Simulations Of The Climate Changes In The Antarctic Ice Sheet Based On Polar WRF

In context of global warming,as one of the important factors driving global climate,much attention has been drawn on the Antarctic climate changes.As one of the most advanced polar Limited Area Models(LAMs),Polar WRF is a powerful tool for Antarctic climate change diagnosis and forecast.However,it's essential to test the performance of Polar WRF before application because there are still some shortcomings in the simulating climate changes.In this paper,the latest version of Polar WRF(Polar WRF 3.9.1 at a resolution of 50 km were used to simulate Antarctic climate during 2004-2013.This model were forced by ERA-interim reanalysis as lateral boundaries and initial field.Based on meteorological observations from 28 automatic and manual weather staions,we examined the performance of the model for temperature,near-surface wind speed and surface pressure over the Antarctic Ice Sheet on annual,seasonal and monthly time scales.The main conclusions are as follows:1.Polar WRF exhibits a good performance on simulating the 2m temperature in Antarctica.The inter-annual variation trend of simulations during 2004-2013 is broadly similar to the obervations,and the MAE(mean absolute error),RMSE(root mean square error)are 2.33?,2.49?and 0.99,respectively.Furthermore,the model represents inter-annual varations of spring,summer,autumn and winter temperature well.In addition,the RMSE and the MAE are all less than 4 ? in 12 months,the smallest of which are January and December,and the largest are July,October and November.The correlation between simulations and field observations is higher than 0.97 at each month,with the lowest correlation appeared in November and December and the highest correlation occoured in March and April.However,in terms of the multi-year and seasonal mean simulations,a cold bias is found at the coast areas,a warm bias in the inland and both warm bias and cold bias in Antarctic Peninsula and the simulation error in the coastal area is less than that in the interior zones in quantity.In comparison of the simulation accuracy of four seasons,the lowest occurs in winter,followed by spring,the higher present to summer and autumn.Apart from that,the precision of Antarctic coast is smaller than that of the inland areas in spring and summer and similar to inland perision in autumn.Nevertheless there are a high precision over the East Antarctic coast in winter,the accuracy in the Antarctic Peninsula is slightly lower than that in the Inland areas.2.On the whole,the simulations of surface wind speed are consistent with the measurements,and it can reflect the annual,seasonal and monthly wind speed features to some extent,even though larger errors exist in some staions.The interannual variation trend of model results for the period 2004–2013 as the same as the observed data,and the MAE,RMSE and correlations between simulations and obversations are 3.91m/s,4.75m/s and 0.62,respectively.There is less error between the seasonly mean values of observed and modeled wind speed(<4m/s),and the model skillfully captures the inter-annual trend of seasonal wind speed.Meanwhile,the monthly errors ranged from 2.5m/s to 4.5m/s,the maximum and minimum of errors occured in July and November respetively.The correlation was greater than 0.5 in all months,with the lowest of which in January and the highest in November and December.Nevertheless,the multi-year and seasonal mean simulated wind speed are underestimated.In autumn and summer,the simulation errors in coastal areas are smaller than those in inland areas,while in winter and spring,they show slight difference and the errors among individual sites are significantly different.3.The surface pressure is accurately reproduced by Polar WRF at annual,seasonal and monthly scales.The variation trend in simulations agrees well with those from observations in the period of 2004-2013,and the RMSE and MAE and correlations are 0.59 hPa,0.47 hPa and 0.99,respectively.Moreover,trends in the time series of seasonal mean pressure are well represnted by the PolarWRF,and the simulated pressure errors in all seasons are less than 1.8 hPa.Simulation errors are limited to 2.4 hPa in all months,and the smallest errors present to January and November and largest present to August.The correlation coefficients of all months ranged from 0.79 to 0.99,with the lowest in March and the highest in July and September.Whereas,the model overestimates the multi-year average and seasonal average pressure over the Antarctic ice sheet.The lowest error appears in spring,the second one is winter and fllowed by the summer and autumn.In addition,the accuracy of inland area is lower than that of inshore area in autumn,while in winter and spring they are roughly the same,and in summer the accuracy of interior region is lower than that of inshore region.