Estimation of atmospheric predictability for multivariable system using information theory in nonlinear error growth dynamics

Abstract

To estimate atmospheric predictability for multivariable system, based on information theory in nonlinear error growth dynamics, a quantitative method is introduced in this paper using multivariable joint predictability limit (MJPL) and corresponding single variable predictability limit (SVPL). The predictability limit, obtained from the evolutions of nonlinear error entropy and climatological state entropy, is not only used to measure the predictability of dynamical system with the constant climatological state entropy, but also appropriate to the case of climatological state entropy changed with time. With the help of daily NCEP-NCAR reanalysis data, by using a method of local dynamical analog, the nonlinear error entropy, climatological state entropy, and predictability limit are obtained, and the SVPLs and MJPL of the winter 500-hPa temperature field, zonal wind field and meridional wind field are also investigated. The results show that atmospheric predictability is well associated with the analytical variable. For single variable predictability, there exists a big difference for the three variables, with the higher predictability found for the temperature field and zonal wind field and the lower predictability for the meridional wind field. As seen from their spatial distributions, the SVPLs of the three variables appear to have a property of zonal distribution, especially for the meridional wind field, which has three zonal belts with low predictability and four zonal belts with high predictability. For multivariable joint predictability, the MJPL of multivariable system with the three variables is not a simple mean or linear combination of its SVPLs. It presents an obvious regional difference characteristic. Different regions have different results. In some regions, the MJPL is among its SVPLs. However, in other regions, the MJPL is less than its all SVPLs.