|Abstract:||Terrestrial evapotranspiration (ET) is an important eco-hydrologic process the couples the land surface water and energy budgets, links the water, carbon and nutrient cycle, and represents the largest water consumption from agricultural sector. Although advances have been made in monitoring and simulating terrestrial ET in last decades, there are still challenges in reconciling and cross-validating ET observation and numerical model simulation results. In particular, due to human interferences (such as agricultural irrigation), existing knowledge obtained under natural conditions is inapplicable to intensively managed watersheds. Therefore, there is a pressing need to develop hydrologic theory that depicts watersheds as coupled nature-human systems, and to apply knowledge derived from the complex system to validate and diagnose existing hydrologic observations and models, and explore the inter-connects of hydrologic dynamics across scales.
This dissertation focuses on the ET temporal variability as a signature of watersheds as coupled nature-human systems, since ET variability is driven by the climatic fluctuations and modulated by hydrologic processes such as vegetation, snow dynamics and human water use. Based on general hydrologic laws on land surface water-energy coupling, this dissertation derives an Evapotranspiration Temporal VARiance Decomposition (ETVARD) framework for better understanding of both the climatic and hydrologic controls on ET temporal variability. Utilizing best available hydrologic observations, ETVARD quantifies the contributions from the variances and co-variances of climatic and terrestrial water storage change factors to ET variance at various temporal scale (e.g., monthly, seasonal and annual) for watersheds across a wide spectrum of climatic conditions (from humid to arid) under both natural and managed conditions.
As such, we derive hydrologic knowledge from the congruence among theories, observations and models. For multi-variable and multi-source hydroclimatic observations, ETVARD provides an independent diagnosis tool to detect the possible biases and uncertainties in observations and land surface models. Using ETVARD as a benchmark for inter-comparison of observation and models and through five systematically designed experiments, this dissertation identifies the inconsistencies in ET variance estimates among theories, observations and models, assesses the quality of multiple ET products, and provides guidelines to improve land surface model structure in capturing ET variance for the contiguous United States.
In particular, ETVARD identifies the temporal and spatial ET pattern changes due to extensive groundwater-based irrigation through a rea-world case study in the High Plains. The relation between ET and crop yield signatures (i.e., mean and variability) in rain-fed and irrigated crops reflects farmers’ irrigation behavior heterogeneity in the formation of ET patterns, depending on farmers’ preferences between profit-maximization and risk-aversion. In addition, a power-law statistical relationship between ET mean and variability is developed from independent ET observations. While the differences in climate conditions and vegetation structures are reflected by ecosystems’ water use preferences between consumption and variability, these water use preferences cluster on the same a power-law statistical relationship.
The comprehensive assessment on ET variance in this dissertation provides a synthesis from existing theories, observations and simulations towards improved understanding of ET variance at the watershed system level. The knowledge discovered in the dissertation also provides guidelines for conjointly managing the mean and variability of watershed responses to both natural and human driving forces in the context of coupled nature-human systems.