연구보고서
- 저자
- 김가영 연구원
- 작성일
- 2016.01.23
- 조회
- 279
- 요약
- 목차
The Arctic Oscillation (AO) is well known as the most dominant mode in the Northern Hemisphere. AO is one of the factors which affects the East Asian winter monsoon (EAWM). In this study, we examine the predictability of winter AO using seasonal prediction models from the APEC Climate Center and WMO Lead Centre for Long-Range Forecast Multi-Model Ensemble. Also, we show whether or not the climate models can predict the impact of AO on EAWM.
To represent the prediction skill of climate features in the boreal winter, the climatological means and standard deviations of sea-level pressure (SLP) for observation and individual models were compared. The observed climatology shows the Siberian high and Aleutian low around East Asia. It is the most important pressure pattern which determines the intensity of EAWM. The temperature over the Eurasian continent is relatively cool due to the differential heating between the land and ocean. The anticyclonic flow at the surface, which is induced by the Siberian high, leads predominant cold temperature advection in East Asia. Each model has biases from its own systematic errors, etc. There are positive biases in the Arctic region surrounded by negative biases in most models. Other models have dominant positive biases. In the observed monthly variation during the DJF season, there is more than 5hPa of standard deviation in the North Pacific, Tibetan Plateau, and between the Arctic and Greenland Sea. This means that the monthly variations over these regions are relatively large. The ranges of standard deviation for most of the models are similar with those of the observations, but a few models differ. The variations over the Tibetan Plateau are smaller than those of over the North Pacific or Greenland Sea. Observations and models represent the AO mode as the first of the EOF modes of wintertime SLP over the Northen hemisphere. The observed AO explains 25.2% of the total variance, however the models have much larger variances than the observations. Models can predict AO patterns for not only the annular mode of AO, but also the three centers of action over the North Pacific, Arctic region and North Atlantic, as well. While only 5 models have significant temporal correlation between the observed and simulated AO indices, the others cannot predict the variability of AO indices well. This means that current seasonal climate models can predict the AO patterns but not the sub-seasonal variabilities of AO indices well.
Thus, the 4 best models were selected based on the PCCs (pattern correlation coefficients) and TCCs (temporal correlation coefficients) for the AO patterns and indices. The first mode of the EOF using the best models’ MME (BMME) explains 47.5% of the total variance. However, it is still around twice as large as the observed variance. The intensities of the centers of action over the Arctic and North Atlantic are weaker than those in the observations, while the intensity of the North Pacific is similar to the observations. The TCC of the AO index between those observed and simulated by BMME is 0.70, which exceeds a 99% confidence level. The PCC and TCC of BMME are higher than not only the mean of the prediction skill of the individual models, but also the mean of the skill of the best models.
We then assessed how well BMME predicts the circulation features associated with AO by regressing both the temperature at 850hPa and zonal wind at 200hpa onto the AO index compared to observations. The observed regression field shows the opposite temperature correspondingto the AO phase over Eurasia and North America. BMME represents these features well but the magnitudes are smaller than those of the observations. The jet stream over the North Pacific is weak associated with AO in both the observations and BMME. The weakening of the jet stream in BMME is also smaller than in the observations. Nevertheless, BMME captures the circulation over the Northern hemisphere well. Due to reductions of noise and errors, the variance of the MME is smaller than the variance of a single model, but the prediction skill of the MME is higher than the mean skill of the individual models (Yoo and Kang, 2005).
To examine the predictability of the impact on EAWM in relation to the AO phase, composite fields corresponding to the positive and negative phases of AO were shown. In the observations for the positive AO phase, EAWM is weakened by the warming signal due to the southerlies, weakening East Asian trough, and meridional gradient of the jet stream. The opposite atmospheric patterns can be seen in the negative phase of AO. BMME did not capture the weakening of EAWM in the positive AO, while the patterns of BMME are similar with those of the observations in the negative AO. The PCCs of the negative AO phase are much higher than those of the positive AO. This means that the models cannot reproduce the opposite patterns associated with the AO phases but the models do capture the strengthening EAWM in the negative phase of AO well.
The EAWM had experienced a climate regime shift due to global warming (Koide and Kodera, 1999; Nakamura et al., 2002) and has been weakened since the late 1980s, along with a reduced intensity of the Siberian high (Panagiotopoulos et al., 2005). Analysis of 40 years European Centre for Medium-Range Weather Forecasts (ECMWF) and NCEP/NCAR reanalysis by Wang et al. (2009) suggests that EAWM was strong from 1978-87 and has experienced a significant weakening since 1988c . It is known that the variation of EAWM is closely related to variabilities of mean atmospheric circulation at mid-latitudes and that its changes come from the different variabilities of the Siberian high, Aleutian low and AO. However, we could not consider the climate regime change in the scope of this study. The common hindcast period among the models is not long enough to consider the decadal change of EAWM. It may be useful to investigate the predictability of the impact of AO on EAWM in association with the climate regime shift in further studies.