UCLA

Hydrological prediction depends on both accurate models of system dynamics and observations that constrain hydrologic model predictions. This dissertation investigates how advances in hydrologic modeling and satellite remote sensing can improve forecasting and support water-management decisions across data-rich and data-limited regions. First, I evaluate short-range flood forecasting using the Noah-MP land-surface model (which is the hydrologic core of the U.S. National Water Model) in three California watersheds spanning contrasting hydroclimatic conditions from north to south. Model calibration is well known to be essential for accurate flood forecasts, and I confirmed that careful calibration substantially improved retrospective forecasts (reforecasts) of flood hydrographs and peak flows. In addition, correction of precipitation forecast biases and adjustment of antecedent soil moisture through discharge-informed data assimilation further improved reforecast skill, particularly beyond two-day lead times in the southernmost semi-arid basin among Russian River Basin, Yuba-Feather Basin, and Santa Ana Basin. The resulting Noah-MP forecasts were comparable to, and in several metrics slightly more accurate than, archived operational forecasts provided by the U.S. National Weather Service California-Nevada River Forecast Center. The second part of this dissertation focuses on reservoir monitoring and forecasting using observations from the Surface Water and Ocean Topography (SWOT) satellite mission, a cooperative mission of the U.S. and France launched in late 2022. Across 12 western U.S. reservoirs, SWOT water-surface elevations had a median absolute error below 20 cm, while derived storage errors were generally below 5% compared with in-situ observations. Initializing SWOT-derived storage into a reservoir-operation model enabled temporally continuous storage estimates between satellite overpasses. I then integrated SWOT-derived initial storage, seasonal inflow forecasts, and reservoir-operation modeling into a seasonal forecasting framework. Evaluation at two California reservoirs, New Bullards Bar reservoir and Don Pedro reservoir, showed that storage initialization is most influential at short lead times, whereas inflow uncertainty increasingly dominates at longer leads. Application to Tendaho Reservoir, Ethiopia, demonstrated the potential of combining satellite observations and globally available forecasts to assess seasonal water availability in a setting where continuous in situ observations are limited. Together, these studies show that integrating improved models and observations can advance forecast-informed water management across multiple timescales and observational settings.