Authors: Amin Aghababaei, Norman L. Jones, and Gustavious P. Williams – Brigham Young University mins – University of Vermont
Title: Reliability Aware Multi-Source Fusion for QPE Bias Correction in Radar-Blocked and Gauge-Sparse Regions
Presentation Type: Poster
Abstract: Accurate characterization of baseflow, the groundwater sustained component of streamflow, is essential for water resource planning, drought assessment, and low flow forecasting. Yet tools that combine physically based separation, uncertainty quantification, and operational forecasting at continental scale remain scarce. We present Low Flow Analyst (LFA), an open source, interactive web application that integrates PyBFS, a coupled surface and subsurface reservoir model, with live USGS streamflow data and drought monitoring products across 8,729 gages in the conterminous United States.
PyBFS decomposes total discharge into three components (baseflow, surface flow, and direct runoff) by calibrating a 15 parameter hydraulic model that characterizes hillslope geometry, porosity, saturated hydraulic conductivity, and groundwater recession behavior. Parameter optimization minimizes a composite loss function evaluated over recession and nonrecession periods using the Nash Sutcliffe Efficiency criterion, with recession coefficients derived from observed streamflow hydrograph analysis. Calibrated results are compared against four traditional digital filter methods (Chapman, Lyne Hollick, Eckhardt, and strict baseflow).
LFA extends static baseflow separation in two distinct ways. First, the application provides a 90 day operational forecast by projecting baseflow forward from the last observed hydrologic state using the calibrated model. Second, forecast skill is assessed retrospectively by identifying naturally occurring baseflow dominated sequences through a modified strict baseflow mask, running the model over those periods, and quantifying accuracy with RMSE and MAE across all qualifying sequences. This skill assessment enables spatial comparison of model performance independent of the operational forecast. Results are displayed through an interactive Leaflet map that can be symbolized by baseflow index, forecast error, or drainage area, and a side panel delivers on demand time series, flow fraction summaries, annual baseflow index trends, and county level US Drought Monitor context. LFA provides a scalable, reproducible platform for researchers and practitioners to analyze low flow dynamics and assess forecast reliability at gages across the continental United States.