Forecasting streamflow on a continental scale is a challenge, especially in ungauged basins where in-situ data is rare. We can combine numerical modeling with information from remotely-sensed data to make best-estimates of discharge. The body of this dissertation is split into three chapters. Chapter 2 presents a method for characterizing river channels, a key input for hydraulic models, using information purely from remote sensing data. Chapter 3 describes a high resolution, near-global dataset of inputs to the Variable Infiltration Capacity model, a land surface model that has been widely used for hydrologic forecasting, trend analysis, and coupling with climate models. This dataset, "VICGlobal," represents a major improvement on past land surface parameterizations for VIC, with higher resolution than existing VIC inputs, and greatly simplifies the process for running VIC anywhere in the world. Lastly, Chapter 4 focuses on a procedure called "Inverse Streamflow Routing (ISR)," a method for runoff estimation and discharge interpolation, originally proposed by researchers at Princeton in 2013. We build upon their method and use an update borrowed from the Ensemble Kalman smoother to reduce uncertainty in the a priori runoff estimate.