A Maxwell-Based Dual-Morphology Wet Snow Microstructure Model for Liquid Water Amount Retrieval in Greenland’s Percolation Zone Using SMAP L-Band Radiometry

Abstract. Liquid water amount (LWA) in seasonal snow controls meltwater retention, refreezing, and runoff. The retrieval of LWA from L-band brightness temperature (TB) observations remains highly sensitive to the effective permittivity of wet snow. Classical wet snow microstructure studies indicate that low liquid water content (LWC) wet snow contains both interfacial/contact-scale liquid and thicker localized pore-space water, whereas existing mixing formulas assume a single liquid water morphology. In this paper, we introduce a Maxwell-based dual-morphology wet-snow model that partitions liquid water between thin-coated interfacial water and localized thicker Tri-continuous (Tri-C) water through a morphology parameter (α) defined as the fraction of the total liquid water assigned to the thin-coated component. A permittivity database generated from computer-generated microstructures and numerical Maxwell solutions is used to train a neural-network emulator for retrieval. For a representative case with LWC = 2 % and snow density 400 kg/m³, varying α from 0 to 1 increases the imaginary part of the effective permittivity by a factor of 37, demonstrating first-order morphology-control on L-band dielectric loss. We combine this model with a two-layer radiative transfer framework and a temporally constrained inversion of the vertically and horizontally polarized SMAP TB at six Greenland percolation-zone AWS sites (CP1, DY2, KAN_U, NSE, SDL, and SDM) to retrieve α, LWC, wet layer thickness, and LWA. The retrieval reproduces the observed morning-pass seasonal evolution and suggests relatively larger thin-coated contributions at melt onset, Tri-C dominance during peak melt, and an incomplete late-season return toward thin-coated dominance, likely due to incompletely refrozen earlier meltwater. These results show that explicit liquid-water morphology can substantially improve L-band retrieval of wet snow compared with traditional mixing models.