| El Ni(?)o-Southern Oscillation(ENSO)is the strongest interannual variability of the air-sea coupled system.Its occurrence often has important effects on global weather and climate anomalies.It is of great significance to study its predictability and improve its prediction skill.The spring predictability barrier(PB)seriously undermine the forecasting skills of ENSO.A new type of El Ni(?)o that is different from the traditional Eastern Pacific(EP)-El Ni(?)o event that has occurred frequently since the 1990 s,namely the Central Pacific(CP)-El Ni(?)o event,further increase the uncertainty of ENSO prediction.It also puts forward higher requirements for ENSO prediction and necessary initial observation.A new data analysis method for predictability dynamics based on the CMIP5 model data is proposed in this thesis,and be used to study the season-dependent PB for two types of El Ni(?)o events.The target observation sensitive areas in the Pacific Ocean for two types of El Ni(?)o are identified,and the particle filter assimilation methods are used to verify the effectiveness of the target observations in the sensitive areas for improving the two types of El Ni(?)o prediction.The main conclusions are as follows:First,a theoretical framework of a new data analysis method for predictability dynamics is developed.Making full use of the characteristics of the climate seasonal cycle and considering the feature of the solution to differential equations of the dynamic system,the thesis improves and develop a theoretical framework of a new data analysis method for predictability dynamics that can use the offline data model to examine the predictability dynamics of the climate events from the perspective of error growth.The advantage of this method is that it only uses offline data of model long-term integration,instead of actual prediction data,to carry out analysis on predictability dynamics.This method can eliminate the interference of model errors and only focus on the effect of initial errors on the uncertainty of forecast results.It can be used to explore the evolution dynamics of initial errors,reveal the initial errors that have great influence on prediction skills,and identify sensitive areas of targeted observation.Second,it is revealed that CP-and EP-El Ni(?)o events have different season-dependent PB,and their occurrence depends to a large extent on some specific initial error structure.Applying the new data analysis method to CMIP5 model data,it is found that the CP-El Ni(?)o prediction often interfered with the season-dependent PB,but the occurrence time of the season-dependent PB is one season later than the spring PB for EP-El Ni(?)o.That is,CP-El Ni(?)o prediction often have summer PB.In terms of the strength of PB,the spring PB for EP-El Ni(?)o prediction is stronger than the summer PB for CP-El Ni(?)o prediction.And the former tend to bring larger prediction errors to the EP-El Ni(?)o prediction.By exploring the initial errors that cause PB for two types of the El Ni(?)o individually,it is found that there are two types of SST initial errors that often trigger their PB.For CP-El Ni(?)o,the first type of SST initial error(denoted as CP-type-1)include a dipole pattern with positive error in the east and negative error in the west of the tropical pacific along the thermocline,a Victoria Mode(VM)-like error pattern in the upper north Pacific and a South Pacific Meridional Mode(SPMM)-like error pattern in the upper south Pacific.While the second type of SST initial error(denoted as CP-type-2)mainly locates in the upper north Pacific,presenting a VM-like patten,but with opposite signs.The evolution of CP-type-1 initial error is similar to the decay of a El Ni(?)o,transitions to the cold phase and the development of a La Ni(?)a event,eventually leading to the negative SST error in the central-eastern tropical Pacific,which affects the CPEl Ni(?)o prediction in terms of intensity and structure.While the evolution of CP-type-2 initial error is similar to the formation and development of a CP-La Ni(?)a event,eventually leading to the negative SST error in the central tropical Pacific,which affects the CP-El Ni(?)o prediction in terms of intensity.For EP-El Ni(?)o,its two types of SST initial errors(i.e.EP-type-1 and EPtype-2)that are likely to cause spring PB have similar structure to CP-type-1 initial errors.And the former has the same sign as CP-type-1,while the latter has the opposite sign as CP-type-1.The evolution of the EP-type-1 initial error is similar to that of CP-type-1 initial errors.That is,it experiences a process similar to the decay of El Ni(?)o and transitions into a La Ni(?)a event,leading to the negative SST error in the central-eastern tropical Pacific.The evolution of EPtype-2 initial errors is similar to a La Ni(?)a event changing from weak to strong phase,also lead to the negative SST error in the central-eastern tropical Pacific.Both EP-type-1 and EP-type-2initial errors both cause the EP-El Ni(?)o event to be underestimated in term of the intensity prediction.Third,the sensitive areas of targeted observation for two types of El Ni(?)o events are determined.After considering the structures of the SST initial errors and their evolution mechanism,the sensitive areas of targeted observation for the two types of El Ni(?)o predictions are determined as three regions,including: tropical western Pacific subsurface area(10°S-10°N,130°E-150°W,105-155m),the upper VM-like region of the North Pacific(20°N-65°N,170°E-100°W,0-95m),and the upper SPMM-like region of the South Pacific(40°S-20°S,150°W-90°W,0-60m).Using particle filter assimilation method,the effectiveness of the target observation in the above three sensitive areas is explored in terms of improving the two types of El Ni(?)o prediction skills.The results indicate that for CP-El Ni(?)o prediction,the effect of assimilating the target observation of the tropical Pacific sensitive area is better than that of the south Pacific sensitive area,and is more better than that of the north Pacific sensitive area.For EP-El Ni(?)o prediction,the effect of assimilating the target observation of south Pacific sensitive area or the tropical Pacific sensitive area are better than that of the north Pacific sensitive area.It indicate that the target observations in the tropical Pacific sensitive area are very important for improving both types of El Ni(?)o predictions.The target observations in the south Pacific sensitive area are more important for EP-El Ni(?)o prediction than for CP-El Ni(?)o prediction.While the target observations in the north Pacific sensitive area tend to have a weaker effect on the two types of El Ni(?)o predictions that that of tropical Pacific and south Pacific sensitive areas.However,the impact of target observations in the north Pacific sensitive area become stronger when they are assimilated simultaneously with the target observations in the tropical Pacific or south Pacific sensitive areas.Assimilating the target observations in three sensitive areas simultaneously has the greatest improvement on the two types of El Ni(?)o predictions.Therefore,in order to distinguish the two types of El Ni(?)o events in the forecast,it is necessary to pay attention to the accuracy of the initial SST in the tropical Pacific and the extra-tropical Pacific(including the south Pacific and the north Pacific),especially in the three sensitive areas. |
PDF Full Text Request