computationally expedient downscaling approach that provides a probability distribution of flood heights at a high spatial resolution using information from low resolution models. My third project develops a dimension-reduced Bayesian spatial model for relating climate to corn yields. We use this model to project corn yields under climate change to the end of the twenty-first century. My fourth project uses mismatches between historic corn yields and corn yields estimated using climate models in order to improve both the estimates and uncertainties surrounding future corn yield projections.