A SCALE-ADAPTIVE FRAMEWORK TO PREDICT WATER AND NUTRIENT FLUXES ACROSS LAND-WATER INTERFACES

Watershed hydrobiogeochemical (HBGC) models play an important role in quantifying water and nutrient fluxes across land-water interfaces (LWI) and predicting downstream river water quality. However, such models are computationally expensive and require scale-adaptive numerical methods to represent multi-scale, multi-physics watershed processes appropriately. We developed a scale-adaptive framework that melds process fidelity with computational tractability with the overarching goal of computing water and nutrient fluxes across LWI. The framework is a scale-dependent cascade of process integration from a meander to the floodplain to the sub-catchment. First, we develop reactive transport models and derive machine learning-assisted HBGC exchange functions at the LWI for meanders based on their characteristic features such as sinuosity and amplitudes to quantify subsurface geochemical exports. Then, these HBGC exchange functions are integrated with the drainage network. This scale-adaptive framework was tested and applied at the East River Mountainous Watershed, Colorado, a Berkeley Lab's Watershed Function Scientific Focus Area site. The modeling framework demonstrates that meander bends form hot spots of nitrogen species, irrespective of high and low water-level conditions. We further found that these hot spots are produced due to spatial and temporal variations in the river stage, bathymetry, and meander geometry. Our modeling framework sets the stage to quantify river water quality at the watershed scale. Efforts are underway to integrate ML-assisted HGBC exchange functions with the East River Watershed's large drainage network.