| Planning for natural hazards and climate change requires that we develop the best possible understanding of future hydroclimatic conditions. Modeling tools have become essential in meeting this goal, but selecting the most appropriate models and distilling actionable findings from their outputs is still a major challenge. In this research, we investigate this issue from two perspectives: (1) the evaluation of hypothesized natural hazard models, with a focus on predicting flooding events (Chapters 1 and 2); and (2) patterns of agreement and uncertainty in water availability projections derived from a wide array of climate model ensembles (Chapter 3). The first chapter evaluates the appropriateness of traditional metrics of `goodness-of-fit' as measures of the performance of a hypothesized natural hazard model (i.e., the applicability of a selected probability density function). We find that goodness-of-fit can be quite misleading, causing us to reject the correct model and generate potentially large errors in design event (e.g., 1000-year flood) estimation in the process. We propose an alternative metric that gives a more balanced assessment of goodness-of-fit. In the second chapter, we introduce a property called transformational concordance. This property can be used to evaluate whether a hypothesized model and its distributional behavior are consistent with observations. Through our analysis of concordance, we reveal systematic bias in GEV parameter estimation, which is cause for significant concern given the wide application of the model for flood and other natural hazard modeling. The third chapter focuses on improving our understanding of the timing, location, and magnitude of climate change impacts on water needs and availability. Using a wide range of recently available climate model ensembles, we explore the spatial and temporal patterns of inter-model agreement and uncertainty in projected river runoff, irrigation water requirements, and basin storage yield. Cost estimates of adapting global water supply systems are developed for each ensemble, and implications for water management are discussed. |
