Parameterization of distributed hydrological models: learning from the experiences of lumped modeling

The Hydrology Lab (HL) of the National Oceanic and Atmospheric Administrationʹs National Weather Service (NOAA/NWS), Office of Hydrologic Development (OHD) is currently developing and testing the HL Research Modeling System (HL-RMS). Currently, the system has one snow model (SNOW-17) and two runoff models: the Sacramento Soil Moisture Accounting (SAC-SMA) and the Continuous Antecedent Precipitation Index (CONT-API). The NWS lumped CONT-API model is operational in one of the NWS River Forecasting Centers (RFCs), the Middle Atlantic RFC (MARFC) in the United States. This study deals with the derivation of a priori distributed parameters for the CONT-API model. In our strategy, initial distributed parameters based on the calibrated lumped CONT-API parameters of 67 basins in the Susquehanna River Basin are derived. This study shows that the CONT-API model six-hourly calibrated parameters can be used in an one-hourly lumped model with only minor changes in total runoff volume (less than 5%). However, to obtain the timing of simulated hydrographs, appropriate one-hourly unit hydrographs need to be derived. A priori-distributed model parameters were based on relationships between soil properties and calibrated lumped CONT-API parameters. Multiple linear regressions with coefficients of determination ranging from 0.39 to 0.63 were obtained for 10 lumped model parameters. Using these predicted parameters, the lumped model produced simulations having Nash–Sutcliffe efficiency, Neff, statistics ranging from 0.69 to 0.78 for five of the 67 basins. These are commensurate with goodness-of-fit statistics from lumped model calibrations. Furthermore, application of the method in deriving a priori parameters gave a promising result in distributed model simulations.