Study On Pentad-Scale Predictability Of Temperature And Precipitation Over China During Summer Using BCC_CSM1.2 And CFSv2

This paper analyzed multi-pentad prediction skill of temperature and precipitation from the Beijing Climate Center Climate System Model version 1.2(BCC_CSM1.2)and the Climate Forecast System version 2.0(CFSv2)over China from June to August,as well as its connections with Intraseasonal Oscillation(ISO).Further more,using nonlinear dynamical system theory,the predictability of pentad temperature and precipitation was quantitatively estimated.The main results are summarized as follows:(1)By using Temporal Correlation Coefficient(TCC),Anomaly Correlation Coefficient(ACC),and Root-Mean-Square Error(RMSE),the model performances in temperature and precipitation prediction of BCC_CSM1.2 and CFSv2 are assessed,which shows the relatively lower skills in summer both temperature and precipitation prediction whether BCC_CSM1.2 or CFSv2,and the temperature skills are lowest in southern China.The skills of both temperature and precipitation decline with lead time.The effective temperature prediction length is about three pentads(15 days)for BCC_CSM1.2 and four pentads(20 days)for CFSv2.As for precipitation prediction,the effective times are two pentads(10 days)and three pentads(15 days).The above three skills show two same characteristics: the temperature prediction is better than precipitation both BCC_CSM1.2 and CFSv2,and the skills of CFSv2 are higher than those of BCC_CSM1.2 both temperature and precipitation prediction.(2)This paper calculates the ACC between the observed and predicted indices for both the MJO and BSISO,and find that the corresponding lead time of CFSv2 is longer that of BCC_CSM1.2.Meanwhile,the skill of MJO is found to be better than BSISO1 and BSISO2 in terms of ACC and RMSE.(3)The impacts of BSISO1 initial phases are analyzed on the ACC oftemperature and precipitation.The author also research the observation development of temperature and precipitation for pentad 1-6 with some BSISO1 initial phases,in which the prediction skills are relatively higher or lower,and the corresponding model predictions are also showed to verify the performances of these two models.The temperature skill of BCC_CSM1.2 in phase 8 is higher and relatively lower in phase 4,and the difference between them is not significant.On the contrary,the difference of temperature skill between phase 4 and phase 8 is large for CFSv2.The performance of CFSv2 is between than BCC_CSM1.2.Even we focus on phase 8,which is the highest skill phase of temperature prediction for BCC_CSM1.2 but relatively lower for CFSv2,BCC_CSM1.2 can not capture the development of temperature anomaly after pentad 2,and it also can be well reflected in CFSv2 until pentad 3.For precipitation prediction,CFSv2 also performances better than BCC_CSM1.2,which mainly can attribute to the prediction of dry conditions,especially for phase 4.The precipitation skill of BCC_CSM1.2 in phase 5 is higher and relatively lower in phase 4.On the other hand,CFSv2 has the highest skill in phase 4 and the relatively lower skill appears in phase 6.Based on the nonlinear Lyapunov exponent nonlinear error growth dynamics,the spatiotemporal distribution of the pentad temperature predictability limit(PTPL)in China is quantitatively estimated.Meanwhile,the non-normal precipitation is transform into normal standardized precipitation index(SPI),thus calculating the spatiotemporal distribution of the pentad standardized precipitation index predictability limit(PPPL).No matter which season,PTPL is obout 3 pentads(15days),with greater values in high latitude.The spatial distribution of PTPL varies obviously with seasons.PTPL is greater in Autumn and in Winter but less in Spring and in Summer,with the maximum in Autumn and minimum in Summer.Further more,the author divide precipitation into three kinds: drought state,normal state and flood state according to the values of SPI.Later,the annual mean PPPL and seasonal mean PPPL of these three kinds are estimated respectively.It implies that the annual mean PPPL is aboout 2 pentad(10 days),and the values in Summer is lowest.The annual mean PPPL of normal(flood)is highest(lowest)with the normal in the middle.Unlike the lowest annual mean PPPL in flood state,the Summer mean PPPL of flood is relatively higher than other two states.The above research provide some references for further understanding the theoretical basis of subseasonal prediction and improving predictability of S2S models.