Estimating records of daily streamflow at ungaged locations in the southeast United States

Accurate assessment and measurement of streamflow is essential to responsible and sustainable management of the freshwater resources upon which our civilization depends. Streamflow is typically monitored via streamgages, but these provide only a finite record of streamflow in space and time. Continued economic development requires reliable hydrologic information beyond the spatial and temporal limits of physical streamgages. Several methods are developed here to estimate daily streamflow records, at sparsely gaged and completely ungaged locations. Though applicable globally, all tools are developed on and analyses are conducted in unregulated basins in the southeast United States. First, a comprehensive analysis of the spatial structure of daily streamflow shows that the full range of the distribution of streamflow can only be assessed when both positive and negative moment orders are considered. An exploration of the spatial scaling of streamflow indicates that multiple regressions are needed to understand hydrologic response. Coupling these findings with the behavior of flow duration curves (FDCs) provides a hydrologic spatial scaling signature that characterizes regional hydrologic response. In a parallel study, a robust rank-based evaluation technique is introduced as a tool for assessing the performance of 19 alternative estimators of spatially and temporally continuous daily streamflow records. Statistical, direct-transfer techniques are contrasted with mechanistic rainfall-runoff models. Methods for prediction are evaluated in terms of day-by-day performance as well as the ability to reproduce streamflow signatures and other streamflow statistics. Streamflow records estimated with a non-linear spatial interpolation using FDCs are shown to perform better than existing alternatives. Finally, a new method of hydrologic kriging is introduced; this technique leverages the spatial structure of daily streamflow to provide continuous estimates of daily streamflow. Hydrologic kriging is shown to be superior to previous, standard methods, while additionally providing uncertainty estimates and a framework for understanding the mechanistic connections across hydrologic landscapes. Motivating future research, initial analyses of the temporal variability of spatial structure are explored. Coupling of hydrologic kriging and mechanistic rainfall-runoff models is presented as an avenue towards improved hydrologic understanding of ungaged regions.