Toward optimal rainfall for flood prediction in headwater basins – improving soil moisture initialization to close the water budget within observational uncertainty

Study region: 28 headwater basins along the latitudinal range of the Appalachian Mountains across diverse hydroclimatic and physiographic regions. Study focus: The objective of this manuscript is to address errors in the initialization of hydrologic models by introducing a physics-based methodology to correct soil moisture (i.e., Initial Condition Correction, ICC) in a manner consistent with the Inverse Rainfall Correction (IRC) methodology proposed by Liao and Barros (2022) to improve Quantitative Precipitation Estimates (QPE). The coupled IRC-ICC framework is demonstrated using a high-resolution hydrologic model for 215 flood-producing events from 2008 to 2024 in 28 headwater basins in the Appalachians. New hydrological insights for the region: Flood simulations using IRC-ICC QPE, and uncorrected QPE products show a median Kling-Gupta Efficiency (KGE, calculated at 15-minute intervals) of 0.86 versus 0.19, reduction of flood peak timing errors with 90 % versus 20 % of events having peak timing errors within 60 min, and median flood volume errors of 2 % versus −17 %. While the average total precipitation shows a modest increase of 6 % due to ICC, the most significant impact is on spatial variance and on the average maximum rainfall that increases by 68 %. This study establishes the coupled IRC-ICC as a robust general framework for orographic QPE correction and provides a pathway to characterizing and modeling soil moisture uncertainty ahead of extreme precipitation events at high spatial resolution O(100 m).

Duke Scholars

Author Ana P. Barros Civil and Environmental Engineering