Variation Of Global Sea Ice Simulation Errors And Study Of Its Attributions

In this study,three versions of CICE are used for long-term historical simulation,and quantitatively analyzes the characteristics of their errors from multiple angles such as absolute error,cumulative error,and error of multi-year ice and seasonal ice.Afterwards,the latest version of CICE(CICE6.0)is used to study the causes of errors from three aspects: ice-air interface parameters,sea ice internal temperature and salt parameters,and ice-sea interface parameters.Meanwhile,Monte Carlo sampling is used to optimize CICE6.0.Finally,the ability of each model in CMIP6 to simulate the global sea ice concentration is analyzed,and the spatial and temporal distribution characteristics of the simulation errors in each model of CMIP6 are quantitatively studied.The conclusion is as follows:First,the characteristics of the three versions of CICE simulation errors are quantitatively analyzed.The results show that there is a certain linear relationship between the simulation error and sea ice concentration.And the difference in the distribution of the sea and land between Arctic and Antarctic significantly affects the simulation error in different seas.In general,the simulation error shows a clear increase trend,and the increase trend in Arctic is significantly faster than that in Antarctic.In terms of cumulative error,there are two peaks in the seasonal variation.Among them,the peak of CICE4.0 and CICE5.0 in the first half of the year appears from February to March,and the second half of the year is from August to September.Meanwhile their peak in the first half of the year is moving toward March,and that in the second half of the year is moving toward August,But the peak of CICE6.0 in Arctic is more consistent with the other two versions,but its amplitude in Antarctic is greater.We focus on the simulation errors of multi-year ice and seasonal ice.The errors of the Antarctic seasonal ice simulated by CICE4.0 and CICE5.0 have a tendency to decrease,while that of the multi-year ice and seasonal ice in Arctic and the multi-year ice in Antarctic simulated by are increasing.At the same time,the errors of multi-year ice and seasonal ice simulated by CICE 6.0 have an increasing trend.In addition,three versions of CICE overestimate the seasonal ice and multi-year ice in areas with high sea ice concentrationand and underestimate them in areas with low ice concentration.Secondly,using the CICE6.0 sea ice model to conduct sensitivity test to study the source of sea ice simulation errors and optimize the multi-parameters of CICE6.0.The results of the study show that the ice-air interface parameters mainly affect the summer simulation errors,and they mianly affect the sea ice in the Barents Sea,the central sea area and the Weddell Sea.The influence of the internal salinity parameter of sea ice also mainly occurs in summer,and the spatial effect mainly in the central Arctic sea area.Meanwhile,the internal temperature parameter of affects the Arctic sea ice throughout the year,while the influence in Antarctic is mainly from August to March.And it mianly affect Chukotka Sea,the central sea,Beaufort Sea area,Weddell Sea and Ross Sea.The ice-sea interface parameters basically affect the simulation error for the whole year,and mainly affect East Greenland Sea,Baffin Bay,Barents Sea,Bering Sea and Okhotsk Sea.Finally,by evaluating the ability of 25 CMIP6 climate system models to simulate Arctic sea ice,it is found that Had GEMS-GC31-MM and MPI-ESM-1-2-HAM have the best simulation results in Arctic,while ACCESS-ESM1-5 have the best simulation results in Antarctic.The errors of Arctic sea ice simulated by each model in CMIP6 have obvious spatial non-uniformity,and they have a clear relationship with the polar sea ice density and offshore distance.Among them,there are obvious positive centers in the East Greenland Sea,Barents Sea,Bering Sea,and Okhotsk Sea,while the other marginal seas and central sea areas are mainly negative errors.As far as the Antarctic is concerned,there are mainly few simulations.Through the analysis of the error trend,we found that the excessive Arctic simulated by each model has been alleviated,but the situation where the simulated Antarctica is too small is getting worse.