Demonstrating the Added Values of Suction Losses for Channel Infiltration in WRF-Hydro Hydrologic Model

For the southwestern US, the WRF-Hydro hydrological model configured as the NOAA National Water Model (NWM) is challenged to reproduce hydrologic responses. Currently, it does not properly account for infiltration losses (channel transmission losses) from episodic runoff flowing in ephemeral channels into the underlying unsaturated soil. These infiltration losses are an important physical process in the semi-arid domains and have a large impact on the hydrological response. Representing their impact into the NWM is necessary in order to produce realistic streamflow simulations.

In previous work (NWM v1.1 and v1.2) we added a conceptual channel loss infiltration function to the NWM WRF-Hydro Muskingum-Cunge channel routing scheme utilized the Dynamically Dimensioned Search (DDS) calibration algorithm to optimize model parameters, including the channel bed saturated. However, these adjusted and optimized NWM simulations still tend to overestimate streamflow in many semi-arid catchments, thereby affecting the reliability of calibrated parameter sets and thus the performance of the NWM. The focus of the current study is two-fold, first we extend the consideration of channel infiltration by incorporating the effect of suction loss, based on the KINEROS2 hydrologic model representation specifically designed for semi-arid regions. This additional suction term causes the initial infiltration rate to become much larger than the solely using the saturated conductivity of the channel bed to estimate transmission losses. Second, the model is calibrated using distributed stream flow observations, instead of only focusing on the basin outlet.

We examine the efficacy of this advanced infiltration architecture as implemented in NWM v1.2 configuration, using multiple hypothetical numerical experiments by focusing on the 149 km2 USDA-ARS Walnut Gulch Experimental Watershed (WGEW). For WGEW, long-term distributed precipitation and runoff observations are available. These modifications provide more satisfying results as (1) the water balance representation is further improved from a physical perspective and (2) multiple best parameter sets corresponding to different sub-regions in the WGEW can be obtained. This study may also suggest a possible way of the future improvement of WRF-Hydro hydrologic model and demonstrates an example for the real case application.