Advancing Water Cycle Prediction and Projections: An Investigation of the Effects of Model Resolution, Meteorological Forcing Uncertainty, and Land-Use Change Feedback

Abstract

Accurate predictions of water cycle components, including evapotranspiration, runoff, and soil moisture, are vital for sustainable water resource management, risk mitigation, and ecological preservation, particularly in light of ongoing climate change and anthropogenic land-use modifications. Despite substantial advances in climate and hydrological modeling, significant challenges persist due to inherent uncertainties associated with model resolution, meteorological forcing data, and complex feedback mechanisms between land surface processes and atmospheric dynamics. This dissertation systematically investigates key factors influencing hydrological cycle predictions and projections by analyzing the impacts of model resolution, meteorological forcing uncertainties, and land-use feedback mechanisms. The first part of the research focuses on evaluating the role of spatial resolution in hydrological modeling. High-resolution simulations (12.5 km) using the Community Land Model version 5 (CLM5) and Noah land surface model with multi-parameterization options (Noah-MP) were conducted and compared against traditional low-resolution simulations (100 km) across the diverse continental United States. Model performance was rigorously assessed utilizing observational datasets from the National Ecological Observatory Network (NEON). The analyses emphasized critical hydrological variables, such as evapotranspiration, sensible heat flux, runoff, and soil moisture. Findings revealed that high-resolution models exhibited moderate but meaningful improvements in simulation accuracy, evident in higher correlation coefficients and lower normalized root mean square errors (NRMSE). Specifically, evapotranspiration and runoff predictions benefited significantly from increased resolution, though soil moisture improvements were spatially heterogeneous. These findings highlighted that increased resolution alone does not universally enhance model performance, underscoring the necessity of integrating detailed surface representations and accurate forcing data to realize the full potential of high-resolution hydrological models. The second segment of this dissertation addresses uncertainties related to meteorological forcing and their influence on soil moisture dynamics, particularly Soil Moisture Memory (SMM). CLM5 simulations were driven by two distinct meteorological datasets: the Climate Forecast System Reanalysis version 2 (CFSR) and the Global Soil Wetness Project Phase 3 (GSWP3). Comparative analyses indicated substantial sensitivity of SMM to the choice of forcing data, with notable disparities between the two datasets. Simulations utilizing CFSR data demonstrated significantly higher soil moisture memory in tropical and subtropical regions, reflecting pronounced soil moisture-precipitation feedback. Randomizing atmospheric forcing from the CFSR dataset resulted in decreased SMM, aligning results more closely with GSWP3-driven simulations and indicating the crucial role atmospheric persistence plays in shaping soil moisture dynamics. These results underscore the complexity of land-atmosphere interactions and highlight the importance of accurately characterizing atmospheric forcing data in hydrological models to improve predictions of soil moisture and associated feedback processes. In the third part, this dissertation explores the implications of land-use and land-cover changes (LULCC) on regional hydrology and climate, utilizing fully coupled simulations with the Community Earth System Model version 2 (CESM2). Simulations compared two contrasting scenarios: one including historical and projected land-use changes (CESM2-LU) and the other maintaining static land-use conditions (CESM2-noLU). Analysis identified strong regional disparities in hydrological responses to LULCC. In tropical regions such as South America and Africa, deforestation resulted in substantial reductions in evapotranspiration and precipitation, while simultaneously increasing surface runoff and altering soil moisture patterns. Conversely, agricultural expansion and afforestation efforts in mid-latitude regions, particularly North America and Eurasia, were associated with increased local precipitation and evapotranspiration, thus significantly affecting soil moisture dynamics and reducing runoff. These divergent hydrological outcomes further correlated with notable temperature anomalies: deforestation intensified surface warming due to reduced evapotranspiration and enhanced sensible heat flux, whereas mid-latitude agricultural zones exhibited localized cooling driven by moisture availability. These findings emphasize the profound influence of land-use practices on regional climate and hydrology. Collectively, the research presented in this dissertation advances our understanding of hydrological predictability and highlights critical areas for improving model accuracy. Enhanced model resolution, accurate meteorological forcing datasets, and land-use impacts emerge as essential elements for robust hydrological forecasting and water cycle projections. By systematically addressing these factors, the research contributes to refining existing hydrological modeling frameworks and informs better water resource management strategies. This dissertation thus underscores the importance of comprehensive, integrated modeling approaches to predict and manage water resources sustainably under evolving climatic and anthropogenic changes, providing valuable insights for policymakers, resource managers, and climate scientists.