How Much Winter Stratospheric Polar-cap Warming Is Explained By Upward-propagating Planetary Waves In CMIP5 Models?: Part 1. An Indirect Approach Using A Wave Interference Index

Abstract

The breaking of upward-propagating planetary (typically characterized by the combination of zonal wave number 1 and 2) waves in the stratosphere is regarded as one of the factors that provoke the sudden stratospheric warming (SSW) and the accompanying collapse of stratospheric polar vortex during winter. It is also known that if the anomalous stationary wave pattern is in phase with that of the climatology during a certain period, this period is dynamically favorable for the upward propagation and amplification of planetary waves. This kind of phenomenon that amplitude of resultant wave increases by combining two or more waves in phase is called the constructive interference. Our research evaluates whether and to what degree the Coupled Model Intercomparison Project Phase 5 (CMIP5) models simulate such a relation between tropospheric wave interference and Northern polar stratosphere temperature anomaly during winter. Here the 500-hPa wave interference index (WII500) is defined as the coefficient that is obtained by projecting the anomaly of wave number 1 and 2 components of 500-hPa geopotential height onto its climatology. Using monthly outputs of the CMIP5 historical runs currently available to us, we examine the lagged relationship (R-square) between the WII500 during November-December-January (NDJ) and the polar-cap temperature anomaly at 50 hPa (PCT50) during December-January-February (DJF) on an interannual timescale. By sampling uncertainty in R-squares of 33-yr samples (chosen fit with the modern reanalysis period, 1980-2012) with bootstrap resampling, we obtain the sampled medians for all models. The observed relations are then calculated using six reanalyses (ERA-40, ERA-Interim, JRA-25, MERRA, NCEP-R1, and NCEP-R2), and the 5-95% confidence interval of their observed R-square is obtained again with bootstrap resampling of all six reanalyses blended. Then we evaluate which CMIP5 model simulates the WII500-PCT50 relation within the probable range of observed R-squares. It is found that R-squares for 16 (8) of 21 (19) high-top (low-top) models are within the probable range. Our results seemingly suggest that high-top models simulate the WII500-PCT50 relation better than low-top models. Furthermore, it is found that the finer version within a same model suite tends to have the WII500-PCT50 relationship closer to that in the observation.