Higher Frozen Soil Permeability Represented in a Hydrological Model Improves Spring Streamflow Prediction From River Basin to Continental Scales

Despite plentiful evidence of frozen ground effects on snowmelt infiltration from lab experiments at pedon scales, streamflow observations show a weaker or no effect in terms of timing and magnitude at larger scales. We aim to understand this conflicting phenomenon through modeling using the Noah land surface model with multi-physics (MP; Noah-MP) options and the Routing Application for Parallel computatIon of Discharge (RAPID) over the Mississippi River Basin. We conduct 16 experiments with combinations of two supercooled liquid water (SLW) parameterization schemes and four soil hydraulic property schemes in Noah-MP driven by two gridded precipitation products of the North American Land Data Assimilation System (NLDAS) and the Integrated Multi-satellitE Retrievals for GPM (IMERG) Final. We then use RAPID to route Noah-MP modeled surface runoff and groundwater discharge to predict daily streamflow at 52 United States Geological Survey gauges from 2015 to 2019. A model with the highest permeability performs better than other schemes on daily streamflow predictions by 20%-57% throughout a water year and 29%-113% for the spring as measured by the mean Kling-Gupta Efficiency of the 52 gauges. Different SLW schemes demonstrate negligible effects on streamflow predictions. Models forced by IMERG show a better prediction skill compared with those forced by NLDAS at most of the gauges. Both precipitation products confirm that a scheme of higher permeability yields more accurate streamflow predictions over frozen ground. Future models should represent preferential flows through macropore networks to improve the understanding of frozen soil effects on infiltration and discharge across scales.