Abstract. Recession analysis is a classical method in hydrology to assess watersheds' hydrological properties by means of the receding limb of a hydrograph, frequently expressed as the rate of change in discharge (-dQ/dt) against discharge (Q). This relationship is often assumed to take the form of a power law -dQ/dt=aQb, where a and b are recession parameters. Recent studies have highlighted major differences in the estimation of the recession parameters depending on the method, casting doubt on our ability to properly evaluate and compare hydrological properties across watersheds based on recession analysis of -dQ/dt vs. Q. This study shows that estimation based on collective recessions as an average watershed response is strongly affected by the distributions of event inter-arrival time, magnitudes, and antecedent conditions, implying that the resulting recession parameters do not represent watershed properties as much as they represent the climate. The main outcome from this work highlights that proper evaluation of watershed properties is only ensured by considering independent individual recession events. While average properties can be assessed by considering the average (or median) values of a and b, their variabilities provide critical insight into the sensitivity of a watershed to the initial conditions involved prior to each recharge event.

Traditional risk‐based analysis for levee planning focuses primarily on overtopping failure. Although many levees fail before overtopping, few planning studies explicitly include intermediate geotechnical failures in flood risk analysis. This study develops a risk‐based model for two simplified levee failure modes: overtopping failure and overall intermediate geotechnical failure from through‐seepage, determined by the levee cross section represented by levee height and crown width. Overtopping failure is based only on water level and levee height, while through‐seepage failure depends on many geotechnical factors as well, mathematically represented here as a function of levee crown width using levee fragility curves developed from professional judgment or analysis. These levee planning decisions are optimized to minimize the annual expected total cost, which sums expected (residual) annual flood damage and annualized construction costs. Applicability of this optimization approach to planning new levees or upgrading existing levees is demonstrated preliminarily for a levee on a small river protecting agricultural land, and a major levee on a large river protecting a more valuable urban area. Optimized results show higher likelihood of intermediate geotechnical failure than overtopping failure. The effects of uncertainty in levee fragility curves, economic damage potential, construction costs, and hydrology (changing climate) are explored. Optimal levee crown width is more sensitive to these uncertainties than height, while the derived general principles and guidelines for risk‐based optimal levee planning remain the same.

Surface water and groundwater are intimately connected by a two‐way flux between the stream and underlying aquifers. The National Water Model (NWM) currently only considers a one‐way flux, where groundwater can enter a stream but cannot return to the aquifer. The Northern High Plains Aquifer, USA is used as a case study to investigate the consequences of omitting two‐way stream–aquifer fluxes on streamflow prediction capabilities of the NWM during hydrologic extremes. Instead of traditional field techniques to identify stream–aquifer fluxes, this study presents an integrated approach to classify likely stream regimes using three identification methods: United States Geological Survey (USGS) gage data, simulated stream–aquifer fluxes from an existing USGS Groundwater Availability Model, and the normalized difference vegetation index from remote sensing. For flood events, the modeled flood response for losing streams is characterized by statistically significant earlier peak discharges and an overestimate of the observed flood volume when compared to gaining streams. For drought events, our study found no statistical difference between losing and gaining streams, however, modeled streamflow from the NWM overestimated the observed USGS hydrograph. The systematic overestimate of streamflow by the NWM could be, in part, due to the lack of a losing stream mechanism which was on average 0.1% streamflow loss per streamwise km along the river.