The role of intra-watershed processes in improving the understanding of physical processes, hydrologic trends, and restoration measures across the Mobile Bay watershed-AL

Abstract

The southeast United States (SE-US) stands out as a global hotspot of aquatic biodiversity, with the Alabama being a beacon within this rich ecosystem. The Mobile Bay-AL, an expansive estuarine system along the Gulf of Mexico coast, holds national significance under the U.S. Environmental Protection Agency (EPA) National Estuary Program. The Mobile Bay watershed (MBW), covering about 65% of Alabama and parts of neighboring states, is responsible for approximately 95% of the flow, sediment, and nutrient loadings discharged to the Mobile Bay estuary. Past research has raised concerns about declining water quality and streamflow trends across the MBW, posing significant threats to the aquatic biodiversity of the Mobile Bay area and to Alabama's fishing industry. However, to date, there is no quantitative analysis of water quantity and quality trends for this system and the magnitude and drivers of these changes remain poorly understood. This dissertation integrates modeling techniques with remote sensing products and global databases to enhance our understanding of physical processes, hydrologic and water quality trends, and the impacts of restoration strategies across the Mobile Bay watershed. Semi-distributed watershed-scale hydrologic models can offer a comprehensive framework for simulating hydrological and biochemistry processes in the terrestrial-aquatic interface. Despite their utility, these models are subject to uncertainties arising from sources such as input data and model parameterization. The increasing availability of Earth system science information presents a great opportunity to improve the representation of watershed processes often ignored or overlooked in watershed modeling applications, thereby reducing uncertainties associated with model parameterization and input data. By realistically representing intra-watershed processes in modeling frameworks, researchers can improve the realism of model simulations, refine model assumptions, and ultimately enhance the reliability of model predictions, advancing watershed modeling as a science-based approach for decision-making. Throughout four standalone studies, this dissertation explores how we can leverage freely available data from remote-sensing or global databases to better inform hydrologic models in depicting real-world conditions and enhance their reliability as tools for decision-making. The first study investigates the long-term trends in water quantity and quality within the MBW from 1982 to 2020, considering the impacts of climate change (CC) and land-use/cover change (LUC). Utilizing a combination of observed data and modeling techniques, the study reveals a consistent decreasing trend in average annual streamflow across the MBW, alongside varying trends in water quality variables. LUC analysis highlights urbanization and shifts in land cover types, while CC variables such as rainfall and air temperature exhibit increasing trends. Using the Soil and Water Assessment Tool (SWAT) model, the study elucidates the dominant drivers of these changes, attributing decreasing streamflow to CC-induced evapotranspiration increases, and water quality changes to LUC. Results also show that a dynamic land-use (DLU) model configuration outperformed static land-use (SLU) in capturing observed trends, highlighting the importance of incorporating evolving land-use data in hydrological modeling. The second study introduces a model-data approach to incorporate land-atmosphere interactions within watershed modeling frameworks. Leveraging the SWAT model and a global local moisture recycling ratio (LMR) database, the study establishes a relationship between model-simulated evapotranspiration (ET) and LMR. Forest restoration scenarios are then designed based on data from a global tree restoration potential database. Results reveal that forest restoration increases ET and precipitation (P), while reducing runoff, with differing impacts depending on drainage areas and the inclusion of moisture recycling. Despite small magnitude impacts of atmospheric moisture recycling on streamflow predictions, LMR offsets percent streamflow reductions resulting from large-scale forest restoration, highlighting the potential benefits of considering land-atmosphere interactions in watershed modeling in areas highly influenced by atmospheric moisture recirculation and witnessing substantial afforestation rates. The third study addresses the importance of accurately representing canopy evaporation (Ei) in ecohydrological models like the SWAT, particularly in forest ecosystems. Flawed predictions resulting from inaccurate representation of Ei may compromise model outputs related to water availability, soil erosion, nutrient transport, and ecosystem productivity. The study proposes an alternative approach to simulate forest canopy interception and evaporation in SWAT. Using remote-sensing estimates of Ei and plant transpiration (Et) to inform and calibrate the model, the new approach better matched observations of forest Ei, Et, ET, daily streamflow, and ecologically relevant flow metrics compared to the default model. Additionally, the modified model led to reduced sediment, nitrate, and organic nitrogen loadings and improved agreement between model simulated net primary productivity (NPP) and remote-sensing estimates. The fourth study evaluates the representation of bankfull channel width in watershed models and its implications for water quantity and quality estimations. Bankfull channel width has implications for streamflow dynamics, erosion, and nutrient transport, yet its representation in watershed models is often oversimplified. Using the SWAT model, this study finds that the regression equation commonly used in the model's GIS interface (ArcSWAT) can overestimate bankfull channel width by up to three times. The study illustrates how alternative data sources derived from aerial measurements, empirical models, LiDAR, and a global database can be used to better inform the model in capturing bankfull channel width information. Testing the effects of bankfull width on model simulations reveals substantial impacts on maximum flows and water quality.