Satellite gravity measurements as a benchmark for intercomparison of water balance methods and Earth system model performance




Krichman, Benjamin

Journal Title

Journal ISSN

Volume Title



As a driver of both ecosystems and natural disasters, as well as a uniquely critical resource for human civilization, water holds a singular status among natural resources. Accurate quantification of the water cycle is integral to tracking changes in global freshwater resources, understanding drivers of climate, and characterization and forecasting of both acute and protracted adverse events such as flood and drought. The ability to precisely quantify both global and regional water cycles as well as characterize future trends is of paramount importance to humanity, particularly in light of the rapidly changing climate of the 21st century. Numerical models representing varied domains and dynamics have been capable of producing data assimilated representations of the climatic state for the last half century. Herein, numerical models representing various domains of the Earth system are generally referred to as “Earth system models” where specificity is not required. Many of these models have the potential for prognostic use as well as determination of past and current state. However, models vary greatly in their ability to accurately capture the mass flux that drives the water cycle. In order to diagnose relative deficiencies and drive improvements in these models, empirical data and techniques for direct comparison of this data with model outputs is necessary. The Gravity Recovery and Climate Experiment (GRACE) and its successor GRACE Follow On (GRACE-FO) are low Earth orbit remote sensing satellites that comprise measurement systems for precise characterization of spatiotemporal variations in the Earth’s geopotential. The variations measured by the GRACE(-FO) missions are largely the result of mass movement below, on, or above the Earth’s surface, and at the spatial and temporal resolution of these measured variations the majority of this mass movement is the result of liquid water, ice, and water vapor moving through the atmosphere, oceans, land surface and subsurface, and the cryosphere. Thus GRACE(-FO) data may be treated as a measurement of the variability in the mass of the entire water column which is produced in a globally homogeneous manner at a standard interval of one month. The difference between two GRACE(-FO) solutions provides a measure of change in water storage over the water column, which may be directly compared to Earth system model mass flux outputs by means of a water balance. The water balance is a full accounting of mass fluxes and storage changes describing the evolution of water, ice, and water vapor in a closed system. It is an ideal method for comparison between relatively low resolution gravimetry and higher resolution numerical models, and it is the focus of this dissertation. In this dissertation, the methodology of the water balance is analyzed in the context of the performance of land-atmosphere models in accurately capturing mass flux processes of the water cycle. The uncommon surface approach to the calculation of atmospheric moisture flux divergence is formulated and evaluated for its practical value in studies of mass flux in modern atmospheric reanalyses. Contour tracing is evaluated for its use in geographic applications as a means to discretize regional boundaries in order to enable the surface approach in conjunction with modern high resolution model grids. Existing contour tracing algorithms are detailed and evaluated for feature capture in geographic applications as well as speed and effectiveness in practical use, and improvements are proposed to aid in feature capture for geographic applications. The published open source geocontour software is developed and presented as a tool for enabling this type of analysis in future studies. Over 18 hydrological regions in the conterminous United States, water cycle mass fluxes are derived from a land surface model (NLDAS Noah) and an atmospheric reanalysis (NARR) and are evaluated for their accuracy against GRACE(-FO) data by means of various formulations of the water balance. Through uncertainty quantification, GRACE(-FO) data is found to be a suitable empirical measurement that serves as a benchmark for the characterization of model performance in accurate capture of mass flux. The performance of the surface approach in conjunction with modern atmospheric reanalyses is evaluated and found to be at a similar level to the more common volume approach while utilizing significantly fewer model data and enabling a more varied range of studies on atmospheric moisture flux divergence. NLDAS Noah is found to provide better performance in accurate capture of water cycle mass flux than NARR over two decades of GRACE(-FO) data. This disparity in performance is examined across a range of geographic regions and climatic conditions and spatiotemporally localized to show that NARR performance is particularly degraded in regions with low amplitude water cycle variations as well as during heavy drought periods, especially in arid regions. This discrepancy in performance suggests that increasing the quality and quantity of empirical forcing data as well as considering mass conservation in the land-atmosphere system as opposed to the atmospheric system alone are both crucial actions to take in the pursuit of atmospheric reanalyses that accurately capture mass flux


LCSH Subject Headings