Evaluating single-column thermodynamic sea ice models for simulating landfast ice in Prydz Bay, East Antarctica

To evaluate the contributions of different physical processes to sea ice thermodynamic simulation uncertainties, four single-column models were employed to simulate the growth and melting of landfast sea ice: high-resolution snow/ice model (HIGHTSI), energy-conserving model (BL99), mushy layer model (MUSHY), and 0-layer model (0LAYER). These simulations were forced by the observations obtained at Zhongshan Station and validated against in situ measurements collected between July and December 2015 in Prydz Bay, Antarctica. The results indicate that all the models exhibit significant biases in terms of the simulated snow depth because snow redistribution by wind is neglected, leading to errors in ice thickness simulations of more than 5 %. Shortwave radiation absorption parameterizations and sublimation processes crucially determine the onset timing of snowmelt. Among all the models, HIGHTSI results in the smallest ice thickness bias from the observations (0.04 m), which is attributed primarily to differences in snow thermal conductivity. This single factor causes BL99, MUSHY, and 0LAYER to simulate 7.3 % thinner ice than does HIGHTSI. Models incorporating coupled growth enthalpy and salinity (e.g., BL99) showed high sensitivity to initial salinity profiles. This study quantifies the relative importance of various physical processes in sea ice thermodynamics through comparing numerical simulations with in situ observations. The results clarify the feasibility of different models and provide solid references for model optimization in the future.