Global Sea Ice Simulation Uncertainty,Parameter Optimization And Application

In this study,simulation bias in sea ice model(CICE)and ice-ocean coupled model(MOM)are analyzed with the satellite observation and the reanalysis data.And based on the characteristics of ice surface flux in MERRA2 reanalysis,we have investigated the different influences of ice-air fluxes and ice-ocean fluxes on sea ice simulation.Meanwhile,the effects of external forcing on sea ice have been discussed.Then we focus on the physical processes in sea ice model,to study their ways of changing sea ice distributions.Sensitivity and importance of different parameters in CICE,as well as the model uncertainty,are also investigated.On the other hand,we have evaluated the prediction capabilities of four models from S2S project.In addition,assimilation of sea ice concentration is applied in both stand-alone sea ice model and ice-ocean coupled model.Conclusions of this paper will be stated briefly blow.Firstly,CICE and MOM can reasonably simulate the seasonal cycle of Arctic and Antarctic sea ice.However,both CICE and MOM have underestimated the summer sea ice area but overestimated the winter sea ice area in Arctic and Antarctic.Sea ice concentration bias is characterized by the marginal anomaly and the inappropriate ice extent.In addition,sea ice volume bias of CICE is positive in Arctic and negative in Antarctic,while MOM has consistently underestimated the ice volume in the two polar regions.For the causes of simulation bias,we find that the bias of Arctic and Antarctic winter ice as well as the bias near north pole in summer are related to ice-air heat fluxes.Sea ice bias in the summer East Siberia Sea and summer Antarctic ocean can be explained by the ice-ocean heat fluxes.Moreover,external forcing can largely affect the distribution of sea ice concentration and thickness,even on regional scale.The response of the melt rate of top ice and bottom ice to external forcing changes is obvious in the whole melting process,while the melt rate of top snow shows large differences only on the last half of its melting period.The improvement of stability functions and roughness length in the surface turbulent fluxes parameterization will weaken the turbulent exchange process.Therefore,the marginal upward sensible heat flux increases and the upward latent heat flux decreases,accompanying with weaker wind stress.These changes finally lead to the decrease of total turbulent heat flux as well as the reduction of sea ice area and volume in polar regions.Consequently,the bias of sea ice thickness in the East Siberia Sea and the Beaufort Sea are reduced.Impacts of mixed layer depth on ice melting and ice growth are not identical in different regions and different seasons.Deeper mixed layer will result in less ice area in both polar regions,but more ice volume in Arctic and less ice volume in Antarctic.Besides,ice bulk salinity is related to ice vertical thermal properties,which can affect the melt rate of sea ice.When ice bulk salinity increases,the melt rate of top ice and bottom ice will be promoted,leading to less sea ice and more melt ponds in summer.Among the major parameters of CICE,tuning parameter for snow,melt pond ratio,temperature change for non-melt to melt snow,liquid fraction of congelation ice,and time scale of ridged ice play the most important roles in sea ice simulation.Finally,the results of predicted sea ice from S2S project show that UKMO and KMA have better sea ice prediction capabilities than NCEP and CNR.Moreover,their forecast error in Arctic sea ice present similar spatial distribution,including the obvious overestimate over Atlantic sector in March,and negative bias over the Beaufort Sea and the Chukotka Sea in September.Optimal interpolation is used to assimilate sea ice concentration in CICE and MOM,and the results show good performance.Although the effects of assimilation don't last for long time,frequently assimilating is able to produce lasting effects.In contrast,frequently assimilating is more suitable to ice-ocean coupled model,as it can evidently suppress the accumulation of sea ice error in MOM.